# BIOLOGICAL HAZARDS IN FOOD

EDITED BY: Maria Schirone, Pierina Visciano, Rosanna Tofalo and Giovanna Suzzi PUBLISHED IN: Frontiers in Microbiology

#### *Frontiers Copyright Statement*

*© Copyright 2007-2017 Frontiers Media SA. All rights reserved. All content included on this site, such as text, graphics, logos, button icons, images, video/audio clips, downloads, data compilations and software, is the property of or is licensed to Frontiers Media SA ("Frontiers") or its licensees and/or subcontractors. The copyright in the text of individual articles is the property of their respective authors, subject to a license granted to Frontiers.*

*The compilation of articles constituting this e-book, wherever published, as well as the compilation of all other content on this site, is the exclusive property of Frontiers. For the conditions for downloading and copying of e-books from Frontiers' website, please see the Terms for Website Use. If purchasing Frontiers e-books from other websites or sources, the conditions of the website concerned apply.*

*Images and graphics not forming part of user-contributed materials may not be downloaded or copied without permission.*

*Individual articles may be downloaded and reproduced in accordance with the principles of the CC-BY licence subject to any copyright or other notices. They may not be re-sold as an e-book.*

*As author or other contributor you grant a CC-BY licence to others to reproduce your articles, including any graphics and third-party materials supplied by you, in accordance with the Conditions for Website Use and subject to any copyright notices which you include in connection with your articles and materials.*

> *All copyright, and all rights therein, are protected by national and international copyright laws.*

*The above represents a summary only. For the full conditions see the Conditions for Authors and the Conditions for Website Use.*

ISSN 1664-8714 ISBN 978-2-88945-113-5 DOI 10.3389/978-2-88945-113-5

#### About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals.

#### Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

#### Dedication to Quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public - and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view. By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

#### What are Frontiers Research Topics?

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: researchtopics@frontiersin.org

# **BIOLOGICAL HAZARDS IN FOOD**

Topic Editors: **Maria Schirone,** University of Teramo, Italy **Pierina Visciano,** University of Teramo, Italy **Rosanna Tofalo,** University of Teramo, Italy **Giovanna Suzzi,** University of Teramo, Italy

The ingestion of food containing pathogenic microorganisms (i.e. bacteria and their toxins, fungi, viruses) and parasites can cause food-borne diseases in humans. A growing number of emerging pathogens, changes of virulence of known pathogens and appearance of antibiotic resistance has recently exposed consumers to a major risk of illness. Also infected people and the environment can spread microorganisms on raw or processed food.

Outbreaks of food-borne diseases are often unrecognized, unreported, or not investigated and particularly in developing countries their agents and sources are mostly unknown. Surveillance and analytical methods aiming at their detection are to be hoped, as well as good strategies to struggle against these threats.

This E-book is subdivided in chapters regarding to pathogenic and spoiling microorganisms, chemical hazards produced by biological agents and food safety management systems.

**Citation:** Schirone, M., Visciano, P., Tofalo, R., Suzzi, G., eds. (2017). Biological Hazards in Food. Lausanne: Frontiers Media. doi: 10.3389/978-2-88945-113-5

# Table of Contents

#### *05 Editorial: Biological Hazards in Food*

Maria Schirone, Pierina Visciano, Rosanna Tofalo and Giovanna Suzzi


Xinjie Song, Shruti Shukla, Gibaek Lee, Sunhyun Park and Myunghee Kim

*49 Structural Analysis and Biological Toxicity of Aflatoxins B1 and B2 Degradation Products Following Detoxification by* **Ocimum basilicum** *and* **Cassia fistula** *Aqueous Extracts*

Wajiha Iram, Tehmina Anjum, Mazhar Iqbal, Abdul Ghaffar, Mateen Abbas and Abdul Muqeet Khan


Bingxuan Liu, Haiquan Liu, Yingjie Pan, Jing Xie and Yong Zhao

*88 Rapid Detection and Isolation of* **Escherichia coli** *O104:H4 from Milk Using Monoclonal Antibody-coated Magnetic Beads*

Mirella Luciani, Tiziana Di Febo, Katiuscia Zilli, Elisabetta Di Giannatale, Gisella Armillotta, Laura Manna, Fabio Minelli, Manuela Tittarelli and Alfredo Caprioli

*95 Tracing Back Clinical* **Campylobacter jejuni** *in the Northwest of Italy and Assessing Their Potential Source*

Elisabetta Di Giannatale, Giuliano Garofolo, Alessandra Alessiani, Guido Di Donato, Luca Candeloro, Walter Vencia, Lucia Decastelli and Francesca Marotta


Jingjiao Li, Feng Xue, Zhenquan Yang, Xiaoping Zhang, Dexin Zeng, Guoxiang Chao, Yuan Jiang and Baoguang Li

*133 Recovery and Growth Potential of* **Listeria monocytogenes** *in Temperature Abused Milkshakes Prepared from Naturally Contaminated Ice Cream Linked to a Listeriosis Outbreak*

Yi Chen, Emma Allard, Anna Wooten, Minji Hur, Ishani Sheth, Anna Laasri, Thomas S. Hammack and Dumitru Macarisin

*139 Biofilm-Forming Capacity in Biogenic Amine-Producing Bacteria Isolated from Dairy Products*

Maria Diaz, Victor Ladero, Beatriz del Rio, Begoña Redruello, María Fernández, M. Cruz Martin and Miguel A. Alvarez


YingWang Ye, Jina Gao, Rui Jiao, Hui Li, Qingping Wu, Jumei Zhang and Xian Zhong

*230 An insight into the isolation, enumeration, and molecular detection of* **Listeria monocytogenes** *in food*

Jodi Woan-Fei Law, Nurul-Syakima Ab Mutalib, Kok-Gan Chan and Learn-Han Lee


# Editorial: Biological Hazards in Food

Maria Schirone\*, Pierina Visciano, Rosanna Tofalo and Giovanna Suzzi

*Faculty of Bioscience and Technology for Food, Agriculture and Environment, University of Teramo, Teramo, Italy*

Keywords: food-borne pathogens, food safety, food spoilage, food management

**Editorial on the Research Topic**

**Biological Hazards in Food**

# HISTORY AND INTRODUCTION

Food safety is one of the main objectives related to public health protection. It is expected to prevent, minimize or eliminate risks on different stages of the food chain and in the meantime maintain, provide, and distribute high-quality food to meet consumer demands. However, millions of cases of food-borne diseases occur every year worldwide (Martinovic et al., 2016 ´ ). Their global impact on health and food quality assurance is well-known even if the full health effects, the kinds of unsafe food, and the economic costs are often undervalued or miscalculated, as well as the outbreaks of food-borne diseases are often unrecognized, unreported, or not investigated. The globalization can lead to a widespread distribution of foods with the introduction of new pathogens strictly associated to a specific geographical area. Nowadays foods travel long distances to be consumed worldwide but only in developed countries consumers are aware of potential presence of food-borne pathogens and surveillance and analytical methods for their detection are really effective, while in developing countries the agents and sources of food-borne diseases are mostly unknown (Wang et al., 2016). In addition, the growing number of emerging pathogens, changes of virulence of known pathogens and appearance of antibiotic resistance has potentially exposed consumers to a major risk of illness and on the other hand food industry has been required to improve the strategies to struggle of these threats.

Food-borne diseases can be caused consuming food or water contaminated by pathogenic microorganisms such as bacteria and their toxins, fungi, viruses, and parasites. Food can be contaminated both at the source as raw material, and during food processing up to storage and distribution. Also people (infected persons or carriers of pathogens) and the environment (food contact surfaces and facilities) can spread microorganisms on raw or processed food.

# FOOD-BORNE PATHOGENS AND FOOD SPOILAGE

Among the major food-borne pathogens there are Listeria monocytogenes, Campylobacter spp., Escherichia coli, Staphylococcus aureus, and with respect to these microorganisms the most surveillance attention from both government agencies and food industry is recommended. These pathogens are different in terms of epidemiology, physiology, host association, and virulence, but for their overall persistence a great and continue monitoring is required. Outbreaks of listeriosis in humans are reported worldwide every year, also with a moderate percentage of fatal cases (Scallan et al., 2011). The potential pathogenic profile and virulence characterization were investigated in L. monocytogenes strains isolated from Chinese retail ready-to-eat food (Wu et al.). The genetic variation and phenotypic characteristics of L. monocytogenes isolates from retail raw foods (Chen et al.), as well as the growth potential of isolates from milkshakes, prepared from naturally contaminated ice-cream scoops linked to a listeriosis outbreak (Chen et al.) were studied

#### Edited by:

*Mehdi Razzaghi-Abyaneh, Pasteur Institute of Iran, Iran*

#### Reviewed by:

*Masoomeh Shams-Ghahfarokhi, Tarbiat Modares University, Iran Ioannis S. Boziaris, University of Thessaly, Greece*

> \*Correspondence: *Maria Schirone mschirone@unite.it*

#### Specialty section:

*This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology*

Received: *16 November 2016* Accepted: *21 December 2016* Published: *09 January 2017*

#### Citation:

*Schirone M, Visciano P, Tofalo R and Suzzi G (2017) Editorial: Biological Hazards in Food. Front. Microbiol. 7:2154. doi: 10.3389/fmicb.2016.02154* and reported in this eBook. The methods for cultural detection, enumeration, and molecular identification of this pathogen in different foods were also exposed (Law et al.).

Human campylobacteriosis is one the most commonly reported food-borne diseases connected with the consumption of dairy and poultry products. Campylobacter jejuni, followed by Campylobacter coli and Campylobacter lari are the most common species associated with human infections. The high genome diversity and plasticity within the Campylobacter genus need of genotypic methods for outbreak investigations (Di Giannatale et al.).

Escherichia coli is a facultative anaerobic microorganism commonly found in human intestine but several strains have acquired virulence traits and can cause illness in humans. Among these strains E. coli O104:H4 has been associated with foodborne diseases recently occurred in some countries in the world. The development of an immune-magnetic separation method, by using of beads coated with monoclonal antibodies specific for the lipopolysaccharide of this pathogen from milk samples, was described in detail by Luciani et al.

Some original papers included in this research topic showed the prevalence of different pathogenic microorganisms such as S. aureus (Yang et al.) and Vibrio parahaemolyticus (Xie et al.) in a particular category of food (ready-to-eat) which do not need further processing before consumption and thus, the identification of microbial contamination is critical for assuring food safety. Vibrio parahaemolyticus has emerged as a major food-borne pathogen in China, Japan, Thailand, and other Asian countries. A total of 72 strains were isolated from clinical (i.e. patients with food poisoning) and food samples in China (Li et al.). The environmental parameters, strain sources and genotypes can affect the strain growth variability of V. parahaemolyticus and these researches can be helpful in predictive microbiology and microbial risk assessment (Liu et al.).

Another microorganism increasing importance as a food poisoning pathogen is Bacillus cereus. Some parameters (i.e., enterotoxin gene sequences, transcription, toxin secretion, and cytotoxicity) were studied in different B. cereus strains isolated from food and food poisoning outbreaks (Jeßberger et al.).

Cronobacter sakazakii is an opportunistic food-borne pathogen linked with life-threatening infections in infants (Song et al.). The virulent characterization (biofilm formation and flagella motility) studied by Ye et al. provided novel insides and better knowledge of the pathogenic mechanism of this microorganism.

Human noroviruses are major contributors to acute nonbacterial gastroenteritis outbreaks. A novel approach to characterize their interaction with receptors or ligands was explored by Niu et al.

If the consumption of unsafe foods can cause food-borne illness in consumers, microbial spoilage of food can be considered a greater issue than safety in terms of economic loss. It is estimated that food spoilage results in the loss of high rates of the total food supply (Pinu, 2016). Some acetic acid bacteria growing in wine can be considered spoilage bacteria because they metabolize ethanol to acetaldehyde by alcohol dehydrogenase and then produce acetic acid by acetaldehyde dehydrogenase (Longin et al.). Spoilage microorganisms such as Lactobacillus buchneri are able to produce acetic acid from lactic acid consumption under anaerobic conditions also in table olives, with consequent loss of their quality. In their review, Medina-Pradas and Arroyo-López reported other microbial metabolites of the product, which can also be considered toxic. Bacteria named macergens, which are responsible for plant tissue maceration by releasing pectic enzymes were described in the review of Aremu and Babalola.

# OTHER MICROBIAL TOXINS AS PUBLIC HEALTH HAZARDS

Some chemical hazards are produced by biological agents and affect people with very different symptoms ranging from relatively mild discomfort to serious and life-threatening illness up to a fatal outcome. Among these hazards, marine biotoxins are produced by harmful algal blooms and can accumulate in bivalve molluscs. They are classified on the basis of their poisoning symptoms or their chemical structures. Regulatory limits and official detection methods have been established by the European legislation for these compounds (Visciano et al.).

The presence of biogenic amines in food and food products can be due to microorganisms (mainly bacteria) through the action of decarboxylases and environmental factors such as temperature, salt concentrations, and pH can influence their formation. Histamine and tyramine cause more severe acute effects in consumers. Some technological factors (i.e., addition of starter cultures, additives, effects of packaging) can be used for controlling their production (Gardini et al.).

Mycotoxin contamination in cereals and derived products represents an important problem for agriculture and food industries because it can cause substantial economic losses, mycotoxicoses in farmed animals and also toxic effects for human health. Aflatoxins are the most serious carcinogenic, hepatotoxic, teratogenic, and mutagenic secondary metabolites produced by fungi at any stage of food production from pre-harvest to storage. In recent years researchers reported new interventions to reduce or prevent the growth of fungi and subsequently mycotoxin production. The use of plant aqueous extracts was tested for their antifungal potential against aflatoxigenic isolates of Aspergillus spp. by Iram et al. and Iram et al.

# FOOD SAFETY MANAGEMENT SYSTEMS

A food safety management system includes both control and assurance activities. Preventive measures aiming at avoiding contamination or outgrowth of microorganisms, as well as at their reduction or elimination, can be physical, chemical, and/or biological (Zwietering et al., 2016).

Equipment and food contact surfaces in food industries can provide a substrate for the development of the socalled biofilm, which represent a microbial community where bacteria can live in an extracellular matrix made of polysaccharides, extracellular DNA, and proteins. Cleaning and disinfection chemical products, such as surfactants and alkali compounds, can be used for the prevention of biofilm formation (Campana et al., 2017). Biofilm can be formed both on abiotic and biotic surfaces (Diaz et al.). The adherence of pathogenic microorganisms to surfaces and the formation of biofilm can be associated also to the production of biosurfactants. Rossi et al. reported that the biosurfactants produced by Salmonella Enteritidis SE86 contributed to adherence to slices of lettuce leaf and decreased the antimicrobial action of sanitizers used to sanitize whole lettuce leaves.

Some bacterial strains able to produce antagonistic molecules used as antimicrobials and preservatives can represent biocontrol agents against harmful or pathogenic microorganisms in food. Lactobacillus plantarum is one of the most versatile species used in food industry as microbial starter or probiotic microorganism. Different strains of this species were shown to produce antimicrobial compounds and bacteriocins (Arena et al.).

Among physical approaches for food safety management, Patrignani and Lanciotti reported the application of high and ultra-high pressure homogenization for microbial inactivation

#### REFERENCES


and food safety purpose, even if they are not yet implemented in food industry, because they do not guarantee the sterilization of food.

### CONCLUSIONS

Food-borne diseases represent a significant problem for individuals, communities and food industries. Many factors are involved in these outbreaks such as a lacking quality of raw material, uncorrected handling of food, poor personal hygiene, and improper holding times/temperatures along the food chain. The articles included in this research topic range from food-borne pathogens and their metabolites/microbial toxins to preventive measures and management approaches to control and reduce/eliminate these hazards for the public health.

All the papers published within this eBook are reported in the references.

# AUTHOR CONTRIBUTIONS

MS and PV drafted the editorial, RT and GS contributed to editorial revision. All authors approved the final paper.

to application of new tools to prevent foodborne diseases. Trends Food Sci. Technol. 57, 188–198. doi: 10.1016/j.tifs.2016.09.016

Zwietering, M. H., Jacxsens, L., Membré, J.-M., Nauta, M., and Peterz, M. (2016). Relevance of microbial finished product testing in food safety management. Food Control 60, 31–43. doi: 10.1016/j.foodcont.2015. 07.002

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Schirone, Visciano, Tofalo and Suzzi. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Technological Factors Affecting Biogenic Amine Content in Foods: A Review

Fausto Gardini1,2, Yesim Özogul<sup>3</sup> , Giovanna Suzzi<sup>4</sup> , Giulia Tabanelli1,2 \* and Fatih Özogul<sup>3</sup>

<sup>1</sup> Centro Interdipartimentale di Ricerca Industriale Agroalimentare, Università degli Studi di Bologna, Cesena, Italy, <sup>2</sup> Dipartimento di Scienze e Tecnologie Agro-alimentari, Università degli Studi di Bologna, Cesena, Italy, <sup>3</sup> Department of Seafood Processing Technology, Faculty of Fisheries, Çukurova University, Adana, Turkey, <sup>4</sup> Faculty of BioScience and Technology for Food, Agriculture and Environment, University of Teramo, Mosciano Sant'Angelo, Italy

Biogenic amines (BAs) are molecules, which can be present in foods and, due to their toxicity, can cause adverse effects on the consumers. BAs are generally produced by microbial decarboxylation of amino acids in food products. The most significant BAs occurring in foods are histamine, tyramine, putrescine, cadaverine, tryptamine, 2 phenylethylamine, spermine, spermidine, and agmatine. The importance of preventing the excessive accumulation of BAs in foods is related to their impact on human health and food quality. Quality criteria in connection with the presence of BAs in food and food products are necessary from a toxicological point of view. This is particularly important in fermented foods in which the massive microbial proliferation required for obtaining specific products is often relater with BAs accumulation. In this review, up-to-date information and recent discoveries about technological factors affecting BA content in foods are reviewed. Specifically, BA forming-microorganism and decarboxylation activity, genetic and metabolic organization of decarboxylases, risk associated to BAs (histamine, tyramine toxicity, and other BAs), environmental factors influencing BA formation (temperature, salt concentration, and pH). In addition, the technological factors for controlling BA production (use of starter culture, technological additives, effects of packaging, other non-thermal treatments, metabolizing BA by microorganisms, effects of pressure treatments on BA formation and antimicrobial substances) are addressed.

#### Edited by:

Andrea Gomez-Zavaglia, National Scientific and Technical Research Council (CONICET), Argentina

#### Reviewed by:

Giuseppe Spano, University of Foggia, Italy Elvira Maria Hebert, National Scientific and Technical Research Council (CONICET), Argentina

> \*Correspondence: Giulia Tabanelli giulia.tabanelli2@unibo.it

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 22 March 2016 Accepted: 21 July 2016 Published: 12 August 2016

#### Citation:

Gardini F, Özogul Y, Suzzi G, Tabanelli G and Özogul F (2016) Technological Factors Affecting Biogenic Amine Content in Foods: A Review. Front. Microbiol. 7:1218. doi: 10.3389/fmicb.2016.01218 Keywords: biogenic amines, fermented foods, lactic acid bacteria, decarboxylase activity, aw, pH, temperature

#### INTRODUCTION

Biogenic amines (BAs) are organic bases, which can be present in foods and can cause several adverse reaction in the consumers. They are produced by microorganisms (mainly bacteria) through the action of decarboxylases (carboxy-lyases EC number 4.1.1.1.), which act selectively on specific amino acids in which they remove the carboxyl group with the formation of the correspondent amine and CO2.

In relation to their amounts and their toxicological effects, the most important BAs in foods are histamine (an heterocyclic amine deriving from histidine), tyramine and 2-phenylethylamine (deriving from the aromatic amino acids tyrosine and phenylalanine, respectively), tryptamine

(heterocyclic BA from tryptophan), putrescine (a polyamine obtained through a direct decarboxylation of ornithine or through the agmatine deiminase pathway, which follow the decarboxylation of arginine to agmatine), and cadaverine (a polyamine derived from lysine; Silla Santos, 1996; Landete et al., 2008a; Tanaka et al., 2008; Marcobal et al., 2012; Wunderlichová et al., 2014). In addition, other polyamines (spermine and spermidine) can be produced with a more complex pathway, which starts from putrescine (Bardócz, 2005; Kalac and Krausová, ˇ 2005).

There are two essential physiological reasons leading to the activation of decarboxylative pathways. From one side, decarboxylation is one of the responses of cells to acid stress and the final balance of the pathway, which consists in the loss of a carboxylic group, contributes to the intracellular (and extracellular) pH increase (Connil et al., 2002; Tanaka et al., 2008; Pereira et al., 2009; Perez et al., 2015). Furthermore, it has been demonstrated that these pathways can bring supplementary energy for the cells by energizing the protonmotive force associated to the membrane (**Figure 1**). In fact, a net positive charge is transferred outside the cell during the exchange between precursor (amino acid) and BA (Molenaar et al., 1993; Konings et al., 1997; Konings, 2006; Wolken et al., 2006; Pereira et al., 2009).

The presence of dangerous amounts of BAs is associated with a relevant growth (>7 log cfu/g) of decarboxylating microorganisms. For this reason, some authors proposed microbial quality indices based on food BA content as indirect indicators of excessive microbial proliferation (Karmas, 1981; Ruiz-Capillas and Jiménez-Colmenero, 2004; Baixas-Nogueras et al., 2005; Özogul and Özogul, 2006; Al Bulushi et al., 2009). Nevertheless, fermented foods requires (by definition) a massive growth of microorganisms which are often responsible for noteworthy BA accumulation, especially during the ripening phase (when selected starter cultures can be replaced by wild strains) or when natural (spontaneous) fermentations are adopted (Suzzi and Gardini, 2003; Ancín-Azpilicueta et al., 2008; Rabie et al., 2011a; Linares et al., 2012).

BAs can be produced both by Gram-positive and Gramnegative bacteria (Landete et al., 2008a; Marcobal et al., 2012; Wunderlichová et al., 2014). Also some fungi (yeast and molds) are involved in BA accumulation (in particular cadaverine and putrescine), but their role is debated and, for many aspects, controversial (Caruso et al., 2002; Gardini et al., 2006; Kiss et al., 2006; Tristezza et al., 2013; Qi et al., 2014).

In general, there is no legal regulation about the BA content in food. This is mainly due to the individual toxicological threshold, which can be extremely variable from few mg/kg in sensitive person to some hundred mg/kg in healthy person (Shalaby, 1996; McCabe-Sellers et al., 2006; Fogel et al., 2007; Hungerford, 2010; Knope et al., 2014). The only exception is the scombroid fish, because its richness in histidine and aptitude to support the growth of decarboxylating microorganisms, for which several national and international authorities define a limit for histamine (Anonymous, 2001; The Australian and New Zealand Food Standards Code, 2002; European Community Commission Regulation, 2005; Prester, 2011). In alcoholic beverages (wine, beer, etc.) the toxicity of BAs is enhanced by the presence of ethanol, an inhibitor of mono amino oxidases (Ancín-Azpilicueta et al., 2008). Nevertheless, the presence of high amounts of BA can be related to scarce microbiological quality of raw materials, during the improper storage of the products, as well as an uncontrolled fermentation (Suzzi and Gardini, 2003; Naila et al., 2010; Linares et al., 2012).

Recently, a qualitative risk assessment concerning BA in fermented foods was conducted in the European Union

(EFSA, 2011). According to this study, histamine was present in detectable amounts in the 85% of the dairy fermented products, 83% of fermented vegetables, 45% of fermented meats, and 73% of fish products. The higher concentrations of that were found in fish sauces, dried anchovies, and cheeses. Tyramine was detected in 75% of dairy products and 83% of fish products, but the higher concentrations were found in meat product (fermented sausages and cured meats) and cheeses. Similar results were characterized in 2-phenylethylamine formation in various food products as well. In addition, putrescine, cadaverine, and tryptamine were most frequently detected in dairy food products.

This review summarizes the current state of knowledge of the effects of the major environmental and process factors on the decarboxylase activity of microorganisms. It also highlights the strategies available to reduce the BA accumulation in food products.

### BIOGENIC AMINE TOXICITY AND DECARBOXYLATING MICROORGANISMS IN FOOD

The decarboxylation process can be catalyzed through two biochemical pathways. The first is catalyzed by naturally occurring endogenous amino acid decarboxylases present in animal or vegetable cells and the second by exogenous enzymes produced by various microorganisms under favorable conditions (Halász et al., 1994).

Natural polyamines represent the main amines found in fresh food products (fish, fruits, vegetable, milk, and meat) where they have a physiological role associated with cell growth and proliferation (Bover-Cid et al., 2014). The intracellular biosynthesis of endogenous amines such as the polyamines spermine and spermidine and other amines like histamine involves the incorporation of aminopropyl groups into their precursor putrescine (Bardócz, 2005).

Exogenous BAs derive from decarboxylases secreted by microorganisms, which are present naturally in food products, introduced by contamination, or also added to foods as a starter culture. The enzymatic decarboxylation process depends on various factors such as the availability of the substrate in free form, the presence of decarboxylase-producing microorganisms and the medium conditions (pH, temperature, O2, etc.). The free amino acid are either naturally present in the food or produced via proteolysis both by endogenous proteases in the raw products and microbial enzymes (Danquah et al., 2012). In fact, proteolysis may play an important role in the release of free amino acids from tissue proteins, which offer a substrate for decarboxylases reactions (Shalaby, 1996).

#### Risk Associated to Biogenic Amines in Food

The presence of BAs in food can constitute a risk to the consumer health (Gram and Dalgaard, 2002; Gram et al., 2002; Özogul et al., 2011). Ingestion of food containing high amounts of BAs is implicated in various pharmacological and toxicological reactions. In fact, BA intake can cause headaches, heart palpitations, vomiting, and diarrhea. Moreover, hypertensive crises have been reported after consuming food containing BAs, such as cheese, wine, beer, and vegetables including sauerkraut, broad bean, banana peel, and avocado (Moret et al., 2005; Maintz and Novak, 2007; Hungerford, 2010).

Once the decarboxylase enzymes are synthesized by the bacteria, BA production can continue even if the bacteria are eliminated from the food product by cooking or other technological treatment. The BAs produced are heat stable and, once formed, are not destroyed by cooking, smoking, freezing, or some other type of preservation techniques (Becker et al., 2001).

Under normal conditions in humans, exogenous amines ingested with food are rapidly detoxified (Hornero-Mendez and Garrido-Fernandez, 1997). The enzymes monoamine oxidase (MAO) and diamine oxidase play an important role in this detoxification process. However, the severity of BA toxicological effects depends on the intake with food, on individual allergy and on the consumption of MAO inhibiting drugs, alcohol, and other food amines (Silla Santos, 1996; Sathyanarayana Rao and Yeragani, 2009).

The BAs with the more severe acute effects for human health are histamine and tyramine.

Histamine causes a symptomatology known as "scombroid fish poisoning" consisting in flushing of the face, neck and upper arms, oral numbness and/or burning, metallic taste, headache, itchy rash, heart palpitations, asthma attacks, hives, gastrointestinal symptoms, and difficulties in swallowing (Lehane and Olley, 2000; Maintz and Novak, 2007; Hungerford, 2010; Knope et al., 2014). This type of food intoxication often results from consumption of scombroid fish (such as tuna, sardines, anchovies, bonito, mackerel, etc.), which are rich in histidine, because of the proliferation of histidine decarboxylative Gramnegative bacteria. However, these BAs can also be found in fermented products (wine, cheese, fish sauce, and fermented meat) where it is mainly produced by lactic acid bacteria (LAB).

Tyramine toxicity is known as "cheese reaction" because it was initially observed following the consumption of cheeses with high level of this BA (Shalaby, 1996). Tyramine has a vasoconstrictor effect and causes dietary-induced migraine, increased cardiac output, nausea, vomiting, respiratory disorders, and elevated blood glucose (Shalaby, 1996; McCabe-Sellers et al., 2006; Scheepens et al., 2010; Stadnik and Dolatowski, 2010; Marcobal et al., 2012). This increase in blood pressure due to tyramine can cause heart failure or brain hemorrhage (Naila et al., 2010). Besides to these effects, tyramine has also been determined to have an effect on the gut microbiota. The adherence of some enteropathogens, such as Escherichia coli O157:H7, to intestinal mucosa is increased in the presence of tyramine (Russo et al., 2012). This type of food intoxication is usually associated with the consumption of fermented foods because LAB are the most efficient producers of tyramine (Shalaby, 1996; Ladero et al., 2012).

As far as other BAs, putrescine and cadaverine have low toxicological properties on their own (Košmerl et al., 2013). However, they could potentiate the effects of histamine and

tyramine toxicity by inhibiting their metabolizing enzymes (Landete et al., 2007; Al Bulushi et al., 2009). Moreover, putrescine and cadaverine can act as a precursor to the formation of carcinogenic N-nitrosamines in the presence of nitrite (Rauscher-Gabernig et al., 2012; De Mey et al., 2014). High concentrations of putrescine have been linked to tumor development such as promotion of CT-26 colon tumor cell growth (Farriol et al., 2001).

Despite their cellular functions, the polyamine agmatine, spermine, and spermidine catabolism and their excess levels can lead to toxicity. It has been demonstrated that spermine and spermidine could decrease blood pressure, inhibit blood clotting and provoke respiratory symptoms and neurotoxicity resulting in renal insufficiency (Pegg, 2013).

Even if 2-phenylethylamine is present naturally in several mammalian tissues such as the brain, it is found in certain foodstuffs (chocolate, cheese and red wine) and has been known to trigger migraine attacks and increase blood pressure (Panoutsopoulos et al., 2004; Souza et al., 2005).

# Main Microbial Groups Involved in BA Production in Foods

Among Gram-negative bacteria, spoilage microorganisms belonging to enterobacteria and pseudomonads are known as the major producers of histamine, cadaverine, and putrescine (Ben-Gigirey et al., 2000; de las Rivas et al., 2007; Pircher et al., 2007; Lorenzo et al., 2010). The genus Photobacterium is often involved in the accumulation of histamine in fish and seafood products, together with Aeromonas hydrophila and enterobacteria such as Morganella morganii, Enterobacter aerogenes, Raoultella planticola, and Klebsiella oxytoca (Morii and Kasama, 2004; Veciana-Nogués et al., 2004; Kanki et al., 2007; Al Bulushi et al., 2009; Fernández-No et al., 2010; Küley et al., 2013). BAs produced by enterobacteria were also found in fermented sausages, meat, and cheeses (Ruiz-Capillas and Jiménez-Colmenero, 2004; Bover-Cid et al., 2009; Linares et al., 2011).

The ability to produce BAs is widespread also among Grampositive bacteria. The decarboxylase activity has been found in strains belonging to the genera Staphylococcus and Bacillus (Landeta et al., 2007; Chang and Chang, 2012). However, the attention has been mainly focused on LAB, which are commonly present in the ripening microbiota of several fermented foods. In fact, LAB can produce histamine, cadaverine, putrescine, but, in particular, they are the most efficient producers of tyramine (Arena and Manca de Nadra, 2001; Pereira et al., 2001; Suzzi and Gardini, 2003; Pircher et al., 2007; Bunková et al., ˇ 2009, 2011; Kuley and Özogul, 2011; Ladero et al., 2012). The tyrosine decarboxylase of LAB is often able to decarboxylate also phenylalanine (producing 2-phenylethylamine) even if with a minor efficiency compared to tyrosine (Marcobal et al., 2006a). This ability was confirmed by several authors which also underlined that phenylalanine was used as substrate by the enzyme only when tyrosine was not available (Beutling and Walter, 2002; Pessione et al., 2009; Bargossi et al., 2015b).

Oenococcus oeni, Lactobacillus hilgardii, Lactobacillus fructivorans, Pediococcus parvulus, Lactobacillus brevis are responsible for amines accumulation in wine. Decarboxylating strains of Lactobacillus curvatus, Enterococcus faecalis, Enterococcus faecium, Lactobacillus fermentum, Lactococcus lactis, Streptococcus thermophilus, and Lactobacillus paracasei were isolated from cheese, meat, and sausage with high BA content (Moreno-Arribas et al., 2003; Smit et al., 2008; Marcobal et al., 2012; Russo et al., 2012; Wunderlichová et al., 2014).

Yeast strains belonging to various species including Saccharomyces cerevisiae, Hanseniaspora uvarum, Candida stellata, Kloeckera apiculata, Metschnikowia pulcherrima, and Brettanomyces bruxellensis are also capable of aminogenesis (Romano et al., 2007). Yeast strains isolated from grapes and wines are able to yield high BA amounts (Caruso et al., 2002). In addition, the fungus Botrytis cinerea is considered a producer of BAs in grape must (Bäumlisberger et al., 2015).

The genetic organization of decarboxylase clusters has been reviewed for tyramine (Marcobal et al., 2012), histamine (Landete et al., 2008a), and putrescine (Wunderlichová et al., 2014).

## ENVIRONMENTAL FACTORS INFLUENCING BIOGENIC AMINE FORMATION

The main environmental factors affecting microbial activities in foods are temperature, salt concentration, and pH. These factors can influence the formation of BAs in two ways. In first instance, they are responsible for the overall metabolism of the decarboxylating cells. In addition, the activity of decarboxylases depends on the same parameters. The optimal values of environmental factors for these two different aspects can be different (Bargossi et al., 2015a) and the final amount of BAs is the result of this double influence. In other words, growth of aminobiogenetic bacteria is an essential but not sufficient condition for BA production (Marcobal et al., 2006b). Moreover, some decarboxylase can maintain their activity independently of the integrity of the microbial cells in a wide range of conditions. This has been demonstrated for tyrosine decarboxylase in lactobacilli (Moreno-Arribas and Lonvaud-Funel, 2001), histidine decarboxylase in S. thermophilus (Tabanelli et al., 2012) and Gram-negative bacteria such as Photobacterium phosphoreum, Photobacterium damselae, M. morganii, and R. planticola (Kanki et al., 2007).

In general, the data reported indicate a great variability in the response of the cell decarboxylase systems to the environmental factors, which reflects differences among species and genus metabolic pathways, experimental conditions, type of matrix (food) considered, but also it is the results of the great heterogeneity characterizing decarboxylase activity, even within strains of the same species.

Even if the environmental factors significantly affect the rate and the entity of BA accumulation, in fermented foods their modulation is often limited by the conditions, which allow the fermentation and ripening processes (in turn linked to the "traditional" features of the products) and by health trends, as in the case of the reduction of NaCl content.

#### Temperature

fmicb-07-01218 August 12, 2016 Time: 11:3 # 5

Temperatures close to the optimum growth values, promoting cell metabolism and proliferation, generally favor the production of BAs, which is often related to the number of cells present in the system. However, the presence of high number of decarboxylating cells is not a sufficient condition to explain the final BA amount (Marcobal et al., 2006b, 2012).

Studies carried out in a model system using a E. faecalis EF37, demonstrated that the increase of temperature from 16 to 44◦C coincided with a faster growth and with a more rapid and intense accumulation of tyramine (Gardini et al., 2001). More recently, Bargossi et al. (2015a) studied the effect of temperature on the activity of a pure commercial tyrosine decarboxylase extracted from E. faecalis and found the highest decarboxylation efficiency at a temperature comprised between 30 and 37◦C. By contrast, Zhang and Ni (2014) found that a tyrosine decarboxylase from L. brevis had its optimum temperature at 50◦C, but it was rapidly inactivated at higher temperature. However, the activity of the enzyme at the optimum temperature was rapidly decreased during the permanence at 50◦C for an hour.

Cells of E. faecalis and E. faecium strains suspended in buffered systems containing tyrosine and incubated at different temperatures confirmed the maximum decarboxylase activity at 37◦C after 2 h of incubation. However, after 24 h the maximum tyramine content was surprisingly found in the sample incubated at the less favorable temperature (20◦C; Bargossi et al., 2015a).

Using an experimental design in which several parameters were taken into consideration, Marcobal et al. (2006b) found that the optimum temperature for tyramine production under aerobic condition by E. faecium and L. brevis was 32◦C. By contrast, under anaerobic conditions, the maximum tyramine concentration was obtained at 22.0–24.5◦C.

The histidine decarboxylase of a cell free extract of a S. thermophilus strain had its maximum activity at 50◦C. It decreased rapidly at higher temperature (60◦C), while it maintained a detectable activity at 5◦C (Tabanelli et al., 2012). By contrast, active cells of the same strain produced more rapidly histamine at 40◦C, while the BA accumulation was limited or negligible at 25 and 20◦C within the incubation period considered. In addition, an histamine producing strain (S. thermophilus) incubated at low-temperature (4◦C) in milk produced less histamine than did the same strain kept at 42◦C. This reduction was attributed to a lower activity of the histidine decarboxylase enzyme rather than to a reduction in gene expression or the presence of a lower cell number (Calles-Enríquez et al., 2010).

Regarding Gram-negative bacteria, Morii and Kasama (2004) studied in P. phosphoreum cell free extract the activity of two different histidine decarboxylases and found that the inducible enzyme had its maximum activity at 30◦C while the constitutive one at 40◦C. The optimum growth temperature of the strain was 25◦C while the specific activity of histidine decarboxylase of cell free extracts was extremely high in the cells grown at low temperature (7◦C). The histamine producing potential of the Gram-negative bacterium Mycobacterium psychrotolerans was studied in relation to the temperature in the range 0–20◦C. Increasing temperature enhanced the rate of accumulation but also the final tyramine accumulation (Emborg and Dalgaard, 2008). The optimum temperature for histidine decarboxylase activity of bacteria belonging to different species (M. morganii, R. planticola, P. phosphoreum, and P. damselae) ranged between 30 and 40◦C. It was still active at 5◦C but not at 60◦C (Kanki et al., 2007).

Scarce reports are available concerning the relation between other BAs (putrescine, cadaverine, and tryptamine) and temperature. Generally, the accumulation of BAs, among which cadaverine and putrescine increased with temperature (Wunderlichová et al., 2014); nevertheless, prolonged storage at low temperatures can result in accumulation of putrescine explained by the metabolism of psychrotrophic pseudomonads (Paulsen and Bauer, 1997).

Bubelová et al. (2015) studied the production of putrescine and cadaverine in relation to temperature in Serratia marcescens. They found that the maximum amount of these two BAs was reached at 20–30◦C. If the production was compared to the biomass ("yield factor"), the decarboxylase activity of the single cells was maximum at 10◦C.

In general, the ability to produce BAs is limited by the decreasing of the temperature. This implies that the control of the cold chain during storage and commercialization is a main tool to avoid the accumulation of undesired products after manufacturing, especially in not fermented foods, such as fishery products (Knope et al., 2014). Several authors stressed the crucial effect of the storage temperature on the histamine formation (and other BAs) in fish such as tuna (Veciana-Nogués et al., 2004; Emborg and Dalgaard, 2006; Kanki et al., 2007) and anchovies (Veciana-Nogués et al., 1997; Visciano et al., 2007).

Regarding fermented foods, the temperature of fermentation and during ripening has to allow the microbiological activity of the desired microbiota and the range within they can be modulated is rather strict, defined by the protocols for the production of the different fermented food typologies. The temperature applied during the first 3 days of fermentation of dry sausages influenced the BA accumulation (tyramine, 2 phenylethylamine, cadaverine, and putrescine) during all the ripening period (1 month) with increasing values in the presence of higher temperature, which was between 15 and 25◦C (Gardini et al., 2008; Bover-Cid et al., 2009). Higher fermentation temperatures (and higher relative humidity) favored tyramine and 2-phenylethylamine accumulation in the Spanish fermented sausages Fuet and Llonganissa inoculated with L. curvatus (Latorre-Moratalla et al., 2012). By contrast, no differences were found in Turkish sausages ripened at 22 or 26◦C (Gücükoglu and ˘ Küplülu, 2010).

Ruiz-Capillas et al. (2007) found that tyramine was accumulated in higher amounts in pressurized sliced cooked ham packaged under vacuum when the storage temperature was higher. Also fermented sausages stored after production under room temperature were characterized by higher BA content than refrigerated products (Komprda et al., 2001).

A higher tyramine content was found in salted duck inoculated with E. faecalis stored at 20◦C compared with the samples incubated at 4◦C; however, no differences were found in the tyrosine decarboxylase gene expression indicating an univocal effect of temperature on enzymatic activity (Liu et al., 2014). Also in green fermented olives, the adoption of a low fermentation temperature reduced the accumulation of cadaverine and tyramine together with the formation of "zapatera" defect (García García et al., 2004).

The application of thermal treatments (when possible) to raw material such as milk before fermentation (pasteurization) can contribute to the elimination of the wild decarboxylating microbiota. The Gram-negative BA producers (enterobacteria and pseudomonads) are rapidly inactivated by temperatures higher than 60◦C. LAB are more resistant and require more drastic thermal treatments. For this reason, usually cheeses from pasteurized milk are characterized by lower BA content (Schneller et al., 1997; Novella-Rodríguez et al., 2003; Novella-Rodríguez et al., 2004; Marino et al., 2008). However, Ladero et al. (2011) observed the survival in skim milk of a tyraminogenic strain of L. curvatus after a treatment at 78◦C. Independently of the cell viability, the information concerning the thermal stability of decarboxylase is scarce. The histamine decarboxylase produced by S. thermophilus maintained a residual activity after treatments at 70 and 75◦ for 10 min (Tabanelli et al., 2012). In any case, the pasteurization of milk does not avoid the presence of a ripening microbiota, which can contribute to BA accumulation. Pinho et al. (2001) found a relevant increase in BA content of a Portuguese cheese (Azeitao) when the storage temperature increased from 4 to 25◦C, related to the higher protease activity. In addition, it has been observed that the BAs content was higher in Dutch type cheese when the temperature of ripening and storage increased (Bunková et al., 2010 ˇ ; Pachlová et al., 2012).

#### Salt Concentration

In general, increasing salt concentrations contribute to the reduction of BA accumulation in foods, mainly reducing the metabolic activities of decarboxylating microorganisms. In particular, Gram-negative bacteria are more inhibited by increasing salt concentrations than Gram-positive microbiota. However, the health trend to reduce NaCl concentration is in contrast with this possible tool to reduce BA accumulation in foods.

The pure tyrosine decarboxylase tested by Bargossi et al. (2015a) demonstrated a limited loss of its relative activity in the presence of increasing amount of NaCl up to 10%. Only the addition of 15% of salt determined a more marked reduction of the decarboxylation potential, which remained, however, higher than 50% of the activity recorded in the absence of salt added.

In an E. faecalis strain (EF37) grown on a synthetic medium, the ability to accumulate tyramine and phenylethylamine was inversely related to the NaCl concentration, in a range comprised between 2 and 6%, while the proteolysis showed an optimum between 2 and 3% (Gardini et al., 2001). In fermented sausages inoculated with the same tyraminogenic E. faecalis strain, increasing amounts of salt reduced the concentration of tyramine, 2-phenylethylamine produced by enterococci, but also limited cadaverine and putrescine production by enterobacteria (Gardini et al., 2008; Bover-Cid et al., 2009).

The tyramine production of different strains of E. faecalis and E. faecium was studied in buffered systems containing tyrosine; the results indicated that E. faecalis partially reduced its tyraminogenic potential of cells passing from 0 to 5% of NaCl but the decarboxylation activity did not change significantly increasing NaCl concentration up to 15%. On the other hand, the same enzymatic activity in cells of E. faecium remained quite constant independently of the NaCl concentration (Bargossi et al., 2015a). Strains of Lactococcus lactis ssp. lactis and L. lactis ssp. cremoris reached their maximum tyramine production in synthetic medium only in the presence of the higher salt concentration used in the trials (2%; Bunková et al., 2011 ˇ ); in addition the same condition allowed the maximum tyramine production rate as well as the lowest time for the production of detectable amounts of the BA.

A different role of salt on the histidine decarboxylase activity in a strain of S. thermophilus was highlighted in viable cells and in cell free extract. While the production of histamine was almost completely prevented in living cells by a salt concentration of 2.5% (by limiting or inhibiting the growth potential of S. thermophilus), the activity of the decarboxylase in cell free extract was unaffected up to 5% NaCl and then slowly decreasing, maintaining an activity, even if reduced, at 20–30% NaCl (Tabanelli et al., 2012). The presence of NaCl led to an up-regulation of histidine decarboxylase gene in the same strain grown on skim milk, suggesting a potential role of this enzyme also in osmoprotection mechanisms (Rossi et al., 2011). This confirmed that the activation of decarboxylase systems is a part of complex metabolic responses in the presence of different stress conditions (Pessione et al., 2009).

A halophilic LAB strain of Tetragenococcus muriaticus isolated from fish sauce produced histamine during the late exponential growth phase, reached a maximum production of this BA at 5– 7% of NaCl, and was able to maintain a histidine decarboxylase activity also in the presence of 20% of salt (Kimura et al., 2001). On the other hand, the histamine decarboxylase activity of two strains of P. phosphoreum decreased rapidly with the increase of salt (2–5% of the optimum activity in the presence of 10% of NaCl), while more resistant (40–50% of the optimum activity) resulted the activity of the same enzyme in strains of R. planticola, P. damselae, and M. morganii (Kanki et al., 2007). Using a strain of P. phosphoreum, Morii and Kasama (2004) found that histidine decarboxylase activity in cell free extract was higher at level of 5% NaCl, while the cells were not able to multiply (and produce BA) under this salt concentration. Similar effects were described also for Bifidobacteria (Lorencová et al., 2014). The hypothesis of this possible enhancing effect on the BA production of NaCl has been found by some authors in the essential role of Na<sup>+</sup> ion in the sodium/proton antiport system through which H<sup>+</sup> ions are removed from the cell (Pereira et al., 2009; Bunková et al., 2011 ˇ ; Lorencová et al., 2014; Bubelová et al., 2015).

Serratia marcescens produced putrescine and cadaverine with more efficiency in the presence of 1–3% NaCl (3–5 in the yield factor was applied; Bubelová et al., 2015). Cells of M. psychrotolerans produced more histamine when grown in the

presence of 4% NaCl compared with lower salt concentration (Emborg and Dalgaard, 2008) even if this salt concentration slowed the growth of the strain. In other words, stressed cells seem to activate the decarboxylating pathways in the framework of more complex response systems. This make the potential of BA production by each single cell more efficient. Also E. aerogenes produced the maximum amounts of cadaverine, putrescine, and histamine in the presence of the 3% of NaCl (Greif et al., 2006).

In fermented sausages, BAs are accumulated during ripening. However, the rate of accumulation decreases with the decrease of a<sup>w</sup> due to the water losses. Products packaged under modified atmosphere packaging (MAP), in which the weight losses were inhibited, continued to accumulate BAs when the packaging was carried out at high a<sup>w</sup> (0.92 and more; González-Tenorio et al., 2013; Tabanelli et al., 2013). The Greek cheese Feta, characterized by a high salt content, with a ripening carried out in brine and with a low pH, was characterized by a noteworthy amine concentration (about 200 mg/kg of tyramine, 90 mg/kg of histamine, and 200 mg/kg of putrescine; Valsamaki et al., 2000).

#### pH

Since the decarboxylation is a mechanism of cells to counteract acidic stress, it is clear that several studies were focused on the study of the relationships between pH and BA accumulation. Also in this case, the effect of pH is different if the focus is directed toward the activity of the pure enzyme or the decarboxylase activity of the living cells. In any case, it has been extensively demonstrated that the transcription of genes of many decarboxylase clusters are induced by low pH and improves the fitness of cells subjected to acidic stress (Pereira et al., 2009; Pessione et al., 2009; Marcobal et al., 2012; Romano et al., 2012, 2014; Perez et al., 2015)

A commercial pure tyrosine decarboxylase had its maximum activity in buffered systems at pH between 5 and 6 (Bargossi et al., 2015a), while at pH 4 the same activity was extremely weak. The tyrosine decarboxylase obtained from a strain of L. brevis had its maximum relative activity at pH 5, while it maintained the higher stability at pH 7.4, e.g., 92% activity retained after 7 days of incubation (Zhang and Ni, 2014).

By contrast, a strain of E. faecalis under the same conditions showed the maximum tyramine accumulation at pH 4 after 2 h of incubation and pH between 4 and 5 after 24 h. In the same system, a strain of E. faecium did not show relevant differences in its decarboxylase activity at pH between 4 and 6. No pH differences in relation to the production of tyramine were observed in whole cells or cell free extract of L. brevis, with optimum activity at pH 5 (in the range 2–9); however, the cell-free extract had a higher activity compared with the whole cells (Moreno-Arribas and Lonvaud-Funel, 2001). Two tyrosine decarboxylases from E. faecalis and E. faecium (heterologously expressed in E. coli) had their optimum pH for the activity at 5.5 and 6, respectively (Liu et al., 2014).

The enhancing effects of lower pH on tyramine production (as responses to acidic stress) was observed also in Enterococcus durans (Fernández et al., 2007), E. faecium (Marcobal et al., 2006a; Pereira et al., 2009). Similar effects were observed for the histidine decarboxylase in L. lactis (Trip et al., 2012) and L. brevis (Marcobal et al., 2006b).

The histidine decarboxylase of S. thermophilus has its optimum pH at pH 4.5, measured in cell free extract, while histamine accumulation by viable cell cultures was very low at the same pH, due to the negative effect of acidity on the overall metabolism of the strain (Tabanelli et al., 2012).

Regarding Gram-negative bacteria, the pure histidine decarboxylase from P. phosphoreum had its higher activity at pH 7; this value decreased at 6.0 for P. damselae and 6.5 for M. morganii and R. planticola (Kanki et al., 2007). Morii and Kasama (2004) found that an optimal histidine decarboxylase activity was slightly lower (pH 6) in P. phosphoreum, while BA production by cells of Enterobacter cloacae and E. aerogenes was higher at pH 6 (Greif et al., 2006).

Cid et al. (2008) suggested that in fermented sausages, the accumulation of tyramine by LAB (L. curvatus) started immediately at the end of fermentation, when pH of the sausages has already reached its minimum value. Other BAs, produced by the same LAB, were produced more gradually only at a later stage of ripening.

#### TECHNOLOGICAL FACTORS FOR CONTROLLING BIOGENIC AMINE ACCUMULATION

## Use of Starter Culture

The addition of selected starter cultures is one of the main tools able to counteract BA accumulation in fermented foods. In first instance, the microorganisms used have to be characterized by the absence of any decarboxylating activity. Then, they had to be active in inhibiting the growth performances and the aminobiogenetic potential of wild decarboxylating bacteria. In addition, aspects such as the production of antimicrobial compounds like bacteriocins, as well as the ability to degrade BAs have to be considered (Ayhan et al., 1999; Bover-Cid et al., 2000). The possibility of some microorganisms to metabolize BA will be specifically discussed in Section "Microorganisms Able to Metabolize Biogenic Amine." The use of LAB cultures producing bacteriocins can have an important potential in limiting BA accumulation, even if further researches are needed to clarify this potential. Recently it has been demonstrated that using bacteriocinogenic strains of L. lactis it is possible to limit BA production by S. thermophilus and E. faecalis (Tabanelli et al., 2014).

The use of starter cultures has a consolidated application in cheese and the selection of not decarboxylative cultures for dairy products is well established since long time (Cogan et al., 2007). The use of selected starter cultures aimed to limit BA accumulation in dairy products has been recently reviewed by Linares et al. (2012). The use of autochthonous starter has been tested successfully in ewe's milk cheeses by Renes et al. (2014) who found a significantly lower BA content when L. lactis starter cultures were used. These results confirmed the observation of Novella-Rodríguez et al. (2002a) regarding goat cheeses. Also in

Manchego cheese, the use of autochthonous starter cultures of L. paracasei decreased the accumulation of BA, even if compared with commercial starter cultures (Poveda et al., 2015).

European Food Safety Agency (EFSA) recommends that socalled autochthonous strains (i.e., selected strains originating from each specific fermented product) with suitable technological profiles and reduced tendencies to produce BA should be selected as starter cultures (EFSA, 2011). Recently, it was suggested to use autochthonous starter cultures for fermentation of artisanal sausages (Talon et al., 2007; Casquete et al., 2011). Latorre-Moratalla et al. (2012) have extensively reviewed many of these aspects relevant in fermented sausage production. The addition of pure or mixed selected starter cultures can decrease BAs accumulation in sausages. Mixed starters were reported to perform better than single starters to control the growth of different bacterial groups (Latorre-Moratalla et al., 2010b; Naila et al., 2010). In fermented sausages, autochthonous starter cultures determined BA reductions higher (Latorre-Moratalla et al., 2010a; Casquete et al., 2011; Renes et al., 2014) than commercial mixed starter cultures (Ayhan et al., 1999; Gücükoglu ˘ and Küplülu, 2010). In fact, Latorre-Moratalla et al. (2012) reported that mixed starter cultures of amine negative strains of LAB and coagulase-negative staphylococci, well adapted to the meat fermentation environment, were the best choice to reduce BA content in sausages. In order to avoid formation of high level of BAs during fermentation of sausages, the use of raw materials with low microbial counts is recommended. The main level of BA production was during the first 3 days, when a sharp pH decrease and the development of LAB occurred during the fermentation process of dry sausages (Bover-Cid et al., 1999). LAB used as starter cultures can induce rapid acidification, thus inhibiting the growth of decarboxylating microorganisms, resulting in decreasing formation of BAs (Zhang et al., 2013). However, it is well known that the use of selected starter cultures alone cannot assure the reduction or inhibition of BA production (Parente et al., 2001).

The production of BA in wine is mainly associated with the activity of LAB (Lonvaud-Funel, 2001). For this reason, particular attention has been posed on the selection of starter cultures used for malolactic fermentation. Strains of O. oeni, the main responsible for the conversion of malic acid into lactic acid in wine, can produce histamine (Lonvaud-Funel and Joyeux, 1994; Guerrini et al., 2002) and tyramine (Gardini et al., 2005) and selected strains without this potential should be used for the conduction of this process in winemaking (Moreno-Arribas et al., 2003).

Also in the production of sauerkraut, the use of starter cultures can contribute to the reduction of BA content (Kalac et al., 2000 ˇ ; Rabie et al., 2011a).

#### Technological Additives

The use of some additives is widespread to improve the appearance and the quality of the final product such as development of the typical color of the cured meat, inhibition of mold growth and decrease of toxic compounds in the product. Thus, the effects of these additives on BA generation are important.

The scarcity of sugars, and in general poor nutritional environments, has been often associated to higher BA accumulation, being decarboxylative pathways secondary transport system providing metabolic energy (Konings, 2006). For this reason, the in vitro production of BAs has been tested by several authors in relation to sugar supply. Bunková et al. (2012) ˇ found the maximum tyramine accumulation by E. durans in the presence of the higher lactose concentration (5%). A less clear effect of this sugar was observed for L. lactis: increasing lactose concentration did not result in higher tyramine concentration and this was attributed to the excess of metabolic energy obtained by primary fermentation (Bunková et al., 2011 ˇ ). Higher amount of tyrosine were produced by S. thermophilus in the presence of limiting amounts of lactose (0.1%; La Gioia et al., 2011). Also Landete et al. (2008b) observed that increasing concentration of fructose and glucose progressively inhibited the histamine accumulation of enological LAB belonging to the species L. hilgardii, P. parvulus, and O. oeni.

In industrial formulations, especially in fermented sausages, sugars (mainly glucose, sucrose, and lactose) are added in order to improve the LAB fermentation process. González-Fernández et al. (2003) investigated the influence of three decarboxylase negative starter cultures in relation to different composition and concentration of sugars, on the presence of BAs in chorizo, a typical Spanish dry fermented sausage. The highest concentrations of BAs were found at the end of the ripening process in the control sausage with no starter culture irrespective of the use of different sugar concentrations. However, when a starter culture and sugar concentrations equal to 0.5% or 1% were used, the presence of BAs in the sausage decreased considerably in comparison with control and low sugar concentration sausages. The production of high amounts of putrescine (223–252 mg/kg) and tyramine (64–102 mg/kg) was observed when the concentration of sugar in the sausage was only 0.1%, even in sausages with a starter culture added. Bover-Cid et al. (2001a) also found contents of tyramine and cadaverine significantly higher in sausages without sugar in their formulation. They concluded that sugar omission is not recommended since it might increase BA accumulation during the manufacture and storage of slightly fermented sausages. The amount of sugar added can be a key factor in determining the equilibrium among the microbial communities during the fermentation step of sausages favoring the accumulation of different BAs. The modulation of sugar addition between 0 and 1.4% determined the maximum tyramine content at intermediate level, in correspondence with the maximum LAB and enterococci growth. By contrast, at the extreme levels (0 and 1.4%), the accumulation of putrescine and cadaverine was higher, associated with the best performance of enterobacteria (Bover-Cid et al., 2009). High accumulation of cadaverine and putrescine were observed also in chorizo in the presence of low sugar addition (0.1%) even in the presence of starter cultures added (González-Fernández et al., 2003).

The use of essential oils as aroma and flavor ingredients has increased recently because of the growing consumer demand for natural products as natural food preservatives and the replacement of synthetic additives in the food industry.

Wendakoon and Sakaguchi (1993) reported that antibacterial activity of essential oil such as eugenol present in clove might delay the BA formation of E. aerogenes in mackerel muscle extract. No amines were observed in blue fish burgers treated with thymol, lemon extract, and grapefruit seed extract combination with MAP (Del Nobile et al., 2009). Özogul et al. (2015) investigated the impact of carvacrol at different level (0.1, 0.5 and 1 ml/100 ml) on BAs production by foodborne pathogens including, Staphylococcus aureus, E. coli, Klebsiella pneumoniae, E. faecalis, Pseudomonas aeruginosa, Listeria monocytogenes, A. hydrophila, and Salmonella Paratyphi A in histidine decarboxylase broth. The results of this study showed that all bacteria tested were able to decarboxylate more than one amino acid and that carvacrol was able to reduce BAs formation depending on its concentration and bacterial species. Cai et al. (2015) reported low BAs content, especially histamine, putrescine, cadaverine in red drum (Sciaenops ocellatus) filets treated with 4 ml/l clove, cumin, and spearmint oils. In addition, Bozkurt (2006) reported that the levels of histamine, putrescine, and tyramine in sucuk (Turkish dry-fermented sausage) were lower in sausages with green tea extract than control. In Gouda cheese, the addition to curd of Zataria multiflora essential oil, whose major constituent was carvacrol (71.1%) determined a significant reduction of tyramine and histamine in the final product (Gorji et al., 2014).

In this perspective, spices can affect BA contents in food. Spices are often defined as aromatic, dried plant substances used in foods for flavoring and coloring. Komprda et al. (2004) reported that high content of red pepper, together with starter cultures, contributed to the lower BA content in dry fermented sausages. Red pepper contains capsaicin, known to prevent the growth of some bacteria (Hirasa and Takemasa, 1998). The effects of a variety of spices including ginger, garlic, green onion, red pepper, clove, and cinnamon were investigated to reduce BA contents in myeolchi-jeot, Korean salted and fermented anchovy (Mah et al., 2009). The greatest inhibitory effect on BA production was found in the culture treated by garlic extract while the other spice extracts showed minor effect in reducing BA contents. In particular, ginger extract reduced putrescine contents while red pepper extract decreased cadaverine accumulation. Since garlic contains allicin, antimicrobial component, its efficacy is probably related to the presence of this molecule. Other studies (Shakila et al., 1996) reported that clove and cinnamon reduced histamine production of M. morganii by 95%.

# Effects of Packaging on Biogenic Amine Formation

There is evidence that oxygen can affect the BA production. For example, it has been demonstrated that E. durans (Bunková et al., ˇ 2012) and L. lactis (Bunková et al., 2011 ˇ ) tyramine accumulation is favored by anaerobic conditions. A similar trend was reported also by Cid et al. (2008) using a strain of L. curvatus. However, the main technologies for food preservation based on atmosphere modification are focused on oxygen exclusion. Nevertheless, in such strategy, the principal aim, in relation to BA presence, is not the inactivation of decarboxylase activity but the inhibition of microbial population with decarboxylating properties.

In this perspective, the atmosphere used for packaging can affect the qualitative and quantitative formation of BAs. MAP and vacuum packaging (VP) play an important role in the selection of spoilage microorganisms and, particularly, on decarboxylating bacteria (Curiel et al., 2011).

CO<sup>2</sup> is the main gas used as bacteriostatic agent. Different concentrations of this gas in MAP have been applied to prolong shelf-life of foods by inhibiting microbial growth of Enterobacteria and H2S-producers bacteria (Lopez-Caballero et al., 2002), histamine forming bacteria (Özogul and Özogul, 2006), Pseudomonas spp. (Li et al., 2014), Lactobacillus sakei (Devlieghere et al., 1998) and psychrophilic microorganisms (Arashisar et al., 2004). The CO<sup>2</sup> concentration (>60%) in the MAP caused significant reductions in the contents of total BAs in barramundi filets (Yassoralipour et al., 2012) and sardine (Özogul and Özogul, 2006). Rodrigues et al. (2016) also indicated that lower production of putrescine and cadaverine was observed in MAP (80% CO2/20% N2) and VP samples of rainbow trout. Yew et al. (2014) investigated the major BA profile in Indian mackerel packed in different carbon dioxide compositions (30, 60, 80, and 100% CO2) with content 5% O<sup>2</sup> and corresponding N<sup>2</sup> level. Each amine responded differently to different CO<sup>2</sup> levels. Histamine concentration was reduced by 6.4, 8.5, 70.3, 78.8, and 90.2% in fish packed under VP, 30, 60, 80, and 100% CO2, respectively. In particular, histamine and tyramine increased rapidly in fish packed under VP and 30% CO2. This was attributed to the presence of histidine and tyrosine decarboxylase bacteria that cannot grow in the presence of increasing CO<sup>2</sup> concentration (Barakat et al., 2000).

LAB, due to their capacity to grow under high concentrations of CO2, constitute a substantial part of the natural microbiota of MAP meats (Chouliara et al., 2007). Curiel et al. (2011) demonstrated that under vacuum conditions or MAP (20% CO<sup>2</sup> and 80% N2) the enterobacteria producing putrescine, cadaverine, and agmatine were inhibited. In addition, the packaging under vacuum did not reduce significantly the tyraminogenic potential of strains of Carnobacterium divergens.

Zhang et al. (2015) evaluated the effect of air and MAP on the shelf life of chilled chicken. The putrescine and cadaverine content of gas mixture-packaged samples was significantly lower than that of the air-packaged samples during the storage period. The production of these BAs was slowed as the CO<sup>2</sup> content increased, indicating that the increasing concentration of CO<sup>2</sup> inhibited the growth of putrescine and cadaverine producing bacteria. The results are in agreement with those reported by Rodriguez et al. (2015) who investigated the effect of CO<sup>2</sup> concentration on the formation of BAs in shredded cooked chicken breast filet packed in modified atmosphere. They reported that the putrescine and cadaverine are affected by storage time and CO<sup>2</sup> concentration. Putrescineproducing bacteria are more CO2-resistant. The correlation between BA formation and bacterial growth showed that putrescine and cadaverine concentrations could be used as quality indicators as their formation is related to total viable count.

Nowadays the most widely used active packaging technology for food is oxygen scavengers which eliminate oxygen in the packaging and in the product or permeating through the packaging material during storage (Alvarez, 2000). Mohan et al. (2009) investigated the effect of O<sup>2</sup> scavenger on the formation of BAs during chilled storage of seer fish (Scomberomorus commerson) and indicated that the use of O<sup>2</sup> scavenger with the proper maintenance of chilled storage temperature helped in reducing the formation of BAs and also reduced the risk of Clostridium botulinum toxin. Similar results using commercial O<sup>2</sup> scavengers were reported by Mohan et al. (2008) for catfish (Pangasius sutchi) steaks and Goncalves et al. (2004) for gilthead sea bream (Sparus aurata).

Naila et al. (2010) indicated that active packaging, VP and MAP inhibit formation of BAs more effectively than air packaging, through inhibition of BA forming bacteria or enzyme activity.

#### Other Non-thermal Treatments

Food irradiation can be used to increase the safety and shelf life of foods by reducing microbial growth. Gamma-irradiation has been reported to reduce BAs contents in food due to the delay in the initial growth of adventitious microorganisms (Kim et al., 2003, 2004). Low levels of BAs by gamma irradiation in pepperoni sausage during storage were observed (Kim et al., 2005a). Gamma irradiation at 5, 10, or 15 kGy reduced putrescine, cadaverine, agmatine, histamine, tryptamine, spermine, and spermidine during fermentation of low-salt fermented soy paste (Kim et al., 2005b). Rabie et al. (2010) investigated the effects of different doses (2, 4, and 6 kGy) of gamma irradiation on BA formation in Egyptian fermented sausages. Histamine was detected in irradiated samples, immediately after irradiation, but not afterward and low levels of all BAs were observed during storage period especially in products treated at 6 kGy. Similar results were obtained for Blue cheese (Rabie et al., 2011b). Chub mackerel (Scomber japonicus) in chilled storage, after irradiation followed by vacuum packing slowed down the formation of BAs (Mbarki et al., 2009). Özogul and Ozden (2013) reported that radiation levels of (2.5 and 5 kGy) had similar effects on reducing the BA content except for agmatine and tryptamine. Although radiation caused an increase in spermine, agmatine, and tryptamine content in sea bream muscle, the putrescine and cadaverine contents in this product significantly decreased following the radiation process.

Pulsed electric fields (PEF) is a non-thermal method of food preservation that uses short pulses of electricity for microbial inactivation and causes minimal detrimental effect on food quality attributes (Quass, 1997). Previous studies have proven that PEF treatment can be useful in the sterilization of fruit juice under various conditions (Gurtler et al., 2010; Guo et al., 2014). They observed a clear microbial inactivation, depending on the juice, microorganisms, treatment conditions, and equipment. In addition, increasing the electric field strength (>600 kV/m) improved the inhibition of microorganism growth in tilapia during storage (Ko et al., 2016).

# Microorganisms Able to Metabolize Biogenic Amine

Many microorganisms can produce amino oxidases, which are the enzyme responsible for BA detoxification. These enzymes metabolize BAs firstly by deamination, with the production of NH<sup>3</sup> and H2O<sup>2</sup> in the presence of oxygen. The aldehydes formed is further reduced to the corresponding acids which can be then transferred to the central metabolism of the cells (Cooper, 1997). This metabolic pathway can be used as a source of NH<sup>3</sup> in nitrogen poor media. These enzymatic activities have been evidenced in vitro in several microorganisms. Leuschner et al. (1998) found these abilities widespread among Kocuria (former Micrococcus) varians. In addition, they isolated amino oxidase positive Brevibacterium linens strains, which were used for the production of a surface ripened Munster cheese, causing the reduction of BA (Leuschner and Hammes, 1998).

Amino oxidase activity was found also in Bacillus amyloliquefaciens and Bacillus subtilis (Zaman et al., 2010, 2011). Among LAB, many strains of Lactobacillus, Pediococcus, and Oenococcus showed these ability in culture media (García-Ruiz et al., 2011). Similar activities were observed in vitro also in Lactobacillus casei, Lactobacillus plantarum (Fadda et al., 2001; Herrero-Fresno et al., 2012), and in L. sakei in ensiled fish slurry (Dapkevicius et al., 2000). Herrero-Fresno et al. (2012) demonstrated also that the use of two strains of L. casei with high degradation rates for histamine and tyramine could reduce the accumulation of these BAs in Cabrales-like mini-cheeses.

Also staphylococci were able to deaminate tyramine and histamine in vitro (Martuscelli et al., 2000; Zaman et al., 2011). However, the activity of selected staphylococci in real food systems was limited by the presence of other nitrogen sources easily available for these bacteria, such in the case of dry fermented sausages (Gardini et al., 2002). In addition, in real complex food systems, the BA reduction by an amino oxidase positive L. casei strain cannot be clearly ascribed to an effective BA deamination or to a specific antagonism of the strains toward decarboxylating microorganism (Nishino et al., 2007). More interesting results were obtained in a nitrogen poor matrix such as wine, in which the use of strains of L. casei and Pediococcus spp. (García-Ruiz et al., 2011), as well as L. plantarum (Capozzi et al., 2012), determined a BA deamination.

Finally, some authors suggested the possibility to use purified amino oxidase in foods with negligible results (Hobson and Anderson, 1985; Dapkevicius et al., 2000). More interesting reductions were observed in wine using fungal amino oxidases (Cueva et al., 2012).

# Effects of Pressure Treatments on BA Formation

Among the alternative processes to thermal treatment for food preservation, the field of high pressure processing, better known as high hydrostatic pressure (HHP) or high pressure homogenization (HPH), is one of the most scientifically explored. Given the antimicrobial effects of HHP, this technology can modify the microbiota of treated foods both quantitatively and qualitatively, affecting also the food matrix characteristics

(Georget et al., 2015). Novella-Rodríguez et al. (2002b) studied the differences in BA formation in goat cheeses by using pasteurized and pressurized milk and found no significant effects between the two trials on tyramine, histamine, and putrescine accumulations. The treatment at 400 MPa or 600 MPa of 21- and 35-day ripened cheeses determined a significant increase of both aminopeptidase activity and free amino acid concentration. By contrast, the total BA concentration was higher in the not treated cheeses, especially in samples treated at 600 MPa (50% and more of BA reduction; Calzada et al., 2013). In other words, HHP treatment reduced the population of potentially decarboxylating microorganisms, limiting BA accumulation also in the presence of a higher concentration of precursors. Novella-Rodríguez et al. (2002a) used HHP to accelerate cheese ripening, founding no differences in tyramine content between treated and not-treated goat cheese samples when 400 MPa were applied; however, in the same work, a prolonged treatment at 50 MPa increased the tyramine content.

Ruiz-Capillas et al. (2007) studied the BA formation in sliced dry-cured "chorizo" sausage, HHP treated at 350 MPa for 15 min, during chilled storage at 2◦C and found a significant reduction of tyramine, putrescine, and cadaverine concentrations. The reason for this decrease was found in the higher susceptibility of decarboxylating microorganisms. However, during the storage, BAs continued to increase, probably due to the residual decarboxylases, whose activity was independent on the cell viability. Ruiz-Capillas et al. (2007) described a positive effect of HHP treatment of 400 MPa for 10 min on tyramine accumulation in vacuum packaged cooked sliced ham, with an increase of the shelf-life by at least 35 days. Similar results were observed by treating at 500 MPa/10 min Hungarian dry fermented sausages cut in 5 cm long pieces under vacuum (Simon-Sarkadi et al., 2012). Under these conditions, BA content was reduced during the storage.

In addition to HHP, also HPH can be a strategy to control the presence of decarboxylating microorganisms in raw materials such as the milk used for cheese making. Lanciotti et al. (2007) demonstrated that a HPH milk treatment at 100 MPa can significantly reduce the BAs accumulation compared with the thermal treated milk, due to a deep modification of the microbiota during ripening.

#### Antimicrobial Substances

The use of substance with antimicrobial properties can modify the BA profile of foods interfering with the equilibrium among microbial population rather than affecting directly the decarboxylase efficiency. Nitrate and nitrite salts are commonly used in fermented sausages for different purposes. In fact, they influence the color, the flavor and the oxidation of cured meat. In addition, they are used for controlling hazardous bacteria, such as clostridia. Nevertheless, their action can also interfere with BA accumulation. The addition of increasing concentration of nitrite (up to 150 mg/kg) reduced tyramine and cadaverine accumulation in Sucuk, a Turkish fermented sausage (Gençcelep et al., 2008). Opposite results were obtained in a Spanish not fermented cured meat (Lacon) in which the addition of nitrate and nitrite significantly increased the BA content (Lorenzo et al., 2007). This trend was explained by the fact that the addition of nitrite favored the selection of a superficial microbiota mainly constituted by LAB with relevant decarboxylative activity. The addition of nitrite reduced also the BA content in fresh beef, pork, and poultry meats stored at 4◦C (Jastrz˛ebska et al., 2016). These latter authors tested, under the same condition, also a weak acid (sorbic acid) whose effectiveness was rather limited. The use of two weak acids (sorbate and benzoate), in combination with clove, inhibited the decarboxylative action of an aminobiogenetic E. aerogenes strain isolated from fish (Wendakoon and Sakaguchi, 1993). Benzoate and sorbate were also used to limit BA production during the storage of two different type of fishery products (cod roe and pearl mullet filets, respectively; Lapa-Guimarães et al., 2011; Gençcelep et al., 2014).

Also sulfur containing antimicrobials were used to control BA accumulation in foods. Bover-Cid et al. (2001b) added sodium sulfide (maximum concentration 1000 mg/kg) to ripened sausages and observed a contradictory effect on BA accumulation. In fact, while the cadaverine content was inhibited by the addition of sulfide, tyramine, and putrescine accumulation was strongly enhanced. Jastrz˛ebska et al. (2016) observed a contribution of sodium metabisulphite to the reduction of BA content in fresh meats. However, the most relevant matrices in which sulfur compounds play a key role in the reduction of BAs are the alcoholic beverages. In particular, in wine it is known the potential aminobiogenetic potential of the LAB responsible for malolactic fermentation (Lonvaud-Funel, 2001; Landete et al., 2005). Some authors demonstrated that the addition of SO<sup>2</sup> when malate was completely converted into lactate prevented amine formation in subsequent stages (Vidal-Carou et al., 1990; Marcobal et al., 2006c). A similar effect of SO<sup>2</sup> on tyramine accumulation was evidenced in a model system inoculated with an O. oeni strain (Gardini et al., 2005). The control of BA accumulation due to SO<sup>2</sup> is related to the inhibiting effect on cell metabolism rather than to the repression of the decarboxylases activity (Landete et al., 2008b). Interesting results were obtained in wine using lysozyme instead of SO2. In fact, using this enzyme in Rioja wines subjected to malolactic fermentation, a drastic decrease of histamine content was found (López et al., 2009).

#### CONCLUSION

Even if the presence of BAs in food (and the risks associated with them) is known since a long period (Gale, 1946), systematic studies regarding their presence have been carried out only in relatively recent times. The reviews of Shalaby (1996) and Silla Santos (1996) had the merit to collect the fragmented information about this issue and were the starting point for a drastic multiplication of scientific publications regarding the presence of BA in food products and the elucidation of the metabolic and genetic drivers of their production by microorganisms. Combining the words "biogenic amine" and "food" the number of publication selected by the Web of Science passed from about 500 in the year 2000 to more than 4500 in 2015.

This increasing scientific effort allowed obtaining a deeper knowledge about the genetic and biochemical mechanisms

responsible for BA production by foodborne microorganisms, but also furnished important information about the possibility to reduce their accumulation in food and the risks associated with their presence.

The possible ways reviewed here to achieve this goal in food are mainly based on two strategies, which always are strictly interacting each other: the modulation of process and environmental factors including storage and distribution conditions and the control of the microbiota associated with fermented foods.

While the studies regarding the genetic bases of microorganism decarboxylating activity have brought to relevant steps forward and new insights on this topic, the role of environmental and technological factors on the overall activity of aminobiogenic microorganisms and on their decarboxylases requires deeper researches aimed to improve the possibility of intervention on the food processes in the perspective of the reduction of the risks associated to the BA presence.

#### REFERENCES


For the fermented foods, from a strictly microbiological point of view, more information is available regarding the possible role of selected starter cultures aimed to overcome and inhibit decarboxylating microbiota. Further work is required to evaluate the real potential and optimize the use of microbial cultures able to degrade BAs and detoxifying them through the action of amino oxidases. In this perspective, also the use of bioprotective cultures producing bacteriocins or other antimicrobial substances needs greater attention due to their not fully explored potential in this field.

# AUTHOR CONTRIBUTIONS

FG and GS wrote the introduction and conclusion sections; YO and FO wrote chapters 2, 3, 4, and 5; GT coordinated contributions and provided the final draft of the manuscript.

accumulation and tyrDC gene expression. Front. Microbiol. 6:259. doi: 10.3389/fmicb.2015.00259


marcescens CCM 303. J. Environ. Sci. Health Part B-Pestic. Contam. Agric. Wastes 50, 797–808. doi: 10.1080/03601234.2015.1058097





amine content of ewe's milk cheese throughout ripening. Food Microbiol. 44, 271–277. doi: 10.1016/j.fm.2014.06.001



meat: a special breed in Asia. LWT - Food Sci. Technol. 64, 1123–1129. doi: 10.1016/j.lwt.2015.07.039

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Gardini, Özogul, Suzzi, Tabanelli and Özogul. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Applications of High and Ultra High Pressure Homogenization for Food Safety

Francesca Patrignani and Rosalba Lanciotti\*

Department of Agricultural and Food Sciences, University of Bologna, Bologna, Italy

Traditionally, the shelf-life and safety of foods have been achieved by thermal processing. Low temperature long time and high temperature short time treatments are the most commonly used hurdles for the pasteurization of fluid foods and raw materials. However, the thermal treatments can reduce the product quality and freshness. Consequently, some non-thermal pasteurization process have been proposed during the last decades, including high hydrostatic pressure, pulsed electric field, ultrasound (US), and high pressure homogenization (HPH). This last technique has been demonstrated to have a great potential to provide "fresh-like" products with prolonged shelf-life. Moreover, the recent developments in high-pressure-homogenization technology and the design of new homogenization valves able to withstand pressures up to 350–400 MPa have opened new opportunities to homogenization processing in the food industries and, consequently, permitted the development of new products differentiated from traditional ones by sensory and structural characteristics or functional properties. For this, this review deals with the principal mechanisms of action of HPH against microorganisms of food concern in relation to the adopted homogenizer and process parameters. In addition, the effects of homogenization on foodborne pathogenic species inactivation in relation to the food matrix and food chemico-physical and process variables will be reviewed. Also the combined use of this alternative technology with other non-thermal technologies will be considered.

#### Edited by:

Maria Schirone, University of Teramo, Italy

#### Reviewed by:

Luca Settanni, University of Palermo, Italy Lucilla Iacumin, University of Udine, Italy

#### \*Correspondence:

Rosalba Lanciotti rosalba.lanciotti@unibo.it

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 23 March 2016 Accepted: 07 July 2016 Published: 03 August 2016

#### Citation:

Patrignani F and Lanciotti R (2016) Applications of High and Ultra High Pressure Homogenization for Food Safety. Front. Microbiol. 7:1132. doi: 10.3389/fmicb.2016.01132 Keywords: high pressure homogenization, ultra high pressure homogenization, pathogenic species, inactivation, food matrices, non-thermal technologies, spores, enzymes

# INTRODUCTION

For several decades, the food research has been focused on finding new solutions to old problems to ensure food safety without losing nutrients and more recently, in response to market demand, to preserve the sensory and the freshness of the products as well as to produce innovative foods, using safe, quick, economical, and environmental friendly processes (Zamora and Guamis, 2015). In a globalized market, the solution to these goals is fundamental for the survival of food and equipment enterprises, especially the small and medium enterprises (SMEs). The enterprise competition makes crucial the innovation in processes and products and user-friendly processes have been designed, focusing the interest on the application of new processing technologies including high pressure homogenization (HPH) and ultra high pressure homogenization (UHPH; Dumay et al., 2013).

The word "homogenization" is referred to the ability to produce a homogeneous size distribution of particles suspended in a liquid, by forcing the liquid under the effect of pressure through a specifically designed homogenization valve. Homogenizer able to process fluid matrices at pressure ranging between 20–100 MPa are nowadays employed in the dairy beverage, pharmaceutical, and cosmetic industries mainly to reduce particle size and consequently increase stability of emulsions in order to avoid creaming and coalescence phenomena (**Figure 1**). However, HPH was firstly employed as an useful method for cell disruption and recovery of intracellular bio-products (Keshavarz Moore et al., 1990; Shirgaonkar et al., 1998). The successful results obtained on cell disruption of dense microbial cultures stimulated researches on the application of HPH for food safety and shelf-life improvement. In fact, in the food industry, there is a growing interest in mild non-thermal processes, which combine an efficient microbial reduction with a maximal retention of physic-chemical product properties, as well as nutritional and sensory characteristics of the raw materials and ingredients used. Among the non-thermal treatments studied and proposed, HPH is regarded by a wide literature as one of the most encouraging alternatives to traditional heat treatments for food preservation and diversification dairy products, emulsions, egg based foods and fruit juices. Its effectiveness in the deactivation of spoilage microorganisms in model and real systems is well documented since 1994 (Lanciotti et al., 1994, 1996; Guerzoni et al., 1999; Kheadr et al., 2002; Wuytack et al., 2002; Vannini et al., 2004; Diels and Michiels, 2006; Bevilacqua et al., 2009; Patrignani et al., 2009a, 2010, 2013a,b; Zhao et al., 2014; Ferragut et al., 2015). However, HPH effectiveness for microbial inactivation is affected by several parameters such as process and microbial-physiological factors and aspects related to the characteristics of the treated fluid.

The HPH has been proposed also for the bacterial spore inactivation, generally in combination with other physicchemical hurdles for spore surviving and/or germination due to the ability of bacterial spores to survive to the most prohibitive conditions (Wuytack et al., 2002; Bevilacqua et al., 2007, 2012; Chaves-López et al., 2009; Chen et al., 2013; Roig-Sagues et al., 2015). In addition to the effects on microbial cells, the HPH treatment is reported to act on food constituents, especially proteins and enzymes, modifying their functional properties and activities (Kheadr et al., 2002; Hayes and Kelly, 2003; Vannini et al., 2004; Iucci et al., 2008; Navarro et al., 2014). In fact, it is reported to improve food microstructure, rheology and availability of food bioactive compounds (Guerzoni et al., 1997, 2002; Lanciotti et al., 2004a,b; Sandra and Dalgleish, 2005; Betoret et al., 2015). Moreover, it has been proposed as a useful tool to enhance the cheese yield and reduce the cheese ripening times due to the enhancement of proteolysis and lipolysis of milk naturally occurring or microbial enzymes (Guerzoni et al., 1999; Lanciotti et al., 2004b, 2006, 2007a; Burns et al., 2008; Vannini et al., 2008; Burns et al., 2015). Moreover, the effects of HPH, applied at sub-lethal level (50 MPa), was studied on several strains of Lactobacillus spp., and particularly non-starter LAB (NSLAB), inoculated in milk as adjunct and involved in dairy product ripening (Lanciotti

et al., 2007b). The literature data demonstrated that this cold technology is able to affect the strain metabolic activity and enzymes leading to a modification of the cheese ripening patterns, especially linked to the breakdown of proteins. The modification of enzyme and protein activity and functionality can greatly affect not only the food quality and ripening patterns but also the food safety feature due to the increased activities of egg and milk naturally occurring antimicrobials such as lysozyme, lactoperoxidase systems, and lactoferrin (Vannini et al., 2004; Iucci et al., 2007; Patrignani et al., 2013b) or reducing the biogenic amine content in cheeses obtained from HPH treated milk (Lanciotti et al., 2007a). In addition, HPH has been proposed with several roles in the functional food sector, for the production of probiotic dairy products with improved sensory or functional properties, such as probiotic strain viability over refrigerated storage and accelerated fermentation kinetics with less environmental impact with respect to the traditional heat treatment (Burns et al., 2008, 2015; Patrignani et al., 2009b, 2015a). In fact, sub-lethal pressure levels (50 MPa) applied directly to microbial cells, increased some functional properties (i.e., hydrophobicity, resistance to simulated gastric conditions, and stomach-duodenum passage) in some probiotic strains associated to an increased maintenance of viability during refrigerated storage (Tabanelli et al., 2012, 2013, 2014). Also, Muramalla and Aryana (2011) reported low homogenization pressures (up to 13–80 MPa for five passes) to improve certain probiotic characteristics of yogurt bacteria and Lactobacillus acidophilus LA-K (i.e., strain acid and bile tolerance) without effects on protease activity and strain growth potential. Additionally, these low homogenization pressure treatments, applied directly to probiotic strains, modified their interaction

with the small intestines of BALB mice and induced a higher IgA response compared with untreated mice in a strain- and feeding period-dependent way (Tabanelli et al., 2012). Burns et al. (2015) studied the effects of a sub-lethal HPH on a probiotic strains used as adjunct for producing Caciotta cheese demonstrating that the HPH-treated probiotic strain maintained high viability for 14 days whilst the physico-chemical analyses on Caciotta cheese containing HPH treated cells showed a faster ripening, compared to cheeses containing not HPH treated cells. In addition, these Authors demonstrated that the 50 MPa treatment increased the L. paracasei gastric resistance in Caciotta, maintaining high strain viability without any significant effects on IgA production in mice.

The HPH technology has shown a great impulse both at industrial level and research field during the last decades also with the support of some important EU projects. The most relevant EU projects relative to the effects of HPH on food safety can be envisaged in the EU project HighQ RTE FP6- FOOD-023140 "Innovative non-thermal processing technologies to improve the quality and safety of ready-to-eat (RTE) meals-HighQ RTE," EU Craft project "UHPH 512626, Development and Optimisation of a Continuous Ultra High Pressure Homogenizer for Application on Milks and Vegetable Milks" and the EU project FUNENTECH 232603 "study of functionality, nutritional and safety aspects of liquid foods, liquid food preparations, and cosmetics processed by ultrahigh-pressure homogenization". The last had the aim to reinforce transfer the HPH processing to SMEs of fluid or pumpable ingredients and foods as well as cosmetic sectors.

The research performed within national and international projects stimulated the improvement of the HPH equipments and devices and their adaptation to food processing lines. Several HPH equipments are nowadays available and some equipment producers, such as Microfluidics (USA), Bee International (USA), Avestin (Canada), Gea Niro Soavi (Italy), and APV (UK), have proposed pilot devices able to exert a pressure of 100– 200 MPa, with average flow of 160 L/h at 200 MPa (Zamora and Guamis, 2015). However, Stansted Fluid Power Ltd (United Kingdom) have developed from the year 2006 pilot devices able to work up to 400 MPa giving origin to the UHPH technology applied to food sector. Also Gea Niro Soavi (Italy) developed within the project made in Italy ATENA a pilot homogenizer able to work at 400 MPa, having an average flow of 5 L/h (patent no. US-2015-0314254-A1, 11/05/2015). This prototype was used to obtained apple and tomato based formulations having a shelf-life of 30 days when stored at environmental temperature and treated at 300–400 MPa. At the expire date the levels of Listeria monocytogenes, Escherichia coli, Staphylococcus aureus, Bacillus cereus and Salmonella spp. were under the detection limits even in the samples inoculated before the HPH treatments at levels ranging between 3 and 4 log CFU/ml.

Since the safety features are prerequisite both for traditional and innovative products, this review is focused on the effects of HPH on foodborne pathogen inactivation in relation to the food matrix and food physico-chemical and process variables. Moreover, the review takes into consideration also the opportunity to inactivate the spores from pathogenic bacteria by using dynamic pressure. Also the potential of HPH to increase the activity of antimicrobial enzymes to inactivate pathogenic microorganism will be taken into consideration. Finally, the combined use of this alternative technology with other non-thermal technologies will be considered within this review.

## MECHANISM OF ACTION OF HIGH PRESSURE HOMOGENIZATION

From a technological point of view, an homogenizer consists principally of a pump and a homogenizing valve (**Figure 2**). The pump is used to force the fluid into the valve where the homogenization happens (Diels and Michiels, 2006). In the homogenizing valve the fluid is forced under pressure through a small orifice between the valve and the valve seat. The operating pressure is controlled by adjusting the distance between the valve and the seat. Among the process parameters, the level of pressure applied and the temperature reached in the food matrix during the process are the variables able to affect the food components and the eventually occurring microbial cells. The microbial inactivation caused by the application of HPH, although affected by several factors and mainly by the physic chemical features of the food matrix and the sensitiveness of different microorganisms, increases with the pressure level (Diels and Michiels, 2006). The effects induced by the temperature have to be necessarily taken into account in HPH, since, during homogenization, a rise of the temperature (about 2.5◦C per 10MPa), related to the fluid food employed, is observed in the fluid downstream of the valve. This is generally attributed by the literature to the viscous stresses caused by the high velocity of the fluid flow, which is then impinging on the ceramic valve of the homogenizer, leading to the dissipation of a significant fraction of the mechanical energy as heat in the fluid (Floury et al., 2000; Zamora and Guamis, 2015). However, such temperature increase did not result in the appearance of heat indicators in HPH treated food samples probably due the flash time of treatment of the food matrices (lower than 1 s; Floury et al., 2004; Pinho et al., 2011). For example HPH treated milk samples suffered less Maillard reaction, less whey protein denaturation and did not present lactose isomerisation compared to a commercial pasteurized milk with a concomitant preservation of essential amino acids and as a consequence with a better nutritional value desired by consumers (Pereda et al., 2009).

The significant improvement of HPH equipment functionality and flexibility (capacity to work in a food processing line), the design of new valves able to work at 300–400 MPa and the significant research efforts make HPH technology ready for the scaling up at industrial level for the development of new products differentiated from traditional ones by sensory and structural characteristics or functional properties. Also patents regarding continuous systems and procedure of sterilization and physical stabilization of pumpable fluids by means of HPH are available (i.e., EP 2409583 A1).

# POTENTIAL AND APPLICATIONS OF HIGH AND ULTRA HIGH PRESSURE HOMOGENIZATION FOR PATHOGENIC SPECIES INACTIVATION IN VITRO AND FOOD SYSTEMS

The first application of homogenization for the stabilization of food and dairy emulsion was presented at the Paris World's Fair in 1900 where Auguste Gaulin presented an invention for "intimately mixing milk" using pressures up to 30 MPa (Gaulin, 1899). Since then conventional homogenization extended the pressure range until 50 MPa. HPH, also known as dynamic HPH, has been frequently highlighted for its potential for food cold pasteurization (Diels and Michiels, 2006; Donsì et al., 2009a,b,c; Patrignani et al., 2009a, 2010, 2013a,b; Pereda et al., 2009; Panozzo et al., 2014). Modern high pressure homogenizer enables pressures 10–15 times higher than traditional ones and covers pressure ranges between 300 and 400 MPa. These last ranges have been referred to as UHPH. The progression toward UHPH has also opened the view to new sterilization opportunities, including also the inactivation of spores by HPH. The inlet temperature of pumpable products and the level of pressure, which both determine the temperature reached during the UHPH treatment, have been considered as the main factors of the microbial inactivation (Zamora and Guamis, 2015). UHPH warrants the destruction of microorganisms reaching sterilization of liquid food products, and it has a positive influence on food stability, with few effects on nutritional value and sensory characteristics of the processed fluids. UHPH technology allows more efficient particle reduction than the classical homogenization, and its effect is the results of several mechanisms such as sudden pressure drop, torsion and shear stresses, turbulence, impingement, cavitation phenomena, shock waves, and temperature increase, with a concomitant reduction of microbial load and interesting consequences in emulsion properties (**Figure 3**). Since its first appearance, HPH and, later, UHPH have been tested on several matrixes and in vitro systems by using different microbial targets to demonstrate their effectiveness. The microbial targets deal with pathogenic and spoilage microorganisms but in this review only the researches regarding the pathogenic species will be reviewed.

In **Table 1**, the studies regarding the microbial inactivation in model systems by HPH or UHPH, in relation to the microbial targets, inoculation level and conditions adopted, are reported.


TABLE 1 | High pressure homogenization (HPH) microbial inactivation in relation to the model system, species and process conditions adopted.

Lanciotti et al. (1994) studied the effectiveness of pressures ranging from 15 to 200 MPa, applied with a continuous homogenizer (Niro Soavi, Parma, Italy), on the cell viability of spoilage and pathogenic microorganisms in two model systems made of water 50%, oil 25%, egg yolk 25%, the first, and water 25%, oil 53%, egg yolk 22%, the second. Moreover, the effects of microstructural modifications of food systems associated with the treatments on cell viability and on its subsequent evolution were investigated. Modulation of the homogenization pressure allowed strong instantaneous reductions of the initial cell loads of L. monocytogenes and Yersinia enterocolitica. The combined effects of the initial pressure treatment and space reduction, resulting from the microstructural modifications of food systems, increased the safety and the shelf-life both of water in oil and oil in water emulsions. Moreover, Lanciotti et al. (1996) investigated also the effects of chemico-physical growth conditions such as pH, temperature and water activity (a) on lethal high homogenization pressure effects on L. monocytogenes, S. aureus, and E. coli. The results, based on standard medium such as Brain Heart Infusion (BHI), highlighted the importance of food system composition and its thermal history on the high pressure tolerance of the microbial population. In previous work, Lanciotti et al. (1994) showed that the relationship between survival cells and instantaneous pressure applied in medium appeared to be log-linear in a range between 40 and 200 MPa. According to the author findings, S. aureus was the most resistant pathogenic species while L. monocytogenes was resistant to pressure only when grown at highest aw. Also Guerzoni et al. (1997) studied the growth of L. monocytogenes in a model system, simulating a dairy product, when treated at 25 MPa, in relation to the lipid content NaCl and pH values, modulated according to a Central Composite Design (CCD). Polynomial equations describing the effects of such variables on the aw, microstructural features, and L. monocytogenes growth, were obtained. The three variables and their interactions, in addition to a direct effect on microbial growth, played an indirect role

due to their influence on microstructural features, such as diameter of water droplets and total water phase availability. In particular, the pH value affected the a(w) and the total space available for microbial growth, while the NaCl content had a prevalently indirect effect on space availability and on the diameter of the water droplets. The results suggested that the microstructural changes induced by HPH affected the growth of L. monocytogenes which was dependent on the total water phase space availability. Guerzoni et al. (2002) studied the survival of Salmonella enteritidis after pressure treatments, ranging between 0.1 and 140 MPa, in relation to compositive variables (NaCl content, pH). The study was performed both in model (BHI) and real systems consisting of an egg-based mayonnaise type product. Moreover, the fate of the surviving cells of S. enteritidis was monitored during storage at 10◦C and the growth or death parameters were calculated and modeled in relation to pH, NaCl concentration of the medium and the level of pressure treatment applied. From this study, it was evident that the salt content and pH displayed a synergistic effect with pressure, whose extent was higher in the mayonnaise based products than in BHI. In fact, while in the model systems the cell recovery and growth during the subsequent incubation at 10◦C was allowed in many combinations of the CCD, in the real systems no recovery or growth of S. enteritidis were observed. According to the Authors, this viability loss, which was maximum at pH 4.00 or 2% NaCl, is not be attributed merely to the interactions of such variables, but it probably involved the naturally occurring antimicrobial enzymes of the raw material, whose activity can be enhanced by the pressure treatment. Reviewing the literature, one of the most studied in vitro system is represented by PBS buffer. Different Authors have investigated in this medium the potentialities of HPH treatment in pathogenic species inactivation. For example, Vachon et al. (2002) studied the effect of an homogenization treatment performed at 200 MPa, repeated for five cycles, on the inactivation of S. enteritidis ATCC 13047, E. coli O157:H7 ATCC35150 and L. monocytogenes LSD 105-1 when inoculated in PBS buffer at pH 7. The results showed that E. coli and L. monocytogenes reached 8 log reduction cycles after the HPH treatment after three passes while five passes were necessary to reach the same inactivation level for S. enteritidis. Also Wuytack et al. (2002) found that a treatment of 300 MPa was able to inactivate S. aureus, Enterococcus faecalis and L. innocua of 6, 3, and 5 log reduction cycles in PBS buffer pH 7, using a treatment of 300 MPa and a fluid inlet temperature of 25◦C. Diels et al. (2003) conduced a detailed study of the inactivation of S. aureus and Y. enterocolitica in PBS buffer by HPH at, respectively, 25 and 35 different combinations of process temperature and process pressure covering a range of 5–50◦C and 100–300 MPa. In the entire studied, it was clear that S. aureus was more resistant to HPH than Y. enterocolitica, as already demonstrated by Lanciotti et al. (1996). Also, temperature between 5 and 40◦C did not affect inactivation of S. aureus by high-pressure homogenisation, while Y. enterocolitica inactivation was affected by temperature over a much wider range. Later, Diels et al. (2005) investigated the resistance of E. coli K12 when treated in PBS buffer at 300 MPa by using a axial-flow through orifice valve, outlining reduction of 6 log cycles. Donsì et al. (2009c) studied the inactivation of E. coli by HPH in model systems, under a wide range of operating conditions (temperature, pressure, number of homogenization passes, cell concentration) in a lab-scale and a pilot-scale unit (Stansted Fluid Power) utilizing single or multiple passes. Results highlighted that the inactivation kinetics did not depend linearly on pressure, due to the distribution of individual cell resistance in the sample. The efficacy of the treatment at higher pressures or upon multiple passes was reduced. Moreover, the scale of the apparatus, which in this case differed of several order of magnitude (from 0.7 to 120 l/h), and consequently of the width of the gap of the valve did not affect the extent of microbial inactivation at a given pressure. Taylor et al. (2007) studied the inactivation of E. coli K-12 cells, grown statically or in chemostat, when exposed to HPH processing pressures of 50 to 350 MPa in the absence or presence of the antimicrobial nisin. These Authors found that pressure and temperature exhibited a quadratic relationship. Significant HPH-induced inactivations of the Gram-negative microorganism was observed in the range of 100 to 250 MPa. Above 300 MPa, heat was the main factor promoting microbial inactivation, regardless of whether cells were grown in chemostat or statically. Chemostat-grown cells were significantly more resistant to HPH processing than were statically grown cells. Moreover, the data indicated a potential synergistic effects of nisin and HPH on the inactivation of bacterial contaminants, although this antimicrobial is generally active against Grampositive bacteria.

However, although a huge amount of data collected from model systems were available and helpful to understand the mechanisms of action of HPH against pathogenic species, many researchers tested potentialities of this technology against pathogenic species inoculated in food matrixes because the composition and viscosity of the treated food also have an indirect effect on the microbial inactivation (**Table 2**). The available literature reports about the use of HPH for the reduction of foodborne pathogens in several food matrices, such as milk (Lanciotti et al., 1994; Diels et al., 2005; Hayes et al., 2005; Brinez et al., 2006a,b, 2007; López-Pedemonte et al., 2006; Roig-Sagues et al., 2009), egg-based products (Velazquez-Estrada et al., 2008; Patrignani et al., 2013b), orange juice (Brinez et al., 2006b; Kumar et al., 2009; Pathanibul et al., 2009; Maresca et al., 2011), mayonnaise type products (Guerzoni et al., 2002). Lanciotti et al. (1994) found that the relationship between surviving cells and pressure applied was log-linear in milk inoculated with Y. enterocolitica and L. monocytogenes and processed at different pressures (40–150 MPa) at 25◦C. In order to establish the fate of the surviving cells, the growth of Y. enterocolitica, and L. monocytogenes, in the samples stored at 3–4◦C, was followed over 220 h. The data were analyzed according to the Gompertz equation. For Y. enterocolitica and L. monocytogenes, the treatment apparently did not induce irreversible damage to the surviving cells; in fact, although the lag phase was prolonged when pressures higher than 400 bar were used, the µmax increased with the applied pressure and the maximal cell numbers attained (A + K) were independent on the level of the applied pressure. Diels et al. (2005) studied the inactivation of E. coli MG1655 in skim, soy and strawberry-raspberry milk


#### TABLE 2 | High pressure homogenization microbial inactivation in relation to the food matrix, species and process conditions adopted.

subjected to 300 MPa founding that the highest inactivation was reached in skim milk with a Tin and Tout temperature of 25 and 18◦C, respectively, by using a counterjet dispergator valve. Hayes et al. (2005), utilizing an axial-flown through orifice valve, found in milk treated at 250 MPa, with a Tin of 45◦C and a Tout of 76.8, a 6 log reduction of Pseudomonas fluorescens AFT 36, resulting from the application of pressure and temperature increase. Lowest reduction (2.7 log) was found by Brinez et al. (2006b) for L. innocua ATCC33090 in bovine milk or S. aureus ATCC 13565 (3.6 log reduction) and S. aureus CECT 4491 in

milk (Brinez et al., 2007). López-Pedemonte et al. (2006) found high inactivation for S. aureus CECT 976 inoculated in milk treated at 330 MPa. Roig-Sagues et al. (2009) reported about the inactivation of L. monocytogenes CCUG 15526 when inoculated at 7.0log CFU/ml in milk samples having 0.3, 3.6, 10, and 15% of fat contents. The samples were subjected to a single cycle of UHPH treatment at 200, 300, and 400 MPa. Microbiological analyses were performed 2 h after the UHPH treatments and after 5, 8, and 15 days of storage at 4◦C. Maximum lethality values were observed in samples treated at 400 MPa with 15 and 10% fat (7.95 and 7.46 log CFU/ml), respectively, while in skimmed and 3.6% fat milk samples, complete inactivation was not achieved and, during the subsequent 15 days of storage at 4◦C, L. monocytogenes was able to recover. In milk samples with 10 and 15% fat, L. monocytogenes recovered to the level of initial counts only in the milk samples treated at 200 MPa but not in the milk samples treated at 300 and 400 MPa. According to these data, fat content increase enhanced the maximum temperature reached during UHPH treatment and this could have contributed to the lethal effect achieved. In addition the HPH treatments of milk is reported to enhance the release of free fatty acids (due to the rupture of fatty globule membranes and the activation of lipases), and mainly short and medium chain ones, having an antimicrobial effects (Lanciotti et al., 2006; Vannini et al., 2008). On the contrary, some Authors attributed to the fat content in milk a protective role against microbial species during the high pressure treatment performed at 100 MPa for several cycles. Although several Authors have tested the same matrixes, and in some case the same microbial species, different inactivation results was achieved, particularly using the multi-pass approach. For example, Patrignani et al. (2013b) found that Salmonella inactivation in eggs, resulting from the application of HPH at 100 MPa, seems to be linearly correlated to the number of passes. This result, although using a different substrate and microorganism, is in agreement with the findings of Patrignani et al. (2010) who demonstrated for the spoiling Zygosaccharomyces bailii, inoculated in apricot and carrot juice, that the effect of each pass is additive and, therefore, each homogenization pass causes almost the same reduction of the microbial load. Several other authors have also found first order inactivation kinetics as a function of the number of passes (Wuytack et al., 2002; Diels and Michiels, 2006; Tahiri et al., 2006), although the literature data concerning the inactivation kinetics by HPH are still conflicting. In fact, Donsì et al. (2009c) observed that HPH inactivation of E. coli, produced a nonadditive trend for multiple pass processes at a given pressure level. These Authors attributed this behavior mainly to the physiological diversity within a microbial population and to the existence of resistant cells able to survive after repeated passes at the pressure applied. On the other hand, it is important to take into consideration the additional effect of HPH on the antimicrobial activity of naturally occurring enzymes, such as lysozyme, lactoperoxidase system and so on. Velazquez-Estrada et al. (2008) have published data regarding the use of HPH for Salmonella inactivation in liquid whole egg (LWE). In particular, these Authors proposed the single-pass treatment of inoculated LWE with ultra HPH at 100, 150, 200, and 250 MPa, demonstrating that the level of pressure applied can influence the S. enterica lethality attained. Moreover, Panozzo et al. (2014) outlined the potentialities of HPH as a promising alternative to thermal pasteurization of eggwhite. These Authors showed that a HPH at 150 MPa for multiple passes was able to decontaminate egg white inoculated with S. enterica SDMZ 9898.

Many applications of HPH or UHPH for inactivation of spoiling and pathogenic species were tested in juices and vegetable drinks, where HPH and UHPH find a great application potential. For example, Brinez et al. (2006a) tested the inactivation of E. coli 058:H21 ATCC 10536 and E. coli 0157:H7 CCUG 44857 inoculated in orange juice and treated at 300 MPa founding a log reduction of 3.9 and 3.7, respectively. Kumar et al. (2009) investigated on the inactivation of E. coli K12 in apple juice and apple cider obtaining an inactivation of 7 log applying a pressure of 250 MPa with a Tin of 25◦C and Tout of 70◦C, thus avoiding the use of a thermal exchanger. Also Pathanibul et al. (2009) inoculated L. innocua ATCC51742 and E. coli, as surrogate for foodborne pathogens, in apple and carrot juice, containing or not nisin (0-10 IU), and treated from 0.1 to 350 MPa. At 50 MPa homogenization pressure intervals, juice samples were collected, immediately cooled to 10◦C, and then serially diluted and plated on non-selective recovery media. As processing pressure increased, inactivation of E. coli increased, and a 5 log reduction of cells was achieved following exposure to pressures in excess 250 MPa. In contrast, little inactivation was observed for L. innocua with pressures ranging between 250 and 350 MPa. However, several authors have been demonstrated the major efficacy of HPH against Gramnegative bacteria. The addition of 10 IU nisin, together with HPH, did not exhibit significant additional E. coli inactivation, but interactions were observed with L. innocua. On the other hand nisin is reported to be active against Gram-positive bacteria (Siroli et al., 2015).

## POTENTIAL AND APPLICATION OF HIGH PRESSURE HOMOGENIZATION FOR SPORE INACTIVATION IN VITRO AND FOOD SYSTEMS

Bacterial spores represent one of the major hazard in food safety due to their high resistance to most hurdles. In particular, they are resistant to elevated temperatures (80◦C), which distinguishes them from vegetative cells. Thermal sterilization is the method to eliminate spores in most food applications, as it provides the highest guarantee of sterility. From a safety point of view, spores of Bacillus spp. and Clostridium spp. are greatly resistant to several treatment such as heat, desiccation, lack of nutrients, exposure to UV and gamma radiation, organic chemicals, and oxidizing agents. In general, the heat resistance of spores depends on conditions such as elevated sporulation temperature, the presence of minerals and dipicolinic acid (DPA), and core dehydration. The endospores are composed of a central core, which is surrounded by several protective layers. The outermost

layer, the exosporium, is not present in spores of all species, and is the primary site of contact with the environment. Between the outer and inner membrane, there is the cortex. Since the spore structure plays a major role in spore resistance, the spore inactivation in foods requires high levels of heat treatments, which can in turn have negative effects on the sensory and nutritional profile (Schubert and Beaudet, 2011; Reineke et al., 2013; Georget et al., 2014a,b; Dong et al., 2015). For this reason, alternative and convenient methods have been studied in recent years (Chaves-López et al., 2009). In order to increase antimicrobial effectiveness and reduce side effects on food quality, the application of combined hurdles has also received great attention. Extensive literature indicates that the effects of combined stresses on microbial growth and survival may be additive or synergistic, when the outcome is usually significantly greater than the additive response (Tapia de Daza et al., 1996; Leistner, 2000; Ross et al., 2003), because of the disturbing action on microbial homeostasis in several respects (Leistner, 2000).

Because of its great potential for microbial inactivation, several authors have studied the effects of HPH or UHPH, when applied individually or in combination with other mild physical or chemical stresses (heat and H2O2), on the inactivation of Bacillus and Clostridium spores, whose genera, from a safety point of view, the most important species belong. The analysis of the literature shows that the major application for HPH and UHPH regards the inactivation of spores of spoiling bacteria while the reports dealing with the inactivation of spores from pathogenic species are sporadic. Focusing on pathogenic species, Chaves-López et al. (2009) evaluated the influence of HPH treatment, applied individually (one, two, or three cycles) or in combination with other mild physical or chemical stresses (mild heat treatment, H2O2, and low pH), on the capability of B. cereus and B. subtilis spore, suspended in sterilized double distilled water, to form colonies. These Author determined also the effects of the treatments applied on the release of DPA from the spores, since spore resistance to stresses such as temperature and pressure has been correlated to their ability to retain DPA, present in the core region of the dormant spores (Setlow, 2000; Cortezzo et al., 2004). The Authors outlined how the application of specific stress sequences can significantly inactivate B. cereus spores. The remarkable efficacy of repeated cycles at 150 MPa suggested that dynamic high pressure, particularly applied in combination with other sub-lethal stresses, could be a useful and innovative tool for B. cereus control in fluid foods. In particular, the Authors showed that, although plate count only slightly decreased in all the strains when one cycle of HPH at 150 MPa was applied alone, the spores released significant levels of DPA (up to 28%) that could indicate a possible disruption of spore layers. Three consecutive cycles of HPH determined high reduction of colony count (about 5 log CFU/ml) and high DPA release (52%). Among the stress conditions applied, it was observed that only the thermal shock after one HPH cycle reduced the colony count of 2.3 log CFU/ml and induced a DPA release up to 57%. Pinho et al. (2011) evaluated the inactivation of Clostridium sporogenes PA 3679 spores (considered as harmless twin of C. botulinum) by HPH in model system such as skim milk, showing that that pressures up to 300 MPa were not able to cause any reduction on spore counts or promote changes on their thermal resistance. The application of heat shock (100◦C/15 min) before HPH treatment and the homogenization process realized at mild inlet temperature (45◦C), which results in homogenization temperature of around 84◦C at 300 MPa, also did not cause reduction on viable spores counts. A few spores reduction (0.67 logarithmic cycles) were only observed when the milk samples were subjected to homogenization treatment for 16 cycles (multiple passes) at 300 MPa. Therefore, although HPH be recognized as an effective method for milk pasteurization, in this specific case, HPH process is not able to guarantee the commercial sterility of milk, being necessary the association of the homogenization with another preservative method, as refrigeration.

Also Amador Espejo et al. (2014) tested the ability of Ultra High-Pressure Homogenization treatments at 300 MPa with inlet temperatures of 55, 65, 75, and 85◦C to inactivate B. cereus spores inoculated into commercial ultra high temperature treated whole milk in order to evaluate the inactivation level achieved. These Authors provided important evidence of the suitability of UHPH technology for the inactivation of spores in high numbers, leading to the possibility of obtaining commercially sterile milk. In fact, UHPH conditions at 300 MPa with a inlet temperature of 75 and 85◦C were capable of a spore inactivation of 5 log CFU/ml. Furthermore, under these processing conditions, commercial sterility (evaluated as the complete inactivation of the inoculated spores) was obtained in milk treated at 300 MPa with inlet temperature of 75◦C.

### POTENTIAL OF HIGH PRESSURE HOMOGENIZATION TO INCREASE THE ANTIMICROBIAL ACTIVITY OF ENZYME OF FOOD INTEREST

Because of the great interest within the food industry in aldehydes, ketones, esters as natural antimicrobial compounds (Burt, 2004; Patrignani et al., 2015b) or enzymes such as lysozyme, lactoperoxidase system and lactoferrin, many Authors have tried to find strategies to enhance their efficacy in foods. In general, the approaches used are aimed to destabilize the microbial outer membranes of Gram-negative bacteria or to modify the chemical structure of the enzyme. For example, Ibrahim et al. (1993) and Bernkop-Schnurch et al. (1998) showed that the antimicrobial spectrum of lysozyme can be extended, including also Gram-negative bacteria, by slight chemical modification of the protein with hydrophobic ligands throughtout thermal treatment. These modifications generate an amphitropic protein able to spread the cytoplasmic membrane (Vannini et al., 2004; Iucci et al., 2007). Peptic digestion or heat treatment are reported to augment the antimicrobial activity of lactoferrin. In this perspective, several authors tried to modify the enzyme structure, and

consequently its activity, by using HPH in order to increase food safety. On the other hand the effect of HPH on naturally occurring food enzymes involved also in shelf-life, ripening and functionality of several matrices has been demonstrated (Kheadr et al., 2002; Lanciotti et al., 2004b, 2006; Vannini et al., 2008).

From a safety point of view, Vannini et al. (2004) evaluated the effect of HPH on the activity of antimicrobial enzymes such as lysozyme and lactoperoxidase against a selected group of Gram-positive and Gram-negative species inoculated in skim milk, formulating the hypothesis that the interaction of HPH and enzymes is associated to conformational modifications of the two proteins with a consequent enhancement of their activity. L. monocytogenes was the most pressure resistant species while S. typhimurium, S. aureus, and S. enteritidis were found to be very sensitive to the hyperbaric treatment. The enzyme addition enhanced the immediate pressure efficacy on almost all the considered species as indicated by their instantaneous viability loss following the treatment. Moreover, the combination of the enzyme and HPH significantly affected the recovery and growth dynamics of the considered species. Although L. monocytogenes was slightly sensitive to pressure, the combination of the two stress factors induced a significant viability loss within 3 h and an extension of lag phases in skim milk during incubation at 37◦C. Iucci et al. (2007) investigated the effects of HPH treatment at 100 MPa, in comparison to different heat treatments, 70◦C for 30 s, 70◦C for 5 min or 100◦C for 5 min, on the activity of lysozyme and lactoferrin. Their antimicrobial activities were tested on L. monocytogenes inoculated in milk or cultural medium. The results indicated that antimicrobial activities of lactoferrin and lysozyme were enhanced and/or accelerated by HPH treatment. Particularly, the highest immediate inactivation values were recorded when L. monocytogenes cells were added to HPHtreated lactoferrin, processed simultaneously or separately with the target microorganism. Although to a lesser extent than HPH treatment the heat treatments applied also were able to increase the antimicrobial activity of lysozyme. The Author suggested that the large supramolecular structure is disrupted under pressure, allowing the components to move freely and become independent of the original structure. Interactions can reform when the pressure instantaneously decreases but the original structure is not reformed because of the independent movements of the components.

Patrignani et al. (2013b) found for S. enteritidis, inoculated in LWE, HPH inactivation linearly correlated to the number of passes at 100 MPa underling as for LWE, it is important to take into consideration the additional effect of HPH on the antimicrobial activity of naturally occurring enzymes, such as lysozyme. In fact, dynamic pressure is reported to enhance antimicrobial activity of lysozyme and lactoferrin, probably due to the change of the exposure of hydrophobic regions (Iucci et al., 2007). Guerzoni et al. (2002) hypothesized that the inactivation of S. enteritidis in egg based formulation treated at different HPH pressure and its inability to re grow in the system at 10◦C was due to the increase of activity of HPH treated lysozyme from eggs. Also Velazquez-Estrada et al. (2008) attributed higher inactivation of S. enterica serovar Senftenberg 775Win HPH LWE to effects of pressure on cells and naturally occurring antimicrobial enzymes. Diels et al. (2005) observed that, above 150 MPa, E. coli became more sensitive to lysozyme when this compound was added before HPH treatment compared with adding the enzyme after the treatments, indicating that HPH treatment increases effectiveness and action spectrum of lysozyme.

# CONCLUSION

The critical evaluation of the available literature data has showed the great potential of HPH for microbial inactivation and food safety purpose. Since 1990 several Authors have been tested its potentialities in vitro and real systems, demonstrating its different ability for pathogenic species inactivation in relation to the strains considered, the food matrix and technological procedures adopted. However, until the introduction of new valve design and ultra high pressure homogenizers, able to reach pressures of 400 MPa, this technology was implemented in food industry only for fat globule reduction, for juice treatment and emulsion creation. The introduction of these new variables have opened new field in food sector, also for food decontamination and permitting to replace or minimize the traditional thermal treatments generally applied for the safety purposes. The new main applications regard the treatment of milk (for consumption or dairy product manufacture), fruit and vegetable juice, vegetable milks, and food component (such as enzymes) obtaining more stable and safer products without detrimental effects on quality properties. In fact, in the most of the cases, the literature data have underlined an improvement of the sensory and nutritional properties and stability of the HPH and UHPH treated products. The replacement of traditional thermal treatment can represent an advantage for the industry since this HPH is a cold technology with a lower impact on the environment, more sustainable, saving energy, time and additional costs. Moreover the literature data have demonstrated that HPH, when applied to the milk for cheesemaking, can increase the cheese yield, reduce the cheese hydrolytic patterns, reducing the costs for ripening.

Moreover, another important improvement of the state of the art was done for the inactivation of resistant endospores, which still represent a great challenge for food industry. The literature data have pointed out that the combination of UHPH and several chemico-physical hurdles can be regarded as tool for spore inactivation especially for milk based products. Further improve of food safety and functional properties could be achieved exploiting the well recognized dynamic pressure potential to obtain nanoparticles of antimicrobial molecules or functional ingredients. However, also in this case the reduction of thermal treatment can represent an advantage for food industry and contribute to the maintenance of the quality and nutritional properties of foods.

However, although the achieved aims, UHPH has not been yet implemented in food industry since conventional and ultra

homogenizer do not guarantee by themselves the sterilization and, subsequently, packaging of foods in aseptic conditions is needed. This results in a great disadvantage for food industry because it limits the implementation of this technology in food process as full alternative to thermal treatment.

#### REFERENCES


#### AUTHOR CONTRIBUTIONS

FP and RL put in writing this review together consulting the available literature, patents on the topic and trying to investigate several aspects on high pressure homogenization and safety.


chemico-physical characteristics of goat cheeses. J. Dairy Sci. 82, 851–862. doi: 10.3168/jds.S0022-0302(99)75303-8



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Patrignani and Lanciotti. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Detection of Cronobacter Genus in Powdered Infant Formula by Enzyme-linked Immunosorbent Assay Using Anti-Cronobacter Antibody

#### Xinjie Song<sup>1</sup>† , Shruti Shukla<sup>1</sup>† , Gibaek Lee<sup>1</sup> , Sunhyun Park<sup>2</sup> and Myunghee Kim<sup>1</sup> \*

<sup>1</sup> Department of Food Science and Technology, Yeungnam University, Gyeongsan, South Korea, <sup>2</sup> Food Standard Research Center, Korea Food Research Institute, Seongnam, South Korea

Cronobacter species (Cronobacter spp.) are hazardous foodborne pathogens associated with baby food, powdered infant formula (PIF). To develop a rapid and sensitive method for simultaneous detection of seven Cronobacter spp. in PIF, an indirect non-competitive enzyme-linked immunosorbent assay (INC-ELISA) was developed based on a novel immunoglobulin G (IgG), anti-Cronobacter IgG. The developed INC-ELISA was able to detect seven Cronobacter spp. at concentrations ranging from (5.6 ± 0.30) × 10<sup>3</sup> to (2.1 ± 0.01) × 10<sup>5</sup> colony forming unit (CFU)/mL in pure culture. Further, INC-ELISA employing anti-Cronobacter IgG was applicable for analysis of PIF samples contaminated with less than <10 cells of Cronobacter spp. per 25 g of PIF in 36 h. The developed antibody showed slight cross-reactivity with Franconibacter pulveris (LMG 24057) at high concentration (10<sup>8</sup> CFU/mL). The INC-ELISA method displayed excellent specificity without compromising cross-reactivity with other foodborne pathogens. The INC-ELISA assay method developed in this study using a novel anti-Cronobacter IgG facilitated highly sensitive, efficient, and rapid detection of Cronobacter spp. in baby food.

Keywords: Cronobacter species, immunoglobulin G, indirect non-competitive enzyme-linked immunosorbent assay, genus-specificity, powdered infant formula

# INTRODUCTION

Bacteria of the genus Enterobacter are increasingly important human pathogens that cause neonatal meningitis and sepsis (Hunter and Bean, 2013). Enterobacterial infections have become a worldwide health problem partially due to the lack of development of preventive treatments and the emergence of new multi-drug resistant strains. Enterobacter sakazakii has recently been reclassified as a Cronobacter (Strydom et al., 2012). Cronobacter species (Cronobacter spp.), previously known as E. sakazakii, are foodborne pathogens that pose a high risk of infection to neonates as well as immuno-compromised individuals (World Health Organization [WHO], 2007). Cronobacter spp. affect the central nervous system of infants, and survivors often suffer from severe neurological impairments such as hydrocephalus, quadriplegia, and developmental delays (Lai, 2001; Strydom et al., 2012; Jaradat et al., 2014). Species level classification of the Cronobacter genus, which includes C. sakazakii, C. muytjensii, C. malonaticus, C. turicensis, C. dublinensis, C. universalis, and C. condimenti, was recently defined by Iversen et al. (2008)

#### Edited by:

Maria Schirone, University of Teramo, Italy

#### Reviewed by:

Zhao Chen, Clemson University, USA Gonçalo Nieto Almeida, Catholic University of Portugal, Portugal

#### \*Correspondence:

Myunghee Kim foodtech@ynu.ac.kr †These authors have contributed equally to this work.

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

> Received: 26 April 2016 Accepted: 06 July 2016 Published: 21 July 2016

#### Citation:

Song X, Shukla S, Lee G, Park S and Kim M (2016) Detection of Cronobacter Genus in Powdered Infant Formula by Enzyme-linked Immunosorbent Assay Using Anti-Cronobacter Antibody. Front. Microbiol. 7:1124. doi: 10.3389/fmicb.2016.01124

and Joseph et al. (2012b). Cronobacter spp. have been isolated from plant-based food products, including cereal, fruits, vegetables, legumes, herbs, and spices, as well as from animalbased food materials such as milk, meat, and fish (Friedemann, 2007; Lee et al., 2012).

All Cronobacter spp., except C. condimenti, have been associated with human infections (Cruz-Córdova et al., 2012). Historically, Cronobacter spp. have been implicated in newborn and infant infections, causing meningitis, necrotizing enterocolitis, and bacteremia (Healy et al., 2010). However, not all Cronobacter spp. are linked to infantile infections, and it is thought that virulence among strains may vary. C. sakazakii, C. malonaticus, and C. turicensis are most often isolated from infantile cases (Joseph and Forsythe, 2011). Recent research based on international microbiological standards suggested that all species of Cronobacter must be absent in 10 grams of powdered infant formula (PIF) (Odeyemi and Sani, 2016). For an understanding of recent unresolved issues persisting with respect to taxonomy, sources and clinical relevance, and for suggestions on how to safely feed premature neonates (Holy and Forsythe, 2014), it is plausible that virulence determinants have evolved in certain lineages (Joseph et al., 2012a).

Although reservoirs of Cronobacter spp. and their modes of transmission are still unknown, Cronobacter spp. have been suggested as a source of food contamination, with rodents and flies serving as a secondary route of contamination (Jung and Park, 2006; World Health Organization [WHO], 2007). According to a report by Jung and Park (2006), 20% of PIF samples were found contaminated with Cronobacter in the Republic of Korea. Lee et al. (2012) also reported similar results in which Cronobacter spp. were isolated from 18.6% of detected food samples. Although a number of reported cases of Cronobacter infection are quite low, sequelae can occur with high mortality rates (Lai, 2001; United States Food and Drug Administration [USFDA], 2002; Friedemann, 2007). World Health Organization [WHO] (2007), classified Cronobacter together with Salmonella as group A pathogens associated with PIF with clear evidence of illness in infants. These categories of organisms were based on their risk of illness to infants.

Culture-dependent isolation and assay methods for complete analysis of Cronobacter spp. from PIF usually require 5 to 7 days (United States Food and Drug Administration [USFDA], 2002). A further method was recommended by the International Organization for Standardization and the International Dairy Federation as ISO 22964 (Anonymous, 2006). This method includes a pre-enrichment in buffered peptone water (BPW), a selective enrichment in modified lauryl sulfate tryptose broth containing vancomycin and isolation of presumptive C. sakazakii colonies colored in blue-green on E. sakazakii isolation agar. These colonies should be streaked on tryptic soy agar and resulting yellow colonies are indicatory for C. sakazakii. However, this method was also time- and labor-consuming, as it usually requires 5 to 6 days to obtain a positive result (Norberg et al., 2012).

Lampel and Chen (2009) previously described a new method for the isolation and detection of Cronobacter spp. from PIF using a real-time PCR-based assay and chromogenic agar. In their study, suspended cells were isolated from enrichment culture, streaked onto chromogenic agar, and confirmed by real-time PCR assay. Mullane et al. (2006) also developed a method using cationic-magnetic beads to capture Cronobacter spp., and subsequent identification was performed after washing off bound cells from the capture phase and plating them onto Druggan–Forsythe–Iversen formulation agar to detect 1 to 5 colony forming unit (CFU)/500 g of PIF within 24 h. However, these PCR-based methods have significant technical requirements of ultra-pure reagents and chemicals along with high equipment costs. Therefore, a rapid, sensitive, and inexpensive method is needed for the detection of Cronobacter spp.

Polyclonal antibody can be obtained within a short time (4 to 8 weeks) with minimal financial investment, whereas it takes about 3 to 6 months to produce monoclonal antibodies (Leenaars and Hendriksen, 2005). Polyclonal antibody is commonly used in immunological methods, including enzyme-linked immunosorbent assay (ELISA), for the detection of foodborne pathogens (Brigmon et al., 1992; Kumar et al., 2008; Velusamy et al., 2010). It is also critical for developing a rapid and genusspecific method for identification of Cronobacter spp. in the context of food safety issues. Previously, a sandwich ELISA method was developed for the detection of C. muytjensii with a detection limit of 6.3 × 10<sup>4</sup> CFU/mL using a species-specific anti-C. muytjensii immunoglobulin G (IgG) (Park et al., 2012). In addition, Xu et al. (2014) also developed a polyclonal and monoclonal antibodies-based indirect ELISA and a sandwich ELISA for the detection of Cronobacter spp. The indirect ELISA detected all species of Cronobacter assayed, and the limit of detection (LOD) was established as 10<sup>5</sup> CFU/mL. In contrast, sandwich ELISA was specific for C. sakazakii and exhibited greater sensitivity than indirect ELISA (LOD of 2 × 10<sup>4</sup> CFU/mL). Following 10 h of enrichment, Cronobacter spp. were detected using either of the two analytical methods in samples inoculated with 1 CFU/100 g PIF. The results from this study demonstrated that both of these novel ELISAs were specific, sensitive, and rapid assays for the screening of pathogenic Cronobacter spp. in PIF.

A widely available polyclonal antibody capable of detecting Cronobacter spp. could be used to avoid expensive and time-consuming methods. Thus, we decided to produce and characterize a polyclonal antibody for seven Cronobacter spp. To our knowledge, there has been no report on the ELISA-based detection of multiple Cronobacter spp. In the current work, an indirect non-competitive ELISA (INC-ELISA) method based on anti-Cronobacter IgG was developed in pure culture and applied for the detection of seven Cronobacter spp. in PIF.

#### MATERIALS AND METHODS

#### Strains and Reagents

In the present study, seven Cronobacter strains were used to produce antibody and to assess the cross-reactivity of developed method, while eight other bacterial strains belonging to different genera were selected to check the cross-reactivity

of developed assay due to their partial gene sequence similarity and to confirm the accurate detection of pathogen contaminants such as Salmonella spp., Citrobacter spp., and Bacillus cereus (Iversen et al., 2004; World Health Organization [WHO], 2007; Pinto et al., 2013). In the present study, B. cereus and S. Enteritidis were selected since it was found that the chances of contamination in PIF were highly concerned due to the presence of B. cereus and S. Enteritidis (Giammanco et al., 2011; Pinto et al., 2013). Giammanco et al. (2011) also reported that Salmonella spp. and Cronobacter spp. (formerly E. sakazakii) are the microorganisms of greatest concern in PIF. In addition, regarding the use of Citrobacter freundii, Giammanco et al. (2011) observed that molecular epidemiological survey of C. freundii was misidentified as Cronobacter spp. isolated from PIF. The phylogenetic tree also showed that C. freundii and Cronobacter spp. are very closely related species based on their smaller amounts of mahalanobis distances than other species. Hence, it was suggested that C. freundii may also be an under-reported cause of bacterial infection, especially in high risk neonates, due to misidentification. B. cereus (KCCM 40935), Buttiauxella noackiae (ATCC 51713), C. condimenti (LMG 26250), C. dublinensis (LMG 23823), C. malonaticus (LMG 23826), C. muytjensii (CDC 3523-75), C. turicensis (LMG 23827), C. muytjensii (ATCC 51329), C. sakazakii (ATCC 29544), C. sakazakii (ATCC 29004), C. universalis (LMG 26249), C. freundi (ATCC 8090), Escherichia coli (ATCC 39418), Franconibacter helveticus (LMG 23732), Franconibacter pulveris (LMG 24057), and Salmonella Typhimurium (ATCC 13311) were used in this study. C. muytjensii (CDC 3523–75) was donated by Dr. Carol Iversen from University College Dublin, Ireland. Other strains used in this study were purchased from the American Type Culture Collection (ATCC; Manassas, VA, USA) and Korean Culture Center of Microorganisms (KCCM; Seoul, Korea). Strains indicated with LMG were purchased from the Belgian Coordinated Collections of Microorganisms (BCCM; Gent, Belgium). All strains were cultured in nutrient broth (NB) for 18 h at 37◦C in a shaking incubator (150 rpm).

NB, nutrient agar (NA), peptone, and skim milk were purchased from Difco (Franklin Lakes, NJ, USA). Enterobacteriaceae enrichment (EE) broth and violet red bile glucose (VRBG) agar were purchased from MB cell (Seoul, Korea). Sodium carbonate, sodium bicarbonate, sodium azide, potassium phosphate monobasic, potassium phosphate dibasic, sodium chloride, caprylic acid, ammonium sulfate, and alkaline phosphatase yellow liquid substrate p-nitrophenyl phosphate (pNPP) were purchased from Sigma (St. Louis, MO, USA). Phosphatase-labeled goat anti-rabbit IgG was purchased from Kierkegaard & Perry Laboratories, Inc. (Gaithersburg, MD, USA). Lipopolysaccharide (LPS) extraction kit was purchased from Intron Biotechnology (Seongnam, Korea).

#### Animal Care Ethics

Animal use protocol was reviewed by the committee members of Yeungnam University and approved by Korea Food and Drug Administration, Republic of Korea (Animal Ethics License No. 2013-012 and 2012-010).

# Preparation of Immunogen and Immunization

Three types of immunogens were prepared for the development of a genus-specific antibody against Cronobacter spp. A sonicated cell protein (SCP) mixture of seven Cronobacter spp. was used in this study. The seven Cronobacter spp. were separately cultured in NB at 37◦C for 18 h to a concentration of 1 × 10<sup>9</sup> CFU/mL. Each culture (30 mL) was centrifuged at 3,000 × g at 4◦C for 30 min. The pellets were washed three times with 30 mL of 0.01 M phosphate-buffered saline (PBS) and then suspended in 10 mL of 0.01 M PBS for sonication under 20 KHz power for 5 min on ice. Supernatant was collected after centrifugation at 12,000 × g, 4◦C for 20 min. Protein concentrations of supernatants were checked with a Bradford kit, after which SCP solutions were adjusted to a concentration of 0.5 mg/L with 0.01 M PBS. SCP solutions were stored at −20◦C before use. For immunization, equal volumes of the seven SCP solutions were mixed for use as an immunogen in New Zealand white rabbits.

LPS mixture of the seven Cronobacter spp. was prepared using an LPS extraction kit following the instructions of the manufacturer. Each Cronobacter spp. culture was centrifuged at 10,000 × g at room temperature for 20 min to harvest bacterial cells. Lysis buffer was then added and vortexed vigorously, followed by addition of chloroform, vortexing for 20 s, and incubated at room temperature for 5 min. The mixture was then centrifuged at 10,000 × g for 10 min at 4◦C. An aliquot of supernatant was transferred to a new tube, mixed well with purification buffer, and kept at −20◦C for 10 min. The solution was then centrifuged at 10,000 × g for 15 min at 4 ◦C, after which the upper layer was removed to obtain an LPS pellet. The pellet was washed three times with 1 mL of ethanol, centrifuged for 3 min at 10,000 × g at 4◦C, and then collected and dried at room temperature. For the LPS solution, 50 µL of Tris-HCl buffer (10 mM, pH 8.0) was used to dissolve the LPS pellet. LPS solution of the seven Cronobacter spp. was prepared by mixing an equal volume of each LPS solution for use as an immunogen in New Zealand white rabbits.

Formalin-killed cell (FKC) mixture was also prepared as an immunogen. Cultures of the seven Cronobacter spp. were applied as treatments with 0.5% formalin for 24 h, followed by centrifugation (3,000 × g, 4◦C, for 30 min) according to the method of Song and Kim (2013). First injection was administered with a mixture of immunogen and Freund's complete adjuvant (1:1, v/v). Second and third injections were administered using the same mixture of immunogen and Freund's incomplete adjuvant (1:1, v/v) at 4 weeks after the first injection. Immunogen was injected into the back of rabbits at four sites at a concentration of 0.25 mL/site. Blood samples were taken every week until 18 weeks after the first injection.

#### Preparation and Purification of IgG

Blood samples were centrifuged (10,000 × g) for 30 min at 4◦C to separate the anti-sera. Anti-Cronobacter IgG

was purified from the anti-sera by caprylic acid and ammonium sulfate precipitation as described by McKinney and Parkinson (1987) and according to further modified methods of Shukla et al. (2012) and Song and Kim (2013).

### Sodium Dodecyl Sulfate-Polyacrylamide Gel Electrophoresis (SDS-PAGE) for Purified Anti-Cronobacter IgG

The purity of various anti-Cronobacter IgGs preparations was checked using SDS-PAGE (Atto, Tokyo, Japan) under reduced conditions as described by Park et al. (2012) with some modifications.

# Development of INC-ELISA for Detection of Cronobacter spp.

Rabbit anti-Cronobacter IgG was used to develop an INC-ELISA method for detection of Cronobacter spp. Standard response curve of the seven Cronobacter spp. was constructed using the following procedure. First, fresh cultures of all tested Cronobacter spp. were diluted decimally with 0.05 M carbonate buffer (pH 9.5) in order to coat 96-well plates (SPL Life Sciences, Gyeonggi-do, Korea), whereas only carbonate buffer was used as a negative control. Simultaneously, coated cultures of Cronobacter spp. were counted on NA and VRBG agar plates for determination of concentration. After coating at 4◦C overnight, 96-well plates were washed three times with 0.01 M PBS (pH 7.0) and blocked with 200 µL of 5% skim milk at 37◦C for 2 h. Ninety six-well plates were then washed with 0.01 mol/L PBS-0.05% Tween 20 (PBST), added with rabbit anti-Cronobacter IgG, and incubated at 37◦C for 1 h. Then, plates were washed again with 0.01 M PBST, after which 100 µL of phosphatase-labeled goat anti-rabbit IgG was added and the plates incubated at 37◦C for 1 h. Plates were washed again with 0.01 M PBST, after which 50 µL of pNPP liquid substrate was added to each well for a 30 min enzyme-substrate reaction. Finally, 50 µL of 0.01 M NaOH solution was added to stop the reaction, after which the final reaction mixture was analyzed using an Infinite M200 (Tecan; Seestrasse, Switzerland). A detection limit of the developed INC-ELISA was determined by using a data linearity function and standard deviation of responses for the calibration standard and samples, respectively (Hubaux and Vos, 1970).

## Specificity and Sensitivity of INC-ELISA Assay

Seven common Cronobacter foodborne pathogens (**Table 1**) were used for the specificity test of the developed INC-ELISA. Cultures of tested foodborne pathogens were serially diluted to final concentrations of 10<sup>0</sup> to 10<sup>8</sup> CFU/mL and then used to check sensitivity of the developed INC-ELISA. All experiments were repeated three times.

# Food Trial Using Cronobacter-Spiked PIF Sample (Artificial Inoculation of Cronobacter into PIF Sample)

Firstly, the seven Cronobacter strains were separately cultured in NB at 37◦C for 20 h, after which the cultures were serially diluted with BPW to obtain the lowest final concentration (<10 CFU/mL) of Cronobacter live cells for use in further experiments. To examine the applicability of the developed anti-Cronobacter IgG against each Cronobacter strain, 25 g of PIF sample was aseptically placed into a flask, mixed with 225 mL of BPW as a pre-enriched broth, and separately spiked with 1 mL (less than 10 cells/25 g of PIF) of each Cronobacter species, followed by pre-enrichment incubation at 37◦C for 8 h. Following this, 1 mL of the pre-enrichment culture was inoculated into EE broth and incubated for 8 h of enrichment (Korea Food and Drug Administration [KFDA], 2014). Further, enriched aliquots of pre-treated food samples were tested using the developed INC-ELISA and standard USFDA methods. For control, non-spiked PIF samples were analyzed using the developed INC-ELISA and standard microbiological USFDA methods (United States Food and Drug Administration [USFDA], 2002). All experiments were performed in three replicates in order to maintain reliability of the work.

#### Statistical Analysis

The detection limit was calculated as the average value of absorbance at zero concentration with three standard deviations (Park et al., 2012). All experiments and results were analyzed in at least three trials for proving reliable and reproducible data.

TABLE 1 | Detection of Cronobacter species in pure culture using the developed indirect non-competitive enzyme-linked immunosorbent assay.


# RESULTS

#### Characteristics of Developed Antibodies

As shown in **Figure 1**, anti-serum containing anti-Cronobacter IgG developed by the LPS mixture showed no titer for any Cronobacter species. **Figure 1** provides the fact that LPS antigen of Cronobacter is not appropriate for the production of anti-Cronobacter antibody. For antibodies developed using FKC and SCP mixtures of Cronobacter spp., titers of the developed rabbit anti-Cronobacter serum containing IgG were determined by INC-ELISA after 18 weeks of immunization, as shown in **Figures 2** and **3**. Titers of the developed anti-Cronobacter seru containing IgG increased after the first injection as well as gradually increased after the second and third injections. Titers of anti-Cronobacter serum containing IgG developed from the FKC mixture of Cronobacter spp. (**Figure 2**) were in the following order: C. muytjensii, C. turicensis, C. malonaticus, C. condimenti, C. sakazakii, C. dublinensis, and C. universalis. Comparatively,

cells mixture of Cronobacter spp., lane 5: anti-Cronobacter IgG developed from lipopolysaccharide mixture of Cronobacter spp., lane 6: anti-serum developed from sonicated cell protein mixture of Cronobacter spp., lane 7: anti-serum developed from formalin killed cells mixture of Cronobacter spp., lane 8: anti-serum developed from lipopolysaccharide mixture of Cronobacter spp.

anti-Cronobacter serum containing IgG developed from an SCP mixture of the seven Cronobacter spp. showed higher titers against the seven Cronobacter spp. (**Figure 3**), which were in the following order: C. muytjensii, C. condimenti, C. turicensis, C. sakazakii, C. malonaticus, C. dublinensis, and C. universalis. It is well known that an increased titer of developed antibody results in effective performance during immunoassay. The developed and purified anti-Cronobacter IgG showed very high purity as compared to the commercial rabbit IgG (**Figure 4**). Developed antibody as well as commercial rabbit IgG (lanes 2, 3, 4, and 5) showed a strong band around 51 kDa and a light band around 25 kDa, respectively. These results confirm that our developed rabbit anti-Cronobacter IgG with high purity is comparable to commercial rabbit IgG.

# INC-ELISA for Detection of Cronobacter spp. Using Developed Anti-Cronobacter IgG

An INC-ELISA method was developed to detect the seven Cronobacter spp. using rabbit anti-Cronobacter IgG developed from the aforementioned SCP mixture. The developed INC-ELISA method showed excellent sensitivity toward Cronobacter spp., including C. condimenti, C. dublinensis, C. malonaticus, C. muytjensii, C. sakazakii, C. turicensis, and C. universalis, confirming its ability to detect all seven Cronobacter spp. in a single test (**Table 1**). Furthermore, the developed INC-ELISA was highly sensitive to all seven different Cronobacter spp. with a detection limit ranging from (5.6 ± 0.30) × 10<sup>3</sup> to (2.1 ± 0.01) × 10<sup>5</sup> CFU/mL (**Table 1**).

# Specificity of Developed INC-ELISA

According to **Table 2**, the developed INC-ELISA showed no cross-reactivity with any of the other tested bacterial genera. The recognition difference between targeted bacteria (positive value; P value) and non-targeted bacteria (negative value; N value) is an important parameter affecting the results of the

TABLE 2 | Cross-reactivity of the developed indirect non-competitive enzyme-linked immunosorbent assay with non-Cronobacter species in pure culture.


immunoassay. In the present study, non-targeted bacteria (non-Cronobacter) showed a P/N value of <2 (**Table 2**), which is considered as a negative result, whereas targeted strains showed a P/N value of >2, which is considered as a positive result (**Table 1**). In support of this, the P/N value for F. pulveris was around 1.942 (**Table 2**) where a P/N value >2 is considered as a positive result. Compared to several other methods, our newly developed INC-ELISA based on rabbit anti-Cronobacter IgG is less expensive and easier to perform since it does not require expensive reagents or probes and has an assay time of only 36 h comprising 8 h of pre-enrichment, 8 h of enrichment, and 20 h of INC-ELISA. During assay development, accuracy of the rapid detection method was confirmed by standard plate colony counting techniques. Moreover, the developed rapid assay is accurate enough to avoid repeated colony counting steps to save detection time.

#### Detection of Cronobacter spp. in Spiked PIF Samples

To detect small amounts of Cronobacter spp. cells using the developed method, PIF samples were separately spiked with each of the seven Cronobacter spp. at low concentrations (<10 cells/25 g). All PIF samples spiked separately with C. condimenti, C. dublinesis, C. malonaticus, C. muytjensii, C. sakazakii, C. turicensis, and C. universalis showed the presence of Cronobacter spp. in both the INC-ELISA and USFDA/KFDA methods, whereas unspiked samples showed the absence of Cronobacter spp. Specific enrichment was carried out in EE broth as per KFDA standard methods (Korea Food and Drug Administration [KFDA], 2014). In the present study, the detection limit of the developed INC-ELISA against Cronobacter spp. was found to be <10 cells/25 g of PIF after 16 h of enrichment (**Table 3**).

Further, to confirm the results obtained using the developed antibody and INC-ELISA for Cronobacter detection, conventional standard plate count technique using the spreading method on NA and VRBG agar plates was carried out. The PIF samples artificially contaminated with each Cronobacter spp. separately (<10 cells/25 g) after 8 h of pre-enrichment and 8 h of enrichment in EE broth showed similar bacterial counts for each Cronobacter spp. in the range of 1.1 × 10<sup>5</sup> – 2.9 × 10<sup>5</sup> CFU/mL on specific VRBG agar plates (**Table 3**). These results can give an idea to reconfirm the detection limit obtained in pure culture of Cronobacter spp. using the developed antibody and INC-ELISA assay (**Table 1**).

#### DISCUSSION

In this study, an INC-ELISA method based on our laboratoryproduced anti-Cronobacter IgG was developed and applied for the detection of seven Cronobacter spp. in PIF. This study employed three types of immunogens (LPS, FKC, and SCP) to produce three types of antibodies in rabbit for the detection of seven Cronobacter spp. As recommended by the Canadian Council on Animal Care [CCAC] (2002), a second injection of immunogen should be performed 3 to 6 weeks after the first immunization. Therefore, in the present work, two immunogen injections were made after the first injection at an interval of 4 weeks during the immunization period. As a result, IgG developed by LPS immunogen showed no titer for any Cronobacter spp. These results agree with previous findings of Luk and Lindberg (1991) and Jaradat et al. (2011), who were

TABLE 3 | Food trial for detection of Cronobacter and indirect non-Cronobacter species in artificially contaminated PIF using the developed non-competitive enzyme-linked immunosorbent assay.


+: Cronobacter spp. detected; −: Cronobacter spp. not detected.

unable to obtain stable antibodies against LPS from Cronobacter and Salmonella due to the structure and composition of LPS obtained from bacterial cells of Cronobacter and Salmonella.

Previously, Hochel and Skvor (2009) developed a serotypespecific polyclonal antibody against E. sakazakii (Cronobacter spp.) capable of detecting 13 E. sakazakii spp. from different isolations. Compared to the results of Hochel and Skvor (2009), the developed rabbit anti-Cronobacter IgG in this study showed a higher titer, purity, and specificity, which makes it suitable for developing a fast and simple immunoassay method for the detection of Cronobacter spp. Therefore, anti-Cronobacter IgG developed from an SCP mixture of the seven Cronobacter spp. was chosen to develop an INC-ELISA method for genusspecific detection of Cronobacter spp. Similarly, Liang et al. (2001) developed a monoclonal antibody for the detection of Bartonella spp. capable of specifically detecting all tested species of Bartonella.

In addition, Huang et al. (2013) designed two pairs of primers based on gyrB sequences for specific identification of C. sakazakii and C. dublinensis. Although these PCR-based methods can be used to rapidly analyze Cronobacter spp., their strict requirements in terms of technique, equipment, and probes limit their application (Ruan et al., 2013). Further, Zhang et al. (2010) reported a cross-priming amplification method combined with immunoblotting analysis for genus-specific detection of Cronobacter spp. under isothermal conditions.

Similarly, we developed a polyclonal antibody for C. muytjensii as well as fluorescence-based liposome immunoassay for the detection of C. muytjensii in our previous study (Song et al., 2015). In this study, the developed antibody was also able to detect other strains of C. muytjensii such as C. muytjensii ATCC 51329 and C. muytjensii CDC3523–75 (Song et al., 2015). Furthermore, as shown in **Table 2**, F. pulveris (LMG 24057) showed slight cross-reactivity with INC-ELISA at a high bacterial concentration (10<sup>8</sup> CFU/mL), which might be due to close resemblance between multi-locus sequence typing loci for Cronobacter genus members and Franconibacter genera (previously recognized in Enterobacter genus) (Forsythe et al., 2014). Recently, Xu et al. (2014) developed a polyclonal antibody for Cronobacter spp. by using heat-killed antigen preparation and evaluated its cross-reactivity against various Cronobacter strains of the same species. Their results confirmed cross-reactivity for all Cronobacter strains of the same species based on positive signals. The detection limit was observed in the range of 104–10<sup>5</sup> CFU/mL (Xu et al., 2014). Therefore, it can be hypothesized that the developed antibody and assay in the present study could be applicable for the detection of several Cronobacter strains of similar species (Xu et al., 2014).

The above findings confirm that the developed assay method in this study can detect Cronobacter spp. in PIF samples with high sensitivity within 36 h. Similarly, Blazkova et al. (2011) developed a method using an immunochromatographic strip for the detection of Cronobacter spp. with a detection limit less than 10 cells/10 g of PIF. In the present study, all PIF samples artificially contaminated with Cronobacter spp. were positively detected using the developed genus-specific anti-Cronobacter IgG antibody in 25 g of PIF after 16 h of sample pre-treatment (8 h of pre-enrichment and 8 h of enrichment) and 20 h of assay time. In contrast, detection and confirmation of this pathogen by the USFDA culture-based method takes up to 3– 5 days.

However, the developed method has limited differentiation ability for detection of each Cronobacter spp., which must be overcome to validate its practical and industrial usefulness. Hence, development of additional immunoassays based on the produced Cronobacter genus-specific antibody multiplexing format such as ELISA are in progress in our laboratory for the purpose of developing a novel assay method with reduced detection time, high sensitivity, and multiplexing detection ability.

In summary, this study focused on the development of a simple, quick, and sensitive genus-specific detection assay for Cronobacter spp. using rabbit anti-Cronobacter IgG. For this purpose, three immunogens, LPS mixture of Cronobacter spp., FKC mixture of Cronobacter spp., and SCP mixture of Cronobacter spp., were prepared and used to develop antibodies against Cronobacter spp. The newly developed rabbit anti-Cronobacter IgG was purified using caprylic acid and ammonium sulfate precipitation from anti-sera and showed high purity similar to the commercial rabbit IgG. The rabbit anti-Cronobacter IgG developed from the SCP mixture was used to develop a new INC-ELISA method for the detection of Cronobacter spp. The developed INC-ELISA showed excellent reactivity with all seven Cronobacter spp., excluding other non-Cronobacter foodborne pathogens. These results confirm that the developed INC-ELISA is highly sensitive, efficient, and rapid for the detection of Cronobacter spp. with a markedly reduced total detection time (from 5 days to 36 h), costs, and handling procedure. This is the first report of a genus-specific antibody and immunoassay method for the detection of Cronobacter spp. with the goal of reducing risk of Cronobacter spp. contamination in food.

#### AUTHOR CONTRIBUTIONS

XS, GL, and SP performed experiments and drafted manuscript; SS contributed interpretation, analyzed data, and wrote paper, MK contributed for conception, designed experiment, analyzed data, and provided technical support.

# ACKNOWLEDGMENTS

This research was supported by the Basic Science Research Program through the National Research Foundation, Republic of Korea (NRF), funded by the Ministry of Education (NRF-2014R1A2A1A11053211).

#### REFERENCES

fmicb-07-01124 July 21, 2016 Time: 11:19 # 9


monoclonal antibodies and MALDI-TOF mass spectrometry. BMC Microbiol. 11:148. doi: 10.1186/1471-2180-11-148



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Song, Shukla, Lee, Park and Kim. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Structural Analysis and Biological Toxicity of Aflatoxins B1 and B2 Degradation Products Following Detoxification by Ocimum basilicum and Cassia fistula Aqueous Extracts

Wajiha Iram1,2 \*, Tehmina Anjum<sup>1</sup> \*, Mazhar Iqbal<sup>2</sup> \*, Abdul Ghaffar<sup>3</sup> , Mateen Abbas<sup>4</sup> and Abdul Muqeet Khan<sup>4</sup>

1 Institute of Agricultural Sciences, University of the Punjab, Lahore, Pakistan, <sup>2</sup> Health Biotechnology Division, National Institute for Biotechnology and Genetic Engineering, Faisalabad, Pakistan, <sup>3</sup> Department of Chemistry, University of Engineering and Technology, Lahore, Pakistan, <sup>4</sup> Quality Operating Laboratory, University of Veterinary and Animal Sciences, Lahore, Pakistan

#### Edited by:

Giovanna Suzzi, Universitá degli Studi di Teramo, Italy

#### Reviewed by:

Robert Lawrence Brown, United States Department of Agriculture – Agricultural Research Service, USA Rosalba Lanciotti, University of Bologna, Italy

#### \*Correspondence:

Tehmina Anjum tehminaanjum@yahoo.com Wajiha Iram wajiha\_mppl@yahoo.com Mazhar Iqbal hamzamgondal@yahoo.com

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 25 April 2016 Accepted: 01 July 2016 Published: 14 July 2016

#### Citation:

Iram W, Anjum T, Iqbal M, Ghaffar A, Abbas M and Khan AM (2016) Structural Analysis and Biological Toxicity of Aflatoxins B1 and B2 Degradation Products Following Detoxification by Ocimum basilicum and Cassia fistula Aqueous Extracts. Front. Microbiol. 7:1105. doi: 10.3389/fmicb.2016.01105 This study showed the comparison between Ocimum basilicum and Cassia fistula (leaves and branch) aqueous extracts for their ability to detoxify of aflatoxins B1 and B2 (AFB1; 100 µg L−<sup>1</sup> and AFB2; 50 µg L−<sup>1</sup> ) by In Vitro assays and decontamination studies. Results indicated that O. basilicum leaves extract was found to be highly significant (P < 0.05) in degrading AFB1 and AFB2, i.e., 90.4 and 88.6%, respectively. However, O. basilicum branch, C. fistula leaves and branch extracts proved to be less efficient in degrading these aflatoxins, under optimized conditions, i.e., pH 8, temperature 30◦C and incubation period of 72 h. Moreover the antifungal activity of these plants extracts were also tested. The findings depicted that O. basilicum leaves extract showed maximum growth inhibition of aflatoxigenic isolates, i.e., 82–87% as compared to other tested plants extracts. The structural elucidation of degraded toxin products by LCMS/MS analysis showed that nine degraded products of AFB1 and AFB2 were formed. MS/MS spectra showed that most of the products were formed by the removal of double bond in the terminal furan ring and modification of lactone group indicating less toxicity as compared to parent compounds. Brine shrimps bioassay further confirmed the low toxicity of degraded products, showing that O. basilicum leaves extract can be used as an effective tool for the detoxification of aflatoxins.

Keywords: plant extract, aflatoxin, degradation, LCMS/MS, toxicity

# INTRODUCTION

Mycotoxins are chemically and biologically active secondary metabolites (of molecular weight ≤ 700), produced by filamentous fungi that readily colonize crops in the field or after harvest (Richard, 2007; Turner et al., 2010). Currently, more than 400 compounds are recognized as mycotoxins, among them aflatoxins have assumed significance due to their deleterious effects on humans, poultry, and livestock (Sur and Celik, 2003; Wild and Montesano, 2009). Aflatoxins are group of heterocyclic, oxygen containing mycotoxins that possess the bisdifuran ring system.

About 18 different types of aflatoxins are identified among them most commonly occurring ones are aflatoxins B1, B2, G1, G2, M1, and M2 which are mostly produced by Aspergillus flavus and A. parasiticus. Aflatoxins were first identified as the portable toxin that destroyed more than 100,000 turkey poults (turkey X disease) in England in the early 1960s (Kuhn and Ghannoum, 2003). In Kenya, (April 2004) an outbreak of jaundice with a high fatality rate was reported in humans due to aflatoxin poisoning from eating contaminated home grown maize, and resulted in 317 cases and 125 deaths (Baumgartner et al., 2005). There are strict regulations limiting mycotoxins levels in 77 countries, with specific regulatory levels for aflatoxins in food and feed stuffs (Bhatnagar et al., 2006). This shows that contamination of agricultural commodities with aflatoxin has been the subject of serious concern on an international level. Such contamination occurs by the invasion of aflatoxigenic strains ubiquitously found before and during harvesting, or by improper storage of agricultural commodities (Hwan and Choi, 2007). The food and agriculture organization (FAO) estimates that about 1000 million metric tons of foodstuff could be contaminated with mycotoxins each year (Bhat et al., 2010).

In recent years, the need to develop disease control measure as alternative to chemicals has been a priority of scientists worldwide (Reddy et al., 2009). Therefore it is important to find a practical, cost effective, and non-toxic method to prevent fungal and mycotoxin contamination in foods and feeds. Natural plant products are of interest as a source of safer or more effective substitutes for synthetic antimicrobial agents and may provide alternative way to prevent food or feed from fungal or mycotoxin contamination. Large-scale application of plant products have attracted the attention of several microbiologists as they are biodegradable, biologically safe, cost effective, renewable in nature and at the same time, conveniently used as ecofriendly technique to detoxify mycotoxins (Varma and Dubey, 1999). Previously documented literature showed that powder and extracts of various herbs and medicinal plants found to be effective in detoxifying aflatoxins and growth inhibition of toxigenic fungi. Krishnamurthy and Shashikala (2006) worked on inhibition of aflatoxin B1 production by Aspergillus flavus, isolated from soybean seeds by incubation with certain natural plant products at different time interval. The outcome of their study showed that captan, neem cake, leaf powder of Withania somnifera (Linn.), peel powder of Camellia sinensis (Linn.), Citrus medica (Linn.), and pongamia cake controlled the aflatoxin B1 production. They concluded that all the natural product treatments applied were significantly effective in inhibiting aflatoxin B1 production. In a more comprehensive study in Satish et al. (2007) scrutinized aqueous extract of 52 plants for their antifungal potential against eight important species of Aspergillus. They observed that among tested plants, aqueous extracts of Acacia nilotica (Linn.), Eucalyptus globules Labill, Achras zapota (Linn.), Datura stramonium (Linn.), Emblica officinalis (Gaertn.), Lawsonia inermis L. (henna), Mimusops elengi (Linn.), Peltophorum pterocarpum (DC.) Baker ex. K. Heyne, Polyalthia longifolia Benth. and Hook, Prosopis juliflora (Sw.) DC, Punica granatum (Linn.), and Syzygium cumini (Linn.) exhibited significant antifungal activity against Aspergillus species. Correspondingly Reddy et al. (2009) explored the potential of certain plant extracts and biocontrol agents for the reduction of aflatoxin B1 (AFB1) in stored rice. Among the plant extracts tested, Syzygium aromaticum (L.) Merr. Et Perry, Curcuma longa (L.), Allium sativum (L.), and Ocimum sanctum (Linn.) significantly inhibited the A. flavus growth and AFB1 production. Velazhahan et al. (2010) evaluated various medicinal plants extracts for their ability to detoxify aflatoxin G1 (AFG1) by thin-layer chromatography and enzyme-linked immunosorbent assay (ELISA). Of the various plant extracts, the seeds extract of Trachyspermum ammi showed maximum degradation of AFG1 after 24 h of incubation. Another study by Vijayanandraj et al. (2014) also demonstrated the effect of different parameters on aqueous extracts of various medicinal plants for detoxification of aflatoxin B1 (AFB1). They concluded that leaf extracts of Adhatoda vasica Nees showed the maximum degradation of AFB1 (≥98%) after incubation for 24 h at 37◦C. Moreover, Kannan and Velazhahan (2014) explored the potential of some indigenous medicinal plants extracts for detoxification of aflatoxins. The study showed that among several tested plants, Barleria lupulina Lindl. leaves extract exhibited maximum detoxification of aflatoxins B1, B2, G1, and G2 at pH 10 whereas detoxification percentage decreased at pH 7 and 3. Time course study of aflatoxin detoxification by B. lupulina extract showed that degeneration of aflatoxin occurred within 10 min and this percentage was increased with increase in incubation period.

In current investigation, Ocimum basilicum L. (Family: Lamiaceae, Common name: Sweet basil) and Cassia fistula L. (Family: Fabaceae, Common name: Amaltas) were used to evaluate their aflatoxin detoxifying potential. Ocimum basilicum is a common herb that is known for its ornamental and therapeutic importance while C. fistula is a moderate size deciduous tree, commonly known as golden shower. Both plants have been reported to possess hepatoprotective, antitumor, antitoxic, anti-inflammatory, antibacterial, and antifungal effects. Their Phytochemical studies revealed the presence of terpenoids, alkaloids, flavonoids, tannins, saponin, glycosides, and ascorbic acid (Neelam et al., 2011; Khair-ul-Bariyah et al., 2012). In this present study, these plants were selected to develop a cost effect and environment friendly strategy for aflatoxin detoxification.

#### MATERIALS AND METHODS

#### Chemical and Reagents

Aflatoxins B1 and B2 purified from A. flavus were prepared in laboratory under optimized conditions and compared through HPLC with standard AFB1 and AFB2 purchased from (Sigma– Aldrich, St. Louis, MO, USA). Stock solutions of AFB1 (1000 µg L −1 ) and AFB2 (500 µg L−<sup>1</sup> ) were prepared in methanol and stored at 4◦C. The working solutions of AFB1 (100 µg L−<sup>1</sup> ) and AFB2 (50 µg L−<sup>1</sup> ) were prepared by diluting the stock solution.

#### Plant Extract Preparation

Ocimum basilicum and Cassia fistula samples (leaves and branches) were collected from Quaid-e-Azam campus, University

#### TABLE 1 | Effect of temperature and incubation period on aflatoxin detoxification by O. basilicum and C. fistula aqueous extracts.


Values are mean of three replicates. Small alphabetic letters with different values indicate significant differences (p < 0.05) in toxin reduction at different temperature and incubation periods in each column. Capital alphabetic letters with different values indicate significant differences (p < 0.05) among controls and tested plant extracts at different temperature and incubation period in each row.

of the Punjab, Lahore. Samples were surface-sterilized using 1% sodium hypochlorite for 10 min and washed several times with sterile distilled water. After surface sterilization, aqueous plants extract were prepared according to the method described by Velazhahan et al. (2010) with some modifications. For extract preparation, 10 g of leaves/branches were homogenize with 10 mL of sterile distilled water. Then homogenate was filtered through muslin cloth and centrifuged at 14,000 rpm for 20 min. Supernatant was sterilized using syringe filter assembly and used for further detoxification studies.

#### Fungal Growth Inhibition Assay

Antifungal activity of plants extracts against toxigenic isolates of A. flavus was determined by fungal growth inhibition assay as described by Fiori et al. (2000). To 80 ml of malt extract agar medium 20 ml of 10% each filter-sterilized plant extract was added. Then aflatoxigenic isolates of A. flavus from 5 days old culture were inoculated centrally into the plates and incubated at room temperature (28 ± 2 ◦C). The colony diameter of test fungi was measured after 4 days by calculating the average radial growth. The percentage inhibition of mycelial growth in relation to the control treatment was calculated by following equation. Three replicates of each treatment were experimented. In control, isolates were inoculated in MEA plates without any plant extract and incubated as described above.

Percent growth inhibition = ((a − b)/a) × 100

Where,

a = diameter of fungal colony (mean) in control

b = diameter of fungal colony (mean) with plant extract

#### Detoxification of Aflatoxins (B1 and B2) by Medicinal Plants Extracts (In Vitro Studies)

For In Vitro detoxification assay, 50 µL of working solution containing (100 µg L−<sup>1</sup> ) AFB1 and (50 µg L−<sup>1</sup> ) AFB2 was mixed with 250 µL of plants extract and incubated for various intervals of time. Subsequently, the reaction was terminated by adding 250 µL of Chloroform to the above mixture. For recovery of residual toxin mixture was vortex thoroughly. Low speed centrifugation was done for the separation of Chloroform fraction. After that organic phase was withdrawn, evaporated to dryness under gentle stream of nitrogen and redissolved in methanol. Control comprised of toxin: water (50 µL: 250 µL) and incubated under same conditions. All experiments were conducted in triplicate.

#### Optimization of Analytical Parameters for Detoxification Assay (In Vitro) Temperature and Incubation Time

Plants extracts were incubated with toxins at various intervals of temperature and incubation periods, i.e., 25, 30◦C. . .. . .60◦C for 6, 24, and 72 h, respectively. Afterward, the toxin content in the reaction mixture was determined as described above.

#### Optimum pH

The original pH of plants extracts was modified in the range of 2.0–8.0 by using 1 N HCl or 1 N NaoH and then assayed for toxin detoxification activity. Control consists of distilled water with same pH range as well as untreated extract.

#### Decontamination of Maize Samples Using Plants Extracts

Maize samples were decontaminated according to method described by Das and Mishra (2000) with some modifications. Ten grams of maize samples were kept in each 250 ml Erlenmeyer flask and spiked with 3 ml of aflatoxins (with concentration B1 100 µg L−<sup>1</sup> and B2 50 µg L−<sup>1</sup> ). These samples were then incubated with 10 ml of aforesaid plants extracts at 30◦C for 72 h. After that aflatoxin was extracted according to the modified method of Stroka et al. (2000). Maize samples were incubated with water: acetonitrile (15: 85%) on shaking water bath for 2 h. Afterward the extracts were filtered through filter paper (Whatman, Inc., Clifton, NJ, USA). Then filtrate was passed through afla immunoaffinity column in solid phase extraction assembly. Toxins were slowly eluted from the column with 1 ml of methanol in a glass vial, which was further analyzed by TLC and HPLC. Control comprised of untreated maize sample, sample with plant extract without toxin and sample with toxin without plant extract. Each experiment was performed in triplicate.

#### TLC and HPLC Analysis of Aflatoxin

The qualitative analysis of residual toxin was determined by thin layer chromatography (TLC) according to the method described by Ramesh et al. (2013) with some modifications. Twenty-five microliters of methanolic fraction of treated and untreated samples were spotted on 0.25 mm silica gel 60F<sup>254</sup> (20 cm × 20 cm, Merck). Then plates were developed on chloroform: acetone (92: 8 v/v) and viewed under UV light at 365 nm.

Before quantitative analysis by HPLC, treated and untreated toxins were derivatized by the method described by Hernandez-Hierro et al. (2008) with some modifications. For this purpose, eluted toxin was evaporated to dryness with gentle stream of nitrogen, redissolved in 200 µL n-hexane, vortexed and 50 µL of trifluoroacetic acid (TFA) was added in it. Then, 950 µL of acetonitrile–water (1:9) was added to above solution and filtered by using syringe filter assembly. The filtrate was analyzed by HPLC system (Agilent 1100 series, Agilent Technologies, Santa Clara, CA, USA) equipped with a reverse-phase C18 column (Merck, Darmstadt, Germany), and a fluorescence detector. Water:methanol:acetonitrile (60:20:20 v/v/v) were used as mobile phase at a flow rate of 1 mL min−1. Toxins were detected at excitation and emission wavelengths of 360 nm and 440 nm, respectively. For HPLC method validation, calibration curves were drawn using series of calibration solutions in methanol. Each standard solution was chromatographed in duplicate.

# Identification of Degraded Toxin Products

Mass spectral studies were carried out for the identification degraded toxin products.

## LCMS Analysis of Degraded Toxin

Degraded toxin products were analyzed by using surveyor LC system equipped with mass spectrophotometer and PDA plus detectors (Thermo Fisher Scientific). System was validated with known standards individually and in mixture form. All analysis were performed in triplicate using luna phenomenex C<sup>18</sup> column (150 mm × 4.6 mm, 3 µm), in isocratic mode. Following are the LC-MS conditions for Aflatoxins. Injection volume was 10 µL. Mobile phase consisted of methanol:acetonitrile:water (22.5:22.5:55.0 v/v). Column temperature was maintained at 30◦C. The total operation time was 25 min with the flow rate of 0.5 mL min−<sup>1</sup> . MS conditions were as follows: capillary temperature was 335◦C, sheath gas flow and auxiliary gas flow was 20 L min−<sup>1</sup> and 4 L min−<sup>1</sup> , respectively. Source voltage, capillary voltage, and tube lens voltage was 5 KV, 49 V, and 120 V, respectively. Toxins incubated with water under optimized conditions (i.e., temp 30◦C and pH 8) were run as a control.

### Electrospray Ionization Mass Spectrometery (ESI – MS/MS) of Aflatoxins through Direct Insertion Pump

In order to predict the molecular formulae as well as elemental composition of AFB1 and AFB2 degradation products, samples were further analyzed by electrospray ionization mass spectrometer. MS/MS was performed on was performed on a Thermo Scientific LTQ XL System fitted with electrospray ionization (ESI) source operating in positive ionization mode with optimum conditions set as follows: capillary voltage to 49.0 V, source voltage to 5.0 KV, Tube lens voltage to 110 V, and capillary temperature to 275◦C. Sheath and auxiliary gas flow were adjusted to get stable spray, i.e., 3 L min−<sup>1</sup> and 0.4 L min−<sup>1</sup> , respectively. Data was collected in positive mode within the range of 100 to 500 m/z. The final identification of an unknown compound was based on the accurate measurement of mass of parent ions and fragments, as well as other useful MS/MS spectrum information (Wang et al., 2011). In MS/MS experiments, untreated aflatoxins and water treated toxins were run as a control.

# Brine Shrimps Bioassay for Toxicity Assessment of Degraded Products

Brine shrimps (Artemia salina) bioassay was carried out for testing the biological toxicity of degraded toxin products by the method developed by Solis et al. (1993) with some modifications.

Shrimp eggs (100–200 mg) were hatched in artificial sea water (34 g sea salt/L of deionized water) by incubation under 60 W lamp, providing direct light and warmth (26◦C). Throughout hatching period, the same conditions of light sensitivity and temperature were maintained. After an incubation period, the

Three hundred microliters of treated and untreated AFB1 and AFB2 (100 and 50 µg L−<sup>1</sup> ) were added to 96 well plate separately, dried and redissolved in 200 µL of sea water. After that, 200 µL of sea water containing 40–45 organisms were pipetted into each well, resulting in a final volume of 400 µL and incubated for 24– 96 h at 26◦C. Percentage mortality was determined by counting the immobile (dead) larvae under stereo microscope. Toxicity of each solution was evaluated in triplicate.

#### Statistical Analysis

Statistical analysis was carried out by using DSSTAT software. Data were analyzed by analysis of variance (ANOVA) and differences among the means were determined for significance at P ≤ 0.05 using Tukey's multiple range test.

# RESULTS

## Inhibitory Effects of Aqueous Plants Extracts on the Growth of Aflatoxigenic Isolates of Aspergillus flavus

Ocimum basilicum and Cassia fistula aqueous extracts were tested to determine their inhibitory effect on growth of aflatoxigenic isolates of Aspergillus flavus. **Table 2** depicted that O. basilicum leaves extract showed 82.8–87.7% inhibition of growth of tested isolates. This was followed by O. basilicum branch extract with percentage inhibition of 57–68.3% in growth of all tested isolates. While 23.6–49.8% and 16.4–42.3% inhibition in mycelial growth was recorded by aqueous extracts of C. fistula leaves and branch, respectively.

### Influence of Temperature and Incubation Period on Toxin Detoxification by Plants Extracts

The aflatoxins (B1 and B2) detoxifying efficacy of Ocimum basilicum and Cassia fistula (leaves and branches) aqueous extracts were tested at different temperature and incubation time. The degree of detoxification was compared with control experimented under same conditions. Results showed that percentage of detoxification increased with increase in incubation time to 6–72 h (**Table 1**). At lowest tested temperature (25◦C), maximum detoxification of AFB1 and AFB2 was shown by O. basilicum leaves extract, i.e., 77.9 and 76.7%, respectively, after 72 h of incubation. At this temperature, 36.6–65.7% and 48.9– 57.0% reduction in AFB1 and AFB2 level was recorded in samples treated with C. fistula leaves and O. basilicum branch extract. While least significant detoxification of AFB1 and AFB2 was observed after treatment with C. fistula branch extracts, i.e., 30.3 and 43.9%, respectively.

Similarly, at 30◦C most significant detoxification of AFB1 and AFB2 was shown by O. basilicum leaves extract which was followed by O. basilicum branch, C. fistula leaves and branch extracts, after 72 h of incubation. In control samples at 30◦C only



Values are mean of three replicates. Values with different alphabetic letters indicate significant differences among tested plants extract as determined by Tukey's multiple range test.


Values are mean of three replicates. Small alphabetic letters with different values indicate significant differences (p < 0.05) among tested plant extracts. Capital alphabetic letters with different values indicate significant differences (p < 0.05) among pH treatments.

4.19 and 2.41% AFB1 and AFB2 were found to be inactivated (**Table 1**).

The findings of this study depicted that the percentage detoxification of AFB1 and AFB2 by plants extracts was gradually increased with the consequential increase in temperature to 35– 55◦C. Similarly highest detoxification of AFB1 and AFB2 was observed at 60◦C but this detoxification may be due to synergistic effect of heat and moisture. In present study, O. basilicum leaves extracts showed highest detoxification of aflatoxins at all tested temperatures and incubation periods. However, further experimentations were carried out at 30◦C because it is more or less near to room temperature and also found to be prevailing temperature in most of the storehouses of Punjab especially in summer, which may provide a cost effective approach to detoxify these toxins.

#### Effect of pH on Detoxification of Aflatoxins by Plants Extracts

This study was aimed to observe the detoxifying effect of pH using O. basilicum and C. fistula aqueous extracts. Distilled water with pH adjusted to 2, 4, 6, and 8 was used as a control. Data revealed that least significant detoxification was observed at pH 2. At this pH, maximum detoxification AFB1 and AFB2 was observed in samples treated with O. basilicum leaves extract, i.e., 78.8 and 72.7%, respectively (**Table 3**). This was followed by O. basilicum branch, C. fistula leaves and branch extracts.

The percentage of detoxification increases with increase in pH from acidic to alkaline range. Hence, the highest degradation of AFB1 and AFB2 was recorded at pH 8 after treatment with O. basilicum leaves extracts, i.e., 90.4 and 88.6%, respectively. While 40.7–74.9 and 49.4–62.5% detoxification of AFB1 and AFB2 was shown by O. basilicum branch, C. fistula leaves TABLE 4 | Decontamination of maize samples by C. fistula and O. basilicum aqueous extracts under optimized conditions.


Values are mean of three replicates. Figures in the parenthesis are detoxification % age. Values with different alphabetic letters indicate significant differences (P < 0.05) as determined by Tukey's multiple range test.

and branch extracts, respectively. At this pH respective control showed 16.4 and 12.8% degradation of AFB1 and AFB2 (**Table 3**). So, further studies were done at pH 8 as the results clearly depicted that the efficacy of plants extracts to detoxify AFB1 and AFB2 significantly enhanced at this pH.

Plants Extracts

Decontamination of Maize Samples by

This study was conducted under optimized condition of In Vitro assays, i.e., pH (8), temp (30◦C) and incubation time (72 h). According to results, only 0.5–0.3 µg L−<sup>1</sup> AFB1 and AFB2 were found to be present in unspiked control maize samples. However, 97.3 µg L−<sup>1</sup> of AFB1 and 47.6 µg L−<sup>1</sup> of AFB2 were recovered from control maize samples spiked with 100 µg L−<sup>1</sup> AFB1 and 50 µg L−<sup>1</sup> AFB2 (**Table 4**).

The findings of this study depicted a similar detoxifying trend as that was recorded in In Vitro studies. Results showed that aqueous extracts of O. basilicum leaves found as most significant

**55**

fmicb-07-01105 July 13, 2016 Time: 16:26 # 7

source to detoxify AFB1 and AFB2 as they showed detoxification percentage of 86.9 and 83.5%, respectively in spiked maize samples. While in spiked samples, 37.7–67.3% and 42.7–56.1% detoxification of AFB1 and AFB2 was recoded after treatment with O. basilicum branch, C. fistula leaves and branch extracts (**Table 4**).

In this present investigation, overall results of various assays demonstrated that O. basilicum leaves extracts degraded AFB1 and AFB2 most significantly under optimized conditions as compared to O. basilicum branch, C. fistula leaves and branch extracts.

Moreover HPLC chromatograms that obtained after treatment with O. basilicum leaves extracts, showed the presence of various peaks whose footprints were not found in chromatogram of control which may be attributed to toxin degradation products (**Figure 1**). So, aflatoxins degraded with O. basilicum leaves extracts were used in further structural characterization studies for the identification of degradation products.

# Structural Characterization of Aflatoxin and Their Degradation Products

Aflatoxins showed good ESI ionization in positive ion mode. A molecular base ion at m/z 313.17 and m/z 315.17 was detected for protonated adduct [M+ H]<sup>+</sup> of aflatoxins B1 and B2 while sodium adduct [M+ Na]<sup>+</sup> was formed at m/z 335 and m/z 337, respectively. The identity of parent ion was validated by its fragmentation into daughter ions. The protonated molecule was chosen as the precursor ion for aflatoxins in the product ion scan mode because the sodium adduct did not exhibit specific fragmentation for any compound. The fragmentation pattern of AFB1 and AFB2 has an important reference value in analyzing fragmentation pathway of degraded products.

### MS/MS Analysis of AFB1 and AFB2

AFB1 is composed of 17 carbon atoms, 6 oxygen, and 12 hydrogen atoms while AFB2 contains 14 hydrogen atoms in their formulae along with 17 carbon and 6 oxygen atoms. MS/MS spectrum revealed that continuous loss of carbon monoxide (CO) was the main fragmentation pathway of AFB1. Methyl and methanol losses occurred on methoxy group located on side chain of benzene. The double bond equivalence (DBE) of AFB1 was 12 (**Figure 2A**). However, MS/MS fragmentation pathway of AFB2 illustrated that daughter ions were formed by loss of carbon monoxide, oxygen, hydrogen, and methyl group (**Figure 2B**). The DBE of AFB2 was 11. Identification of

degradation products were based on accurate mass measurement of ions and similar fragmentation pathways with that of AFB1 and AFB2.

It is evidenced from the results that nine degraded products were obtained after detoxification of AFB1 and AFB2 by using O. basilicum leaves extract. Among them, some were formed by the modification of lactone ring while others were produced as a result of addition reaction in furan rings of aflatoxins. Structural formulas of possible degraded products of AFB1 and AFB2 are shown in **Figures 3A,B**.

# MS/MS Analysis of Degraded Products of AFB1

The degradation product at m/z 269.08 corresponded to molecular formula C16H12O<sup>4</sup> formed by the loss of carbon dioxide from lactone ring of AFB1. The DBE of C16H12O<sup>4</sup> was one less than AFB1. Loss of oxygen, methyl group and carbon monoxide was the main fragmentation pathway (**Figure 4**).

The degradation products C16H22O<sup>5</sup> (m/z 295.08) and C16H20O<sup>5</sup> (m/z 293.25) were formed due to the loss of carbon monoxide by the opening of lactone ring and addition of hydrogen atom to AFB1 molecule. The DBE content of both products was same, i.e., six while difference between them was of two hydrogen atoms. Losses of CO, CH2, and O were the main fragmentation pathway of both products (**Figures 5** and **6**).

Degradation product C17H10O<sup>6</sup> with ion peak at m/z 311.08 was formed by the loss of hydrogen atoms implying greater DBE content than that of AFB1. The fragmentation pathway of C17H10O<sup>6</sup> was different from that of AFB1. The precursor ion yielded a series of product ions represented by 283.08[M-CO]+, 255.08[M-C2O2] <sup>+</sup>, 241.0[M-C3H2O2] <sup>+</sup>, 215.08[M-C4O3] <sup>+</sup>, and 197.0[M-C4H2O4] <sup>+</sup> (**Figure 7**).

The degradation products C17H8O<sup>7</sup> (with m/z 325.25) and C17H10O<sup>7</sup> (with m/z 327.0) were formed by the addition of oxygen atom on the double bond of terminal furan ring on the left side. The difference between these two products was only of two hydrogen atoms. The DBE content of both products was greater than AFB1, i.e., 14 and 13, respectively. The precursor ion C17H8O<sup>7</sup> yielded a series of product ions represented by 307.08[M-H2O]+, 281.17[M-CO2] <sup>+</sup>, 267.08[M-C2H2O2] +, and 239.17[M-C3H2O3] <sup>+</sup>. While the fragmentation pathway of

C17H10O<sup>7</sup> showed losses of O, CO, CH2, and CH3O groups (**Figures 8** and **9**).

The degradation product C17H22O<sup>7</sup> at m/z 339.17 produced as a result of addition reaction on furan rings. Double bonds were replaced by the addition of hydrogen atoms with lower DBE content than that of AFB1 i.e., seven. The fragmentation pathway was different from that of AFB1. Fragments of C17H22O<sup>7</sup> showed losses of carbon monoxide, water, oxygen, and methyl groups. More details on fragmentation pathway are shown in **Figure 10**.

The degradation product C17H16O<sup>8</sup> (with 349.0 m/z) had one more H4O<sup>2</sup> molecule and less DBE content than AFB1, i.e., 10. The result is most likely caused by additional reaction of two hydroxyl groups on the double bond of terminal furan ring. The precursor ion yielded a series of product ions represented by 331.17[M-H2O]+, 321.33[M-CO]+, 305.25[M-CO2] +, 287.25[M-CH2O3] <sup>+</sup>, 261.08[M-C2O4] <sup>+</sup>, 257.33[M-C2H4O4] +, 231.00 [M-C3H2O5] <sup>+</sup>, and 217.08[M-C3O6] <sup>+</sup> (**Figure 11**).

# MS/MS Analysis of AFB2 Degradation Product

The degradation product at m/z 301.25 with molecular formula C16H12O<sup>6</sup> was formed by the replacement of methoxy group with hydroxyl group on the side chain of benzene ring. The DBE of C16H12O<sup>6</sup> was same as that of AFB2. Fragmentation pathway involves the loss of carbon dioxide, carbon monoxide, and water molecules (**Figure 12**).

# Estimation of Toxicity of Degradation Products

The biological toxicity of untreated and treated aflatoxins B1 and B2 were compared through brine shrimps (Artemia salina) bioassay to confirm the safety of degraded products. Hatched shrimps larvae were incubated with treated and untreated aflatoxins at 26◦C for 24–96 h to observe the mortality response

of larvae. Natural mortalities were determined in blank sea water and in well treated with solvent only. The percentage of mortality was calculated from average number of dead larvae per concentration.

Larvae incubated with untreated aflatoxins showed significant increase in larval mortality as compared to control, i.e., 83%. This percentage of mortality was increased to 91.7% with increase in incubation time. As compared to untreated toxins, significant reduction in larval mortality was recorded after incubation with treated toxins. Data derived from tests conducted with AFB1 (100 µg L−<sup>1</sup> ) and AFB2 (50 µg L−<sup>1</sup> ) degraded by O. basilicum leaves extracts showed only 9.2–22.5% mortality of shrimps larvae after 96 h of incubation (**Table 5**). The outcome of this study clearly implies that aflatoxins degradation products showed much lower toxicity toward brine shrimps larvae than untreated aflatoxins

#### DISCUSSION

Use of certain plant extracts and biocontrol agents as a source of safer and more effective control on the growth of aflatoxigenic fungi and aflatoxin production have been under investigation by many authors (Gowda et al., 2004; Joseph et al., 2005; Suleiman et al., 2008; Reddy et al., 2009). Much emphasis was given to herbal, medicinal, and aromatic plants for their antifungal activities against food spoilage and aflatoxigenic fungi (Soliman and Badeaa, 2002; Maraqa et al., 2007; Gandomi et al., 2009; El-Nagerabi et al., 2012). Plants are rich source of bioactive secondary metabolites such as tannins, terpenoids, alkaloids, and flavonoids, reported to have antifungal properties (Nwachukwu and Umchuruba, 2001).

In the present study, aqueous extracts of O. basilicum and Cassia fistula were tested for their antifungal potential against aflatoxigenic isolates of A. flavus. All tested plants extracts exhibited diverse degree of antifungal activity against A. flavus isolates. The maximum antifungal activity was shown by aqueous extract of O. basilicum leaves followed by O. basilicum branch, C. fistula leaves and branch extract. Similarly a study conducted by Tanackov et al. (2011) also showed that extracts of O. basilicum significantly inhibit the growth of the Fusarium oxysporum, F. proliferatum, F. subglutinans, and F. verticillioides. Antifungal potential of aforementioned plants used in this study, were also described by many scientists (Abo et al., 1999; Phongpaichit et al., 2004; Pujo et al., 2009; Bhalodia and Shukla, 2011).

According to previously documented literature, essential oils and extracts of various spices and herbs like cinnamon,

TABLE 5 | Effect of detoxified toxins on percent mortality of Brine shrimp larvae at 26◦C after 24–96 h after incubation.


Values are mean of three replicates. Data was analyzed by analysis of variance (ANOVA).

peppermint, basil, and lemongrass can be suggested as plant based safe food additive in protecting food and feed from fungal and mycotoxin contamination (Montes-Belmont and Carvajall, 1998; Burt, 2004; Yang et al., 2007). In this study, In Vitro assays were performed with O. basilicum and Cassia fistula aqueous extracts under optimized conditions of temperature, pH and incubation time. In case of temperature, highest detoxification percentage was observed at 60◦C but this detoxification could be due to synergistic action of heat and moisture (Basappa and Shantha, 1996; Rustom, 1997). Similar findings were recorded in a study conducted by Hajare et al. (2005) who worked on aflatoxin inactivation by using Ajwain seeds extract under optimized conditions. Their results showed that highest inactivation was obtained at 60◦C but further studies were conducted on 45◦C to eliminate the effect of heat and moisture on toxin inactivation. Likewise in this present investigation 30◦C was selected for further studies.

The results of present study depicted that percentage of detoxification increase with increase in incubation time to 72 h. These observations are in close agreement to the findings of earlier workers (Hajare et al., 2005; Velazhahan et al., 2010; Kannan and Velazhahan, 2014; Vijayanandraj et al., 2014). Furthermore in In Vitro assays, pH of reaction mixture was optimized and found that percentage of detoxification increases as the pH change from acidic to alkaline range. In this study highest detoxification of AFB1 and AFB2 was shown by O. basilicum leaves extract at pH 8. Similar finding were obtained by Kannan and Velazhahan (2014) who explored the potential of Barleria lupulina leaf extract on detoxification of aflatoxins. In addition, Méndez-Albores et al. (2004) found that aflatoxin florescence, attributed to the coumarin moiety, diminish or even disappear in alkaline treatment.

In the past, numerous researchers observe the structural changes in aflatoxin molecule after detoxification by various means, i.e., micro-organisms, physical and chemical agents, ultraviolet (UV) rays, Gamma rays, and plant products

(Alberts et al., 2006; Albores et al., 2008; Guan et al., 2010; Velazhahan et al., 2010; Wang et al., 2011; Farzaneh et al., 2012; Inoue et al., 2013; Luo et al., 2013; Samuel et al., 2014; Vijayanandraj et al., 2014). Lee et al. (1981) observed that lactone ring plays an important role in fluorescence of aflatoxin molecule. On its cleavage, the molecule becomes non-florescent with subsequent significant reduction in toxicity. The toxicity of aflatoxins has been studied by various scientists (Guengerich, 2001, 2008; Hussein and Brasel, 2001). Their toxicity data demonstrated that aflatoxins have cyclopentene ring and furan moiety in their chemical structure. In AFB1 presence of double bond in the terminal furan ring is key factor for its toxic and carcinogenic activities (Wang et al., 2011). As compare to AFB1, the toxicity of AFB2 is hundreds times less due to presence of saturated furan ring (Dvorackova, 1990). The degraded products of AFB2 may be active but were less potent than that of parent compound. Thus removing the double bond of terminal furan ring and modification of lactone ring are major aims of detoxification.

In this present study, data recovered from HPLC revealed that after treatment with O. basilicum leaves extracts AFB1 and AFB2 were degraded into number of other compounds whose properties are different from parent toxins. So, the presence of these degraded products was further confirmed by LCMS/MS studies. It is evidenced from the results that eight degraded products were obtained after detoxification of AFB1 by using O. basilicum leaves extract. Among them, 50% of degraded products (with m/z 339, 325, 349, and 327) were formed by removal of double bond in furan ring whereas in 25% of products obtained at m/z 293 and 295 modification of lactone ring and removal of double bond in furan ring was occurred. Other studies in literature also supported the similar findings. Velazhahan et al. (2010) reported detoxification of aflatoxin G1 by seed extract of Ajowan (T. ammi) and suggested the modification of lactone ring structure of AFG1 as mechanism of detoxification. Similar findings were observed by Vijayanandraj et al. (2014) after detoxification of aflatoxin B1 by an aqueous extract from leaves of Adhatoda vasica Nees. A study conducted by Wang et al. (2011) on structure elucidation of radiolytic products of aflatoxin B1 in methanol water solution revealed that in most of radiolytic products addition reaction occurred on the double bond of terminal furan ring, resulting significantly reduced toxicity as compared to that of AFB1. Experiments by Luo et al. (2013) showed that aflatoxin B1 can be effectively degraded by using ozone in aqueous system. According to them, due to conjugate addition reaction on the double bond of terminal furan ring for AFB1, the toxicity of degradation products was significantly decreased compared with that of AFB1.

Brine shrimps (Artemia salina) larvae appears to be as susceptible as biological indicator of toxicity of some mycotoxins in foods and feeds (Hartl and Humpf, 2000; Favilla et al., 2006). Several studies in the past were conducted on brine shrimps larvae to evaluate the toxic effects of aflatoxins (Harwing and Scott, 1971; Schmidt, 1989; Logrieco et al., 1996; Durakovic et al., 2002; Moretti et al., 2007). In this study hatched shrimps larvae were incubated with treated and untreated aflatoxins at 26◦C for 24 to 96hrs to observe the mortality response of larvae. As compared to untreated toxins, significant reduction in larval mortality was recorded after incubation with treated toxins. Therefore, toxicity of most of degraded products compared with that of aflatoxin was reduced to a much lower level. The reason behind this is that most of the degradation products obtained after treatment with O. basilicum leaves extracts were formed as a result of addition reaction on the double bond of terminal furan ring and modification of lactone ring as indicated by mass spectral analysis which is most significant determinant of toxic and carcinogenic activities of aflatoxins. These results were in close agreement with those of Samuel et al. (2014) who worked on detoxification of aflatoxin B1 by Pseudomonas putida. He compared the toxicity of treated and untreated AFB1 toward HeLa cells and concluded that degraded products are non-toxic (D1) or much less toxic (D2 and D3) than AFB1 to the cells at the tested concentrations.

# CONCLUSION

The efficient detoxification of aflatoxins by using O. basilicum leaves extracts under optimized conditions. These extracts are easily available, cost effective, biologically safe and provide an excellent alternative source of toxin detoxification instead of physical and chemical methods. Direct spray of aqueous plant extract is easy to prepare and convenient for the farmers as no technical knowledge is involved but the formulations may vary and cause errors in treatment. This trait should be properly investigated. Another most important aspect is the shelf life of the plant extract preparations that should sustain variations in environmental conditions and remain feasible for the effective detoxification of aflatoxins. Future implications of this approach are quite promising in order to overcome world food hunger improving the quality of food items.

# AUTHOR CONTRIBUTIONS

TA is the supervisor and MA is co-supervisor of this work. WI has worked as research scholar for this project. AG and MI has provided technical guidance for various analyses. AMK has helped in data interpretation.

# FUNDING

This work was not funded by any agency and was done using indigenous resources of Fungal biotechnology lab, Institute of Agricultural Sciences, University of the Punjab, Lahore, Pakistan.

# ACKNOWLEDGMENTS

Authors are extremely thankful to MI (National Institute of Biotechnology and Genetic Engineering), AG (University of Engineering and Technology), and MA and AM (University of Veterinary and Animal Sciences) for the provision of specialized apparatus and technical guidance for successful accomplishment of this study.

# REFERENCES

fmicb-07-01105 July 13, 2016 Time: 16:26 # 17


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Iram, Anjum, Iqbal, Ghaffar, Abbas and Khan. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

fmicb-07-01105 July 13, 2016 Time: 16:26 # 18

# Marine Biotoxins: Occurrence, Toxicity, Regulatory Limits and Reference Methods

Pierina Visciano<sup>1</sup> , Maria Schirone<sup>1</sup> \*, Miriam Berti<sup>2</sup> , Anna Milandri<sup>3</sup> , Rosanna Tofalo<sup>1</sup> and Giovanna Suzzi<sup>1</sup>

<sup>1</sup> Faculty of Bioscience and Technology for Food, Agriculture and Environment, University of Teramo, Teramo, Italy, <sup>2</sup> Istituto Zooprofilattico Sperimentale dell'Abruzzo e del Molise "G. Caporale", Teramo, Italy, <sup>3</sup> National Reference Laboratory for Marine Biotoxins, Fondazione Centro Ricerche Marine, Cesenatico, Italy

Harmful algal blooms are natural phenomena caused by the massive growth of phytoplankton that may contain highly toxic chemicals, the so-called marine biotoxins causing illness and even death to both aquatic organisms and humans. Their occurrence has been increased in frequency and severity, suggesting a worldwide public health risk. Marine biotoxins can accumulate in bivalve molluscs and regulatory limits have been set for some classes according to European Union legislation. These compounds can be distinguished in water- and fat-soluble molecules. The first group involves those of Paralytic Shellfish Poisoning and Amnesic Shellfish Poisoning, whereas the toxins soluble in fat can cause Diarrheic Shellfish Poisoning and Neurotoxic Shellfish Poisoning. Due to the lack of long-term toxicity studies, establishing tolerable daily intakes for any of these marine biotoxins was not possible, but an acute reference dose can be considered more appropriate, because these molecules show an acute toxicity. Dietary exposure assessment is linked both to the levels of marine biotoxins present in bivalve molluscs and the portion that could be eaten by consumers. Symptoms may vary from a severe gastrointestinal intoxication with diarrhea, nausea, vomiting, and abdominal cramps to neurological disorders such as ataxia, dizziness, partial paralysis, and respiratory distress. The official method for the detection of marine biotoxins is the mouse bioassay (MBA) showing some limits due to ethical restrictions and insufficient specificity. For this reason, the liquid chromatography–mass spectrometry method has replaced MBA as the reference technique. However, the monitoring of algal blooms producing marine biotoxins should be regularly assessed in order to obtain more reliable, accurate estimates of bloom toxicity and their potential impacts.

Keywords: shellfish poisoning, toxicity, symptoms, human health, detection method

### INTRODUCTION

Harmful algal blooms (HAB) are natural phenomena carried out by the overgrowth of marine phytoplankton (Ferreiro et al., 2015). Over the last decades the occurrence and intensity of HAB appear to be increasing on a global scale due to rising ocean temperatures and growing coastal eutrophication (McCarthy et al., 2015). The geographical expansion of HAB can also be associated with ballast waters transporting encysted algae to new environments or massive algae spreading

#### Edited by:

Mehdi Razzaghi-Abyaneh, Pasteur Institute of Iran, Iran

#### Reviewed by:

Leda Giannuzzi, University of La Plata – Consejo Nacional de Investigaciones Científicas y Técnicas, Argentina Andrew K. Ottens, Virginia Commonwealth University, USA

> \*Correspondence: Maria Schirone mschirone@unite.it

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 28 April 2016 Accepted: 23 June 2016 Published: 06 July 2016

#### Citation:

Visciano P, Schirone M, Berti M, Milandri A, Tofalo R and Suzzi G (2016) Marine Biotoxins: Occurrence, Toxicity, Regulatory Limits and Reference Methods. Front. Microbiol. 7:1051. doi: 10.3389/fmicb.2016.01051

caused by aquaculture practices (Anderson et al., 2002; Maso and Garcés, 2006; Smayda, 2007). Among the thousands of microalgal species known in nature, about 300 are involved in harmful events and more than 100 (of these species) produce persistent natural toxins that can cause intoxication or even death in humans and animals. Moreover, such toxic outbreaks can have consequences on other components of human wellbeing both in terms of socio-economic impact and costs. The main factors affecting HAB occurrence and their influence on the environment, shellfisheries and consumers are shown in **Figure 1**. The direct impact of HAB on human health is linked to poisoning after consumption of contaminated seafood, skin contact with contaminated water, and/or inhalation of aerosolized biotoxins. Nowadays the increasing of information about human exposure hazards as well as the strategies able to prevent HAB occurrence in seafood need to be understood and improved, because such phenomenon results from complex interactions among physical, chemical, and biological processes in the marine environment (Berdalet et al., 2015). Moreover, even if the classic description of intoxication due to marine biotoxins is represented by acute symptoms, there is also little information about the impacts of long-term low-level exposure together with epidemiological studies regarding the populations at risk (Lefebvre and Robertson, 2010). The toxicological mechanisms of some marine biotoxins are yet incompletely understood and no observations of adverse effects in humans have been reported.

It is well known that marine biotoxins can accumulate in the tissues of some marine organisms, particularly filterfeeding bivalves. In all cases, they are de novo produced by certain photo- or mixo-trophic microalgae not by the shellfish, and filter-feeding transfers them to the molluscs (Nielsen et al., 2016). Mussels filter approximately 20 L water/h and during HAB, waters may contain several million algal cells per liter. Some phytoplankton species in these blooms produce phycotoxins that can accumulate through the marine food webs (Ferreiro et al., 2015). Outbreaks of intoxication in humans due to marine biotoxins (**Figure 2**) are caused by the ingestion of contaminated shellfish and can have a wide range of symptoms linked to the specific toxic compound (Richter and Fidler, 2015; Turner and Goya, 2015). Damage to nervous or intestinal system rather than loss of memory have been observed depending on the type of algal bloom. Among them some dinoflagellate and diatom species, such as Noctiluca scintillans and Skeletonema costatum, are responsible only for discoloration of water and death of fish and other marine organisms, whereas species belonging to the genera Alexandrium, Gymnodinium, Dinophysis, and Pseudo-nitzschia are the main producers of marine biotoxins for humans (Berti and Milandri, 2014; Bruce et al., 2015).

Marine biotoxins can be distinguished in water- and fatsoluble according to their solubility. On the basis of their poisoning symptoms, they are also classified as toxins causing paralytic shellfish poisoning (PSP), amnesic shellfish poisoning (ASP), diarrheic shellfish poisoning (DSP), neurotoxic shellfish poisoning (NSP), and ciguatera fish poisoning (CFP) (Poletti et al., 2003).

According to their own chemical structure, marine biotoxins are classified into eight groups – namely the azaspiracid (AZA), brevetoxin (BTX), cyclic imine (CI), domoic acid (DA), okadaic acid (OA), pectenotoxin (PTX), saxitoxin (STX), and yessotoxin (YTX) groups. Two additional groups, palytoxin (PlTX) and ciguatoxin (CTX), are also considered (EFSA, 2009a). Marine biotoxins and toxic effects for consumers after the ingestion of contaminated seafood are reported in **Table 1**.

Human exposure to biotoxins generally refers only to occasional consumption and is characterized by acute and shortterm events, so that acute reference doses (ARfD) have been established for these toxic compounds rather than tolerable daily intakes that could not be determined due to the lack of appropriate toxicological data. The ARfD are derived from the corresponding concentration of marine biotoxins per kg of shellfish meat when consuming a large portion (400 g) of shellfish. A crucial issue, when deriving both these values from the most relevant toxicological information, is the size of safety factors. Generally, default values of 10 and 100 on the basis of human and animal data are applied, respectively (FAO/IOC/WHO, 2004). Furthermore, the safety factor is usually increased if, due to the critical effect, there is a lowest observable adverse effect level (LOAEL) instead of a no observable adverse effect level (NOAEL). The ARfD, safety factors, LOAEL and NOAEL for each individual marine biotoxin, as well as regulatory limits and maximum levels based on consumption of 400 g of shellfish are reported in **Table 2**.

The present review focuses on the specific toxicity and epidemiology of marine biotoxins, the regulatory limits set by European Union (EU) legislation for the most toxic compounds as well as the methods of detection and quantification.

#### PARALYTIC SHELLFISH POISONING

Paralytic shellfish poisoning, caused by 58 closely related compounds based on a tetrahydropurine skeleton (Burrell et al., 2013), is one of the most studied intoxications with serious symptoms in humans. In particular, it is the result of exposure to saxitoxin (STX) and gonyautoxin (GTX). In 1957, a PSP toxin was isolated in clams (Saxidomus giganteus) living in Alaska coastal areas and in 1975 the chemical structure was assigned to STX. The main producers of PSP toxins are dinoflagellates of the genus Alexandrium occurring along the Atlantic and Pacific coast (Bernd and Bernd, 2008) but also in the Mediterranean Sea, where other species such as Gymnodinium catenatum can be present (Berti and Milandri, 2014).

More than 30 STX analogs have been identified and grouped into four subgroups: carbamate, N-sulfo-carbamoyl, decarbamoyl, and hydroxylated saxitoxins (EFSA, 2009a).

The intake of biotoxins necessary to cause human poisoning varies greatly due to the differences in susceptibility among individuals. The symptoms are similar to paralytic phenomena (cramp, signs of paralysis, and blocking of respiration) because PSP toxins are potential neurotoxins blocking the excitation current in nerve and muscle cells (Schirone et al., 2011). Human PSP outbreaks can be distinguished in mild, moderately

FIGURE 1 | Factors involved in the occurrence of harmful algal blooms and the consequences on human health and wellbeing (modified from Berdalet et al., 2015).

severe and extremely severe. The symptoms occurring in the mild form include tingling sensation or numbness around the lips, gradually spreading to the face and neck, a prickly sensation in fingertips and toes, headache, dizziness, and nausea. The moderately severe illness is characterized by incoherent speech, progression of prickly sensation to arms and legs, stiffness and non-coordination of limbs, general weakness and feeling of lightness, then slight respiratory difficulty and rapid pulse plus backache as late symptoms. In the extremely severe form muscular paralysis, pronounced respiratory difficulty and a choking sensation may occur. In fatal cases, death is caused by respiratory paralysis occurring within 2–12 h after the consumption of contaminated shellfish, in absence of artificial respiration. Patients who survive PSP for 24 h, with or without mechanical intervention, have a high probability of a full and rapid recovery (FAO/IOC/WHO, 2004).

Blooms of toxic microalgae producing PSP toxins represent an expanding threat to both human health and fishery resources all over the world. In particular, PSP has been recognized for over a century as a clinical entity in the austral part of South America. A study regarding an outbreak in the Patagonia fjords reported that some fishermen were intoxicated by consumption of bivalve Aulacomya ater and two of them died after 3–4 h; in that case up to 8575 µg of STX, equivalent/100 g of shellfish meat, were detected by MBA (García et al., 2004).

The highest number of PSP cases (2124 with 120 deaths) was reported in the Philippines from 1983 to 2002 (Ching et al., 2015). The human fatality proportion from PSP varies according to the ability of the local medical system to treat such intoxication. In a short report, the authors referred that 45 people living along the coast of Nicaragua developed symptoms of PSP and a person died. In similar outbreaks in Southeast Asia and Latin America, the case fatality proportion ranged between 2 and 14% whereas


TABLE 1 | Classification of marine biotoxins and main adverse effects in humans.

no death occurred among more than 200 cases in Europe and North America (Callejas et al., 2015).

# AMNESIC SHELLFISH POISONING

Domoic acid and other toxic DA isomers, produced mostly by marine diatoms of the genus Pseudo-nitzschia, are responsible for ASP. This compound is a cyclic tricarboxylic amino acid with many structural and functional similarities with kainic acid, an analog to glutamic acid. In particular, DA binds with glutamate receptors in central nervous system causing over-stimulation of these receptors with production of reactive oxygen species and sometimes cell death (Schwarz et al., 2014).

In humans, symptoms range from gastrointestinal effects (nausea, vomiting, diarrhea, or abdominal cramps) and/or neurological signs (confusion, lethargy, disorientation, paresthesia, and short-term memory loss) and in extreme cases coma or death.

The accumulation of DA in bivalve tissues depends on several factors, such as the presence of Pseudo-nitzschia species and their toxic content as well as on the balance of DA accumulation and depuration by shellfish (Mafra et al., 2010).

There are no reported cases of human illness associated with DA in any European countries or regions other than North America. However, in the absence of formal reporting systems, it cannot be assumed that mild cases have not occurred. Furthermore, the data relating to cases of human poisoning caused by DA are limited, except for a unique ASP outbreak in Canada in 1987. Such event involved 150 people with 19 hospitalization and 4 deaths after consumption of contaminated mussels (Jeffery et al., 2004; Álvarez et al., 2015).

Effective seafood monitoring programs for the detection of DA in the shellfish and coastal waters worldwide have been implemented by many regulatory agencies and therefore, human ASP events have not been documented since the first outbreak in 1987.

Domoic acid has also a toxic effect on marine wildlife and many poisoning events have been described in marine birds and mammals. A chronic DA epileptic syndrome was characterized in sea lions between 1998 and 2006. Therefore, sea lions are used as a sentinel species able to predicting potential hazard for human health (Lefebvre and Robertson, 2010).

# DIARRHEIC SHELLFISH POISONING

This syndrome is one of the known intoxications caused by marine biotoxins and represents a frequent concern in shellfish industries, because it can cause a prolonged closure of mussel


harvesting activity. The interruption of mussel sales and early public announcements are highly effective in controlling DSP outbreaks (Chen et al., 2013).

The accumulation of DSP toxins in any shellfish species is still little known. These toxins are produced by dinoflagellates belonging to the genera Dinophysis and Prorocentrum, even if the first genus is considered the main source, whereas Prorocentrum species are benthic and thus unavailable for suspension-feeding mussels (Nielsen et al., 2016).

Diarrheic shellfish poisoning toxins are polyether compounds with distinctive chemical structures grouped into four structural classes (Berti and Milandri, 2014): okadaic acid (OA) and its derivatives (dinophysistoxin or DTX); pectenotoxin (PTX); yessotoxin and its derivatives (YTX) and azaspiracid (AZA). The symptoms caused by OA group include diarrhea, nausea, vomiting, and abdominal pain, starting 30 min to a few hours after consumption of contaminated shellfish, with complete recovery within 3 days (Trainer et al., 2013). Among the other mentioned classes, PTX and YTX were first believed to be relevant in DSP syndrome due to their co-occurrence in shellfish with OA group, but they have not be implicated in human illness (Li et al., 2012), whereas AZA causes a form of poisoning characterized by nausea, vomiting, diarrhea, and stomach cramps.

Regarding to OA group toxins, it has been reported that they are soluble in fats and easily cross the cell membrane determining an inhibition of serine and threonine phosphoprotein phosphatases (EFSA, 2008a).

Pectenotoxins are cyclic polyethers of marine origin and PTX-2 is the main toxin of such group. They have been associated with cases of diarrhea and other symptoms similar to those induced by OA and its derivatives (Poletti et al., 2003). They disrupt actin in the cytoskeleton, and may cause cell cycle arrest and cell death (Espiña et al., 2008).

Yessotoxin and its analogs are polyether toxins produced by dinoflagellates Protoceratium reticulatum, Lingulodinium polyedrum, and Gonyaulax spinifera. Even if they have been associated with DSP group, they do not cause diarrhea or inhibition of protein phosphatases (Paz et al., 2008) and their symptoms are still unknown in humans (Visciano et al., 2013).

Azaspiracids are nitrogen-containing polyether toxins comprising an unique spiral ring assembly containing a heterocyclic amine and an aliphatic carboxylic acid moiety. Even if 21 different analogs have been identified, AZA1, AZA2, and AZA3 are the most important ones depending on occurrence and toxicity (EFSA, 2008b).

Vale (2012) reported that the contamination of Portuguese shellfish with DSP toxins is an annually recurrent phenomenon, influenced by meteorological parameters affecting the blooming of the causative toxic microalgae, belonging to genus Dinophysis.

High levels of DSP toxins were also detected in shellfish harvested along the Chinese coast causing illness in more than 200 people in the year 2011 (Li et al., 2012). In the same year, 57 cases were described from Zheijang Province, China (Chen et al., 2013).

A DSP outbreak involving three people after the consumption of contaminated mussels was reported in Washington State

fmicb-07-01051 July 4, 2016 Time: 14:58 # 5

(Trainer et al., 2013). Coincidentally 62 DSP illnesses occurred in British Columbia due to the ingestion of Pacific coast mussels, representing the first report of DSP in Western Canada (Taylor et al., 2013).

The DSP toxins are the most frequent and abundant marine biotoxins regularly found in shellfish from Southern European coastal areas (Braga et al., 2016). In France 11 DSP outbreaks involving 45 individuals were described in 2009. The contaminated mussels contained OA group toxin concentrations approximately eight times higher than what the European regulatory limits provide (Hossen et al., 2011). In 2010, more than 300 people in Northern Italy were poisoned by OA contaminated mussels (Bacchiocchi et al., 2015).

Symptoms of intoxication caused by YTX in humans are still unknown due to the fact that no human intoxication has been reported to date. However, toxicological studies carried out in rodents showed that YTX can be highly toxic when injected intraperitoneally (Visciano et al., 2013). Furthermore, PTX do not induce DSP-like symptoms even if acute toxicity can be observed in MBA (Wang et al., 2015). High levels of YTX were reported in shellfish farmed along the North Adriatic coast in 2004 correlated with the increasing presence of G. spinifera (Bacchiocchi et al., 2015).

Among the most recently identified groups of toxins causing human intoxication, AZA has been associated with shellfish poisoning in several people after consumption of mussels in the Netherlands in 1995. After that first event it was found in mussels harvested all over Europe, North-West Africa and Chile.

Some other outbreaks due to AZA were described in North-Western Ireland, affecting 12 people after the consumption of locally cultivated mussels; 10 individuals became ill with DSP symptoms in Ravenna (Italy) and 400 poisonings were reported in Belgium from mussels cultivated in Denmark (Furey et al., 2010).

#### NEUROTOXIC SHELLFISH POISONING

The brevetoxins (BTX) produced by the "Florida red tide" dinoflagellate Karenia brevis are polyether ladder compounds responsible for massive fish and marine mammal mortality above all in the Gulf of Mexico (Cassell et al., 2015) but also along the East coast of the United States and New Zealand (EFSA, 2010a). In humans, BTX are the causative agents of NSP and asthma-like symptoms through inhalation exposure (Sun et al., 2016).

Neurotoxic shellfish poisoning is characterized by both neurological and gastrointestinal effects which include nausea, vomiting, diarrhea, parasthesia, cramps, bronchoconstriction, paralysis, seizures, coma and, in extreme cases, may lead to death (Watkins et al., 2008).

On the basis of their molecular structures formed of 10– 11 trans-fused rings, BTX have 2 skeletal backbones: A- and B-type and a variety of side chain substituents on the rings distal to the lactone (Cassell et al., 2015). They bind to and activate the voltage-gated sodium channels in cell membrane causing depolarization of neuronal and muscle cell membranes (EFSA, 2010a).

Brevetoxins have been implicated in the death of large numbers of fish and in morbidity and mortality of marine mammals (Gebhard et al., 2015).

There has been only a small number of sporadic cases of NSP in the United States in humans, with hospitalization but no fatalities. Outbreaks occurred in Florida with 2 cases in 1995, 3 in 1996, 2 in 2001, and 4 in 2005. In last event, 2 out of the 4 patients were children, more seriously affected than the adults. Another outbreak characterized by severity of symptoms was reported in 2006 in Florida (Watkins et al., 2008).

Respiratory effects associated with aerosolized red tide involve sneezing, throat irritation, burning, and itchy; in the case of asthmatics individuals a significantly increase of these symptoms has been observed (Fleming et al., 2007; Bean et al., 2011).

# CIGUATERA FISH POISONING

Ciguatera fish poisoning (CFP) is the most common foodborne illness worldwide with 50,000–500,000 incidences per annum (Mattei et al., 2014). The causative agents are toxins belonging to the CTX group (Hossen et al., 2015). Such syndrome results from the bioaccumulation and metabolism of precursor toxins along the fish food web. Precursor toxins (named gambiertoxins) are produced by benthic dinoflagellates of the genus Gambierdiscus, whose distribution includes tropical and subtropical coral reef areas, and accumulated by large predatory fishes, such as Spanish mackerels, moray eels, barracuda and snappers (Litaker et al., 2009; EFSA, 2010b; Chan, 2015).

Ciguatoxins are lipid-soluble polyether compounds consisting of 13–14 rings fused by ether linkages into a rigid ladderlike structure. More than 20 CTX analogs have been identified (EFSA, 2010b). At the cellular level, CTX activate the sodium ion channels causing cell membrane excitability and cell disruption (Hidalgo et al., 2002).

With regard to symptomatology, the acute period (24 h) is characterized by gastrointestinal problems (nausea, vomiting, abdominal pain, and diarrhea) whereas cardiovascular (bradycardia and hypertension) and neurological complications may occur within a few hours to 2 weeks after exposure, such as paresthesias, disesthesias, and hyperesthesias (Silva et al., 2015).

Even if the occurrence of CTX is generally restricted to some specific areas, in recent years a spread of ciguatoxic fish has been observed near European coasts and in the Mediterranean Sea. Recently, 6 outbreaks involving 28 people between 2010 and 2011 and one event with 20 cases in 2012 were described in New York and Northern Germany, respectively (Mattei et al., 2014).

A survey about the epidemiology of CFP in Asia reported 3 large outbreaks involving 100–200 patients in China, whereas 11 cases occurred between 1991 and 2008 in Taiwan. Among them, there was one death from ciguatera in 1998. In Malaysia 11 individuals were hospitalized in 2010 after the consumption of head and viscera of contaminated fish and 33 outbreaks affecting 103 subjects were described in Japan (Chan, 2015).

#### CYCLIC IMINES

fmicb-07-01051 July 4, 2016 Time: 14:58 # 7

This group includes spirolides (SPX), gymnodimines (GYM), pinnatoxins (PnTX), and pteriatoxins (PtTX); SPX and GYM are produced by algal species from the genera Alexandrium and Karenia, whereas Vulcanodinium rugosum has been identified as the producer of PnTX (McCarthy et al., 2015). Recent studies suggest that PtTX are biotransformation products of PnTX in shellfish (Selwood et al., 2010). On the basis of their chemical structures, CI are macrocyclic compounds with imine and spirolinked ether moieties (EFSA, 2010c). These moieties, common to all the members of CI, are thought to be the main structural determinants for their toxicity. Recent studies showed that CI antagonized both muscle type and heteromeric and homomeric neuronal nicotinic acetylcholine receptors as well as muscarinic acetylcholine receptors (Hellyer et al., 2013).

No acute symptoms have been recorded in humans and the toxicity of SPX and GYM is still being explored in mammalian models (Marrouchi et al., 2013); they induce a rapid death (within minutes) in laboratory mice injected intraperitoneally but they are often co-extracted with other lipophilic toxins such as OA and its analogs producing false positive in MBA test (Salgado et al., 2015). The first isolation of SPX was reported by Hu et al. (1995) from shellfish collected along the South Eastern coast of Nova Scotia, Canada, whereas GYM was isolated and characterized in the early 1990s from New Zealand oysters (MacKenzie, 1994).

Pinnatoxins and PtTX were originally identified in Japan in shellfish belonging to the genera Pinna and Pteria, respectively (Rundberget et al., 2011).

No acute poisoning events have been linked to contamination by PnTX and only one outbreak was reported associated with the bivalve of genus Pinna. However, PnTX show fast-acting toxicity when injected intraperitoneally into mice (Hess et al., 2013).

Despite the high acute toxicity in MBA, CI are not currently regulated in seafood due to the fact that acute poisoning in humans have not been directly related to shellfish contamination (Harju et al., 2016).

#### PALYTOXIN

Palytoxin are potent non-protein marine compounds produced by corals belonging to the genus Palythoa and dinoflagellates belonging to the genus Ostreopsis. Several analogs have been identified (Biré et al., 2013).

Such group of toxins involves complex polyhydroxylated compounds with both lipophilic and hydrophilic areas.

Several symptoms were described after the consumption of shellfish and included a metallic taste, gastrointestinal malaise, diarrhea, nausea, vomiting, ataxias, dizziness myalgia, dyspnea, convulsion, and bradicardia.

About 200 people showed cutaneous and respiratory problems after exposure to marine aerosols containing PlTX in 2005 in Italy, whereas similar symptoms were reported between 2006 and 2009 in France (Biré et al., 2013). In literature, some cases of intoxication were fatal (Deeds and Schwartz, 2010; Aligizaki et al., 2011).

### THE EUROPEAN UNION LEGISLATION

The EU food legislation focuses also on bivalve molluscs, giving space to them in a specific section in the Annexes to the Regulations of the "Hygiene Package". In particular, Regulation (EC) No 853/2004 provides the maximum limits for marine biotoxins and Regulation (EC) No 854/2004 establishes monitoring and sampling plans in the production areas of live bivalve molluscs.

The monitoring of biotoxins in molluscs and health effects due to their consumption are important tasks for seafood control, because marine biotoxins may cause serious diseases in humans. According to the above mentioned Regulation 854/2004/EC, the production areas are periodically monitored to check the presence of toxins-producing plankton and their occurrence in live bivalve molluscs. Such monitoring is generally applied weekly during the periods when harvesting is allowed, but the sampling frequency must be representative for the considered area, taking into account the possible variations in the presence of plankton containing marine biotoxins. Such frequency may be reduced if the risk assessment on toxins or phytoplankton occurrence suggests a very low risk of intoxication. On the contrary, if the results of monitoring exceed the regulatory limits the production area shall be closed by the competent authority and can be reopened when at least two consecutive results of biotoxin levels in molluscs meet with legislation. The closure of the production area is necessary in order to ensure that molluscs harmful to human health are not placed on the market. In fact the prevention of contaminated seafood reaching the markets is currently an effective way to protect human health (Berdalet et al., 2015).

Regulation (EC) No 2074/2005 highlights that the proposed maximum levels are based on provisional data and should be reassessed once new scientific evidence becomes available. The subsequent regulations amending the last one have provided internationally recognized methods for the different marine biotoxins. The increased complexity of biotoxin classes points out how the development of additional tools in the monitoring of production areas and molluscs is necessary. Testing methods for the detection of marine biotoxins can be implemented and refined as well. Actually these methods can be biological, functional or chemical assays, each of them showing advantages or disadvantages that must be taken into account.

The biological methods use animal models such as rats and/or mice. They are not very sensible and can be subject to interferences. Moreover, they do not give information about the concentration of the different toxins and it is difficult to identify which toxin causes the death of the mice. They are also time consuming and expensive. Even if they can have ethical problems, the MBA allows to detect new or unknown marine biotoxins during the periodic monitoring of production areas.

The functional methods use the capacity of marine biotoxins to trigger specific responses by interacting with a cellular component that selectively recognizes their structures and behaves as a receptor in sensitive systems, thereby transforming the chemical information of the incoming ligands into defined cellular effects. Therefore, the analytical challenge posed by the extreme complexity of toxin profiles is handled by the biological

system, whose responses are specific for any toxin group. The main disadvantages are represented by the employment of viable cell strains, the potential of interferences and a particular expertise of laboratory technicians.

The chemical analytical methods are the current most powerful analytical tools able to identify multiple toxins. They are based on liquid chromatography (LC) to separate marine biotoxins by an extraction step, followed by the toxin-specific detection by UV (LC–UV), fluorescence (LC–FL), or mass spectrometry (LC–MS/MS). Due to the compound specific detection, the obtained results relate to individual concentrations of the investigated compounds and can be transformed into toxic equivalent, using conversion factors.

The MBA used for detecting PSP toxins is an official AOAC method, which has been used for more than 50 years in many Member States. Such method lays down that, initially one, but preferably 2 or 3 mice are injected intraperitoneally with 1 mL of an acid extract of molluscs. The time of death is recorded and if it lasts less than 5 min, dilutions of the extract are necessary, until the time of death becomes equal to 5–7 min. At that point three mice are injected and the average time of death is determined. The toxic concentrations are calculated in Mouse Units (MU) multiplied by the dilution factor and, if necessary, by a weight correcting factor, giving the result in MU/100 g. The conversion from MU into µg of STX equivalent is obtained multiplying by a conversion factor, which is calculated by each laboratory and periodically controlled.

The HPLC chemical method with fluorimetric detection and pre-column oxidation is applicable for the determination and quantification of most PSP toxin groups. Other chemical methods to determine PSP toxins, for instance LC–MS, are still being developed.

Other methods able to detect STX-group toxins are biomolecular methods, suitable however for screening purposes only. The biomolecular methodologies for toxins of the STX group are based on three different strategies by means of receptors, characteristics of cytotoxicity and antibodies. With regard to the latter approach, it is important to point out that, although the antibodies are very sensitive, their main problem to detect toxins of the STX group is the lack of good cross-reactivity to all the members of the group. As the differences of toxicity among the compounds of such group might be very high too, not always the toxicity corresponds to the toxin levels quantified by the antibody.

The official tests for the DA group toxins are based on the LC with UV detection. As a screening system for that group of toxins, it is possible to use also an LC–MS method that is completely intra-laboratory validated. The Regulation (EC) No 1244/2007 suggested as a screening procedure for toxins of the DA group the use of the 2006.02 ASP method based on the Enzyme-Linked Immunosorbent Assay (ELISA). It shows some advantages because it is sensitive and rapid, can be automated, requires a minimum training and does not require expansive equipment. However, it must be considered that in case of conflicting results, the HPLC chemical method with UV detector is the only valid reference method.

A Standard Operating Procedure was validated under the coordination of the European Reference Laboratory for marine biotoxins (EURL-MB) in an inter-laboratory validation study carried out by the National Reference Laboratories of Belgium, France, Germany, Ireland, Italy, Netherlands, Sweden, and the United Kingdom. Such method resulted highly specific and sensitive for the direct quantitative detection of all the four groups of regulated liposoluble toxins by the LC–MS/MS, using the certified reference materials for each compound to be investigated. The chromatographic separation of toxins is performed by gradient elution. Results are reported per toxin group as requested by the EU legislation and the total toxicity is calculated using the toxicity equivalence factors (TEF) recommended by CONTAM group of EFSA.

A series of procedures, differing in the portion (hepatopancreas or whole body) to be tested and in the solvents used for extraction and purification, may be carried out to detect marine toxins as referred to in Regulation (EC) No 2074/2005 and further amendments. In particular, two different analytic procedures have been optimized: for species of molluscs with a digestive gland of large dimensions (mussels, oysters, razorshells, scallops, etc.), the procedure of hepatopancreas (20 g) must be used, whereas for species of molluscs where the digestive gland has small dimensions (clams, tellins, etc.) or is absent (tunicates, echinoderms, marine gastropods, etc.) the procedure of whole body (100 g) must be applied. The standard operating procedure provides a further protocol, whose application is recommended when the presence of YTX in the sample is suspected.

However, the lack of specificity of the biological test for DSP and its inadequacy for satisfying the regulatory limits prompted the ban of the MBA and its replacement with validated alternative methods. The Regulation (EU) No 15/2011 identified the LC– MS/MS as the reference method for liposoluble biotoxins, even if the MBA could be used as a screening method for new and unknown toxins.

#### AUTHOR CONTRIBUTIONS

MS and PV devised and drafted the review; MB, AM, and RT prepared the literature overview; GS contributed to manuscript revision.

# ACKNOWLEDGMENT

The authors thank Prof. Francesca Rosati, University of Teramo, for her support in editing the English text.

#### REFERENCES

fmicb-07-01051 July 4, 2016 Time: 14:58 # 9



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Visciano, Schirone, Berti, Milandri, Tofalo and Suzzi. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Comparison of the Effects of Environmental Parameters on the Growth Variability of *Vibrio parahaemolyticus* Coupled with Strain Sources and Genotypes Analyses

#### Bingxuan Liu1 †, Haiquan Liu1, 2, 3 †, Yingjie Pan1, 2, 3, Jing Xie1, 2, 3 and Yong Zhao1, 2, 3 \*

*<sup>1</sup> College of Food Science and Technology, Shanghai Ocean University, Shanghai, China, <sup>2</sup> Laboratory of Quality and Safety Risk Assessment for Aquatic Products on Storage and Preservation, Ministry of Agriculture, Shanghai, China, <sup>3</sup> Shanghai Engineering Research Center of Aquatic-Product Processing and Preservation, Shanghai, China*

#### *Edited by:*

*Rosanna Tofalo, University of Teramo, Italy*

#### *Reviewed by:*

*Antonio Valero, University of Cordoba, Spain Giorgia Perpetuini, University of Teramo, Italy*

> *\*Correspondence: Yong Zhao yzhao@shou.edu.cn*

*† These authors have contributed equally to this study.*

#### *Specialty section:*

*This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology*

*Received: 26 January 2016 Accepted: 09 June 2016 Published: 23 June 2016*

#### *Citation:*

*Liu B, Liu H, Pan Y, Xie J and Zhao Y (2016) Comparison of the Effects of Environmental Parameters on the Growth Variability of Vibrio parahaemolyticus Coupled with Strain Sources and Genotypes Analyses. Front. Microbiol. 7:994. doi: 10.3389/fmicb.2016.00994* Microbial growth variability plays an important role on food safety risk assessment. In this study, the growth kinetic characteristics corresponding to maximum specific growth rate (µmax) of 50 *V. parahaemolyticus* isolates from different sources and genotypes were evaluated at different temperatures (10, 20, 30, and 37◦C) and salinity (0.5, 3, 5, 7, and 9%) using the automated turbidimetric system Bioscreen C. The results demonstrated that strain growth variability increased as the growth conditions became more stressful both in terms of temperature and salinity. The coefficient of variation (CV) of µmax for temperature was larger than that for salinity, indicating that the impact of temperature on strain growth variability was greater than that of salinity. The strains isolated from freshwater aquatic products had more conspicuous growth variations than those from seawater. Moreover, the strains with *tlh*+*/tdh*+*/trh*<sup>−</sup> exhibited higher growth variability than *tlh*+*/tdh*−*/trh*<sup>−</sup> or *tlh*+*/tdh*−*/trh*+, revealing that gene heterogeneity might have possible relations with the growth variability. This research illustrates that the growth environments, strain sources as well as genotypes have impacts on strain growth variability of *V. parahaemolyticus*, which can be helpful for incorporating strain variability in predictive microbiology and microbial risk assessment.

Keywords: *Vibrio parahemolyticus,* maximum growth rate, growth variability, environmental factor, temperature, salinity, gene heterogeneity

# INTRODUCTION

Vibrio parahaemolyticus is a kind of halophytic, Gram-negative bacterium that can cause headaches, diarrhea, fever, gastroenteritis, and even life-threatening sepsis (Makino et al., 2003). Since the first V. parahaemolyticus was isolated in Japan by Fujino Komiro in 1950 (Fujino et al., 1950), it has been considered as the major food-borne pathogen involving in bacterial seafood poisoning incidents in Asia (Fujikawa et al., 2009). According to annual statistics obtained from the detection network of microbial foodborne illness in China, V. parahaemolyticus has been classified as the major foodborne pathogen (accounting for 65% of the total; Wu et al., 2014). Similarly, cases of foodborne illness caused by V. parahaemolyticus are common in Europe and the United States (Yang et al., 2008; Shen et al., 2009, 2010). In fact, the largest outbreak of V. parahaemolyticus gastroenteritis all over the world did occur in the United States in 1978 and caused up to 1133 cases (Daniels et al., 2000). As demonstrated by recent surveillance data (Ma et al., 2014; Tang et al., 2014), the control of V. parahaemolyticus continues to be challenging worldwide.

It was announced that the strain variability gave the importance as well as the difficulty in controlling V. parahaemolyticus in the previous research (Lianou and Koutsoumanis, 2011). Owing to the fact that V. parahaemolyticus is mainly distributed in estuaries, coastal waters, sediments, and aquatic products (such as shrimp, cod, mackerel, and shellfish), it has become the major sources of food-borne pathogen (Wu et al., 2014). Since multiple strain composites of foodborne pathogens with robust growth or inactivation characteristics are preferred in food safety researches that aimed at assessing the behavior of bacterial pathogens in food products (NACMCF (National Advisory Committee on Microbiological Criteria for Foods), 2005; Scott et al., 2005), the characterizations of a variety of strains with respect to phenotypic responses, such as the growth behavior under different environmental conditions, should be analyzed (Nishina et al., 2004). Additionally, V. parahaemolyticus strains in the environment exhibit a halophilic and seasonal distribution, which are directly related to the salinity and temperature (DePaola et al., 2003; Zimmerman et al., 2007; Johnson et al., 2010; Sobrinho et al., 2014; Esteves et al., 2015). For the purpose of evaluation, the quantitative microbial risk assessment (QMRA) of V. parahaemolyticus should be estimated by at least two factors: temperature (T) and salinity (sodium chloride) (Nauta, 2002; U.S. Food and Drug Administration, 2005). It means that the use of predictive models of growth variability is mainly associated with the T-value and sodium chloride (NaCl) concentration (Ratkowsky et al., 1982; Larsen et al., 2015).

In previous studies of V. parahaemolyticus, only a few of them investigated the relationship between growth environments and strain variability (Fujikawa et al., 2009; Larsen et al., 2015). In west countries such as the USA, aquatic animals are mostly cultured in seawater (Depaola et al., 1990); however, in China, a majority of aquatic farmers practice freshwater aquaculture (Wu et al., 2014). It has been determined that the differences of the source for V. parahaemolyticus strains result in a large amount of diversity in the predictive models of growth variability (McMeekin et al., 1993). Furthermore, most previous research findings of the strain variability of the growth kinetic behavior of foodborne pathogens are based on marine culture (Wong et al., 2000; Alam et al., 2002; Larsen et al., 2015), which would be discrepant from China's actual conditions. Thus, new models with the purpose of developing a safe food production process in China should be built. Moreover, further studies on the influence of gene heterogeneity on growth variability were even less (Lianou and Koutsoumanis, 2013; Lopez-Joven et al., 2015), while as mentioned in Martins and Locke (2015), gene heterogeneity could determine phenotypic heterogeneity including strain growth variability, and therefore this variability might reflect the gene heterogeneity as well.

As the growth variability can introduce the food safety risk, the quantification of the growth variability can better service to the QMRA in microbiology. Aiming at furthering the development of precautionary food safety against V. parahaemolyticus in China, the influences of the T value and NaCl, together with strain sources and genotypes on the growth variability were evaluated in this research. The obtained appropriate data of the growth variability for V. parahaemolyticus could be useful for better characterizing the kinetic behaviors of V. parahaemolyticus in different growth environments (Miles et al., 1997; Yang et al., 2009). In total, 9000 optical density (OD) curves with 50 isolates of V. parahaemolyticus from different sources were generated for four levels temperatures and five levels NaCl concentrations, which will accomplish the following: (1) determine the influences of temperature and salinity on growth variability, and discuss the comparison between this two environmental factors; (2) reveal the growth variability of strains isolated from the aquatic products in freshwater and seawater; (3) demonstrate the effects of gene heterogeneity on the growth variability; and (4) provide a reasonable environmental condition for the storage of preserved food against V. parahaemolyticus.

# MATERIALS AND METHODS

#### *V. parahaemolyticus* Strains

Fifty strains of V. parahaemolyticus were isolated from the shrimps which were cultured in freshwater or seawater. The strain information was shown in **Table 1**. tlh+/tdh+/trh−, tlh+/tdh−/trh+, and tlh+/tdh−/trh<sup>−</sup> genes were used for distinguishing the genotype of the isolates (Bej et al., 1999; Okada et al., 2009), twelve V. parahaemolyticus strains were tlh+/tdh <sup>+</sup>/trh<sup>−</sup> genotype, eleven V. parahaemolyticus strains were tlh+/tdh−/trh<sup>+</sup> genotype, one strain (42) was tlh+/tdh+/trh<sup>+</sup> genotype and others were tlh+/tdh−/trh<sup>−</sup> genotype in **Table 1**. All the strains in the present study were stored frozen (−80◦C) in 25% glycerol test tubes. The V. parahaemolyticus strains were first dispensed onto thiosulfatecitrate-bile salts-sucrose agar culture medium (TCBS; Beijing Land Bridge Technology Company Ltd., Beijing, China) plates, and cultured for 18–24 h at 37◦C. The single green strain on TCBS plates was then transferred into 10 ml tryptic soy broth (TSB; Beijing Land Bridge Technology Company Ltd., Beijing, China) with pH 8.0 and 3.0% (w/w) NaCl concentration. The 18 h cultures were incubated at 37◦C for the preparation of the test inocula. The initial strain concentrations of the inocula were about 10<sup>9</sup> CFU/ml after incubation. The automated turbidimetric system Bioscreen C (Oy Growth Curves Ab Ltd., Raisio, Finland) was used for testing the corresponding Optical density (OD) values. OD measurements were taken at regular time intervals using the wideband filter (420–580 nm) of the instrument, for a total time period such that a considerable OD change was observed.


*"*+*" represents positive genotypic, and "*−*" means negative genotypic.*

#### Growth Experiments

To evaluate the single effect of the T value or NaCl concentration on the growth variability in terms of the two environmental factors, a total of 20 different growth conditions were assessed with 4-levels (10, 20, 30, and 37◦C) of temperature and 5-levels (0.5, 3, 5, 7, and 9%) of NaCl concentrations so as to cover the most probable growth region of the V. parahaemolyticus strains. The maximum and minimum boundaries of the T value (37 and 10◦C, respectively) and the NaCl concentration (9 and 0.5%, respectively) were set up based on the findings of preliminary experiments in which the growth environment approximately reached the minimum growth requirements (V. parahaemolyticus strains approached the minimum growth rate in the condition of 10◦C and 9% salinity) or the maximum growth requirements (V. parahaemolyticus strains attained the maximum growth rate in an optimum environment with 37◦C and 3% salinity condition).The prepared initial inocula of each strain were decimally diluted in the TSB with 5 levels of NaCl concentration separately for five times. With strain concentration of approximately 10<sup>4</sup> CFU/ml, the inoculated TSB were transferred into 100-well microtiter plates, which were then placed in the automated turbidimetric system Bioscreen C for 4 levels of temperatures, respectively. Totally three OD measurement replicates were tested in this process. Additionally, three independent experiments were conducted at each growth condition and therefore there were three samples per strain altogether for testing. In such a way, the total number of the described OD curves would amount to 9000 patterns (3 replicates × 3 independent experiments × 20 growth conditions × 50 types of V. parahaemolyticus). The counted data were analyzed in order to achieve an accurate approximation of the V. parahaemolyticus growth states in different cultured environments. Moreover, it would be more reasonable for the V. parahaemolyticus strain evaluation of QMRA (Vose, 1998).

#### Maximum Specific Growth Rate

The maximum specific growth rate (µmax) (Dalgaard and Koutsoumanis, 2001) of each strain at each growth condition was estimated according to Mytilinaios et al. (2012). By using the decimal dilution approach with Bioscreen C, the novel calculation of the maximum growth rate in the unit of OD∗h −1 can be formulated in the model of Modified Gompertz (Gibson et al., 1987; Zwietering et al., 1990; Gil et al., 2006; Juneja et al., 2007; Yoon et al., 2008), with a little regulation, as the following equation:

$$\mathcal{Y} = A + C \exp\left\{-\exp\left[\frac{\mu\_m}{A}(\lambda - t) + 1\right]\right\}$$

Where A means the initial amount of bacteria, C represents the difference between the initial amount and the maximum amount of bacteria, µ<sup>m</sup> represents maximum specific growth rate and λ is the lag time of the strain growth.

To calculate the maximum growth rate, the obtained data with both OD values and cultured times were taken into the above equation in the place of y and t, respectively. The OD curves were then fitted and the matrix of function was calculated including A, C, λ, and µm.

#### Statistical Analysis

The statistical indicators were used to compare the performance of the models: correlation coefficients (R 2 ), the p values from the Fisher F-test, and root mean square error (RMSE), accuracy factor (A<sup>f</sup> ), and bias factor (B<sup>f</sup> ), whose mathematical expressions are as follows:

$$R^2 = \left[1 - \frac{\sum \left(pred - obs\right)^2}{\sum \left(obs - mean\right)^2}\right]$$

$$RMSE = \sqrt{\frac{\sum \left(obs - pred\right)^2}{n}}$$

$$A\_f = 10 \left(\frac{\sum \left(Log \left(pred / obs\right) \right)}{n}\right)$$

$$B\_f = 10 \left(\frac{\sum \log \left(pred / obs\right)}{n}\right)$$

where obs is observed values, pred is predicted values by models, mean is average values, and the n stands for the number of observations. The RMSE values approaching zero indicate a closer fit with the data for the model (Zhang et al., 2015). A<sup>f</sup> provides the accuracy of the model, which reflects how close the predicted values are to the observed values, while B<sup>f</sup> indicates the mean difference between observed and predicted value. Ideally, predictive models would have A<sup>f</sup> = B<sup>f</sup> = 1 (Wang et al., 2014).

The coefficient of variation (CV) of µmax in different conditions were calculated within the formula as

$$CV = \frac{\text{standard deviation of } \mu\_{\text{max}}}{\text{mean value of } \mu\_{\text{max}}} \times 100\%$$

Significance testing making use of p-values was applied to verify the differences of the strain growth rate in different sources. Values differences were compared using the least significant difference (LSD) method at p = 0.05.Statistical analysis was performed using SPSS statistical package17.0 (SPSS Inc., Chicago, IL).

#### RESULTS

#### Tendency of Maximum Growth Rates

The estimated maximum specific growth rate µmax vs. 50 strains in various growth environments were calculated are presented in Supplementary Material, and almost all of the values were fitted in the equation given above. By statistical analysis, all the correlation coefficients achieved above 93%, and all RMSE values approached zero. Both accuracy factors and bias factors got close to 1. The results showed a satisfactory goodness-of-fit in this study. A fraction of the maximal growth rate values could not yet be evaluated by Modified Gompertz model (Lammerding, 1997; Anderson and Hattis, 1999; Nauta and Dufrenne, 1999). It should be pointed out that the equation still cannot afford the actual growth state (Li et al., 2007), which requires the construction of a microbial macro growth model in multi-parameters.

Based on the µmax in Supplementary Material, the tendency chats in various growth environments are shown in **Figure 1**. As shown in **Figure 1A** for the T of 37◦C, the mean value of µmax (OD∗h −1 ) ranged from 0.03 to 0.24 in the condition of 0.5% NaCl, from 0.02 to 0.44 at 3% NaCl, from 0.01 to 0.26 at 5% NaCl, from 0 to 0.15 at 7% NaCl, and from 0 to 0.12 at 9% NaCl among the 50 strains. While with the same NaCl concentration of 3% in the TSB, the mean µmax (OD∗h −1 ) ranged from 0.02 to 0.44 at 37◦C, from 0.005 to 0.065 at 30 ◦C, from 0.007 to 0.031 at 20◦C, and from 0.001 to 0.014 at 10◦C. Obviously, the average growth rate in the condition of 37◦C and 3% NaCl concentration was found to be the largest (**Figure 1B**). Therefore, 37◦C and 3% NaCl were considered as the optimal growth temperature and salinity respectively, similarly 10◦C and 9% NaCl were considered as the most non-optimal temperature and salinity in this research.

#### Evaluation of Growth Variability in Different Temperatures and Salinities

The optimal growth condition at 37◦C with 3% NaCl concentration was used as the reference in **Figure 1**. In this case, the strains from No. 1 to No. 50 tended to be staged growth with an increasing maximum specific growth rate µmax. While in other conditions, the strains from No. 1 to No. 50 seemed to grow randomly with no fixed growth trend as compared with that of the optimal growth condition. For example, the strains No. 50 and No. 1 at 37◦C with 3% NaCl salinity had the highest growth rate and the lowest growth rate respectively, but in the condition at 20◦C with 3% NaCl salinity, the No. 50 and No. 1 both located in the intermediate range of µmax in all 50 strains, nearly 0.02 OD∗h −1 . Similar situations also appeared in other strains like No. 2, No. 13, No. 28 strains at 37◦C with 5% NaCl salinity compared with those at 10◦C with 3% NaCl concentration.

The curves related to the coefficient of variation (CV) of µmax in different conditions were drawn in **Figure 2**. The CV value of maximum growth rate represented the growth variability for V. parahaemolyticus strains. The CV value among the tested strains at 37◦C-3% NaCl concentration was 12.7%, while at 37◦C-0.5% NaCl concentration and T 37◦C-5% NaCl concentration, it was 13.0 and 15.1%, respectively (**Figure 2A**). The CV value among the tested strains corresponding to a mean µmax of approximately 0.16 OD∗h <sup>−</sup><sup>1</sup> was 12.7% for 37◦C-3% NaCl concentration, while corresponding to a mean µmax of approximately 0.03 OD∗h −1 , the CV value was 16.3% for 30◦C-3% NaCl concentration in **Figure 2B**. The non-optimal T and NaCl concentration led to an increase of CV values in the activation range of 0–5% NaCl and 30–37◦C. On the contrary, in the inactivation range of V. parahaemolyticus strains, since the maximum growth rate dropped to nearly 0 OD∗h −1 , the CV

NaCl).

values of µmax would similarly drop down, with less variance of growth variability in V. parahaemolyticus strains as shown in the points among the 7–9% NaCl concentration and 10–20◦C.

#### Comparison of Growth Variability from Different Sources

From the different types of environmental sources in **Table 1**, the V. parahaemolyticus strains could be roughly divided into two categories: freshwater and seawater. The significant difference analyses between these two categories in four environmental conditions (37◦C-3% NaCl, 30◦C-3% NaCl, 37◦C-9% NaCl, and 10◦C-3% NaCl) with in the box plot method were respectively drawn in **Figure 3**. The 4 environmental conditions represented 4 typical growth kinetics of V. parahaemolyticus strains. These four box plots were counted by the mean of the maximum growth rate µmax in freshwater and seawater accordingly. In addition, the significant differences were calculated by p-value, with 0, 0.063, 0.001, and 0.024 respectively, which demonstrated the growth variability in these two sources. The non-significant difference occurred in the condition at 30◦C-3% NaCl, and other conditions performed as the significant difference.

### Influence of Genotypes on Growth Variability

Further investigation of the growth variability of V. parahaemolyticus strains was studied through gene heterogeneity. For the purpose of research on the effect of genotypes on growth variability, all 50 strains were classified by growth condition and genotype. Based on each genotype with different virulence factors of V. parahaemolyticus strains (Letchumanan et al., 2014), four groups of virulence genes-related V. parahaemolyticus strains, tlh+/tdh−/trh−, tlh+/tdh+/trh <sup>−</sup>, tlh+/tdh−/trh+, and tlh+/tdh+/trh+, were introduced in this research in order to explore the internal causes of the growth variability of V. parahaemolyticus. Among these virulence genes, tlh has been expressed by all clinical and environmental strains of V. parahaemolyticus in previous studies (Bej et al., 1999; Okada et al., 2009); thus, the tlh virulence gene was contained in all four groups of isolates. The inter-specific variability of four genotype factors in the environmental factors of T and NaCl concentration is given in **Figure 4**. In **Figure 4A**, the temperature was fixed at the optimal condition of 37◦C, and the genotype tlh+/tdh+/trh<sup>−</sup> (colored in red) embodied the largest strain growth variability. The associated CV values were set at a high level compared with 3 other genotypes. In contrast, the tlh+/tdh+/trh<sup>+</sup> genotype had the lowest CV values overall. Similar circumstances appeared in **Figure 4B** with the NaCl concentration set at 3% as well. The growth variability of tlh+/tdh−/trh<sup>+</sup> (colored in green) and tlh+/tdh−/trh<sup>−</sup> (colored in blue) performed moderate, overtopping the CV values only in the condition at 30◦C and 3% NaCl concentration.

# DISCUSSION

# Effects of Temperature and NaCl Concentration on µmax

It has been reported that the V. parahaemolyticus strains cannot grow at low temperature in nature since there is growth inhibition for V. parahaemolyticus strains with T-values below 10◦C in freshwater or seawater (Cook and Ruple, 1989; Burnham et al., 2009). Regarding growth rate, the conditions at the lowest temperature (i.e., 10◦C) and the highest NaCl concentration (i.e., 9%) almost approached the minimum growth requirements of this pathogen (Fujikawa et al., 2009). In contrast among all of the conditions, it is obvious that the growth condition of 37◦C and 3% NaCl concentration is considered to be the optimal growth condition with the widest range and maximum mean value of specific growth rate compared to the other combinations (**Figure 1**), which has already been verified the similar completion in accordance with its own growth habits and laboratory experiments (Miles et al., 1997; Liu et al., 2008; Yang et al., 2009; Baker-Austin et al., 2010; Fernandez-Piquer et al., 2011). In this study, however, a much wider range of growth conditions in discussing both temperature and salinity simultaneously will give more comprehensive testimony for understanding the variability of the V.

FIGURE 4 | The influence of the genotype on the growth variability of *V. parahaemolyticus* strains in various conditions. (A) various NaCl concentrations with a fixed T value of 37◦C and (B) various temperatures with a fixed NaCl concentration of 3%.

parahaemolyticus kinetic behavior within the scope of the growth environments.

It seemed that the inter-species growth variability of the 50 strains occurred at different environmental conditions. The trend of maximum growth rate in various conditions indicates that the external environment, such as the temperature and salinity, can affect the growth variability among V. parahaemolyticus strains, and such inter-species growth variability performs randomly (Whiting and Golden, 2002).

Moreover, because of incomplete knowledge of the effects of environmental conditions on model parameters in current microbiological studies (Nauta, 2002), the quantitative information based on a single impact parameter, like only the Tvalue or the NaCl concentration, needs to be evaluated separately in order to analyze the strain variability of V. parahaemolyticus (Fujikawa et al., 2009). Since there are no methods for separating two impact parameters absolutely, the mutual comparison between the one most non-optimal factor (T-value or NaCl concentration in this study) and the other factor in the optimal condition that emerged by the quantitative data is used here for approaching the actual microbial growth model (Lindqvist, 2006). As can be seen in the growth condition of 37◦C-9% NaCl concentration, nearly half of the strains were inactive in such an inappropriate growth environment for V. parahaemolyticus, and it seemed that the rest of the strains still "struggled" in a random range of growth rate, mostly from 0.12 to 0, with a mean µmax value of about 0.06; In another extreme condition with the most non-optimal temperature case while the optimal salinity: 10◦C-3% NaCl concentration, although there were few inactive strains, the mean value of the growth rate µmax could just achieve 0.01 or below, and µmax ranged at a smaller scale from 0.015 to 0. The difference indicates that although the temperature and salinity have the same net effect on strain variability, meaning that the µmax variability among the strains increases as the T-values or NaCl concentrations become more unfavorable for V. parahaemolyticus growth, the extent for this growth variability appears to be different for these two environmental parameters (Nauta, 2000). It is illustrated that as one of the two main impact parameters, the influence of low temperature on the decrease of µmax appears to be greater than that of high NaCl concentration in **Figure 1**.

The analysis in **Figure 1** could provide the formulations of temperature and salt which do not allow growth of V. parahaemolyticus. Normally, up to 9% NaCl concentration leads to inactivation for the majority of microorganisms (Francois et al., 2006); however, for halophilic bacteria like V. parahaemolyticus, such NaCl concentration cannot entirely prevent growth of bacteria (Anon, 1988; David et al., 1997). With a peak µmax value of 0.12 for strain No. 7, 30, and 31, it seemed that the strain growth rate was not suppressed. To avoid the growth of most pathogens, the other impact parameter, the temperature, plays an important role in the suppression of µmax. As it was revealed above in the data from 20◦C or even 10◦C conditions, the V. parahaemolyticus strains in a low temperature could be more easily inactivated with weaker growth behavior. This suggests that preserved foods in a salty environment should be stored in low temperatures below 10◦C, which can aid in avoiding V. parahaemolyticus growth. The present study gives a convincing data basis for the instruction of manufacturing preserved foods.

# Effects of Temperature and NaCl Concentration on Growth Variability

As reviewed by Nauta (2002), the assumption is often made by food microbiologists that strain-to-strain variation is equal to or smaller than experimental variation, thus it is not necessary to determine strain-to-strain variation. The data presented here demonstrated that the strain variability of the estimated µmax values increased as the growth conditions became more stressful both in terms of NaCl (**Figure 2A**) and T (**Figure 2B**). The phenomenon that the non-optimal growth condition has a greater strain variability of growth kinetics than the optimal condition has been pointed in previous studies (Barbosa et al., 1994; Begot et al., 1997; Lianou et al., 2006).

In **Figure 2** with both two cases, the maximum CV value occurred at the environmental condition of 30◦C and 3% NaCl concentration. It meant that the growth variability of 50 strains was larger than that in the optimal growth condition and any other conditions. Thus, the condition might introduce much difficulties for the control of food safety risk. Actually, this condition comes closest to the natural environment, leading to a big challenge for food safety control (Pouillot et al., 2003). Nevertheless, the V. parahaemolyticus strains in this condition had a relative high consistency with the medium-tohigh maximum growth rate, it might therefore maintain a large variety of serotypes as much as possible, which would achieve a diversity of V. parahaemolyticus strains with a similar growth rate when incubated in the same TSB. It exerted favorable effect on strain selection in the growth environment of 30◦C and 3% NaCl concentration. Besides, with the similar maximum growth rate of 0.04 in the condition of T 37◦C-9% NaCl in **Figure 2A** and the condition of T 30◦C-3% NaCl in **Figure 2B**, the CV values in the two cases were quite different, corresponding to 3 and 16.27%, respectively, which meant that the growth variability was larger for decreasing temperature than for increasing salinity. It again proves the fact that temperature variation always leads to a more gradual increase in the growth variability in V. parahaemolyticus strains than NaCl variation.

#### Impact of Different Sources on the Growth Variability

In **Figure 3**, the largest p-value for the difference between the freshwater and seawater occurred in the condition of 30◦C and 3% NaCl, meaning this difference of µmax was not significant. Obviously, this environment condition is the common state found in nature, especially in the subtropical and temperate coastal areas, which means that the V. parahaemolyticus strains that are grown in freshwater and in seawater result in a similar growth rate with relatively consistent growth variability in the normal state found in nature (Larsen et al., 2015). Moreover, the 30◦C-3% NaCl condition could aid in strain selection since the consistency of the growth variability from fresh and sea water was optimal in all of the tested conditions, which offered the largest variety of growth variability. In other words, the natural environment found in the coastal areas will lead to a large growth variability in the V. parahaemolyticus strains, which results in difficulties for the QMRA and food safety control (Pouillot et al., 2003), as discussed in **Figure 2**.

Another interesting point is that there was an extremely significant difference in the growth variability of the freshwater and seawater V. parahaemolyticus strains in the environmental condition of 37◦C and 3% NaCl concentration. It has been evaluated that the V. parahaemolyticus strains reached the largest maximum growth rate in such an optimal growth condition in **Figure 3**. On the contrary, the difference in the NaCl concentration caused the freshwater strains (living in 0.1% salinity in nature) having a larger growth variability than the seawater strains (living in 3% salinity in nature), which resulted in a higher mean value of maximum growth rate with a larger distribution according to an evaluation of its standard deviation. It has been stated that strains in a non-optimal environment condition will have growth variability at a considerably higher level than those in the optimal growth condition, and larger environmental pressure will lead to larger growth variability. In addition, based on the comparison between the condition of 37◦C-9% NaCl concentration and that of 10◦C-3% NaCl concentration, it is confirmed that a decreasing temperature leads to a somewhat more gradual decrease of µmax than an increasing NaCl concentration does.

#### Influence of Different Genotypes on the Growth Variability

According to collected data from **Table 1**, since there was only one strain comprising tlh+/tdh+/trh+, it should have no typical representativeness for the properties of this genotype, while the curve in red gave some reference for the tendency of different genotypes in **Figure 4**. From the results, it was concluded that the genotype of the tlh+/tdh+/trh<sup>−</sup> resulted in the largest variation degree in the growth variability of the V. parahaemolyticus strains in all four groups, whereas the genotype with tlh+/tdh+/trh<sup>+</sup> illustrated the least obvious variation degree from among those cultured in the environment condition with temperature and NaCl concentration, which verified that gene heterogeneity also affected the growth interspecific variability for V. parahaemolyticus. Furthermore, as one of the major virulence genotypes in V. parahaemolyticus, tlh+/tdh+/trh<sup>−</sup> modeled the most non-optimal case in evaluating QMRA, due to there being a large risk of growth variability in reality. In addition, it is suggested that in the food safety control of clinical V. parahaemolyticus strains, more attention should be paid to the genotype of tlh+/tdh+/trh−, which is associated with serious virulence (Miyamoto et al., 1969; Honda and Iida, 1993; Baffone et al., 2005) and large growth variability in most of the environmental conditions. The research exploring the effect of gene heterogeneity on the grow variability in V. parahaemolyticus provides a useful reference for the prevention of pathogenic V. parahaemolyticus in nature.

# CONCLUSION

In the present study, the growth kinetics characteristics of 50 V. parahaemolyticus isolates with different sources and genotypes were assessed at different temperatures (10, 20, 30, and 37◦C) and salinity (0.5, 3, 7, and 9%). From the experimental results, it was concluded that the strain variability increased as the growth conditions became more stressful both in terms of temperature and salinity in the activation range, and temperature has larger impacts than salinity on strain growth variability. Therefore, the preserved foods in a salty environment were suggested to be stored in a low temperature below 10◦C, which could promise the inactivation of V. parahaemolyticus strains. Moreover, the results showed the fact that the strains isolated from freshwater aquatic product had more conspicuous variations than those from seawater. And it was interpreted that gene heterogeneity also affected strain growth variability of V. parahaemolyticus. The findings of this study should be useful in incorporating strain variability in predictive microbiology and microbial risk assessment, and could provide scientific guidance for V. parahaemolyticus verification and prevention in nature as well as strain selection in experiments.

# AUTHOR CONTRIBUTIONS

BL performed the data analyses and wrote the manuscript; contributed significantly to analysis and manuscript preparation; HL helped perform the analysis with constructive discussions; Substantial contributions to the design of the work and analysis the results. YP, JX drafted the work or revising it critically for important intellectual content. Agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. YZ contributed to the conception of the study. Drafting the work or revising it critically for important intellectual content. Final approval of the version to be published.

# FUNDING

This research was supported by the National Natural Science Foundation of China (31271870, 31571917), the project of Science and Technology Commission of Shanghai Municipality (14DZ1205100, 14320502100), Key Project of Shanghai Agriculture Prosperity through Science and Technology (2014, 3-5 and 2015, 4-8), Shanghai Engineering Research Center of Aquatic-Product Processing & Preservation (11DZ2280300), and the "Dawn" Program of Shanghai Education Commission (15SG48).

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.00994

# REFERENCES


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Liu, Liu, Pan, Xie and Zhao. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Rapid Detection and Isolation of Escherichia coli O104:H4 from Milk Using Monoclonal Antibody-coated Magnetic Beads

Mirella Luciani<sup>1</sup> \*, Tiziana Di Febo<sup>1</sup> , Katiuscia Zilli<sup>1</sup> , Elisabetta Di Giannatale<sup>1</sup> , Gisella Armillotta<sup>1</sup> , Laura Manna<sup>1</sup> , Fabio Minelli<sup>2</sup> , Manuela Tittarelli<sup>1</sup> and Alfredo Caprioli<sup>2</sup>

1 Istituto Zooprofilattico Sperimentale dell'Abruzzo e del Molise "G. Caporale", Teramo, Italy, <sup>2</sup> EU Reference Laboratory for E. coli, Department of Veterinary Public Health and Food Safety, Istituto Superiore di Sanità, Rome, Italy

Monoclonal antibodies (MAbs) specific for the lipopolysaccharide (LPS) of Escherichia coli O104:H4 were produced by fusion of Sp2/O-Ag-14 mouse myeloma cells with spleen cells of Balb/c mice, immunized with heat-inactivated and sonicated E. coli O104:H4 bacterial cells. Four MAbs specific for the E. coli O104:H4 LPS (1E6G6, 1F4C9, 3G6G7, and 4G10D2) were characterized and evaluated for the use in a method for the detection of E. coli O104:H4 in milk samples that involves antibody conjugation to magnetic microbeads to reduce time and increase the efficiency of isolation. MAb 1E6G6 was selected and coupled to microbeads, then used for immuno-magnetic separation (IMS); the efficiency of the IMS method for E. coli O104:H4 isolation from milk was evaluated and compared to that of the EU RL VTEC conventional culturebased isolation procedure. Milk suspensions also containing other pathogenic bacteria that could potentially be found in milk (Campylobacter jejuni, Listeria monocytogenes, and Staphylococcus aureus) were also tested to evaluate the specificity of MAb-coated beads. Beads coated with MAb 1E6G6 showed a good ability to capture the E. coli O104:H4, even in milk samples contaminated with other bacteria, with a higher number of E. coli O104:H4 CFU reisolated in comparison with the official method (121 and 41 CFU, respectively, at 10<sup>3</sup> E. coli O104:H4 initial load; 19 and 6 CFU, respectively, at 10<sup>2</sup> E. coli O104:H4 initial load; 1 and 0 CFU, respectively, at 10<sup>1</sup> E. coli O104:H4 initial load). The specificity was 100%.

Keywords: detection, Escherichia coli O104:H4, immuno-magnetic separation, milk, monoclonal antibodies

# INTRODUCTION

Escherichia coli (Enterobacteriaceae) is a Gram-negative, facultative anaerobic bacterium that is commonly found in the lower intestinal tract of healthy animals and humans. However, several E. coli strains have acquired virulence traits that allow them to cause disease in humans and animals. At least six categories of pathogenic E. coli able to affect the human gut have been described: Shiga-toxin-producing E. coli, also called verocytotoxin-producing E. coli (STEC or VTEC), of which enterohaemorrhagic E. coli (EHEC) are a highly pathogenic sub-group causing

#### Edited by:

Maria Schirone, University of Teramo, Italy

#### Reviewed by:

Pendru Raghunath, Texila American University, Guyana Giovanni Normanno, University of Foggia, Italy

> \*Correspondence: Mirella Luciani m.luciani@izs.it

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 18 February 2016 Accepted: 01 June 2016 Published: 15 June 2016

#### Citation:

Luciani M, Di Febo T, Zilli K, Di Giannatale E, Armillotta G, Manna L, Minelli F, Tittarelli M and Caprioli A (2016) Rapid Detection and Isolation of Escherichia coli O104:H4 from Milk Using Monoclonal Antibody-coated Magnetic Beads. Front. Microbiol. 7:942. doi: 10.3389/fmicb.2016.00942

bloody diarrhea and the hemolytic uremic syndrome (HUS), characterized by severe acute renal failure, thrombocytopenia and micro-angiopathic haemolytic anemia (European Centre for Disease Prevention and Control [ECDC] and European Food Safety Authority [EFSA], 2011); enteropathogenic E. coli (EPEC); enterotoxigenic E. coli (ETEC); enteroaggregative E. coli (EAggEC); enteroinvasive E. coli (EIEC), and attaching and effacing E. coli (A/EEC) (European Centre for Disease Prevention and Control [ECDC] and European Food Safety Authority [EFSA], 2011; Farrokh et al., 2013).

Prior to 2011, STEC serogroup O104 was not considered as a major STEC serogroup, although it had been associated with an outbreak of diarrhea in the US and with sporadic cases in European countries and Korea (European Centre for Disease Prevention and Control [ECDC] and European Food Safety Authority [EFSA], 2011; Baranzoni et al., 2014).

The concern about this serogroup increased in May-July 2011, with the occurrence of two outbreaks of bloody diarrhea and HUS in Europe: one in Germany (around 4000 cases of bloody diarrhea, 850 cases of HUS and 50 deaths), and a much smaller outbreak in southwest France (15 cases of bloody diarrhea, 9 of which progressed to HUS). Both outbreaks were caused by a STEC strain belonging to serotype O104:H4 and linked to the consumption of contaminated sprouts from fenugreek seeds (Grad et al., 2012; Baranzoni et al., 2014). The genetic analysis of the outbreak strain revealed that it carried virulence genes associated with both STEC and EaggEC (Bielaszewska et al., 2011; Scheutz et al., 2011; Baranzoni et al., 2014); in addition, all isolates also expressed the phenotypes that define STEC and EaggEC, specifically production of Shiga-toxin 2 (Stx2) and the aggregative adherence pattern on intestinal epithelial cells, and were resistant to all penicillins and cephalosporins and to co-trimoxazole (trimethoprim-sulfamethoxazole). The specific combination of the higher adherence to intestinal cells, physical survival, Stx2 production and antibiotic resistance, shows the high genomic plasticity of E. coli O104:H4 and could explain the high virulence of the epidemic strain (Bielaszewska et al., 2011; Scheutz et al., 2011).

The severity of the oubreaks caused by this foodborne pathogen highlights the need for sensitive screening methods allowing its rapid identification and isolation from food matrices, as sprouts, milk and meat.

Raw cow's and goat's milk provides a potential growth medium for bacteria and its consumption has been frequently associated with STEC infections in Europe, USA and Canada. Most of these cases were associated with STEC O157, although other serotypes or serogroups, including O22:H8, O110:H−, O80:H−, and O145 have been identified as causative agents. Consumption of contaminated soft and semi-soft cheeses has also been implicated in outbreaks: E. coli O157:H7 was linked to the majority of cases, but O27:H20, O103, O26, O145, O119:B14, O27:H20, and O104:H21 have also been implicated (Centers for Diseases Control and Prevention [CDC], (1995); Farrokh et al., 2013). Generally, there are two suggested routes by which potentially pathogenic STEC can contaminate raw milk: rare sub-clinical mastitis causing STEC excretion from the udder and contamination during the milking process, when teats are soiled with feces. STEC could also potentially persist if milking equipment is not adequately cleaned. Contamination of dairy products (cheeses, cream, ice-cream, yogurt and butter) is commonly due to the use of raw/unpasteurized milk, to defective pasteurization of milk and/or post processing contamination (Farrokh et al., 2013).

The aim of this work was the development of an immunomagnetic separation (IMS) method based on the use of beads coated with monoclonal antibodies (MAbs) specific for the lipopolysaccharide (LPS) of E. coli O104:H4 for the rapid and efficient isolation of E. coli O104:H4 from milk samples.

#### MATERIALS AND METHODS

#### Bacterial Strains

The E. coli O104:H4 strain used for the production and the screening of MAbs and for the immunomagnetic capture was isolated from an Italian child with HUS in 2009 (Scavia et al., 2011). Other five enteroaggregative E. coli O104 strains were used to test the MAb 1E6G6. These included two VT2-positive strains and one VT-negative strain kindly provided by the Statens Seruminstitut (Copenhagen, Denmark), a VT2-positive strain kindly provided by the Robert Koch Institute (Berlin, Germany), and another VT2-positive strain kindly provided by the Hopital Robert Debre (Paris, France). The E. coli O157, E. coli O26, E. coli O103, E. coli O111, and E. coli O145 strains used for MAbs characterization, were part of the culture collection of the European Union Reference Laboratory for E. coli (EU-RL-ISS). Campylobacter jejuni ATCC 33291, Listeria monocytogenes ATCC 7644, and Staphylococcus aureus ATCC 6538 used to evaluate the specificity of MAbs, were obtained from ATCC Bacterial Collection LGC.

#### Preparation of E. coli Antigens for MAbs Production and Characterization

Escherichia coli O104:H4, E. coli O157, E. coli O26, E. coli O103, E. coli O111, and E. coli O145 were grown in Brain Heart Infusion Broth (BHI) (Oxoid, Basingstoke, Hampshire, UK) at 37◦C for 14–16 h, inactivated at 100◦C for 1 h and centrifuged at 5,500 g for 30 min. The pellet was washed for three times with 0.01 M phosphate buffered saline, pH 7.2 (PBS), resuspended in PBS and stored at –20◦C. One aliquot of the E. coli O104:H4 suspension was sonicated in ice bath for two cycles of 2.5 min each with an interval of 5 min and immediately used for mice immunization.

The protein concentration of antigen preparations was determined using the BCA Protein Assay Kit (Pierce Rockford, IL, USA).

#### Preparation of the E. coli O104 Lipopolysaccharide

The E. coli O104 LPS was extracted by the hot phenol–water method (Westphal and Jann, 1965). The LPS was precipitated with 3 M sodium-acetate and absolute ethanol, ultracentrifuged, resuspended in deionized water and stored at –80◦C until use. The quantity of LPS was determined by the 2-keto-3 deoxyoctonate (KDO) assay (Karkharis et al., 1978).

#### Immunization of Mice

fmicb-07-00942 June 13, 2016 Time: 15:16 # 3

For MAbs production, 6/8 week-old Balb/c mice were inoculated with heat-inactivated and sonicated E. coli O104:H4 preparations. Animal experimentation was carried out in compliance with Italian national law (Decreto legislativo 27 Gennaio, 1992, n. 116) implementing Directive 86/609/EEC of the Council of the European Communities on the protection of animals used for experimental and other scientific purposes (European Commission [EC], 1986). The protocol was approved by the Italian Ministry of Health with number 5146 of 26.04.2012. The whole antigen, diluted to a protein concentration of 100 µg/ml, was emulsified with complete Freund adjuvant (Sigma, St. Louis, MO, USA) and administered intraperitoneally; 14 days later a second immunization was performed using the same concentration of antigen emulsified with incomplete Freund adjuvant (Sigma). Subsequently, on days 28 and 56, 100 µg/ml of antigen diluted in PBS was given (intravenous booster). Three days later, the mice were sacrificed, the spleen collected and splenocytes subjected to cell fusion with murine myeloma cells Sp2/O-Ag-14 (ATCC CRL-1581TM).

### Characterization of Monoclonal Antibodies vs. E. coli O104:H4

The antibody-secreting hybridomas, cultured in Dulbecco's Modified Eagle's Medium (DMEM) containing 20% bovine fetal serum, 2 mM L-glutamine, 100X Penicillin–Streptomycin– Amphotericin, 50 mg/ml gentamicin and 50X HAT, were screened by i-ELISA. Briefly, 96-well microplates (PolySorp, Nunc Brand Products, Roskilde, DK) were coated with 10 µg/ml of E. coli O104 LPS and of the bacterial whole antigens (E. coli O104:H4, E. coli O157, E. coli O26, E. coli O103, E. coli O111, E. coli O145) diluted in 0.05 M carbonate-bicarbonate buffer, pH 9.6, and incubated overnight at 4◦C. After washing and blocking with 1% yeast extract in PBS containing 0.05% Tween 20 (PBST) at 37◦C for 1 h, 100 µl/well of MAbs supernatants were added and incubated for 1 h at 37◦C. As secondary antibody, ECL anti-mouse IgG conjugated with horseradish peroxidase (GE Healthcare, Little Chalfont, Buckinghamshire, UK) was used; the 3,3<sup>0</sup> ,5,5<sup>0</sup> -tetramethylbenzidine (TMB, Sigma) was adopted as chromogenic substrate. Microplate reading was performed with a biophotometer (Bio-Rad, Hercules, CA, USA) at a wavelength of 450 nm. Clones that showed optical densities (OD450) greater than or equal to 3 times the OD<sup>450</sup> of the negative control (serum of a naive Balb/c mouse) were considered as positives and were cloned by the limiting dilution method (Luciani et al., 2006). MAbs were produced in vitro on a large scale by means of serial cultures of hybridomas and collection of the supernatants and were isotyped using the Mouse-Typer Isotyping Panel (Bio-Rad).

For immunoblotting analysis, MAbs with IgG and IgM isotype were purified on affinity chromatography using, respectively, a HiTrap rProtein A FF column (GE Healthcare, Uppsala, SW) and a HiTrap IgM Purification HP column (GE Healthcare), according to the manufacturer's instructions.

Purified MAbs were concentrated with 100 kDa cutoff centrifugal filters (Millipore, Billerica, MA, USA) and resuspended in PBS. The protein concentration of purified MAbs was determined by spectrophotometry (Absorbance at 280 nm/IgG molar extinction coefficient).

The heat-treated E. coli O104:H4 suspensions (2.5 µg/well) were subjected to SDS-PAGE separation at 200 V with NuPAGE 4–12% Bis-Tris Gels Mini (Novex, Life Technologies, Carlsbad, CA, USA) and transferred onto nitrocellulose membrane with iBlot Dry Blotting System (Life Technologies). After blocking with 5% skimmed milk (Fluka Analytical, Sigma-Aldrich) in PBST for 2 h at room temperature (RT), membrane strips were incubated overnight at 4◦C with the purified MAbs. The detection of immune complexes was performed using the ECL anti-mouse IgG HRP-conjugated (GE Healthcare) and a chemiluminescent substrate (ECL Select Western Blotting Detection Reagent, GE Healthcare). The analysis of the results was performed using the Chemidoc MP (Bio-Rad) and the Quantity One Quantitation Software version 4.3 (Bio-Rad).

#### Magnetic Beads Coupling Procedure

Purified MAbs were coupled to Dynabeads M-450 Tosylactivated (Life Technologies). Twenty-five micro litre of dynabeads were first washed with 0.1 M sodium phosphate buffer, pH 7.4, and then incubated with 5 µg of each purified MAb at 37◦C for 24 h using the Dynabeads MX4 Mixer (Life Technologies). Supernatant was removed and dynabeads were resuspended in 0.1 M Tris containing 0.1% BSA (pH 8.5) and incubated for 4 h at 37◦C with gentle and continuous agitation. Supernatant was discarded and activated dynabeads were resuspended in PBS containing 0.1% BSA, 2 mM EDTA (pH 7.4), and stored at 4◦C until use.

Non-coated dynabeads were also prepared using the same protocol, but without the incubation with the MAbs (noncoated-beads control as blank).

#### Preparation of Bacterial Strains for Immuno-Magnetic Separation

Escherichia coli O104:H4 was grown in Tryptic Soy Broth (TSB) (Becton Dickinson, Franklin Lakes, NJ, USA) for 24 h at 37◦C, and then the cultures were streaked onto Tryptic Soy Agar (TSA) (Becton Dickinson) plates and incubated for 24 h in aerobic atmosphere at 37◦C. A well separated colony from the plate was transferred into 10 ml of TSB and incubated at 37◦C in order to obtain a bacterial suspension with a turbidity between 2 and 3 McFarland standards. The number of cells in the bacterial suspension was also estimated by reading the absorbance at 600 nm (OD600). The final concentration of bacterial suspension was 10<sup>11</sup> CFU. The culture was decimally diluted in sterile PBS (until 10−11); the number of bacteria in each diluted suspension was confirmed by plating onto MacConkey Agar at 37◦C for 24 h.

Campylobacter jejuni, L. monocytogenes, and S. aureus (pool of contaminating bacteria) were grown, respectively, on Karmali Agar, Aloe Agar, and TSA Agar. For each strain, a bacterial suspension in PBS with a concentration of 0.8 McFarland (about 10<sup>8</sup> CFU) was prepared.

# Immuno-Magnetic Separation and Evaluation of MAb-Coated Beads Specificity

The specificity of MAb-coated beads was first evaluated using PBS suspensions, prepared as described above, containing decreasing concentration of E. coli O104:H4 from 10<sup>6</sup> CFU to 10<sup>1</sup> CFU. Tests were repeated using E. coli O104:H4 PBS suspensions containing L. monocytogenes, C. jejuni and S. aureus at the concentration of 10<sup>3</sup> CFU each.

One ml of each PBS-diluted bacterial suspension was added with 25 µl of MAb-activated dynabeads and incubated at 4◦C for 30 min with gentle and continuous agitation. Then, dynabeads were washed four times with PBS containing 0.1% BSA and 2 mM EDTA, (pH 7.4), resuspended in 100 µl of PBS, plated onto MacConkey Agar plates and incubated at 37◦C for 24 h. The same


procedure was repeated using non-coated beads to ensure the absence of aspecific interactions between the non-coated beads and bacteria.

# Inoculation and Enrichment of Milk and E. coli O104:H4 Detection from Artificially Contaminated Milk Samples

Six aliquots (9 ml each) of bovine raw milk were artificially contaminated with 1 ml of each one of the six bacterial suspensions previously prepared (E. coli O104:H4 concentration from 10<sup>6</sup> CFU to 10<sup>1</sup> CFU). One ml of each E. coli O104:H4 contaminated milk sample was incubated with MAb-coated and non-coated beads as described above. In the same time, one ml aliquot of each milk sample was analyzed using the EU RL VTEC official method, which include an overnight enrichment step. Aliquots (100 µl) of the IMS-treated samples and of the enrichment broth cultures prepared according to the EU RL VTEC method were plated onto MacConkey Agar plates. After incubation for 24 h at 37◦C in aerobic condition, the colonies obtained with the EU RL VTEC method and with MAb-coated beads were enumerated.

To verify the absence of cross-reaction between MAbs and bacteria other than E. coli O104:H4, the same tests described above were repeated using four aliquots for each of the six cream milk samples artificially contaminated with the six bacterial suspensions at the concentration of E. coli O104:H4 from 10<sup>6</sup> CFU to 10<sup>1</sup> CFU and with each species of contaminating bacteria at the concentration of 10<sup>3</sup> CFU. Each of four aliquots of the six milk suspensions were, respectively, plated onto MacConkey Agar, onto Blood Agar for 24 h at 37◦C in aerobic condition for L. monocytogenes and S. aureus isolation, and Karmali Agar for 48 h at 42◦C in microaerobic conditions for C. jejuni isolation.

# DNA Extraction and PCR for Strains Typing

Escherichia coli O104:H4 DNA extraction was performed using the InstaGene Matrix (Bio-Rad) according to manufacturer's instructions. Primers used for the PCR assay were specific for the rfbO104 gene, encoding the O-antigen specific for E. coli O104 (O104rfbO-f 5<sup>0</sup> TGAACTGATTTTTAGGATGG 3<sup>0</sup> ; O104rfbOr 5<sup>0</sup> AGAACCTCACTCAAATTATG 3<sup>0</sup> , amplicon size 351 bp) (Bielaszewska et al., 2011). A 50 µl PCR mixture contained: 25 µl 2× TopTaq Master Mix (Qiagen, Venlo, NL) (1.25 units of TopTaq DNA Polymerase, 1× PCR buffer, dNTPs concentration 0.2 mM each, 1.5 mM MgCl2), 1 µl of each primer at the concentration of 0.2 µM, 18 µl of Nuclease-Free Water and 5 µl of template DNA. Amplifications were performed using the GeneAmp <sup>R</sup> PCR System 9700 (Applied Biosystems, Life Technologies) and the following temperature cycling conditions: initial denaturation at 95◦C for 5 min, followed by 35 cycles of denaturation at 95◦C for 30 seconds, annealing at 55◦C for 1 min, extension at 72◦C for 1 min and final extension at 72◦C for 7 min. The PCR products were resolved by electrophoresis on a 1,5% agarose gel.



#### Statistical Analysis

In order to determine differences between the EU RL VTEC conventional culture-based isolation procedure and the IMS method described in this paper, the Wilcoxon test for dependent samples was applied (confidence interval = 95%).

#### RESULTS

#### Characterization of Monoclonal Antibodies vs. the LPS of E. coli O104:H4

A total of 28 hybridomas were positive against the E. coli O104:H4 whole heat-treated antigen in indirect ELISA. Four MAbs (1E6G6, 1F4C9, 3G6G7, and 4G10D2) were found positive for both the E. coli O104:H4 LPS and whole antigen, and negative against the other O serogroups tested. MAbs isotypes are shown in **Table 1**. Characterization by western blotting revealed that the four selected MAbs reacted with the LPS of E. coli O104:H4; **Figure 1** shows the ladder-like pattern of LPS bands with molecular weights between 60 and 15 kDa recognized by the MAbs. One of the four selected MAbs, MAb 1E6G6, was tested by slide agglutination against five additional E. coli O104:H4 strains with positive results and subsequently conjugated to magnetic microbeads, to be used for a specific IMS of E. coli O104:H4 from milk samples.

#### Immuno-Magnetic Separation and Evaluation of MAb-Coated Beads Specificity

The results obtained from E. coli O104:H4 re-isolation from PBS pure bacterial suspensions are shown in **Table 2**. At E. coli concentrations higher than or equal to 10<sup>4</sup> , the number of CFU re-isolated was higher than 200. At the concentrations of 10<sup>3</sup> and 10<sup>2</sup> CFU, 76 and 3 CFU, respectively, were re-isolated following the IMS.

When the E. coli O104:H4 PBS suspensions containing other bacteria (L. monocytogenes, C. jejuni and S. aureus) were used, more than 200 E. coli O104:H4 CFU were re-isolated from suspensions with E. coli concentrations higher than or equal to 10<sup>3</sup> ; 19 and 2 CFU were re-isolated from the 10<sup>2</sup> and 10<sup>1</sup> CFU E. coli O104:H4 suspensions, respectively, (**Table 3**). None of the contaminating bacteria were captured by the MAb-coated beads. No aspecific binding between the non-coated beads and the four bacterial suspensions used in the experiments was observed.

## E. coli O104:H4 Detection in Artificially Contaminated Milk Samples

**Table 4** shows the results of the E. coli O104:H4 isolation from milk samples contaminated with scalar dilutions (from 10<sup>6</sup> to 10<sup>1</sup> CFU) of E. coli O104:H4. With the Dynabeads method, a higher number of E. coli O104:H4 CFU was isolated with respect to the EU RL method. Similar results were obtained with milk samples contaminated with E. coli O104:H4 and with other bacteria (L. monocytogenes, C. jejuni and S. aureus) (**Table 5**). The p-value obtained using the Wilcoxon test for dependent samples was 0.0517 with a confidence interval of 95%.

Moreover, none of the other contaminating bacteria were captured by MAb-coated beads, as shown by the absence of growth in the specific media. MAb-coated beads isolates were further confirmed by PCR: all the bacteria captured by coated beads were E. coli O104:H4 (**Figure 2**).

#### DISCUSSION AND CONCLUSION

Shiga toxin–producing Escherichia coli strains are important foodborne pathogens that have been responsible for numerous

TABLE 3 | Evaluation of MAb-coated beads specificity using E. coli O104:H4 PBS suspensions containing contaminating bacteria (Listeria monocytogenes, Campylobacter jejuni, and Staphylococcus aureus).


TABLE 4 | Escherichia coli O104:H4 isolation from artificially contaminated milk samples in the absence of contaminating bacteria: comparison between the EU RL VTEC official method and the E. coli O104 Dynabeads method.



TABLE 5 | Escherichia coli O104:H4 detection from artificially contaminated milk samples in the presence of contaminating bacteria (L. monocytogenes, C. jejuni, and S. aureus): comparison between the EU RL VTEC official method and the E. coli O104 Dynabeads method.

outbreaks of hemorrhagic colitis and HUS worldwide. E. coli O157:H7 is the most frequently implicated in human disease, although infections can be caused by other serotypes or serogroups, as E. coli O104:H4. Consequently it is necessary to have the availability of rapid, sensitive and specific methods for food control.

In this study, MAbs vs. E. coli O104:H4 were produced andcharacterized. The efficiency of the E. coli O104:H4 isolation from milk samples, using Dynabeads activated with a selected MAb (1E6G6) specific for E. coli O104:H4 LPS, was evaluated and compared with that of the EU RL VTEC isolation procedure (European Reference Laboratory for E.coli [EU-RL VTEC], 2011). The specificity of the adsorbed antibodies was also evaluated by preparing E. coli O104:H4 suspensions containing other pathogenic bacteria that could potentially be found in milk, such as Campylobacter jejuni, Listeria monocytogenes, and Staphylococcus aureus (Farrokh et al., 2013).

The use of MAbs, instead of polyclonal ones (PAbs), could be a valid alternative to improve the specificity of pathogen detection. PAbs may have a lack of homogeneity between lots, due to the variability of the animal immune response. They may also show cross-reactions with related pathogens. Conversely, MAbs are highly specific reagents, easily purified and with batch to batch homogeneity.

Immuno-magnetic separation is a rapid and sensitive method that permit the revelation of low concentrations of pathogens in contaminated matrices. The magnetic beads method has already been applied for the detection of E. coli O157:H7 (Weagant et al., 2011) and E. coli O104 (Baranzoni et al., 2014) from sprouts with good results. In particular, Baranzoni et al. (2014) used commercially available E. coli IMS beads in association with optimized enrichment conditions of sprouts samples. The results indicated that the IMS enrichment enhanced the ability to detect E. coli O104 and reduced the interference

of background microflora. In our work, we used O104-specific magnetic beads prepared in-house to detect this pathogen in artificially contaminated milk samples, skipping the use of a previous enrichment step. The procedure showed an analytical sensitivity higher than that of the isolation procedure included in the official EURL method for E. coli O104. Even if the statistic analysis gave a value slightly higher than the expected one, numeric data show that the immunomagnetic method is more sensitive, particularly at low concentrations of bacteria.

The MAb-coated beads showed a good capability to capture E. coli O104:H4 in artificially contaminated milk samples, even in the presence of other contaminating bacteria, such as C. jejuni, L. monocytogenes, and S. aureus, with an increase in the number of CFU re-isolated with respect to the current official method (>200 and 137 CFU, respectively, at 10<sup>4</sup> E. coli O104:H4 initial load; 121 and 41 CFU, respectively, at 10<sup>3</sup> E. coli O104:H4 initial load; 19 and 6 CFU, respectively, at 10<sup>2</sup> E. coli O104:H4 initial load and 1 and 0 CFU, respectively, at 10<sup>1</sup> E. coli O104:H4 initial load). The Dynabeads method also showed a good analytical specificity, since no contaminating bacteria were captured by the MAb-coated beads in both PBS bacterial suspensions and contaminated milk samples and no

#### REFERENCES


aspecific reactions were obtained with non-coated beads. MAb 1E6G6 was able to retain its antigen-binding activity even in the presence of a fat matrix such as full cream milk. Moreover, the direct application of the O104 IMS to a liquid matrix such as milk, skipping the broth enrichment step that requires an overnight incubation period, allowed a more rapid isolation of E. coli O104, with respect to the current official method.

Thus, magnetic beads linked with the MAb 1E6G6 against E. coli O104:H4 could be a useful tool for improving the detection of this E. coli serogroup in milk samples. Further studies are needed to evaluate the performances of MAb 1E6G6 coatedbeads when matrices other than from milk are tested.

#### AUTHOR CONTRIBUTIONS

ML, TD, and AC designed the study; ML, TD, GA, LM, FM, and MT produced, characterized and purified monoclonal antibodies; AC, FM, KZ, and ED performed microbiological analysis; ML coordinated animal management; ML, TD, and KZ drafted the manuscript; AC, ED, and MT revised the paper critically. All authors read and approved the final version of the manuscript.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Luciani, Di Febo, Zilli, Di Giannatale, Armillotta, Manna, Minelli, Tittarelli and Caprioli. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Tracing Back Clinical Campylobacter jejuni in the Northwest of Italy and Assessing Their Potential Source

Elisabetta Di Giannatale<sup>1</sup> , Giuliano Garofolo<sup>1</sup> , Alessandra Alessiani<sup>1</sup> , Guido Di Donato<sup>1</sup> , Luca Candeloro<sup>2</sup> , Walter Vencia<sup>3</sup> , Lucia Decastelli<sup>3</sup> and Francesca Marotta<sup>1</sup> \*

<sup>1</sup> National Reference Laboratory for Campylobacter, Istituto Zooprofilattico Sperimentale dell'Abruzzo e del Molise "G.Caporale", Teramo, Italy, <sup>2</sup> Department of Statistics and GIS, Istituto Zooprofilattico Sperimentale dell'Abruzzo e del Molise "G.Caporale", Teramo, Italy, <sup>3</sup> Food Hygiene and Safety Department, Istituto Zooprofilattico Sperimentale del Piemonte, Liguria e Valle d'Aosta, Torino, Italy

Food-borne campylobacteriosis is caused mainly by the handling or consumption of undercooked chicken meat or by the ingestion of contaminated raw milk. Knowledge about the contributions of different food sources to gastrointestinal disease is fundamental to prioritize food safety interventions and to establish proper control strategies. Assessing the genetic diversity among Campylobacter species is essential to our understanding of their epidemiology and population structure. We molecularly characterized 56 Campylobacter jejuni isolates (31 from patients hospitalized with gastroenteritis, 17 from raw milk samples, and 8 from chicken samples) using multilocus sequence typing (MLST) and pulsed-field gel electrophoresis (PFGE) in order to trace the source of the disease. We also used a population genetic approach to investigate the source of the human cases from six different reservoirs of infection. MLST identified 25 different sequence types and 11 clonal complexes (CCs) (21, 658, 206, 353, 443, 48, 61, 257, 1332, 354, 574) and these included several alleles not cited previously in the PubMLST international database. The most prevalent CCs were 21, 206, and 354. PFGE showed 34 pulsotypes divided between 28 different clusters. At the fine scale, by means of PFGE and MLST, only two human cases were linked to raw milk, while one case was linked to chicken meat. The investigation revealed the presence of several genotypes among the human isolates, which probably suggests multiple foci for the infections. Finally, the source attribution model we used revealed that most cases were attributed to chicken (69.75%) as the main reservoir in Italy, followed to a lesser extent by the following sources: cattle (8.25%); environment (6.28%); wild bird (7.37%); small ruminant (5.35%), and pork (2.98%). This study confirms the importance of correlating epidemiological investigations with molecular epidemiological data to better understand the dynamics of infection.

Keywords: Campylobacter, PFGE, population genetic approach, epidemiological investigations, MLST genotyping

# INTRODUCTION

Human campylobacteriosis in the European Union (EFSA, 2014) continues to be the most commonly reported zoonosis, and with 214,268 confirmed cases, this disease has considerable socio-economic impact. The Campylobacter species most commonly associated with human infections are Campylobacter jejuni followed by C. coli and C. lari, but other species, including

#### Edited by:

Maria Schirone, University of Teramo, Italy

#### Reviewed by:

Odile Tresse, French National Institute for Agricultural Research, France Issmat Kassem, The Ohio State University, USA

> \*Correspondence: Francesca Marotta f.marotta@izs.it

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 17 February 2016 Accepted: 25 May 2016 Published: 13 June 2016

#### Citation:

Di Giannatale E, Garofolo G, Alessiani A, Di Donato G, Candeloro L, Vencia W, Decastelli L and Marotta F (2016) Tracing Back Clinical Campylobacter jejuni in the Northwest of Italy and Assessing Their Potential Source. Front. Microbiol. 7:887. doi: 10.3389/fmicb.2016.00887

the non-thermophilic C. fetus, also occasionally cause human diseases (EFSA, 2014; Rodrigues et al., 2015). In Italy, 1252 cases of human campylobacteriosis were reported in 2014 (EFSA and ECDC, 2015). However, the data are likely to grossly underestimate the real number of cases (EFSA and ECDC, 2015) because the Italian reporting system for human infectious illnesses does not differentiate between gastroenteritis caused by Campylobacter and gastroenteritis caused by the other agents listed in the National Legislation in Italy (Calistri and Giovannini, 2008). Therefore, campylobacteriosis is not a statutory notification illness and the only data available on these infections are those reported voluntarily by Enter-Net, the international network for the surveillance of human gastrointestinal infections.

Generally, campylobacteriosis infections are self-limiting and only last for a few days. However, post-infection complications or extra-intestinal infections such as reactive arthritis and neurological manifestations can also arise (Haddad et al., 2010). Outbreaks caused by thermophilic Campylobacter species are most commonly connected with dairy and poultry products, as well as Campylobacter-contaminated food and untreated water (Meldrum et al., 2005; EFSA, 2008, 2014; Taylor et al., 2013). Unpasteurized or inadequately pasteurized cow's milk has been implicated as the sources of infection in some outbreaks (Schildt et al., 2006; Heuvelink et al., 2009), and recently, the European Union summary report on food-borne disease outbreaks confirmed the importance of milk as a source of human campylobacteriosis (EFSA, 2013).

In a recent survey conducted in Northern Italy, C. jejuni was detected in 12% of the bulk tank milk samples that were examined (Bianchini et al., 2014), and some disease outbreaks have been reported in the same Italian regions (two in Emilia Romagna, one in Veneto, and one in Marche) following the consumption of raw milk from self-service automatic vending machines (Amato et al., 2007; Arrigoni et al., 2009; Petruzzelli et al., 2011). Despite no official data existing for the incidence of campylobacteriosis associated with raw milk consumption, quantitative risk assessment modeling has estimated that the worst possible scenario for human infections with C. jejuni linked to the consumption of raw milk in Italy would exceed 300,000 cases per year (Giacometti et al., 2015).

The high genome diversity and plasticity within the Campylobacter genus hampers the surveillance and the outbreak detection. Nevertheless, multilocus sequence typing (MLST) and pulsed-field gel electrophoresis (PFGE) are important and well-established tools that can be used to elucidate the epidemiology of Campylobacter cases. MLST is the most common genotyping method successfully applied in population genetic models for clarify the reservoirs of infection, while PFGE is a well standardized technique very useful in localized outbreak investigation (Dingle et al., 2001; Fitzgerald et al., 2001; Sheppard et al., 2009).

The objective of this study was to undertake an epidemiological investigation of Campylobacter infections in 31 patients hospitalized in the summer of 2012 from the Piemonte region in Northern west Italy. Source investigation using MLST and PFGE comprised C. jejuni isolates from raw bovine milk and chicken meat. A polymorphism in the hip0 gene was also identified in some Campylobacter strains from raw bovine milk, and new PCR primer set for PCR identification were designed and successfully implemented in this study.

# MATERIALS AND METHODS

# C. jejuni Isolates

A total of 31 cases of campylobacteriosis were registered in August, 2012 in the local health district of Turin, Northern Italy. A temporal and spatial proximity was suspected because of a cluster of infections; consequently, active surveillance was carried on suspicious retail food items. The monitoring found 17 raw milk samples from vending machines and 8 chicken meat samples from a retail market were positive for Campylobacter. Here, the 56 C. jejuni isolates we investigated were cultured on Columbia blood agar and incubated at 42◦C for 48 h in a microaerobic atmosphere.

# DNA Isolation

Genomic DNA was extracted using an UltraClean Microbial DNA Isolation Kit (MO BIO Laboratories, CA, USA) according to the manufacturer's instructions. The DNA concentration was quantified using a Nanodrop spectrophotometer (Nanodrop Technologies, Italy).

#### Strain Identification

The isolates were typed using a PCR method described previously (Marshall et al., 1999; Di Giannatale et al., 2014).

Six strains isolated from raw milk were non-typable using the PCR method described above. Nevertheless, the strains were determined to be C. jejuni by 16S-rRNA gene sequencing (Zhou et al., 1997). Furthermore, a new PCR protocol targeting the hip0 gene was developed to check for miss-typed isolates. The sequences of the primers, which were designed against the regions upstream and downstream of the original amplification sites, were as follows: P3Fs: 5<sup>0</sup> -GGAAAAACAGGCGTTGTGGGG G-3<sup>0</sup> and P3Rs: 5<sup>0</sup> -CCGAAGAAGCCATCATCGCACC-3<sup>0</sup> ; P3Fi: 5 0 -CCTGCTTGAAGAGGGTTTGGGTGG-3<sup>0</sup> and P3Ri: 5<sup>0</sup> -TG CAACCTCACTAGCAAAATCCACA-3<sup>0</sup> . The PCR was performed using Master Mix 2X (Promega, Italy), the final concentration of the primers was 1 µM, the annealing temperature was 55◦C, and the procedure involved 35 cycles in a thermal cycler 9700 Applied Biosystems (Applied Biosystems, Foster City, CA, USA). Sequencing of the amplified fragment was performed using an ABI PRISM BigDye Terminator 3.1 Cycle Sequencing Kit (Applied Biosystems, Foster City, CA, USA) according to the manufacturer's instructions, and the sequences were analyzed with an ABI PRISM 3500 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA).

#### PFGE

Pulsed-field gel electrophoresis was conducted according to the instructions from the 2009 USA PulseNet protocol for Campylobacter (CDC, 2009). The isolates previously identified by PCR were grown on Columbia agar (48 h at 42◦C) in a

microaerophilic atmosphere and embedded in agarose blocks (Seakem Gold agarose, Lonza, Rockland, MD, USA). Following DNA purification, 1 mm of agar plugs slices was digested 18 h with SmaI restriction enzyme (Promega, Milan, Italy) and the DNA fragments were separated by PFGE (Chef Mapper II, Biorad Laboratories, Hercules, CA, USA) in 1% agarose gel (Seakem Gold agarose, Lonza).

Salmonella Braenderup H9812, digested with XbaI enzyme (Promega, Milan, Italy), was used as the standard molecular weight marker. The gel was stained with SYBR Safe DNA gel stain (Invitrogen, Cergy Pontoise, France) and photographed on a UV transilluminator (Alpha Innotech Corporation, San Leandro, CA, USA). The image analysis was performed using Bionumerics program v. 6.6 (Applied Maths NV, Sint-Martens-Latem, Belgium). The similarity analysis was carried out using the Dice coefficient (position tolerance, 1%). The unweighted pair group mathematical average was used to cluster patterns. Isolates with <90% similarity were clustered as separate pulsotypes.

#### MLST Analysis

Multilocus sequence typing was performed as reported by Dingle et al. (2001) for all C. jejuni isolates. MLST amplified a segment of seven housekeeping genes: aspA (aspartase, 477 bp), glnA (glutamine synthase, 477 bp), gltA (citrate synthase, 402 bp), glyA (serine hydroxyl methyl transferase, 507 bp), pgm (phosphor glucomutase, 498 bp), tkt (transketolase, 459 bp), and uncA (ATP synthase, alpha subunit, 489 bp) to yield a total composite sequence length (all seven loci) of 3309 bp. PCRs and sequencing reactions were carried out according to the guidelines on the Campylobacter MLST website. Briefly, purified PCR products were sequenced using an ABI PRISM BigDye <sup>R</sup> Terminator 3.1 Cycle Sequencing Kit (Applied Biosystems, Foster City, CA, USA) according to the manufacturer's instructions and then analyzed by the ABI PRISM 3500 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA). The alleles, STs, and CCs were identified using the MLST database<sup>1</sup> . Novel alleles were submitted to the PubMLST C. jejuni/C. coli database curators for number assignment.

#### Source Attribution Analysis

According to a previous study conducted in the Netherlands (Mughini Gras et al., 2012), Campylobacter isolates from humans were found to share significant similar ST frequency distribution with those from Europe (the Netherlands, the UK, and Switzerland), but were genetically dissimilar from abroad isolates (from the USA and New Zealand). We have assumed this situation is also true for Italy and the details in our reference database included 558 C. jejuni strains from previously published Italian data; Piccirillo et al., 2014) and from our previous survey (Marotta et al., 2015), as well as the 6854 C. jejuni from European countries accessible in PubMLST<sup>2</sup> .

The reservoir data we identified were pooled and arranged in six groups: (i) chicken, (ii) cattle, (iii) environment, (iv) wild bird, (v) small ruminant, and (vi) pork. Environmental strains comprised those from water, sand, and soil. The population genetics approach used for attributing human campylobacteriosis isolates to the six putative reservoirs was the asymmetric island model described by Wilson et al. (2008). This model estimates the recombination rates within the reservoirs, between the reservoirs and from each reservoir in the human population to estimate the posterior distribution used to infer the fraction of human cases attributable to each source.

### Diversity Index and Evaluation of the Combined Typing Methods

The discriminatory ability of the two typing systems (PFGE and MLST) was measured according to Simpson's diversity index. The concordance of the methods was determined by calculating the adjusted Rand and Wallace coefficients. The first coefficient is used to evaluate the extent of agreement between two typing methods, and the second method is an estimate of how much new information is obtained from a typing method in comparison with another one (Behringer et al., 2011). The adjusted Rand and Wallace coefficients were used to calculate the index from the Online Tool for Quantitative Assessment of Classification Agreement<sup>3</sup> .

# RESULTS

A total of 55 C. jejuni isolates were typed with PFGE using SmaI restriction enzyme. One chicken isolate could not be typed. Thirty-four different pulsotypes were obtained from the isolates typed (**Figure 1**). The cluster analysis distinguished the isolates according to 28 main clusters, with a similarity of 90% (**Figure 1**). Among these, only two clusters included non-human sources and three clinical isolates. Cluster VIII contained two human isolates, six raw milk isolates and two chicken meat isolates, and cluster XVII, contained one human and four raw milk isolates (**Figure 1**). MLST typing for the 56 isolates identified 25 STs featured in 11 clonal CCs (**Figure 2**). The most prevalent CCs were 21, 206, and 354. CC21 was the largest one shared by humans, cattle, and chickens, while CC206 and CC354 included human and cattle isolates or human and chicken isolates, respectively. At the fine scale, in three cases human isolates were linked to chicken sources sharing the STs 19 and 2863, and in six cases the human isolates were linked to raw milk sources sharing STs 122 and 21. MLST revealed the presence of 18 STs that were not attributable to any of the sources we investigated, including three new STs (STs 7407, 7418, and 6788), probably suggesting multiple foci of infections (**Figure 2**, **Table 1**).

Overall, the 31 human samples were assigned to 22 STs belonging to 11 CCs; among them, two clonal complexes have not been assigned to the international database yet. By means of the two techniques employed here and as shown in **Figure 1**, only 3 out of 31 cases were truly traceable back to the two cases linked to raw milk from vending machines, to chicken meat and one human case linked to raw milk.

<sup>1</sup>http://pubmlst.org/campylobacter/

<sup>2</sup>http://pubmlst.org/

<sup>3</sup>http://darwin.phyloviz.net/ComparingPartitionsv2/index.php?link=Tool

Frontiers in Microbiology | www.frontiersin.org June 2016 | Volume 7 | Article 887 |

from different sources.

The Simpson's index and the confidence intervals for PFGE (0.949, 95% CI = 0.920–0.979) were higher than for MLST (0.897, 95% CI = 0.834–0.962), p = 0.044. PFGE is a more discriminatory typing method than MLST; however, when comparing the 95% confidence intervals, it is noteworthy that they overlap. Therefore, we cannot exclude the hypothesis that both methods have similar discriminatory power (at the 95% confidence level).

The concordance of the methods based on the calculation of the adjusted Rand coefficient and 95% CI shows a concordance between MLST and PFGE of 0.32 (95% CI = 0.209–0.437). The concordance based on the Wallace coefficient was also higher for the PFGE–MLST combination (Wallace = 0.505, 95% CI = 0.290–0.714), with respect to the MLST–PFGE combination (Wallace = 0.236, 95% CI = 0.120–0.343), p = 0.019.

The attribution analysis revealed that 69.75% of the human cases were predominantly related to chicken (**Figure 3**). Interestingly, only two human cases belonging to CC1332 were predominantly related to wild bird (62.42%, **Figure 3**). In the asymmetric island model, most human cases (69.7%) were attributed to the chicken reservoir (95% CI 47.6–87.5%). Cattle were the second most probable source reservoir (8.2%, 95% CI 0–25.6%) followed by wild birds (7.3% CI 0–19.9%), the environment (6.2% CI 0–22%), small ruminants (5.3% CI 0– 17.8%) and pork (2.9% CI 0–10.7%). Although there was a large credibility interval for the chicken source, the model gave the most probable source at a probability of 99.8%.

For each patient, the model restituted the posterior probability for the source of infection as shown in **Figure 4**. One dominant color confirmed that chicken is the likely source for the cases under consideration. It is worth noting that, for the generalist CC21 and CC48 regularly isolated from multiple reservoirs, the model showed a minor probability of chicken in favor of ruminants (cattle and small ruminants). CC1332 featured two clinical isolates and resulted with wild birds as the most probable source of infection. Overall, the 31 human cases investigated were probably infected in 29 instances by chicken and in the last two cases by wild birds. A minor probable source was confirmed for pork.

#### DISCUSSION

The percentage of human cases with campylobacteriosis ascribed to eating or handling raw poultry differs between countries and studies. Estimates of cases of a food-borne origin range from 30 to 58% (Mullner et al., 2009), and up to 80% can be attributed to the chicken sources as a whole (EFSA, 2011; Mughini Gras et al., 2012). Furthermore, the sources of the notified cases have not been able to be determined, indicating that the real number of food-borne cases is unknown (Meinersmann et al., 2005; EFSA, 2013). It is well known that the presence of C. jejuni in bovine raw milk samples and fresh dairy products is increasing (Fernandes et al., 2015). Recent data show that the consumption of undercooked chicken meat and raw milk are the most important sources of human campylobacteriosis in Europe (EFSA, 2014), accounting for hundreds of cases.



<sup>a</sup>PFGE profiles for the restriction enzyme SmaI. Bold font: the two identified clusters shared among humans and cattle or humans and chicken.

§ New allelic combination.

‡No clonal complex applicable.

Outbreaks associated with the consumption of raw or pasteurized milk contaminated with Campylobacter were recently recorded in Italy (Giacometti et al., 2012). A quantitative risk assessment focused on one region of Northern Italy evaluated that 1–2% of the population were consumers of raw milk from vending machines and only 57% of users boiled the raw milk before consumption; therefore, the estimated proportion of the people consuming unboiled raw milk is 0.5–0.9% (4548–9096 people; Giacometti et al., 2012).

The present study aimed to investigate an outbreak of campylobacteriosis among humans and to define the infection sources. An active surveillance was undertaken that sampled raw milk from vending machines and chicken meat from a retail market. The PFGE and MLST results suggested that the human cases we studied were not caused by a single-point source, but were likely triggered by multiple foci of infection instead. The evidence supporting this conclusion was the presence of 26 different pulsotypes discovered by PFGE and 22 STs for MLST. The MLST results found nine clinical cases with similar profiles for the raw milk and chicken samples. In particular, ST21 and ST122 were shared among humans and cattle, while ST19 and ST 354 were shared between humans and chickens.

Calculation of the two coefficients (adjusted Rand and Wallace coefficients) and their confidence intervals permits quantification of the congruence between the two different method results (Wallace, 1983; Pinto et al., 2008). In our study, the adjusted Rand coefficient gave low congruence between the different methods. The PFGE analysis showed that isolates with the same MLST profile were partially congruent and sometimes grouped inside different clusters, as was also revealed by the adjusted Rand coefficient between MLST and PFGE of 0.32 (95% CI = 0.209– 0.437). However, the Wallace coefficient for PFGE and MLST gave a value of 0.5, which is higher than the 0.2 value from the MLST–PFGE combination, suggesting that PFGE performs better in an outbreak scenario.

Nevertheless, only three cases featured in VIII and XVII groups clustered together with chicken and raw milk samples, a finding in agreement with the MLST results. In contrast, ST

122 and ST 354 did not cluster together in the PFGE profiles, suggesting a possible diverse origin for them. In this study, the clustering concordance of the two molecular techniques is not similar, but we propose that complementation of the results could lead to a higher level of isolate discrimination.

The data suggest that genomic-based techniques like MLST and PFGE can provide useful insight for outbreak investigations. PFGE has been standardized for Campylobacter spp., and has been used in localized outbreak investigations (Fitzgerald et al., 2001), while MLST is useful for investigating the DNA sequence

Frontiers in Microbiology | www.frontiersin.org June 2016 | Volume 7 | Article 887 |

diversity via the index of variations in housekeeping genes (Dingle et al., 2001).

Recently, MLST has commonly been employed to increase the analysis resolution of outbreak-associated isolates, leading to quicker and much more accurate source identification, as well as investigating different epidemiological hypotheses.

Our findings revealed three human cases that had similar genetic profiles to chicken and cattle isolates. However, in this study, there were no clues for tracing the remaining isolates investigated and we did not retrieve similar profiles from the source strains isolated during the active surveillance. Therefore, to try to understand the source of infection, we applied a genetic population approach using MLST data from our database and the data from other studies conducted in Europe.

From our analysis, chicken was estimated to be the most dominant source of infection for the human isolates we studied. Using asymmetric island model, the MLST profiles associated with human disease were most similar to those from a chicken source in 69.75% of the cases of C. jejuni infection. Ruminants (cattle and small ruminants) contributed far fewer cases of C. jejuni infection (13.6%) and the contribution from environment, wild bird, and pig sources was very low (6.28, 7.37, and 2.98%, respectively). This finding is in line with many other studies performed in industrialized countries, even if divergence in the percentage of cases attributable to the chicken source varied among these studies (Wilson et al., 2008; Mullner et al., 2009; Sheppard et al., 2009). Our results should indicate the importance of chicken as the main reservoir of human campylobacteriosis and lend to the suggestion that this disease could be greatly reduced by focusing interventions on chickens. The cases studied are not representative of the Italian epidemiological context but seem to be congruent with other European scenarios.

Notably, inspecting the three cases that were traced back to a potential source revealed that the model estimated chicken as the most probable source of infection for all the cases as well. Contrastingly, our investigation found that one case featured in

#### REFERENCES


PFGE cluster VIII and ST 21 was attributable to raw milk instead of chicken. In fact, CC21 is a multi-host lineage shared among different sources, and this can lead to an incorrect assignment in the genetic model. The model calculates the likely source of the infection but cannot rule out completely the minor reservoir as the actual source. In this case, the back-tracing analysis using PFGE and MLST is likely to be the best approach for drawing the correct reconstruction of the food-borne illness. In this scenario, whole-genome*ı* sequencing (WGS) would probably provide the best fit for the epidemiological analysis, but this approach is not yet ready to be applied routinely because of the lack of standardization of the analysis of the WGS data in the context of the molecular epidemiology of food-borne pathogens.

#### CONCLUSION

The number of clinical isolates, albeit limited to only 56 samples from a restricted area and in a short time window, confirms the high genetic diversity present in Campylobacter. Genetic assessment of Campylobacter spp. is fundamental to our understanding of its epidemiology. Additional analyses of isolates from various sources will consent major progresses in the knowledge of the epidemiology and population structure of C. jejuni in Italy. This study confirms the interest of molecular epidemiology as a powerful support for epidemiological investigations.

#### AUTHOR CONTRIBUTIONS

EG has contributed to the conception and design of the work and wrote the paper. AA performed experiments. FM performed experiments, analyzed data and wrote the paper. GG performed the source attribution analysis, analyzed data and wrote the paper. LC performed the source attribution analysis. GD helped draft the manuscript. WV and LD contributed strains.

meat in two Italian regions. Int. J. Food Microbiol. 128, 274–287. doi: 10.1016/j.ijfoodmicro.2008.08.021



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Di Giannatale, Garofolo, Alessiani, Di Donato, Candeloro, Vencia, Decastelli and Marotta. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Prevalence of *Staphylococcus aureus* and Methicillin-Resistant *Staphylococcus aureus* in Retail Ready-to-Eat Foods in China

Xiaojuan Yang † , Jumei Zhang † , Shubo Yu † , Qingping Wu\*, Weipeng Guo, Jiahui Huang and Shuzhen Cai

*State Key Laboratory of Applied Microbiology Southern China, Guangdong Provincial Key Laboratory of Microbial Culture Collection and Application, Guangdong Open Laboratory of Applied Microbiology, Guangdong Institute of Microbiology, Guangzhou, China*

#### *Edited by:*

*Pierina Visciano, University of Teramo, Italy*

# *Reviewed by:*

*Jyoti Prakash Tamang, Sikkim University, India Adrian Canizalez-Roman, Autonomous University of Sinaloa, Mexico*

> *\*Correspondence: Qingping Wu wuqp203@163.com*

*† These authors have contributed equally to this work.*

#### *Specialty section:*

*This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology*

*Received: 30 March 2016 Accepted: 13 May 2016 Published: 07 June 2016*

#### *Citation:*

*Yang X, Zhang J, Yu S, Wu Q, Guo W, Huang J and Cai S (2016) Prevalence of Staphylococcus aureus and Methicillin-Resistant Staphylococcus aureus in Retail Ready-to-Eat Foods in China. Front. Microbiol. 7:816. doi: 10.3389/fmicb.2016.00816* *Staphylococcus aureus*, particularly methicillin-resistant *S.*aureus (MRSA), is a life-threatening pathogen in humans, and its presence in food is a public health concern. MRSA has been identified in foods in China, but little information is available regarding MRSA in ready-to-eat (RTE) foods. We aimed to investigate the prevalence of *S. aureus* and MRSA in Chinese retail RTE foods. All isolated *S. aureus* were tested for antimicrobial susceptibility, and MRSA isolates were further characterized by multilocus sequence typing (MLST) and staphylococcal cassette chromosome *mec* (SCC*mec*) typing. Of the 550 RTE foods collected from 2011 to 2014, 69 (12.5%) were positive for *S. aureus*. Contamination levels were mostly in the range of 0.3–10 most probable number (MPN)/g, with five samples exceeding 10 MPN/g. Of the 69 *S. aureus* isolates, seven were identified as MRSA by cefoxitin disc diffusion test. Six isolates were *mecA*-positive, while no *mecC-*positive isolates were identified. In total, 75.8% (47/62) of the methicillin-susceptible *S. aureus* isolates and all of the MRSA isolates were resistant to three or more antibiotics. Amongst the MRSA isolates, four were identified as community-acquired strains (ST59-MRSA-IVa (*n* = 2), ST338-MRSA-V, ST1-MRSA-V), while one was a livestock-associated strain (ST9, harboring an unreported SCC*mec* type 2C2). One novel sequence type was identified (ST3239), the SCC*mec* gene of which could not be typed. Overall, our findings showed that Chinese retail RTE foods are likely vehicles for transmission of multidrug-resistant *S. aureus* and MRSA lineages. This is a serious public health risk and highlights the need to implement good hygiene practices.

Keywords: *Staphylococcus aureus*, methicillin-resistant, ready-to-eat foods, prevalence, antibiotic resistance

#### INTRODUCTION

Staphylococcus aureus is an important cause of food poisoning worldwide. It is estimated that ∼20–25% of foodborne bacterial outbreaks are caused by S. aureus in China (Wang et al., 2014). In addition, it is the leading cause of infection in both healthcare facilities and communities, causing illnesses ranging from mild skin and soft tissue infections to life threatening diseases such as septicemia, necrotizing fasciitis, endocarditis, and necrotizing pneumonia (Lowy, 1998; Chen and Huang, 2014; Rodríguez-Lázaro et al., 2015). Furthermore, the increasing antimicrobial resistance rates of this bacterium pose a serious threat to public health. Methicillin-resistant S. aureus (MRSA) strains exhibit resistance to all β-lactam antibiotics through acquisition of the mobile staphylococcal cassette chromosome mec (SCCmec), which carries the antibiotic-resistant gene mecA. Together, healthcareacquired MRSA (HA-MRSA), community-acquired MRSA (CA-MRSA), and livestock-associated MRSA (LA-MRSA) strains constitute a major health concern.

Epidemiological studies have revealed differences between HA-MRSA, CA-MRSA, and LA-MRSA strains, including antimicrobial resistance profiles, SCCmec types, and clonal complexes (CCs) identified by multilocus sequence typing (MLST). HA-MRSA isolates typically carry relatively large SCCmec elements (types II or III), and are resistant to many classes of antimicrobials, including β-lactams (Yamamoto et al., 2013). CA-MRSA isolates usually harbor smaller SCCmec elements (types IV or V) and are only resistant to β-lactam antibiotics (Rodríguez-Lázaro et al., 2015). Interestingly, specific MRSA clones have spread across different geographical regions worldwide. The New York/Japan (ST5/SCCmecII), Brazilian/Hungarian (ST239/SCCmecIII), EMRSA-15 (ST22/SCCmecIV), and EMRSA-16 (ST36/SCCmecII) clones are pandemic HA-MRSA lineages (Yamamoto et al., 2013), while the Taiwan (ST59/SCCmecIV or V), USA300 (ST8/SCCmecIV), European (ST80/SCCmecIV), and USA400 (ST1) clones are always associated with community-acquired infections (Yamamoto et al., 2013). In China, ST239/SCCmecIII and ST5/SCCmecII are predominant HA-MRSA clones, while ST59/SCCmecIV or V is the most prevalent CA-MRSA clone (Chuang and Huang, 2013; Chen and Huang, 2014).

The first reports of MRSA infections in animals appeared in the 1970s and, on the basis of their putative source, these MRSA strains are referred to as LA-MRSA (Petinaki and Spiliopoulou, 2012). ST398 was the first detected and most widespread LA-MRSA sequence type (ST). The isolates belonging to this clonal lineage are not typeable by pulsed field gel electrophoresis (PFGE) using SmaI, and often exhibit co-resistance to many non-β-lactam antimicrobials, including those commonly used in animal production.

The emergence of LA-MRSA has led to concerns about its transmission via the food chain. In recent years, LA-MRSA has frequently been detected in food-producing animals (Wagenaar et al., 2009; Petinaki and Spiliopoulou, 2012; Visciano et al., 2014), meat (Rodríguez-Lázaro et al., 2015), milk and dairy products (Song et al., 2015), fish (Hammad et al., 2012), and ready-to-eat (RTE) food products (Hammad et al., 2012; Wang et al., 2014), causing a significant public health concern.

RTE foods that are consumed without further treatment, such as cooked meat and poultry, cold vegetable dishes, cold noodles, and fried rice, are popular in China. It has been reported that these food products are associated with the introduction of microbiological hazards, including Listeria monocytogenes (Chen M. T. et al., 2014), Cronobacter (Xu et al., 2015), and Salmonella (Yang et al., 2016). However, fewer data are available regarding the prevalence of S. aureus and MRSA in these foods. The aim of this study was to determine the prevalence, antibiotic resistance, and molecular characteristics of S. aureus and MRSA isolated from RTE food samples collected in China.

# MATERIALS AND METHODS

# Sample Collection

From December 2011 to May 2014, a total of 550 samples were collected from RTE foods from retail markets. Samples included cooked pork (119), cooked chicken (153), cooked duck (127), cold vegetable dishes in sauce (53), cold noodles (52), and fried rice/sushi (46). The retail markets were located in 24 cities across most of the provincial capitals of China (**Figure 1**). Each sample was weighed, labeled, placed in a separate sterile bag, and then immediately transported to the laboratory in an ice box.

## Detection and Enumeration of *S. aureus*

The prevalence and bacterial load of S. aureus in the samples was determined using the most probable number (MPN) method according to National Food Safety Standards of China document GB 4789.10-2010. A 25-g sample was randomly collected from each RTE food product and placed into a sterile glass flask containing 225 mL of saline solution (Huankai, Guangzhou, China). Following homogenization, trypticase soy broth (Huankai) supplemented with 10% NaCl was inoculated in triplicate with 1-ml aliquots of decimal dilutions of each sample. Broths were incubated at 37◦C for 48 h. Loopfuls of the resulting cultures were streaked onto chromogenic S. aureus agar plates (Huankai), then incubated at 37◦C for 24 h. Putative S. aureus isolates were tested for coagulase activity, and were further confirmed using API STAPH test strips (bioMerieux, Marcyl ′Etoile, France). The MPN value was determined on the basis of the number of positive tube(s) in each of the three sets using the MPN table.

#### Antimicrobial Susceptibility Testing

All isolates were evaluated for antimicrobial resistance using the Kirby–Bauer disk diffusion method according to the (Clinical Laboratory Standards Institute previously National Committee on Clinical Laboratory Standards., 2014). Susceptibility to the following 16 antimicrobial agents was tested: ampicillin (AMP), cephalothin (KF), cefoxitin (FOX), penicillin G (P), chloramphenicol (C), tetracycline (TE), ciprofloxacin (CIP), amikacin (AK), gentamicin (CN), kanamycin (K), trimethoprim-sulfamethoxazole (SXT), erythromycin (E), clindamycin (DA), rifampicin (RD), linezolid (LZD), and quinupristin/dalfopristin (QD) (Oxoid, Basingstoke, UK). The isolates were also examined using a microdilution test according to the CLSI method for vancomycin minimum inhibitory concentrations (MICs) (Clinical Laboratory Standards Institute previously National Committee on Clinical Laboratory Standards., 2014).

#### Detection of mecA and mecC

mecA, which has been shown to confer methicillin resistance on S. aureus, and mecC, a divergent mecA homolog (also called mecALGA251), were detected by PCR using primers

described previously (Pérez-Roth et al., 2001; Stegger et al., 2012).

# MLST and SCCmec-Typing of the mecA-Positive MRSA Isolates

The mecA-positive MRSA isolates were characterized by MLST analysis. MLST was carried out using previously reported primers specific for seven housekeeping genes (arcC, aroE, glpF, gmk, pta, tpi, and yqiL), and the sequence type (ST) was assigned according to the MLST database (http://www.mlst.net/).

The SCCmec type of mecA-positive strains was determined using a multiplex PCR method as previously described (Kondo et al., 2007). Strains with unanticipated fragments or completely lacking fragments were defined as non-typeable (NT).

# RESULTS

# Prevalence and Load of *S. aureus* in Retail RTE Food

Of the 550 retail RTE food samples, 69 (12.5%) were positive for S. aureus according to the MPN method. This included 15 (12.6%) of the 119 cooked pork samples, 15 (9.8%) of the 153 cooked chicken samples, 17 (13.4%) of the 127 cooked duck samples, six (11.3%) of the 53 samples from cold vegetable dishes in sauce, 10 (19.2%) of the 52 cold noodle samples, and six (13.0%) of the 46 fried rice/sushi samples. Overall, 50.7% (35/69) of positive samples had a bacterial load of <1 MPN/g, and 42.0% (29/69) reached 1 MPN/g. Five samples exceeded 10 MPN/g (**Table 1**).

# Antimicrobial Susceptibility of the *S. aureus* Isolates

Of the 69 S. aureus isolates recovered, seven were confirmed as MRSA by cefoxitin disc diffusion test. Six of the isolates were mecA-positive, and none tested positive for mecC by PCR. All of the MRSA isolates were resistant to ampicillin and penicillin G, and 66.7% were resistant to clindamycin, erythromycin, tetracycline, and kanamycin. All MRSA isolates were susceptible to cephalothin, amikacin, linezolid, quinupristin/dalfopristin, and vancomycin (MICs <1 µg/mL). The antimicrobial resistance profiles of the MRSA isolates are shown in **Table 2**.

Among the 62 methicillin-susceptible S. aureus (MSSA) isolates, all were resistant to at least one antimicrobial agent, and 47 isolates (75.8%) were resistant to more than three antimicrobials. The highest levels of resistance were observed for ampicillin (98.4%), penicillin G (98.4%), and tetracycline (43.5%). The antimicrobial resistance profiles of the MSSA strains are shown in **Table 3**.

# MLST and SCCmec-Typing of mecA-Positive MRSA Isolates

In the present study, seven phenotypically MRSA isolates were identified; however, only six isolates were mecA-positive by PCR. Among the six mecA-positive MRSA isolates, two were recovered from different cooked duck samples in the same city (Nanning). These were both identified as ST59 and harbored SCCmec type IVa (2B). Amongst the remaining mecA-positive isolates, one belonged to ST338 and harbored SCCmec type


TABLE 1 | Prevalence and load of *Staphylococcus aureus* in retail RTE foods in China.

TABLE 2 | Prevalence and characteristics of methicillin-resistant *Staphylococcus aureus* isolates.


\**Novel sequence type. NT, not typeable. /, no amplification product. , undescribed SCCmec type.*

*AMP, ampicillin; E, erythromycin; DA, clindamycin; FOX, cefoxitin; RD, rifampicin; P, penicillin G; C, chloramphenicol; TE, tetracycline; CIP, ciprofloxacin; CN, gentamicin; K, kanamycin; SXT, trimethoprim-sulfamethoxazole.*

V (5C2), one belonged to ST1 and harbored SCCmec type V (5C2), and one belonged to ST9 and harbored an unreported SCCmec type (2C2). One novel ST (characterized as ST3239) was identified for a MRSA isolate recovered from sausage. SCCmec typing of this isolate was not possible as multiplex PCR-1, which provides the ccr gene complex type, produced a 937-bp DNA fragment, consistent with type 2, but multiplex PCR-2, which types the mecA gene complex class, did not amplify.

#### DISCUSSION

In recent years, the consumption of RTE foods in China has increased markedly. However, reports on S. aureus and MRSA contamination in these foods are scarce. The prevalence of S. aureus in food samples in the current study was 12.5% (69/550), which is lower than that observed (25.1%) in RTE foods (cooked meat, vegetable salads, boiled peanuts, cold noodles, and dried tofu) from Shaanxi Province, China (Wang et al., 2014). However, the prevalence of MRSA in our study (1.3%) was higher than that (0.6%) of the study by Wang et al. (2014). As our results were obtained from a large number of samples from most regions in China, the data are more comprehensive and systematic, and more representative of China as a whole.

The prevalence of S. aureus and MRSA in RTE foods in the current study differs from that reported in raw food products in China (Wang et al., 2014; Song et al., 2015), which generally have relatively higher levels of S. aureus (19.3–24.2%) and MRSA (1.7– 5.6%). In addition, antimicrobial susceptibility testing revealed that not only all of the MRSA isolates, but a significant number of the MSSA isolates (47/62; 75.8%), were resistant to three or more antibiotics. The high prevalence of penicillin, ampicillin, tetracycline, and erythromycin resistance observed in our study is similar to results from previous studies in raw food products in China (Wang et al., 2014) and other countries (Rodríguez-Lázaro et al., 2015). However, the findings of our study are even more serious in terms of public health because RTE foods are consumed without further cooking, which would eliminate or reduce the microbial load. Consequently, the incidence of S. aureus and MRSA in RTE foods, along with the spread of antibiotic resistant strains, represents a potential health hazard to humans.

All MRSA isolates in the current study displayed resistance to three β-lactams, as well as showing high levels of resistance to clindamycin, erythromycin, tetracycline, and kanamycin. Fortunately, all of these isolates were susceptible to linezolid, quinupristin/dalfopristin, and vancomycin (MICs <1 µg/mL), which are the few remaining effective agents for treatment of MRSA infections.


TABLE 3 | Antimicrobial resistance susceptibility profiles of methicillin-susceptible *Staphylococcus aureus* isolates.

Our study provides evidence for the existence of two different lineages of MRSA in RTE foods in China: CA-MRSA and LA-MRSA. ST59 (2/6, 33.3%) was the predominant ST in our study, and its single locus variant ST338 was also detected. ST59 and ST338 belong to CC59, which is the dominant CA-MRSA CC in Asia and is a significant cause of human infection attributable to MRSA (Li et al., 2013; Chen and Huang, 2014). CC59 MRSA isolates also prevail in China, and are also the dominant clone in healthy carriers (Du et al., 2011) and in patients with community-acquired infections (Li et al., 2013). Amongst CC59 strains, two dominant SCCmec types (IV and V) and different antimicrobial resistance profiles have been described. A previous study showed that ST59 and ST338 were the first and second most common STs in Chinese pediatric community-acquired pneumonia, especially the dominant ST59-MRSA-IVa and ST338-MRSA-V clones (Geng et al., 2010). ST59-MRSA-IV and ST338-MRSA-V have also been associated with cases of bacteremia in China (He et al., 2013).

ST1 isolates are generally considered to be CA-MRSA (Porrero et al., 2013), although they have also been found in animals (Porrero et al., 2013). ST1 MRSA has most commonly been reported in the United States and Canada (Chuang and Huang, 2013), and has only been found sporadically in China. However, multiple ST1-MRSA-IV clones were reported to cause community-acquired pneumonia and skin/soft-tissue infections in Chinese children (Geng et al., 2010). ST1 has also been associated with staphylococcal food poisoning in Korea (Cha et al., 2006) and China (Yan et al., 2012). Some studies have shown that rates of resistance to non-β-lactam agents amongst ST1 MRSA isolates vary between countries and clones. ST1 MRSA strains in the United States usually show resistance to several non-β-lactam agents, including erythromycin and clindamycin, while MRSA strains in Australia are often uniformly susceptible to almost all non-β-lactams (Chen J. et al., 2014). The ST1 MRSA isolate in our study was also susceptible to all non-β-lactams, indicating that it might differ from that prevailing in the United States.

One ST9 LA-MRSA isolate was also identified in the current study. ST9 is the most prevalent LA-MRSA clone in most Asian countries, including China (Cui et al., 2009; Wagenaar et al., 2009; Petinaki and Spiliopoulou, 2012), despite ST398 being the most widespread ST in the rest of the world (Weese and Van Duijkeren, 2010; Petinaki and Spiliopoulou, 2012). ST9 MRSA is the predominant clone in food animals and animal-derived products (pork, chicken, and raw milk) in China (Cui et al., 2009; Wagenaar et al., 2009; Boost et al., 2013), and has also been found in farm workers (Cui et al., 2009) and associated with human infections (Liu et al., 2009; Yu et al., 2014). A previous study described ST9-MRSA-SCCmecV/NT isolates from patients with severe clinical illness (Yu et al., 2014). However, these patients were not livestock handlers and did not keep close contact with livestock (Yu et al., 2014), indicating that MRSA ST9 can pose a threat to humans through the food chain. ST9 MRSA strains with SCCmec types II (Wang et al., 2014), III (Cui et al., 2009), IVb (Boost et al., 2013; Wang et al., 2014; Yan et al., 2014), and V (Yu et al., 2014) have also been reported in China. However, the ST9 MRSA isolate recovered in our study showed considerable heterogeneity, harboring a previously undescribed SCCmec type with a type 2 ccr (A2B2) and a class C2 mec gene complex (2C2). These results suggest that ST9 MRSA acquired novel genomic islands during evolution from ST9 MSSA. Notably, the ST9 MRSA isolate in the current study was multidrug resistant, with resistance to β-lactams as well as eight other antibiotics widely used in chemotherapy. This finding suggests that the isolate originated from livestock.

Human infections caused by foodborne MRSA strains have been reported (Jones et al., 2002). Therefore, the potential role of food in the dissemination of successful MRSA lineages cannot be ignored. While ST59 and ST9 MRSA strains have been detected in food in China (Wang et al., 2014; Song et al., 2015), ST338 and ST1 MRSA isolates from food have never been reported. However, the previously reported ST59 MRSA strains were ST59-MRSA-II and ST59-MRSA-NT clones, not the more virulent ST59-MRSA-IVa clone identified in our study, which is associated with severe infections. Our results suggested that the presence of MRSA in RTE foods is the result of human contamination through poor personal hygiene, or through crosscontamination of carcasses during food processing. On the other hand, the MRSA clonal complexes identified in the present study are prevalent amongst clinical isolates associated with severe infection in China, suggesting food as a potential environmental source of S. aureus isolates with significant clinical relevance.

#### CONCLUSION

To our knowledge, this is the first comprehensive study of the prevalence of S. aureus and MRSA in retail RTE foods from diverse regions of China. The present study revealed a relatively high prevalence of S. aureus and MRSA, and high rates of antimicrobial resistance amongst the isolates. Epidemic CA-MRSA and LA-MRSA clones associated with severe infection were identified amongst the isolates. Our data confirm the potential role of RTE foods in the dissemination of multidrugresistant S. aureusstrains and successful MRSA lineages in China,

#### REFERENCES


and highlights the health risks for consumers. Effective measures should be taken to ensure the safety of our food products.

### AUTHOR CONTRIBUTIONS

Conceived and designed the experiments: XY, QW, and JZ. Performed the experiments: XY and SY. Analyzed the data: XY and JZ. Contributed reagents/materials/analysis tools: XY, JZ, WG, JH, and SC.

# FUNDING

This work was supported by the Science and Technology Projects of Guangdong (2014B050504007, 2013B020312001).

toxin genes and genotypes. Int. J. Antimicrob. Agents 42, 211–219. doi: 10.1016/j.ijantimicag.2013.04.031


China. Appl. Environ. Microbiol. 78, 6637–6642. doi: 10.1128/AEM. 01165-12


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Yang, Zhang, Yu, Wu, Guo, Huang and Cai. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Design and Performance Testing of a DNA Extraction Assay for Sensitive and Reliable Quantification of Acetic Acid Bacteria Directly in Red Wine Using Real Time PCR

Cédric Longin\*, Michèle Guilloux-Benatier and Hervé Alexandre

Equipe VAlMiS (Vin, Aliment, Microbiologie, Stress), Institut Universitaire de la Vigne et du Vin Jules Guyot, UMR Procédés Alimentaires et Microbiologiques, AgroSup Dijon – Université de Bourgogne, Dijon, France

#### *Edited by:*

Rosanna Tofalo, University of Teramo, Italy

#### *Reviewed by:*

Luca Settanni, Università degli Studi di Palermo, Italy Rolf Dieter Joerger, University of Delaware, USA

> *\*Correspondence:* Cédric Longin cedric.longin@u-bourgogne.fr

#### *Specialty section:*

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

*Received:* 16 March 2016 *Accepted:* 17 May 2016 *Published:* 01 June 2016

#### *Citation:*

Longin C, Guilloux-Benatier M and Alexandre H (2016) Design and Performance Testing of a DNA Extraction Assay for Sensitive and Reliable Quantification of Acetic Acid Bacteria Directly in Red Wine Using Real Time PCR. Front. Microbiol. 7:831. doi: 10.3389/fmicb.2016.00831 Although strategies exist to prevent AAB contamination, the increased interest for wines with low sulfite addition leads to greater AAB spoilage. Hence, there is a real need for a rapid, specific, sensitive, and reliable method for detecting these spoilage bacteria. All these requirements are met by real time Polymerase Chain Reaction (or quantitative PCR; qPCR). Here, we compare existing methods of isolating DNA and their adaptation to a red wine matrix. Two different protocols for isolating DNA and three PCR mix compositions were tested to select the best method. The addition of insoluble polyvinylpolypyrrolidone (PVPP) at 1% (v/v) during DNA extraction using a protocol succeeded in eliminating PCR inhibitors from red wine. We developed a bacterial internal control which was efficient in avoiding false negative results due to decreases in the efficiency of DNA isolation and/or amplification. The specificity, linearity, repeatability, and reproducibility of the method were evaluated. A standard curve was established for the enumeration of AAB inoculated into red wines. The limit of quantification in red wine was 3.7 log AAB/mL and about 2.8 log AAB/mL when the volume of the samples was increased from 1 to 10 mL. Thus, the DNA extraction method developed in this paper allows sensitive and reliable AAB quantification without underestimation thanks to the presence of an internal control. Moreover, monitoring of both the AAB population and the amount of acetic acid in ethanol medium and red wine highlighted that a minimum about 6.0 log cells/mL of AAB is needed to significantly increase the production of acetic acid leading to spoilage.

Keywords: acetic acid bacteria, red wine, microbiological internal control, DNA extraction, real time PCR

# INTRODUCTION

Acetic Acid Bacteria (AAB) species typically associated with grapes and must is Gluconobacter oxydans (G. oxydans) which prefers a sugar rich environment (Joyeux et al., 1984; Bartowsky and Henschke, 2008). AAB associated with wine are Acetobacter aceti (A. aceti) and Acetobacter pasteurianus (A. pasteurianus) which prefer ethanol as a carbon source, as does Gluconacetobacter liquefaciens (Ga. liquefaciens) (Joyeux et al., 1984; Drysdale and Fleet, 1985; Yamada et al., 1997). When these AAB are present during winemaking, aging or wine storage, they metabolize ethanol to acetaldehyde by alcohol dehydrogenase and then produce acetic acid by acetaldehyde dehydrogenase. Acetic acid is the main constituent of wine volatile acidity (Bartowsky and Henschke, 2008) and considered to be undesirable in dry wine at concentrations exceeding 0.4–0.5 g/L depending on wine type (Davis et al., 1985; Eglinton and Henschke, 1999a,b). The European regulation (CE 1308/2013) has set out limits for sale at 1.20 and 1.08 g/L acetic acid for red wines and white/rosé wines, respectively. Taking into account the ability of AAB to convert ethanol into acetic acid, these bacteria are considered as spoilage bacteria in the wine industry. Plate counting is typically used in order to quantify AAB in wine. However, culturing and enumerating AAB is challenging despite the availability of various growth media. Many studies have reported that plate counting is not appropriate for estimating AAB populations in stressful environments like wine (Sievers et al., 1992; Sokollek et al., 1998; Millet and Lonvaud-Funel, 2000; Bartowsky et al., 2003; Trcek, 2005). Indeed, this technique often underestimates AAB populations (Bartowsky and Henschke, 2008). According to Millet and Lonvaud-Funel (2000), the difficulties of isolating AAB may be due in part to the existence of a Viable But Non Culturable (VBNC) state which may result from anaerobic conditions in wine (Du Toit et al., 2005). Many techniques can be used for AAB detection like nested PCR (González et al., 2006), AAB gene (adhA) PCR (Trcek, 2005), PCR-Restriction Fragment Length Polymorphism (Ruiz et al., 2000), and many others, however, all these techniques are culture dependent and the VBNC state may lead to underestimating the AAB population. Independent culture quantification techniques using Denaturing Gradient Gel Electrophoresis (De Vero et al., 2006), Temperature Gradient Gel Electrophoresis (Ilabaca et al., 2008), or epifluorescence have been reported (Mesa et al., 2003; Baena-Ruano et al., 2006). These latter authors quantified AAB in vinegar fermentation using viability (i.e., measurement of cell membrane permeability) and vitality (i.e., measurement of cell enzymatic activity) dyes. This technique has a high detection limit and appreciation of fluorescence is operator dependent. Thus, qPCR techniques have been developed (Torija et al., 2010; Valera et al., 2013) to quantify AAB. The qPCR method presented by González et al. (2006) did not use any internal control and was validated with red wine inoculated with a known amount of AAB, but without any growth. However, AAB growth in wine modifies DNA extraction efficiency and qPCR reliability since polyphenols seem to be adsorbed onto the cell walls (Morata et al., 2004). In addition, problems often arise with DNA amplification due to inhibitory substances such as tannins, polysaccharides and pigments (Rossen et al., 1992; Wilson, 1997). Therefore, poor DNA isolation and/or amplification efficiency, leading to falsenegative results, were observed (Tessonnière et al., 2009). The specific species quantification of AAB using TaqMan probes was also reported (Torija et al., 2010; Valera et al., 2013). However, these probes are expensive and could not be used for routine laboratory analysis to determine total AAB population. Furthermore, none of these assays used internal controls to screen for the presence of inhibitors contained in the samples, leading to unreliable quantification. The goal of the current work was to develop a qPCR technique using an internal control that provides efficient and reliable quantification of the AAB naturally present in red wine. Different DNA extraction protocols were compared in order to remove wine inhibitors and assess the sensitivity, specificity, and reproducibility of the method.

# MATERIALS AND METHODS

#### Microorganism Strains

Four different species of AAB were used: Acetobacter aceti DSM 3508 (DSMZ-Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH, Braunschweig, Germany), Acetobacter pasteurianus CECT 7582 (Colección española de cultivos tipo, Universitat de València, Edificio de Investigación, Burjassot, Spain); Gluconobacter oxydans DSM 7145; Gluconacetobacter liquefaciens CIP 103109 (Collection of Institut Pasteur, Biological Resource Center of Institut Pasteur (CRBIP), Paris, France). Oenococcus oeni sabo11 (a biotechnological strain isolated from a South African wine) was also used as the majority LAB present in wine. AAB were adapted to ethanol by growing them in mannitol medium supplemented with ethanol [2.5% (w/v) mannitol; 0.5% (w/v) yeast extract; 0.3% (w/v) peptone, 5% (v/v) ethanol]. O. oeni was adapted in FT80 medium (Cavin et al., 1988). Escherichia coli K12 ER2738 (available from New England Biolabs) was cultivated in LB medium (1% (w/v) peptone, 0.5% (w/v) yeast extract, 1% (w/v) NaCl). Yeasts, namely Zygosaccharomyces bailii MUCL 27812 (Mycothèque, de l'Université Catholique de Louvain, Louvain-la-Neuve, Belgium), Candida vini MUCL 27720, Pichia membranifaciens PMmb2000, P. fermentans PFmb2005, and Saccharomyces cerevisiae FERMOL-PB2023, were grown in modified YPD medium (2% (w/v) glucose; 0.5% (w/v) yeasts extract, 1% (w/v) peptone) with chloramphenicol at 0.02% (w/v) added after sterilization.

#### Growth Conditions

For the artificial contaminations, adapted AAB were inoculated into mannitol medium supplemented with 10% (v/v) ethanol (ethanol medium), white and red wines. Growth was ensured at 28◦C. Red wine made from Pinot Noir grapes was supplied by the vineyard of the University of Burgundy [pH: 3.5; 12% (v/v) alcohol] as was white wine made from Chardonnay grapes [pH: 3.5; 12% (v/v) alcohol]. The wines were filtered through a 0.2µm sterile membrane and dispensed into sterilized Erlenmeyer flasks.

#### Enumeration of Microorganisms

After AAB growth in ethanol medium and white and red wines, bacteria levels were measured by flow cytometry (FCM) in BD Accuri C6 flow cytometers with a single dye DiBAC4(3) [Bis- (1,3-Dibutylbarbituric Acid)Trimethine Oxonol] (ThermoFisher Scientific, Molecular ProbesTM, B-438) (final concentration 6µM) in PBS buffer 1X (Biosolve, 10X concentrate Molecular biology, 162323). This compound was excited by the flow cytometer laser at 488 nm and emitted fluorescence collected by the filter 530 ± 15 nm. This dye was used for counterstaining and circumventing potential culture dependent shortcomings such as the VBNC state of wine microorganisms (Millet and Lonvaud-Funel, 2000). Three enumeration repetitions (20µL at 34µL/min) were performed for each sample.

# Internal Control for DNA Isolation and Amplification

Specific EC23S primers were selected to quantify E. coli K12 ER2738 (Forward: 5 ′ -CATAAGCGTCGCTGCCG-3′ ; Reverse: 5 ′ -AAAGAAAGCGTAATAGCTCACTGGTC-3′ ) (Ludwig and Schleifer, 2000; Chern et al., 2011). A standard curve with these primers using the 23S rRNA gene sequences as target was obtained in LB medium at 10<sup>7</sup> to 10<sup>1</sup> cells/mL. 20µL of internal control at 5.10<sup>5</sup> E. coli/mL were added to each 1 mL sample prior to DNA isolation to obtain a concentration of 10<sup>4</sup> E. coli/mL in each sample.

#### DNA Isolation and Extraction

After AAB growth and the addition of the internal control, two DNA extraction methods were tested: the Ausubel (Ausubel et al., 1992) and Lipp methods (Lipp et al., 1999). The two methods use CTAB as Jara et al. (2008) recommend. The main differences are that the second method use chloroform to purify DNA and use a CTAB precipitation solution after cell lysis to eliminate the remaining polyphenols. Moreover, DNA was re-dissolved in 50µL sterile deionized water for the methods. When indicated, cells were centrifuged and the pellets were resuspended into lysis buffer supplemented with PVPP at a final concentration of 1% (w/v). Each bacterium was tested in growth medium supplemented with 10% (v/v) ethanol, and white and red wines. All the experiments were performed in triplicate and repeated three times.

#### AAB Real-Time PCR Amplification (qPCR)

AAB primers used to amplify the 16S rRNA gene were selected from Valera et al. (2015). The forward primer AAB-F (5′ - TGAGAGGATGATCAGCCACACT-3′ ) and the reverse primer AAB-R (5′ -TCACACACGCGGCATTG-3′ ) were synthesized by Eurogentec <sup>R</sup> (France). The PCR mixture was prepared in a total volume of 25µL with 100 nmol of each primer and 5µL DNA extract. The amplifications were done in triplicate on a CFX90 real-time PCR system (Bio-Rad) under the following conditions: 95◦C for 10 min, 40 denaturation cycles at 95◦C for 15 s, and 62◦C for 1 min. Then a melting curve was produced to check the presence of only one amplification fragment. To test PCR amplification quality, BSA, and PVP treatments during qPCR at 400 ng/µL and 0.5% (w/v), respectively, were performed (Tessonnière et al., 2009). A control without treatment was also performed. The PCR cycle in which fluorescence first occurred (quantification cycle: Cq) was determined automatically using Bio-Rad CFX Manager <sup>R</sup> software after setting the regression method.

### Volatile Acidity According to AAB Population

The acetic acid concentration and AAB population were monitored over time in synthetic medium containing 10% (v/v) alcohol. The amount of acetic acid was measured enzymatically using a Biosentec kit (Cat. No 021) according to the manufacturer's instructions, expressed in gram per liter. The AAB population was determined throughout their growth by plating on mannitol agar (CFU/mL). The same experiment was also performed in red wine from the vineyard of the University of Burgundy but no inoculation was performed; the AAB contamination was natural. In this experiment, the AAB population was determined by both qPCR and FCM. FCM provided the total bacteria population count. LAB enumeration was performed in FT80 Petri dishes to avoid overestimating the AAB population determined by FCM. Moreover, 15 red wines were chosen randomly from different wine regions to analyze the AAB populations (qPCR and FCM). In addition, acetic acid concentration was determined.

# RESULTS

#### qPCR Specificity

The specificity of the AAB primers was tested against a panel of microorganisms known to be naturally present in wine. In silico tests were performed on O. oeni, Pediococcus spp., Lactobacillus spp, Zygosaccharomyces bailii, Candida vini, Pichia membranifaciens, P. fermentans, and S. cerevisiae. AAB primers did not match with the main wine microorganisms. In vitro tests were performed in synthetic medium against O. oeni and a panel of yeasts (see above). The Cq-values were the same as the negative control for all populations of O. oeni and yeast, thus validating the specificity of the primers. The EC23S specific primers targeting the 23S rRNA gene from E.coli (internal) control did not amplify DNA of the wine microorganisms tested. E. coli has been chosen because this microorganism is not naturally present in wine.

#### Comparison of DNA Isolation Methods

DNA extraction for AAB quantification was performed after growth in ethanol medium, and white and red wines using either the Lipp or Ausubel methods with or without PVPP during cell lysis. In order to compare the DNA extraction methods, the C<sup>q</sup> obtained from DNA extracted by both methods with or without PVPP for the three different media were compared for each bacterium.

As shown in **Table 1A**, for ethanol medium, the Lipp method without PVPP gave significantly better results, as shown by a lower C<sup>q</sup> whatever the bacteria, except for G. oxydans, showing no significant difference in the results obtained by both methods. PVPP addition during DNA extraction using either the Lipp or Ausubel method did not improve qPCR efficiency for most of the bacteria tested. PVPP addition during extraction with the Ausubel method significantly improved C<sup>q</sup> only for A. aceti and A. pasteurianus. Thus, the choice of DNA extraction method is essential.

For white wine (**Table 1B**), the effect of PVPP is species dependent, but the results confirmed that the Lipp method gave much better results.

Regarding red wines which are rich in qPCR inhibitor compounds, the Lipp method gave much better C<sup>q</sup> compared to the Ausubel method, except for G. oxydans. When PVPP was added prior to DNA extraction, the Cq-values did not improve whatever the DNA extraction method, except in a few cases (**Table 1C**). However, when using PVPP, the Lipp method was always better than the Ausubel method, as demonstrated by the lower Cq-values. For example, in our study we obtained a C<sup>q</sup> for TABLE 1 | *C*q results according to DNA extraction (Ausubel and Lipp methods) with (+) or without (−) PVPP for *A. aceti*, *A. pasteurianus*, *G. oxydans,* and *Ga. liquefaciens* in (A) growth medium supplemented with 10% (*v*/*v*) ethanol, and (B) white, and (C) red wines.


Values represent the C<sup>q</sup> mean ± Standard Deviation (SD); n = 3; values followed by different letters within a column are statistically different at p < 0.05 (XLStat©).

A. aceti of 27.8 in red wine (Ausubel + PVPP). When using the Lipp method with PVPP, the C<sup>q</sup> was equal to 22.4. Thus, using the suboptimal extraction method, C<sup>q</sup> led to underestimating the real AAB population.

Another way to improve qPCR efficiency is to add either BSA or PVP in the qPCR mix (Jiang et al., 2005; Malorny and Hoorfar, 2005). These compounds are assumed to trap inhibitors in the reaction mix. **Figure 1** shows the results of PCR assays on A. pasteurianus. DNA was extracted using the Lipp extraction method with PVPP, and qPCR was run with or without either the addition of BSA or PVP in the qPCR mix compared to control. The addition of BSA or PVP in the qPCR mix did not lead to any improvement of the C<sup>q</sup> obtained. No improvement of Cq-values was observed using these compounds and, as no difference was highlighted in red wine with the Lipp DNA extraction method between the four species, neither BSA nor PVP were added in the mix to quantify AAB using the Lipp method in red wine. However, BSA added to A. pasteurianus DNA using the Ausubel method with or without PVPP in red wine significantly improved Cq-values compared to control, but they were always significantly higher (less effective) than the Lipp method. The values from

FIGURE 1 | *C*q results for *A. pasteurianus* after DNA extraction by the Lipp method in mannitol supplemented with 10% (*v*/*v*) alcohol (containing 2.4.10<sup>6</sup> cells/mL), white wine (containing 2.7.10<sup>5</sup> cells/mL), and red wine (containing 9.9.10<sup>7</sup> cells/mL). Cell lysis was performed in triplicate with PVPP and BSA (400 ng/µL) ( ) or PVP (0.5% w/v) ( ), added to the qPCR mix and compared to control ( ). An Anova with Tukey's test was performed for each medium independently to analyze the Cq results with BSA and PVP compared to control (\*p > 0.05), (XLStat©). Error bars represent standard deviations.

the Ausubel DNA extraction method with or without PVPP were higher than 11.3 and 12.8 units, respectively, compared to the values obtained with the Lipp method. These results seem to indicate the presence of PCR inhibitor compounds in DNA extract using the Ausubel method with or without PVPP.

#### Use of *E. coli* ER2738 as Microbiological Internal Control

An E. coli standard curve from DNA isolation at various levels between 10<sup>1</sup> and 10<sup>7</sup> bacteria/mL was used to evaluate qPCR efficiency (98.8%) with a correlation coefficient of 0.998. The T<sup>m</sup> of the product had a value of 82.5 ± 0.5◦C. The trend curve is: [E. coli]concentration(log cells/mL) = −0.2977 × C<sup>q</sup> + 11.581. A suspension of the strain E. coli K12 ER2738 was added to obtain 10<sup>4</sup> cells/mL in ethanol medium (10% v/v), and white and red wines samples containing A. aceti, A. pasteurianus, G. oxydans, and Ga. liquefaciens. According to the standard curve, efficient DNA extraction and amplification with EC23S primers should give a Cq-value of 25.5 ± 0.4 for 4 log cells/mL of E. coli. Thus, following DNA extraction performed in triplicate of each AAB in each media with both Ausubel and Lipp methods, with or without PVPP, E. coli was quantified.

The values in **Table 2** are the averages of E. coli populations found after DNA extraction according to methods with or without PVPP. No significant differences (p > 0.05) can be seen between the two techniques used for DNA extraction with or without PVPP for E. coli enumeration in ethanol medium and white wine. No significant cell loss was highlighted for either the Ausubel or Lipp method. However, the standard deviation of the Lipp method was lower than 0.3 log E. coli/mL compared to the Ausubel method for both media.

In red wine, the mean values for E. coli concentration in the Lipp method after DNA extraction with or without PVPP were 3.7 ± 0.2 and 3.3 ± 0.2 log cells/mL, respectively, whereas Ausubel DNA extractions led to 1.8 ± 0.2 and 1.9 ± 0.5 log


TABLE 2 | Averages of *E. coli* K12 ER2738 found after initially adding 4 log cells/mL in ethanol growth medium and white and red wines containing *A. aceti*, *A. pasteurianus*, *G. oxydans,* and *Ga. liquefaciens,* and following DNA extraction.

The Lipp and Ausubel methods were used with (+) or without (−) PVPP during cell lysis. qPCR was done with EC23S primers.

An Anova representation with Tukey's test was applied to the E. coli enumeration results using as control modality of 4.0 log E. coli/mL, (XLStat©). Values followed by different letters within a column are statistically different at p < 0.05 (XLStat©).

cells/mL, respectively, with and without PVPP. Thus, as shown in **Table 2**, the Lipp method with or without PVPP led to the recovery of the internal control population, which was not significantly different from the added population. Although not significant, the Cq-value following Lipp DNA extraction with PVPP tended to be slightly lower in red wine compared to experiments performed without PVPP. On the basis of these results, the Lipp DNA extraction method with PVPP was used for all extractions in wine. AAB can be quantified only if E. coli quantification is not significantly different from the added concentration (10<sup>4</sup> E. coli/mL), since the detection of a loss of internal standard leads to an underestimation of AAB population.

#### AAB Quantification with the Presence of Other Wine Microorganisms

To verify the absence of interference by other microorganisms in naturally contaminated wine, the following procedure was implemented. AAB quantifications in red wine containing 10<sup>5</sup> A. pasteurianus/mL alone or supplemented with O. oeni and B. bruxellensis at 10<sup>5</sup> cells/ml were performed. qPCR with AAB primers was performed after validating the presence of E. coli at 10<sup>4</sup> cells/mL. The results of the samples containing A. pasteurianus alone or with other microorganisms (O. oeni and B. bruxellensis) were identical: 5.1 ± 0.1 and 5.1 ± 0.2 log A. pasteurianus/mL, respectively.

#### Linearity, Repeatability, and Reproducibility in Red Wine

After artificial contamination of the red wine by AAB and incubation allowing their growth, verified over time by flow cytometry, Lipp DNA extraction with PVPP was performed in triplicate with 3 repetitive independent experiments. The T<sup>m</sup> of the product had a value of 82.5 ± 0.5◦C. AAB quantification could be done after confirming the presence of 4 log E. coli/mL. Efficiency, r 2 , and y-intercept are presented in **Table 3**. No significant differences between efficiency, r 2 , and y-intercept following a Tukey's test between red wines, containing each AAB species or a mix of all AAB, were observed. Thus, all the data were compiled to determine the general standard curve allowing AAB quantification in red wine whatever TABLE 3 | Statistical analysis of qPCR parameters obtained from independent DNA isolation experiments performed on *A. aceti, A. pasteurianus. G. oxydans,* and *Ga. liquefaciens* after growth in red wine in triplicate and with three repetitions of the experiment over time.


Three red wines containing a mix of AAB were also analyzed in triplicate. An Anova with Tukey's test were performed according to efficiency, r<sup>2</sup> , and y-intercept, (XLStat©). Values followed by the same letter within a column are statistically identical at p > 0.05 (XLStat©).

the species present (**Figure 2**):

[AAB]concentration(log cells/mL) = −0.2927 × C<sup>q</sup> <sup>+</sup>11.135.

#### Limit of Quantification (LOQ) of AAB in Red Wine

For our study, LOQ was determined using the slope, residue standard deviation, and standard deviation of the intercept obtained from linearity validation experiments (OIV, 2005). Using the Lipp method after AAB growth in red wine, the LOQ value of 5.2.10<sup>3</sup> cells/mL was obtained. These results were the mean of the three independent experiments performed in triplicate. Moreover, the sampling volume of 10 mL instead of 1 mL led to the improvement of LOQ by 0.9 ± 0.1 log cells/mL.

# Acetic Acid Monitoring According to AAB Population

Acetic acid concentration monitoring was performed after A. pasteurianus inoculation in ethanol medium [10% (v/v)] at 10<sup>2</sup> cells/mL in independent experiments performed in triplicate (**Figure 3**). The initial amount of acetic acid was 0.04 g/L. **Figure 3** shows acceptable acetic acid values of 0.04 ± 0.00 g/L for an AAB concentration from 2.2 to 5.4 log CFU/mL. The sample presented 0.52 g/L acetic acid when the A. pasteurianus population reached 6.9 log CFU/mL. The last analysis point has a bacterial population of 7.2 ± 0.1 log CFU/mL and shows acetic acid values from 2.39 ± 0.61 g/L. Thus, acetic acid production increased in ethanol medium and occurred when acetic acid bacteria exceeded 6 log CFU/mL. AAB growth was low between the analysis points of 10 and 14 days, indicating that culture reached a stationary phase. The high acetic acid concentration determined during this time seemed to highlight overproduction during the stationary phase of the culture, i.e., when AAB are in high concentration.

To confirm these hypotheses of acetic acid production beyond acceptable limits at a population higher than 6 log AAB/mL and during the stationary phase of AAB growth, AAB growth was monitored in red wine from the vineyard of the University of Burgundy, contaminated naturally over time (**Figure 4**). Both the AAB population and acetic acid concentration were determined over time. The enumeration of total bacteria in the samples was performed by FCM and LAB quantification in specific medium. Neither yeasts nor LAB were detected in the wine under study, consequently only AAB were present and enumerated. Therefore, it was possible to compare the AAB concentration determined by FCM/qPCR methods and acetic acid production. Acetic acid

concentration as a function of AAB population is presented in **Figure 4**. For the first points of analysis containing a population ranging from 2.9 to 6 log AAB/mL, quantifications by qPCR were well-correlated with quantification by FCM. The acetic acid concentrations at these analysis points did not exceed 0.29 g/L. Then, higher acetic acid concentrations were reached for higher acetic acid bacteria concentrations. The analysis point containing an AAB population of 6.8 ± 0.2 log bacteria/mL had 0.82 ± 0.01 g/L acetic acid. The last point of AAB growth monitoring in red wine presented a high AAB concentration with acetic acid levels exceeding the European limit values and an increase in the mean difference between the qPCR and FCM methods (1.2 ± 0.6 log cells/mL). **Figure 4** validates acetic acid production by AAB during the stationary growth phase, thus when AAB exceeded 6 log bacteria/mL.

The measure of acetic acid and counting of AAB by FCM/qPCR methods were then performed on 15 red wines chosen randomly from wine regions (**Figure 5**). Twelve red wines had a mean amount of acetic acid at 0.52 ± 0.15 g/L. For these same samples, the median amount of acetic acid was 0.57 ± 0.15 g/L. These wines are considered as unspoiled. The mean relative difference between FCM vs. qPCR, allowing the determination of the AAB population, was 0.5 ± 0.3 log cells/mL. Three of the fifteen red wines analyzed presented a high amount of acetic acid and were thus spoiled. These wines had an AAB concentration higher than 6 log AAB/mL.

# DISCUSSION

Cell quantification by qPCR in red wine is difficult since various qPCR inhibitors such as polyphenols and polysaccharides are abundant, thereby increasing the risk of false negative results (Demeke and Jenkins, 2009) and making the amplification of genetic material challenging. qPCR has been used to quantify AAB in wine (González et al., 2006; Andorrà et al., 2008; Torija et al., 2010; Valera et al., 2013), however, the authors of these

studies did not take into account extraction efficiency or the presence of inhibitors, nor did they use any control process (cells added to the matrix). Process control can provide information on the efficiency of extraction and on the PCR procedure. The addition of a "spike" control to ensure accurate and reliable quantification is important (Fukushima et al., 2003; Stoeckel et al., 2009; van Doorn et al., 2009) and has already been reported for food analysis (Josefsen et al., 2010; Krøjgaard et al., 2011; Ishii et al., 2013), but as far as we know only once for quantifying Brettanomyces in wine (Tessonnière et al., 2009). In order to develop an accurate and efficient qPCR method to quantify AAB, DNA extraction procedures were first tested in order to ensure good DNA preparation. Indeed, the extraction efficiency and quality of DNA must be optimized for quantitative PCR. The selection of an appropriate DNA extraction method from those available is thus crucial (Jara et al., 2008).

## DNA Extraction Method

We opted for the Lipp method (Lipp et al., 1999) since this method has proven useful for many foodstuffs, especially those rich in phenolic compounds. By comparing this method with classical CTAB extraction, we clearly demonstrated the advantage of using the former. Using our process control strain, we demonstrated that the DNA extraction method was well-suited for wine and especially red wine. Moreover, as PVPP is known to remove PCR inhibitors (Tessonnière et al., 2009) it was added in our assay and improved Cq-values. The results of our validation protocol proved the specificity of the assay. Indeed, qPCR primers showed good specificity for all the wine AAB tested and did not amplify from the other wine organisms. Moreover, our study showed that AAB population determination by qPCR with the Lipp method using PVPP was not influenced by the other microorganisms present (O. oeni and B. bruxellensis) in high concentration. Therefore, pre-amplification using a nested PCR technique (González et al., 2006) is not required to determine the AAB population in samples.

#### AAB qPCR Limit of Quantification

Our LOQ was 5.10<sup>2</sup> cells with a test sample of 10 ml. This LOQ is similar to those previously reported and was reached with AAB grown in red wines, contrary to González et al. (2006) and Torija et al. (2010). In our study, the correlation coefficient of the standard curve obtained for AAB in red wines was 0.76. This coefficient might appear low but it comes from a calibration curve used for wine. Indeed, it is essential to create calibration curves in food matrices. The creation of standard curves for the quantification of a microorganism by diluting DNA from one extraction, as has been done in most studies, should be avoided (González et al., 2006; Torija et al., 2010; Valera et al., 2013). Consequently, the influence of food matrices is not considered, resulting in an underestimation of microbial load (Cocolin and Rantsiou, 2012). Most studies have obtained a high correlation coefficient between 10-fold serial DNA dilutions, but have not performed DNA extraction from a sample containing a low bacterial population, thus biasing efficiency. DNA extraction must be performed from each dilution before running the qPCR to create a standard curve.

### Evaluation of the Method with Naturally Spoiled Wines

Little is known of the spoilage caused by AAB populations and it probably depends on the matrix (sulfites, polyphenols, etc.). Also, this topic is controversial in the literature. According to Joyeux et al. (1984), a low population can activate significant volatile acidity production when exposed to air. However, Drysdale and Fleet (1985) reported that A. pasteurianus and A. aceti occur at 101–10<sup>3</sup> CFU/mL in many wines during bulk storage in wineries without causing spoilage. Moreover, Bartowsky and Henschke (2008) have shown that 2.10<sup>4</sup> A. pasteurianus/mL in a Shiraz wine leads to an acetic acid concentration of 0.6 ±0.0 g/L. This wine was not considered spoiled because the initial wine, without microorganisms, contained 0.5 ±0.0 g of acetic acid per liter. Another wine containing 9.10<sup>4</sup> A. pasteurianus/mL had 3.5 ±1.7 g/L of acetic acid, and was thus spoiled. These results may underestimate the true population due to the uncertainty in recovering all the bacteria. In the literature, there is no clear consensus regarding AAB concentration leading to spoilage risk. Therefore, we determined the level of acetic acid concentration in ethanol medium and red wine according to the AAB population over time. In the latter, the LAB concentration was determined to avoid overestimation of the AAB population by FCM bacteria enumeration, even if the LAB is not a serious issue because AAB causes wine spoilage only during aging in the cellar and after bottling (Henick-Kling, 1993), normally without the presence of LAB. In our study, neither yeasts nor LAB were detected in these red wines. Thus, the assessment of our improved qPCR assay for AAB quantification in naturally contaminated red wine proved reliable and efficient in comparison with cytometry results. The small deviations between both methods were probably derived from the matrix, which was different to that used to write the equation. In our study, during the monitoring of AAB naturally present in red wine, red wine exhibiting an AAB population lower than 6 log bacteria/mL was not altered. However, for wines presenting an AAB population of about 6-7 log cells/mL, the acetic acid exceeded the aroma threshold (Davis et al., 1985; Eglinton and Henschke, 1999a,b; Swiegers et al., 2005) but not the European limit values for sale. For the other wines with high AAB populations and during the stationary phase, acetic acid exceeded the European limit values. Thus, the method developed in this study had an elevated LOQ without using a higher sample volume (e.g., 10 mL).

#### CONCLUSION

The method allowed AAB detection and quantification before spoilage occurred, which meets the needs and expectations of the wine industry when monitoring AAB populations on a regular basis. Therefore, in our study, we developed a qPCR method which allows the reliable quantification of AAB in red wine. We showed that the previously reported DNA extraction method was not efficient enough for precise quantification. The methods used in these studies probably led to underestimating the AAB population. Use of an internal control allows validating DNA extraction and qPCR efficiencies. No underestimation can be made if the initial concentration of the internal control added in the sample before DNA extraction is found. Moreover, the standard curve was established with AAB that proliferated in red wine. Finally, as far as we know, this is the first time a qPCR protocol allowing AAB quantification in red wines without bias (neither cell loss nor PCR inhibitor presence) has been validated. Moreover, the presence of other microorganisms in the sample did not alter AAB quantification.

Specific AAB species quantification is not possible with our method; however, our main goal was to quantify AAB in wines

#### REFERENCES


whatever the species present in order to evaluate the risk of spoilage. Furthermore, one drawback of the method is that qPCR quantifies live and dead AAB. Propidium monoazide could be used (Vendrame et al., 2013; Rizzotti et al., 2015) but cell exposure to a stress like ethanol, which is known to permeabilize membranes (Alexandre et al., 1994), can result in cells stained by propidium iodide, for example, but that are still alive (Davey and Hexley, 2011). Moreover, Shi et al. (2012) highlighted an underestimation of total yeasts, S. cerevisiae, total LAB, non-O. oeni LAB, and total AAB in wine by realtime PCR coupled with ethidium monoazide. However, they demonstrated that 40 min of incubation in recovery medium could completely cancel the underestimation of viable cell counts performed in wine, but not cell growth, which must be checked for each sample. To avoid this potential error, one way to circumvent the drawback of the method is to monitor the AAB population over time. Under these conditions, a decrease in Cq-value would reflect AAB growth and potentially the risk of spoilage.

#### AUTHOR CONTRIBUTIONS

Conceived and designed the experiments: HA, CL, and MG. Performed the experiments: CL. Generated and analyzed the data: CL, HA, and MG. Wrote the paper: CL, HA, and MG.

#### FUNDING

This work was funded by the Regional Council of Burgundy and the Interprofessional Office of Burgundy Wines.

RNA and single copy genes. Lett. Appl. Microbiol. 52, 298–306. doi: 10.1111/j.1472-765X.2010.03001.x


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Longin, Guilloux-Benatier and Alexandre. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# *Vibrio parahaemolyticus* Strains of Pandemic Serotypes Identified from Clinical and Environmental Samples from Jiangsu, China

Jingjiao Li 1, 2, Feng Xue<sup>1</sup> \*, Zhenquan Yang<sup>3</sup> , Xiaoping Zhang<sup>4</sup> , Dexin Zeng<sup>1</sup> , Guoxiang Chao<sup>5</sup> , Yuan Jiang<sup>1</sup> \* and Baoguang Li <sup>6</sup>

<sup>1</sup> Animal Quarantine Laboratory, Jiangsu Entry-Exit Inspection and Quarantine Bureau, Nanjing, China, <sup>2</sup> Key Laboratory of Veterinary Biotechnology, Department of Animal Science, School of Agriculture and Biology, Shanghai JiaoTong University, Shanghai, China, <sup>3</sup> Jiangsu Key Laboratory of Zoonosis, School of Food Science and Engineering, Yangzhou University, Yanghzou, China, <sup>4</sup> Beijing Kemufeng Biopharmaceutical Company, Beijing, China, <sup>5</sup> Yangzhou Key Centre for Disease Control and Prevention, Yanghzou, China, <sup>6</sup> Division of Molecular Biology, Center for Food Safety and Applied Nutrition, US Food and Drug Administration, Laurel, MD, USA

*Edited by:* Giovanna Suzzi, Università Degli Studi di Teramo, Italy

#### *Reviewed by:*

Adrian Canizalez-Roman, Autonomous University of Sinaloa, Mexico Learn-Han Lee, Monash University Malaysia, Malaysia

#### *\*Correspondence:*

Feng Xue fengxue1219@aliyun.com; Yuan Jiang jiango@yeah.net

#### *Specialty section:*

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

> *Received:* 15 April 2016 *Accepted:* 09 May 2016 *Published:* 31 May 2016

#### *Citation:*

Li J, Xue F, Yang Z, Zhang X, Zeng D, Chao G, Jiang Y and Li B (2016) Vibrio parahaemolyticus Strains of Pandemic Serotypes Identified from Clinical and Environmental Samples from Jiangsu, China. Front. Microbiol. 7:787. doi: 10.3389/fmicb.2016.00787 Vibrio parahaemolyticus has emerged as a major foodborne pathogen in China, Japan, Thailand, and other Asian countries. In this study, 72 strains of V. parahaemolyticus were isolated from clinical and environmental samples between 2006 and 2014 in Jiangsu, China. The serotypes and six virulence genes including thermostable direct hemolysin (TDR) and TDR-related hemolysin (TRH) genes were assessed among the isolates. Twenty five serotypes were identified and O3:K6 was one of the dominant serotypes. The genetic diversity was assessed by multilocus sequence typing (MLST) analysis, and 48 sequence types (STs) were found, suggesting this V. parahaemolyticus group is widely dispersed and undergoing rapid evolution. A total of 25 strains of pandemic serotypes such as O3:K6, O5:K17, and O1:KUT were identified. It is worth noting that the pandemic serotypes were not exclusively identified from clinical samples, rather, nine strains were also isolated from environmental samples; and some of these strains harbored several virulence genes, which may render those strains pathogenicity potential. Therefore, the emergence of these "environmental" pandemic V. parahaemolyticus strains may poses a new threat to the public health in China. Furthermore, six novel serotypes and 34 novel STs were identified among the 72 isolates, indicating that V. parahaemolyticus were widely distributed and fast evolving in the environment in Jiangsu, China. The findings of this study provide new insight into the phylogenic relationship between V. parahaemolyticus strains of pandemic serotypes from clinical and environmental sources and enhance the MLST database; and our proposed possible O- and K- antigen evolving paths of V. parahaemolyticus may help understand how the serotypes of this dispersed bacterial population evolve.

Keywords: *Vibrio parahaemolyticus*, serotyping, MLST, virulence genes, TDH-related hemolysin (TRH), pandemic serotypes, epidemiology, phylogeny

# INTRODUCTION

Vibrio parahaemolyticus is a Gram-negative, halophilic bacterium that inhabits global coastal waters and rivers, and in seafood, such as fish and shellfish (Kelly and Stroh, 1988). V. parahaemolyticus was initially discovered in 1950 (Parveen et al., 2008; Letchumanan et al., 2015b). A novel serotype of O3:K6 clone was discovered in India in 1996 (Okuda et al., 1997), and since then, this clone and its serovariants have disseminated worldwide and become a pandemic clonal group (Ansaruzzaman et al., 2005; Quilici et al., 2005; Nair et al., 2007; Chao et al., 2009; Velazquez-Roman et al., 2013; Li W. et al., 2014; Flores-Primo et al., 2015).

V. parahaemolyticus can cause three major clinical syndromes: gastroenteritis, wound infections, and septicemia (Daniels et al., 2000), and is also considered as the causative agent of the most prevalent food poisoning in Asia since the outbreak in 1959 (Miyamoto et al., 1962). V. parahaemolyticus infections usually resulted from consumption of raw or undercooked seafood, mostly causing gastroenteritis (Miyamoto et al., 1969). Recently, V. parahaemolyticus has been identified as a major foodborne pathogen in food poisoning incidents in China, raising public health concern (Ma et al., 2014; Zhang et al., 2015).

V. parahaemolyticus was initially discovered in 1950 (Parveen et al., 2008). Traditionally, the identification of V. parahaemolyticus is performed by serological and biochemical tests. V. parahaemolyticus is classified by serotyping and the serotypes of V. parahaemolyticus are determined by the combination of somatic (O) antigens and capsular (K) antigens. There are 13 O serotypes and 71 K serotypes that have been reported (Iida et al., 1997; Nair et al., 2007; Chen et al., 2012). V. parahaemolyticus infections are associated with pathogenic strains of numerous serotypes (clinical); whereas non-pathogenic strains comprise an even greater variety of serotypes.

Serotypes are useful distinguishing features to identify clinical isolates (Jones et al., 2012), and serotyping has been widely used in epidemiological research and etiological diagnostics for many decades. However, the increasing genetic diversity such as emerging of new serotypes and STs among V. parahaemolyticus strains poses a challenge to this traditional way of strain identification and differentiation due to the high cost of the antisera and the potential ambiguity encountered during serotyping (Bogdanovich et al., 2003). In the last two decades, numerous DNA-based subtyping methods such as pulsed-field gel electrophoresis (PFGE; Wong et al., 1996), multilocus sequence typing (MLST; Gonzalez-Escalona et al., 2008), repetitive element PCR (Rep PCR; Wong and Lin, 2001), multilocus variable-number tandem-repeat analysis (MLVA; Kimura et al., 2008), clustered regularly interspaced short palindromic repeats (CRISPR; Sun et al., 2015), and microarray analysis (Li et al., 2015) have been developed to investigate the genetic diversity of outbreaks caused by V. parahaemolyticus and other foodborne pathogens (Li B. et al., 2014). Each of these subtyping methods has its advantages and disadvantages with respect to sensitivity, specificity, cost, and speed (Li et al., 2015). In general, MLST is the most commonly used method (Maiden, 2006; Nair et al., 2007; Gonzalez-Escalona et al., 2008), due to its high specificity, repeatability, and portability (Nair et al., 2007).

In this study, in an effort to assess the homogeneity and heterogeneity between the clinical and environmental V. parahaemolyticus isolates from Jiangsu Provence, an east coast area in China, where seafood is widely consumed, we used the traditional classification method, serotyping, to identify 72 V. parahaemolyticus strains from the food poisoning case samples and the environmental samples. The genetic diversity among V. parahaemolyticus strains were further assessed by MLST analysis and the presence of the virulence factors such as tdh, trh, orf8, GS-PCR, PGS-PCR, and HU-α. Furthermore, we analyzed the V. parahaemolyticus isolates by using eBURST and START (http:// eburst.mlst.net) to investigate the relationship between clinical and environmental V. parahaemolyticus strains based on MLST databases. Additional information on genetic variation and the distribution of virulence genes among V. parahaemolyticus strains from various cities in Jiangsu Province would enrich the MLST database and epidemiological archive and be beneficial for the development of an efficient risk assessment of this common foodborne pathogen.

#### MATERIALS AND METHODS

#### *V. parahaemolyticus* Strain Identification

Seventy-two presumed V. parahaemolyticus isolates were collected from nine different cities in Jiangsu province of China between 2006 and 2014, including 21 clinical isolates from patients with food poisoning and 51 isolates from food samples. All strains were characterized according to GB 4789.7–2013 Chinese Food Safety Standards (http://www.foodmate.net). The V. parahaemolyticus strains were inoculated onto Vibrio culture plates (CHROMagar, Paris, France) and Thiosulphate Citrate Bile salt Sucrose [(TCBS) Beijing Land Bridge, China] culture plates and incubated at 37◦C for 16–24 h. The colonies with typical contour were selected and characterized by VITEK automatic biochemical analyzer (Biomerieux, France).

#### Serotyping

The serotype of V. parahaemolyticus strains was determined using two diagnostic kits; 11 antisera targeting the O1–O11 antigens and 71 antisera for the K antigens (Denka Seiken, Tokyo, Japan) and 11 antisera for O1–O13 antigens (Tianjin Biochip Corporation, Tianjin, China). Serotyping was carried out in accordance with the GB4789.7–2013 Chinese Food Safety Standard (http://www.foodmate.net). Single colonies were selected and plated onto the surface of 3% sodium chloride peptone agar plates, incubated at 37◦C for 18 h. Bacterial suspension was obtained by washing the surface of agar with solution containing 3% NaCl and 5% glycerol.

O-antigen identification: The bacterial suspension was autoclaved at 121◦C for 1.5 h followed by centrifugation at 12,000 g for 15 min. The pellets were washed two or three times with normal saline solution and centrifuged at 12,000 g for 15 min. The final suspension was used for O antiserum agglutination and normal saline solution was used as a negative control. If the result was negative, all the above steps were repeated; thereafter, the negative was considered as unknown antigen O.

K-antigen identification: Multi-serum against the K-antigen was added to the bacterial suspension. Positive colonies were selected for further analysis using individual K-antigen antiserum. Solution of NaCl (3%) was used as negative control.

# Identification of the *tdh, trh, orf8, GS-PCR, PGS-PCR,* and *HU*-α Genes

Single colonies of V. parahaemolyticus strains were picked and inoculated into liquid culture medium containing 3% sodium chloride peptone. Cultures were incubated at 37◦C for 16 h, followed by centrifugation at 12,000 g for 10 min. The genomic DNA was isolated using E.Z.N.A.TM kit (OMEGA, Beijing, China), and the concentrations of DNA were determined by spectrophotometry to ensure the OD260/OD280 was between 1.8 and 2.0.

The primers for the tdh, trh, orf8, GS-PCR, PGS-PCR, and HU-α genes were synthesized (Chao et al., 2009; Li W. et al., 2014; **Table 1**) by Sango Biotech Co., Ltd. (Shanghai, China). PCR reactions were performed in a volume of 25µl containing 1µl DNA (50 ng/µl), 12.5µl 10 × PCR mix (Takara, Dalian, China), 1µl of the forward and reverse primers of the virulence genes, and 9.5µl of sterile distilled water. Strains ATCC33847 (tdh+trh−, isolated in US in 1973) and ATCC17802 (tdh−trh+, isolated in Japan in 1965) were used as positive controls. PCR reactions were performed under the following conditions: initial denaturation at 95◦C for 2 min, followed by 30 cycles of denaturation at 95◦C for 30 s, annealing at 50◦C for 30 s, and elongation at 72◦C for 1 min, and ending with elongation at 72◦C for 5 min. PCR products (1µl) were analyzed using Agilient 2100 analyzer (Waldbronn, Germany) to determine the PCR amplicon size.

#### MLST

Seven housekeeping genes, dnaE, gyrB, recA, dtdS, pntA, pyrC, and tnaA, were selected as target genes in this study. The primers for these genes were adopted from MLST website (http://www. pubmlst.net). The MLST primers for recA gene failed in most strains in the current study, and new primers for recA gene were adopted from a previous study (Gonzalez-Escalona et al., 2008). All primers as shown in **Table 2** were synthesized by Sango Biotech Co., Ltd. (China). Primers were diluted to 10µM and stored at −20◦C.

#### PCR Amplification of Housekeeping Genes

To sequence the housekeeping genes, the PCR reaction volume was set at 50µl. The components of PCR reactions as well as the parameters for each cycle were optimized for best performance. PCR products were visualized using Agilent electrophoresis and imaging system. The bidirectional DNA sequencing was conducted by Sango Biotech Co., Ltd.

#### Allele and Sequence Analysis

The alignment of DNA sequences was performed by uploading each of the sequences of the seven housekeeping genes of the 72 V. parahaemolyticus strains to the MLST website (http://

TABLE 1 | Primers used for the detection of *V. parahaemolyticus* virulence genes and pandemic marker genes by PCR.


<sup>a</sup>Forward primer.

<sup>b</sup>Reverse primer.

pubmlst.org/vparahaemolyticus/). If a novel allele or sequence was identified, the forward and reverse sequences were uploaded and submitted to the database administrator to obtain a serial number for the allele or sequence.

### Sequence Analysis by eBURST v3.0 and START v2.0

The ST types of all the 72 strains were analyzed by eBURST v3.0 (http://eburst.mlst.net) to distinguish clonal complex (CC), group, and singleton STs. The most stringent definition was adopted to identify the homeotic complexes, each of which was defined by the presence of at least six or seven identical alleles. The single locus variant (SLV) was defined by the presence of a single allele difference between any two ST types, based on eBURST v3.0 analysis. The evolution of each allele and ST type was analyzed by using START v2.0 (http://pubmlst.org/software/ analysis/start/).

# RESULTS

V. parahaemolyticus colonies were round, translucent and purplish red on CHROMagar plates measuring 2–3 mm in diameter. They were round, translucent, and smooth greencolored colonies on TCBS plates. All 72 V. parahaemolyticus isolates were confirmed by VITEK biochemical analysis.

#### Serotyping

There were 25 serotypes identified among the 72 V. parahaemolyticus isolates. The dominant serotypes were

#### TABLE 2 | Primers of housekeeping genes used in MLST and PCR conditions in this study.


The underlined lower-cased letters represent common sequences that were adapted from M13 for more efficient sequencing.

O3:K6 (n = 8) and O2:K28 (n = 8), followed by O1:KUT (n = 6), O5:K17 (n = 5), O5:KUT (n = 5), O1:K25 (n = 3), O4:K34 (n = 3), O10:KUT (n = 3), O2: KUT (n = 3), O1:K32 (n = 3). For the 21 clinical strains, nine serotypes were identifies where O3:K6 was the dominant serotype (n = 8), followed by O5:K17 (n = 5), O2:KUT (n = 3), and one strain of O1:K25, O11:K40, and O13:KUT (novel serotype). In addition, there was one strain (WX14115) that failed to react to either O or K antiserum (**Table 3**).

Of the 25 serotypes, six were novel serotypes. Out of the six novel serotypes, five new serotypes were recovered from environmental samples and one was from clinical sample. Specifically, the five environmental strains are O2:K25, O4:K9, O4:K42, O8:K39, and O11:31, and the single clinical isolate is O13:KUT (**Table 4**). In the present study, 25 strains of pandemic serotypes were identified from Jiangsu Province, including 16 clinical strains and 9 environmental strains. Specifically, the clinical strains were serotypes O3:K6 (n = 8), O5:K17 (n = 5), and environmental strains were serotypes O1:KUT (n = 6), O1:K25 (n = 3), O3:K29 (n = 1), and O3:KUT (n = 1) (**Table 3**).

# Identification of Virulence Factor Genes *tdh, trh, GS-PCR, PGS-PCR, orf8,* and *HU*-α

The virulence genes in the 72 V. parahaemolyticus isolates were assessed by PCR. There were nine tdh<sup>+</sup> strains and four trh<sup>+</sup>

strains, accounting for 61.9% (13/21) of the clinical strains. No strain was tdh+trh+. The nine tdh<sup>+</sup> strains included seven O3:K6, one O1:K25 strain and one O4:KUT strain. Four trh<sup>+</sup> strains included three O5:K20 and one O13:KUT (**Table 3**). The prevalence of the other four virulence genes, GS-PCR, PGS-PCR, orf8, and HU-α, are 81.9, 38.9, 16.9, and 18.9%, respectively. GS-PCR gene showed the highest prevalence and orf8 gene showed the lowest prevalence for the four virulence genes in our collection (**Table 3**).

# MLST Analysis

All seven housekeeping genes were amplified in the V. parahaemolyticus strains using specific primers, and the PCR products were sequenced. **Seventy-two** V. parahaemolyticus strains were classified into 48 STs by MLST analysis. Of the 48 ST types, 34 were singleton and 14 were ST groups. Each singleton represented only one strain, while each ST group included two to eight strains (**Figure 1**). Twenty-one clinical strains were classified into 13 STs. The dominant ST was ST-3 with eight strains, followed by ST-79 and ST-564, each of which covered three strains. The remainder of STs was singletons. The MLST results indicated that V. parahaemolyticus strains showed genetic polymorphisms with much higher incidence rate in environmental strains than in clinical strains.

# Novel Allele and STs

The ST composition of the 72 V. parahaemolyticus strains included 32 novel STs with 32 new allele values and 34 allele spectra, accounting for 79% of total STs. All of these newly identified allele values and allele spectra were submitted to the PubMLST database (http://pubmlst.org/vparahaemolyticus/) as shown in **Table 3**. There were 13 STs among the 21 clinical strains, including seven newly identified STs accounting 53.85% (7/13) of the ST types. There were 36 STs identified among the 51 environmental strains, including 27 newly identified STs, accounting 75% (27/36) of the STs. It appears that 13 new STs were formed through allele recombination, while the other 21 new STs were created by the newly identified housekeeping gene alleles which included several types of dnaE (N = 4), gyrB (N = 6), recA (n = 7), dtdS (n = 2), pntA (n = 3), pyrC (n = 9) genes, and one tnaA gene.

# Homologous Allele Complex and Systematic Development Analysis

The system evolution diagram was plotted by eBURST v3 software, and 48 STs were divided into one clonal complex (CC), four double combinations (D), and 38 singletons (S). The CC identified in this study was the CC3 which covered seven strains of O3:K3 ST-3 and one strain of O3:KUT ST-3; these eight strains were epidemic strains from clinical samples. The D type included ST-799-ST-1108, ST-212-ST-1002, ST-996-ST-999, and ST-993- ST-968. The ST-1108, ST-1002, ST-996, ST-999, and ST-993 were newly identified in this study. Of the 38 singletons, 28 novel STs were identified in this study, which were genetically distant to the ST-3 and other STs (**Table 3**; **Figure 1**).



(Continued)


TABLE 3 | Continued

<sup>a</sup>A bold-faced number refers a novel ST.

#### Association of STs between Serotypes and Virulence Genes

The strains with identical serotype usually showed similar STs or were clustered together, such as the majority serotype of O3:K6 in the present study belonging to ST-3; only one O3:KUT strain shared the ST-3 with O3:K6 strains. On the other hand, some strains with identical serotypes have different STs, such as strains of O1:KUT. These were identified as ST-1007, ST-212, and ST-276 (**Figure 2**).

There were nine tdh<sup>+</sup> pathogenic strains identified in this study, including eight ST-3 strains (seven O3:K6 serotypes and one O1:K25 serotype), and one ST-332 strain (serotype O4:KUT). All of these 10 strains were epidemic strains from clinical samples. There were four trh<sup>+</sup> strains, including three ST-79 (O5:K17) strains and one ST-1114 (O13:KUT) strain. All the four trh<sup>+</sup> strains seemed to be epidemic strains (**Table 4**).

#### DISCUSSION

V. parahaemolyticus is the major foodborne pathogen. It is widely distributed with high survival and incidence rates, especially in the coastal areas of China (Su and Liu, 2007; Chao et al., 2009; Yan et al., 2015). Based on the epidemiological surveillance data from countries in Southeast Asia, V. parahaemolyticus infections have become a majority of foodborne pathogen (Pan et al., 1997; Wong et al., 1999; Obata et al., 2001; Liu et al., 2004; Cho et al., 2008; Letchumanan et al., 2014). In the present study, 25 serotypes and 48 STs were identified among the 72 V. parahaemolyticus isolates. Of the 25 serotypes and 48 STs, six additional novel serotypes and 34 novel STs were identified, indicating the V. parahaemolyticus population in Jiangsu Province was highly dispersed. The diversity in serotypes and STs of V. parahaemolyticus, especially environmental strains, is attributed to frequent recombination events in the pathogen but not by mutation (Gonzalez-Escalona et al., 2008). One example to support this is that the serotype of epidemic strains (O3:K6), which has been continuously evolving, resulted in 21 derivative serotypes of O3:K6 such as O4:K68, O1:K25, O1:KUT, O4:K12, and O5:K17 (Nair et al., 2007). All those strains demonstrated identical genotypes and molecular spectra, therefore the O3:K6 and its derivative are called O3K6 clones or pandemic strains. MLST analysis confirmed that O3:K6 and its derivatives belong to the same genetic lineage (Chowdhury et al., 2000, 2004; Matsumoto et al., 2000). Surprisingly, the 25 pandemic strains identified in this study were not exclusively from clinical samples (n = 16), rather, some strains were recovered from environmental samples (n = 9) as well.

Serotypes of O3:K6 and O5:K17 were the most common serotypes among the 21 clinical strains whereas some serotypes such as O11:K40, O4:K8, O2:KUT, and O13:KUT are rarely reported globally. It has been shown that serotype O1:KUT is closely related to pandemic strains and is seldom detected as environmental strains (Iida et al., 1997; Mahoney et al., 2010). However, in the present study, O1:KUT was identified as a common serotype with six isolates from environmental samples. More importantly, the six O1:KUT strains demonstrated extremely high genetic diversity with five different STs among



<sup>a</sup>UT refers to untypable.

<sup>b</sup>Bold faced letters and number refer to a novel serotype or ST.

the six strains (**Table 4**). Furthermore, of the five STs possessed by O1:KUT strains, three were novel STs, namely ST-1002, ST-1007, and ST-1105, suggesting these O1:KUT strains were highly dispersed and evolving rapidly in the environment. In addition, some of the O3:K6-specific pandemic markers such as GS-PCR and PGS-PCR were tested positive among most of these strains of pandemic serotypes. However, the pathogenicity of those strains needs to be further confirmed by recovering those strains from patients. Nevertheless, the risk posed to the public health in China by these "environmental" pandemic strains should not be overlooked.

The exact pathogenic mechanism of V. parahaemolyticus remains unclear, but the tdh and trh genes are considered the main pathogenic factors. Thermostable direct hemolysin (TDH), encoded by the tdh gene, manifests hemolytic, intestinal, and cardiac toxicities (Iida et al., 1997; Rosec et al., 2009; Raghunath, 2014). The trh gene is closely associated with the production of urease (Quilici et al., 2005). An epidemiological surveillance of V. parahaemolyticus in Northwest Mexico showed that up to 71.74% of the environmental isolates carried the tdh gene (De Jesus Hernandez-Diaz et al., 2015). However, numerous reports have shown that many pathogenic strains from patients were detected with neither of the two virulence factor genes, indicating more virulence factor genes are needed as markers for identification of pandemic V. parahaemolyticus strains (Garcia et al., 2009; Jones et al., 2012; Liu and Chen, 2015). Additionally, the type

III secretion system (T3SS) of V. parahaemolyticus has been identified as a potential strain virulence factor (Park et al., 2004; Broberg et al., 2011).

V. parahaemolyticus O3:K6 serogroup has group-specific gene sequences in the toxRs operon and orf8, one of the 10 known open reading frames (ORFs) which is unique to the O3:K6 filamentous phage f237. The toxRs and orf8 genes have been used as genetic markers to differentiate O3:K6 from other serogroups (Matsumoto et al., 2000; Nasu et al., 2000). Additionally, the HU-αORF, a specific biomarker for pandemic strain, which has a C-terminal amino acid sequence different from those of other strains of V. parahaemolyticus, was used to identify O3:K6 and other serotypes, such as O1:K25, O1:KUT, and O4:K68 (Matsumoto et al., 2000; Williams et al., 2004).

It has been reported that V. parahaemolyticus that lacked the tdh and trh genes were pathogenic in a study using mice (Rosec et al., 2009). In this study, most of the clinical strains (13/21) were negative for the tdh and trh genes, but possessed at least one of the four other virulence genes, GS-PCR, PGS-PCR, orf8, and HU-α. For instance, two tdh−trh<sup>−</sup> strains, OUT:KUT (WX14115, clinical) and O5:KUT (YZ0608, environmental) were positive for the other four virulence genes (GS-PCR, PGS-PCR, orf8, and HUα). Using a panel of six virulence genes as pathogenic markers, almost all the O3:K6 strains (except for two strains, WX1475 and WX14116) were positive for at least four of the six virulence genes, and only one clinical strain of O5:K17 (WX14102) was negative for all the six virulence genes among the 21 clinical strains. Thus, the results of this panel of virulence genes may more closely reflect the pathogenicity potential of those strains.

A high percentage (90.3%, 65/72) of strains were positive for at least one of the six virulence genes, tdh, trh, GS-PCR, PGS-PCR, orf8, and HU-α. Out of the six virulence genes, the GS-PCR gene showed the highest prevalence 81.9% (59/72). This result is very different from other investigators in China, where GS-PCR gene was seldom detected from environmental samples (Alam et al., 2009; Chao et al., 2009; Zhang et al., 2013). This difference might be a reflection of the genetic diversity between the indigenous isolates in this area (Jiangsu, China) and other areas or an artifact caused by sampling difference.

The GS-PCR gene has been shown to be a specific for genetic marker for the identification of pandemic V. parahaemolyticus strains (Li W. et al., 2014; Pazhani et al., 2014). The trh gene showed the lowest prevalence 5.6% (4/72), which is consistent with the results from other Asia countries (Alam et al., 2009; Chao et al., 2009; Zhang et al., 2013, 2015; Letchumanan et al., 2015a). Of note is that a number of different serotypes of tdh−trh<sup>−</sup> strains (such as O28:K28, O5:KUT, and O11:KUT) were positive for at least three of the four virulence genes, GS-PCR, PGS-PCR, orf8, and HU-α (**Table 3**). Although these strains were isolated from the environment, it is quite possible for a nonpathogenic strain to gain pathogenicity potential after acquiring several virulence genes from pathogenic strains as indicated by the evolving path of the pandemic strain O3:K6 (Nair et al., 2007).

In this study, there were 34 new STs out of the 48 ST types, and the newly defined gene spectra accounted for 70.83%, suggesting this V. parahaemolyticus group is highly dispersed. The cluster diagram of MLST analysis showed that

V. parahaemolyticus strains in Jiangsu area demonstrated a higher level of polymorphisms within environmental strains than clinical strains. ST-3 was the dominant ST among the 21 clinical strains and all ST-3 strains (n = 7) belonged to pandemic O3:K6 serotype, which is in agreement with previous reports from China and other countries (Gonzalez-Escalona et al., 2008).

It is worth noting in this study that there were a number of cases in which strains with identical serotype were classified into different STs; while strains with an identical ST possessed different serotypes. For instance, several strains of O1:KUT serotype were subtyped as ST1007, ST212, and ST276 by MLST; and the same ST strains, such as ST-3 strains, included


TABLE 5 | Possible recombination scenerios of reprentive strains of *V. parahaemolyticus* that share a serotype or a STa.

<sup>a</sup>ST, sequence type; O, O-antigen; K, K-antigen.

<sup>b</sup>An arrow refers to an O- and K-antigen or K antigen exchange between strains of different serotypes or STs.

<sup>c</sup>A blue-highlighted number indicates a new allelic gene.

O3:K6 and O3:KUT serotypes, whereas ST-276 strains included O1K:25 and O1K:KUT serotypes (**Table 5**). This observation not only indicates the advantages and limitations of serotyping and MLST analysis, but also implies that frequent mutation and/or recombination occur on the O- and K-antigens on the chromosome.

In this study, we used serotyping and MLST methods to differentiate the V. parahaemolyticus isolates and assessed the presence of six virulence factors, tdh, trh, orf8, GS-PCR, PGS-PCR, and HU-α. A total of 25 serotypes and 48 STs were identified among the 72 V. parahaemolyticus isolates; six novel serotypes and 34 novel STs were identified; and highly variable profiles of six virulence factor genes were detected among the isolates, suggesting this V. parahaemolyticus group was a highly dispersed group and was evolving rapidly. This information not only can enrich the MLST database but also can serve as a valuable set of matrices to trace the gene mutation and HGT (or recombination) among the V. parahaemolyticus population.

Data from **Table 5** indicate that the O- and K-antigens move together more often than to move independently; and the Kantigen is more likely to be swapped than O-antigen. For example, the ST-3 strains included, in addition to the seven O3:K6 strains, a single strain of O1:K25 (WX1461). TheO1:K25 strain (belonged to ST-3) can serve as an example for O- and K-antigen exchange between serotypes O3:K6 and O1:K25, i.e., the O3:K6 antigens of strain WX1461 were replaced by O1:K25 antigens as evidenced by its identical allele and virulence gene profiles with seven O3:K6 (ST-3) strains. Another example is WX1475 which is serotype O3:K6 but was subtyped as ST-1109 by MLST based on different allele and virulent gene profiles than with those other O3:K6 serotype strains (**Table 5**, lower middle). The third example is that strains with different STs acquired the O2:K28 antigens from a strain with different ST (ST-1003; Top of **Table 5**). Specifically, a serotype O2:K28 isolate (YZ0864) was subtyped as ST-1003 with another serotype O2:K28 isolate (YZ0685).

In this study, there were eight strains with serotype O2:K28 but were differentiated into six different STs, based on the high discriminatory power of MLST (**Table 5**). Serotype O2:K28 was initially found in 1977 (Libinzon et al., 1977) and is generally considered as an environmental strain (Martinez-Urtaza et al., 2004). Drastic differences were found between the results derived from the two subtyping methods. It is hard to pinpoint the exact cause(s) that led to the differences in allele profiles among those isolates during their evolution path. However, a plausible scenario is that the O- and K-antigens of the six strains (O2:K28) of different STs might have been acquired from strains such as YZ0684 (O2:K28; ST-1003) over years (**Table 5**).

Similar virulence gene profiles exist between the O2:K28 strains of ST-1003 and the six different STs strains (lacking tdh and trh genes but are positive for GS-PCR) seem to corroborate this notion (**Table 5**). In addition to the genetic evidence generated in this study, our hypothesis is also supported by the findings on bacterial antigens, virulence genes, and genetic traits of V. parahaemolyticus (Chowdhury et al., 2000; Gonzalez-Escalona et al., 2008; Chao et al., 2009; Mala et al., 2016). As shown in **Table 5**, serotype conversion (from one serotype to a different serotype) occurred more frequently than ST change (from one ST to a different ST) among the listed strains, whose serotype was shared by strains of different STs or whose ST was shared by strains of different serotypes. It seemed that O- and K-antigen conversion occurred simultaneously more often than independently, suggesting that the O- and K-antigens are actively evolving and the two antigens are mostly moving together by horizontal gene transfer (HGT) (or recombination). This presumption is well in agreement with the recent findings on V. parahaemolyticus genomic evolution that HGT is 10–1000 times more attributable than single nucleotide variants to genome diversification. This may be the underlying drive that is responsible for the high diversity among the V. parahaemolyticus studied. This hypothesis is consistent with other researchers on the diversity and pathogenicity of V. parahaemolyticus (Chowdhury et al., 2000; Gonzalez-Escalona et al., 2008; Chao et al., 2009; Mala et al., 2016). Furthermore, other studies on K-antigen and comparative genomic analysis of V. parahaemolyticus (Chen et al., 2010, 2011), have demonstrated that the O- and K-antigens are at two adjacent loci on chromosome II and, thus the O- and K-antigens could be swapped via a single recombination event to create both novel O- and K-antigens (Chen et al., 2010). Moreover, the human upper intestine is believed to be a particularly suitable niche for the intra- and inter-specific lateral transfer of genetic material necessary to enhance bacterial pathogenicity (Larocque et al., 2005; Okada et al., 2009; Hasan et al., 2010; Wang et al., 2011). Therefore, our model is supported not only by multiple lines of genetic evidence from V. parahaemolyticus, but also by the presence of suitable ecological niche that can facilitate HGT. With this notion, we can better interpret the scenarios we encountered here, i.e., the pandemic V. parahaemolyticus strains recovered not only from patients, but also from nine "environmental" seafood samples; and the high genetic diversity among the 72 isolates.

In summary, V. parahaemolyticus in Jiangsu, China, were highly dispersed and widely distributed in the environment. In light of that V. parahaemolyticus has become one of the major foodborne pathogens in China in recent years (Li Y. et al., 2014; Qi et al., 2016), the new threat to the public health posed by these "environmental" pandemic strains should not be overlooked. T. The findings of this study provide new insight into the phylogenic relationship between V. parahaemolyticus strains of pandemic serotypes from clinical and environmental sources; the information on the genetic diversity among isolates enriches the MLST database; and our proposed possible Oand K- antigen evolving paths of V. parahaemolyticus may help understand how the serotypes of this dispersed bacterial population evolve. Our findings also underscores the necessity for more epidemiological studies and more comprehensive surveillances on V. parahaemolyticus in order to efficiently prevent the diseases caused by this organism.

# AUTHOR CONTRIBUTIONS

JL performed the experiments. YJ, FX, and BL conceived and designed the study. XZ, DZ, JL, FX, and BL analyzed the data. ZY and GC provided the isolates. JL, YJ, FX, and BL wrote the manuscript. All the authors reviewed the manuscript.

#### ACKNOWLEDGMENTS

We would like to thank Yanhong Wan, Lu Gao and Haibo Zhou for their useful suggestions to this work. The study was supported by the National Natural Science Foundation

#### REFERENCES


of China (31301460) the National Science and Technology Support Program of 2012BAK17B10, National 10000 Talents— Youth Top-Notch Talent Support Program, and Jiangsu Province Science and Technology Support Program of BE2013733.

strains during seafood-related summer diarrhea outbreaks in southern Chile. Appl. Environ. Microbiol. 75, 7482–7487. doi: 10.1128/AEM.01662-09


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Li, Xue, Yang, Zhang, Zeng, Chao, Jiang and Li. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Recovery and Growth Potential of Listeria monocytogenes in Temperature Abused Milkshakes Prepared from Naturally Contaminated Ice Cream Linked to a Listeriosis Outbreak

Yi Chen<sup>1</sup> \*, Emma Allard<sup>2</sup> , Anna Wooten<sup>3</sup> , Minji Hur<sup>4</sup> , Ishani Sheth<sup>1</sup> , Anna Laasri<sup>1</sup> , Thomas S. Hammack<sup>1</sup> and Dumitru Macarisin<sup>1</sup>

<sup>1</sup> Office of Regulatory Science, Center for Food Safety and Applied Nutrition, Food and Drug Administration, College Park, MD, USA, <sup>2</sup> College of Life Sciences and Agriculture, University of New Hampshire, Durham, NH, USA, <sup>3</sup> Joint Institute for Food Safety and Applied Nutrition, University of Maryland, College Park, MD, USA, <sup>4</sup> Department of Food Science, Gachon University, Seoul, South Korea

Edited by: Giovanna Suzzi, Università degli Studi di Teramo, Italy

#### Reviewed by:

Antonio Valero, University of Cordoba, Spain Alexandra Lianou, Agricultural University of Athens, Greece

> \*Correspondence: Yi Chen yi.chen@fda.hhs.gov

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 29 March 2016 Accepted: 05 May 2016 Published: 18 May 2016

#### Citation:

Chen Y, Allard E, Wooten A, Hur M, Sheth I, Laasri A, Hammack TS and Macarisin D (2016) Recovery and Growth Potential of Listeria monocytogenes in Temperature Abused Milkshakes Prepared from Naturally Contaminated Ice Cream Linked to a Listeriosis Outbreak. Front. Microbiol. 7:764. doi: 10.3389/fmicb.2016.00764 The recovery and growth potential of Listeria monocytogenes was evaluated in three flavors of milkshakes (vanilla, strawberry, and chocolate) that were prepared from naturally contaminated ice cream linked to a listeriosis outbreak in the U.S. in 2015, and were subsequently held at room temperature for 14 h. The average lag phase duration of L. monocytogenes was 9.05 h; the average generation time was 1.67 h; and the average population level increase per sample at 14 h was 1.14 log CFU/g. Milkshake flavors did not significantly affect these parameters. The average lag phase duration of L. monocytogenes in milkshakes with initial contamination levels ≤ 3 CFU/g (9.50 h) was significantly longer (P < 0.01) than that with initial contamination levels > 3 CFU/g (8.60 h). The results highlight the value of using samples that are contaminated with very low levels of L. monocytogenes for recovery and growth evaluations. The behavior of L. monocytogenes populations in milkshakes prepared from naturally contaminated ice cream linked to the listeriosis outbreak should be taken into account when performing risk based analysis using this outbreak as a case study.

Keywords: L. monocytogenes, ice cream, milkshake, growth potential, temperature abuse, outbreak

# INTRODUCTION

In March 2015, a listeriosis outbreak (outbreak I) was reported from a hospital (hospital X) involving five highly susceptible elderly patients who were hospitalized for other medical conditions prior to exposure to Listeria monocytogenes (Centers for Disease Control and Prevention, 2015). DNA fingerprinting [pulsed field gel electrophoresis (PFGE) and whole genome sequencing (WGS)] of L. monocytogenes isolates from these patients and various ice cream products linked four patients to the consumption of three flavors of milkshakes (vanilla, chocolate, and strawberry) served at lunch or dinner in hospital X and prepared with the contaminated

ice cream scoops that were produced in production line A (Centers for Disease Control and Prevention, 2015; Karl Klontz, personal communication). It is expected that very little postcontamination growth occurred in ice cream because this product is kept frozen all along the production and distribution chain. Thus, enumeration of L. monocytogenes in these products would provide a relatively clear estimation of the amount of L. monocytogenes consumed by the case patients and shed some light on the risk associated with L. monocytogenes contamination and its infective dose. Therefore, Chen (2015) enumerated the levels of L. monocytogenes in 2,320 individually wrapped scoops of ice cream produced in production line A. The enumeration study demonstrated that L. monocytogenes was present in 99.4% of all tested products produced between November 2014 and March 2015, and among them L. monocytogenes was present in 100% of tested products produced between November 2014 and January 2015. The levels of L. monocytogenes in individual scoops were below 20 and 50 most probable number (MPN)/g in 92.3 and 98.4% of tested scoops, respectively; and they were homogeneously low among different production dates with a geometric mean concentration of 0.15 to 7.1 MPN/g (Chen, 2015).

However, the case patients consumed milkshakes prepared from contaminated ice cream scoops, and thus, the enumeration data cannot be directly used for risk assessment without taking into account the possible growth of L. monocytogenes in milkshakes prior to human consumption. Another listeriosis outbreak (outbreak II) linked to contaminated ice cream products occurred in another state in 2015 (Rietberg et al., 2015), involving two patients from a hospital (hospital Y) who were served milkshakes prepared from contaminated ice cream. No evidence of temperature abuse of milkshakes in hospital X or Y was reported (Centers for Disease Control and Prevention, 2015; Rietberg et al., 2015), and no samples were available from outbreak II (Rietberg et al., 2015). Milkshake is a commonly consumed commodity, and these two outbreaks called for studies on the behavior of L. monocytogenes in milkshakes prepared from ice cream products, especially those naturally contaminated with L. monocytogenes. However, such studies have never been reported. We were able to obtain ice cream samples produced from the production line that was implicated in outbreak I, and therefore, the objective of the present study was to evaluate the recovery and growth potential of L. monocytogenes in milkshakes prepared from naturally contaminated ice cream scoops linked to outbreak I. Assuming an extreme scenario that a milkshake was unintentionally left at room temperature for an extended period of time, we evaluated the recovery and growth of L. monocytogenes in the course of 14 h at room temperature.

#### MATERIALS AND METHODS

#### Ice Cream Samples

Individually wrapped scoops of ice cream produced from production line A were made available by the company. The scoops produced between November 2014 and January 2015 were used to prepare the milkshakes.

#### Preparation of Milkshakes

Vanilla, strawberry and chocolate milkshakes were prepared using the recipe from hospital X. Vanilla milkshakes were made from two scoops of ice cream (80–85 g/scoop) and 118 mL of 1% fat milk. Strawberry milkshakes were made from two scoops of ice cream, 118 mL of 1% fat milk and 15 mL of strawberry syrup. Chocolate milkshakes were made from two scoops of ice cream, 118 mL of 1% fat milk and 15 mL of chocolate syrup. Milk, strawberry syrup, and chocolate syrup were purchased from a local supermarket. Ice cream scoops were briefly left at room temperature to soften, and milkshakes were then prepared using a sterilized commercial drink mixer (Model, HMD200, Hamilton Beach Inc., Southern Pines, NC, USA). For each milkshake flavor, 10 milkshake samples were prepared from randomly picked ice cream scoops, and held at room temperature (22.5 ± 0.5◦C). To facilitate thorough mixing of the milkshakes, the entire portion of milkshakes were aseptically transferred to sterilized stainless steel laboratory blenders (Model, Waring <sup>R</sup> 7011S, Conair Corporation, East Windsor, NJ, USA) after preparation, and blended prior to each hourly sampling. The temperature increases of 10 randomly selected milkshakes used for growth curve construction were measured every half hour using thermometers certified by U.S. National Institute of Standards and Technology (Cat. #ACC10033BLSFC, Thermo Fisher Scientific Inc., Waltham, MA, USA). Ten additional vanilla milkshakes were prepared as described above, aseptically transferred to plastic cups and incubated at room temperature. Their temperatures were monitored to investigate the effect of container material and the absence of blending on the temperature increase. These 10 additional samples were not used to evaluate the growth potential of L. monocytogenes and thus were not subject to hourly blending.

## Growth Curves

Enumeration of L. monocytogenes in milkshakes was conducted hourly in the course of 14 h. During the first 10 h of sampling 2 g of each milkshake was directly plated onto 5 RAPID'L. mono agar (Cat. No. 3563694, Bio-Rad Laboratories, Hercules, CA, USA) plates (400 µl/plate) using the easySpiral <sup>R</sup> automatic spiral plater (Interscience, Inc., France) set to constant volume plating, and this plating yielded a limit of detection (LOD) of 0.5 CFU/g. After 10 h, 1 g of samples was directly plated onto 5 RAPID'L. mono agar plates (200 µl/plate), and this plating yielded a LOD of 1 CFU/g. All plates were left to dry before incubation. Representative colonies were confirmed according to the L. monocytogenes chapter of the FDA Bacteriological Analytical Manual (Hitchins et al., 2016). The L. monocytogenes level change (in CFU/g) of each sample was then used to construct the growth curve. The lag phase was determined as the time for the initial population level to increase twofold (Buchanan and Solberg, 1972). The data were not transferred to log values due to very low values (<10 CFU/g) of most of the data points. The generation time was calculated as the time required for cells to double during the exponential phase.

#### Statistical Analysis

fmicb-07-00764 May 14, 2016 Time: 12:18 # 3

Comparison of lag phase durations and generation times of L. monocytogenes in different flavors of milkshakes with different initial levels of L. monocytogenes were performed using one way ANOVA analysis (Bewick et al., 2004) or t-test (Whitley and Ball, 2002).

### RESULTS AND DISCUSSION

The growth evaluation was performed on naturally contaminated ice cream products that had varying initial levels of L. monocytogenes. Combining results from milkshakes of all three flavors, the initial L. monocytogenes levels ranged from 1 to 20.5 CFU/g with 50% (15/30) of the milkshakes having initial L. monocytogenes levels ≤ 3 CFU/g and 76.7% (23/30) of the milkshakes having initial L. monocytogenes levels ≤ 5 CFU/g. The final L. monocytogenes population after 14 h of incubation at 22.5◦C ranged from 9 to 422 CFU/g with 73.3% (22/30) of the milkshakes having final L. monocytogenes levels ≤ 100 CFU/g and 93.3% (28/30) of the milkshakes having final L. monocytogenes levels ≤ 150 CFU/g. Two samples having initial L. monocytogenes levels of 12 and 20.5 CFU/g yielded final levels of 233 and 422 CFU/g, respectively. The average level increase per sample at 14 h was 1.23 ± 0.26 (average ± standard deviation), 1.01 ± 0.19, and 1.19 ± 0.15 log CFU/g for vanilla, strawberry, and chocolate milkshakes, respectively, with an average of 1.14 ± 0.22 log CFU/g for all milkshakes (**Table 1**). This indicated that the milkshakes were in the early stage of the exponential phase, and therefore, the growth curve was presented as the change of CFU/g and lag phase duration was determined as the time required for initial cell level to increase twofold (**Figure 1**; Buchanan and Solberg, 1972). The average lag phase durations of L. monocytogenes were 8.85 ± 0.78, 9.50 ± 0.82, and 8.80 ± 0.95 h for vanilla, strawberry, and chocolate milkshakes, respectively (**Table 1**; **Figure 1**), which were not statistically different (P > 0.05) from each other. The average lag phase duration was 9.05 ± 0.88 h for all milkshakes (**Table 1**). The average generation times of L. monocytogenes were 1.65 ± 0.39, 1.79 ± 0.36, and 1.57 ± 0.18 h for vanilla, strawberry, and chocolate milkshakes, respectively (**Table 1**), which were not statistically different (P > 0.05) from each other. The average generation time was 1.67 ± 0.33 h for all milkshakes. Because of the relatively long lag phase, no growth of L. monocytogenes was observed for 7.30 ± 0.79, 8.05 ± 0.96, and 7.40 ± 0.70 h for vanilla, strawberry, and chocolate milkshakes, respectively (**Table 1**; **Figure 1**). To our knowledge, this is the first report of the recovery and growth potential of L. monocytogenes in milkshakes prepared with ice cream products that were naturally contaminated with very low levels of bacteria and produced from a production line that was implicated in a listeriosis outbreak. The lag phase duration and generation time of L. monocytogenes determined in the present study contributed to a better understanding of the behavior of L. monocytogenes in the temperature abused milkshake prepared with naturally contaminated ice cream. The data can be combined with previously generated enumeration data for risk based characterization of L. monocytogenes contamination in this commodity. There is a possibility that the average generation time of the entire exponential phase may be different from the generation time observed in the present study; however, we chose to calculate the lag phase duration and generation time within 14 h of incubation because that value was relevant in assessing the recovery and growth of L. monocytogenes, had the milkshakes been unintentionally left at room temperature during the day of serving. A probabilistic analysis between direct plating and MPN for the ice cream samples showed agreement of the two methods in 96.1% of the samples with direct plating providing an underestimate in 0.8% of the samples and MPN providing an underestimate in 3.1% of the samples. This indicated that the injury status of L. monocytogenes in these ice cream products did not prevent the cells from recovering and growing on RAPID'L. mono agar (Chen, 2015). WGS analysis using the Center for Food Safety and Applied Nutrition (CFSAN) Single Nucleotide Polymorphism (SNP) pipeline (Pettengill et al., 2014; Davis et al., 2015) showed that the randomly picked isolates from these ice cream products all matched the clinical specimen of the patients in hospital X (unpublished data).

It is very important to keep in mind that during the first few hours of the exposure to room temperature, the temperature of the milkshakes was low. Milkshakes were −2 to 0◦C when freshly prepared and reached 17 to 19◦C after being held at room temperature for 3 h (**Figure 2**) in stainless steel cups, and thus, after milkshakes reached room temperature, it took L. monocytogenes another 6.05 h to reach the exponential phase. Milkshakes in plastic cups that were not subject to hourly blending reached 17 to 19◦C after 4 h at room temperature. Therefore the use of stainless steel container and hourly blending facilitated a slightly faster temperature increase of milkshakes than plastic cups without hourly blending.

Studies which examined growth kinetics of L. monocytogenes in various enrichment broths and foods under various conditions often had the technical limitation (Gnanou Besse et al., 2006) set by the sensitivity of enumeration methods. In some cases, the variability in the enumeration of low levels of bacterial cells

TABLE 1 | Lag phase duration, generation time, time before growth was observed, level increase per sample for the 30 milkshakes (10 vanilla milkshakes, 10 strawberry milkshakes, and 10 chocolate milkshakes) analyzed in the present study.


FIGURE 1 | Growth curves of 30 milkshake samples (10 vanilla milkshakes, 10 strawberry milkshakes, and 10 chocolate milkshakes) held at 22.5◦C for 14 h. Each line represents the growth curve of a milkshake sample.

could compromise the reliability of any statistical analysis (Duffy et al., 1994). This is probably why quite a few studies used foods artificially inoculated with 10<sup>2</sup> to 10<sup>4</sup> CFU/g or even higher levels of bacteria for the convenience of subsequent enumeration and statistical analysis (Xanthiakos et al., 2006; Panagou and Nychas, 2008; Schvartzman et al., 2014; Huang et al., 2015; Luo et al., 2015). However, these inoculum sizes do not reflect the low bacterial contamination levels usually found in food. Indeed, if the bacterial levels are low, a large number of agar plates or an MPN scheme with a large number of MPN tubes per level and a large number of biological replicates would have to be used to accurately determine the lag phase duration. This would be impractical for hourly monitoring of bacterial growth, especially for an extended period of time. The milkshake samples in the present study could be homogenized with no dilutions in any buffer; and they were viscous and could stay off the edge of the plates. As a result, as much as 400 µl of milkshake could be evenly spread onto one plate with the help of the automatic spiral plater, and this allowed us to plate up to 2 g per sample and to obtain reliable data to perform statistical analysis.

The average lag phase duration of L. monocytogenes in 15 milkshakes having initial L. monocytogenes levels ≤ 3 CFU/g was 9.50 ± 0.76 h, and it was significantly higher (P < 0.01) than that of 15 milkshakes having initial L. monocytogenes levels > 3 CFU/g, which was 8.60 ± 0.78 h. Among the 15 milkshakes having initial L. monocytogenes levels > 3 CFU/g, eight milkshakes had initial L. monocytogenes levels of 3.5 to 5 CFU/g and seven milkshakes had initial L. monocytogenes levels of 5.5 to 20.5 CFU/g. Their average lag phase durations were 8.44 ± 0.78 and 8.79 ± 0.81 h, respectively, which were not significantly different (P > 0.05) from each other. The enumeration study revealed that 7.7, 1.6, and 0.2% of all tested ice cream scoops had L. monocytogenes of more than 20, 50, and 100 MPN/g, respectively (Chen, 2015). The milkshakes were made of two 80–85 g scoops and then diluted in 118 mL of milk, and thus, in order to obtain milkshakes containing L. monocytogenes of more than 30 CFU/g the addition of numeric values of the L. monocytogenes levels in the two scoops needed to exceed 100. The probability of this happening was very low, and therefore, we would not have been able to obtain sufficient biological replicates of milkshakes containing L. monocytogenes of more than 30 CFU/g even if we had prepared a much larger number of milkshakes. Robinson et al. (2001) showed that when the inoculum levels of L. monocytogenes were below 100 to 1,000 cells per sample, the lag time increased as the inoculum size decreased, especially under suboptimal growth conditions, partially due to the variability in the lag phase of individual cells; and the effect of inoculum size on lag phase duration started to disappear when the inoculum levels were higher. Aguirre et al. (2013) found that under experimental conditions, the lag phase durations of L. innocua with initial levels of 0.7 to 20 CFU/sample were higher than those with initial levels of 20 to 200 CFU/sample, and there was no statistical difference in lag phase durations among populations with initial levels between 20 and 200 CFU/sample. This phenomenon was also observed by Baranyi (1998) and Pin and Baranyi (2006) in other bacteria. The results from different studies may not be directly compared due to differences in the Listeria species/strains, growth media, and applied methodologies, however, they could partially explain why we observed a longer lag phase duration (P < 0.01) of cells in milkshakes with initial levels ≤ 3 CFU/g than that of cells with initial levels between 3 and 20.5 CFU/g, and no significant difference (P > 0.05) in lag phase durations between cells with initial levels of 3.5 to 5 CFU/g and those with initial

levels of 5.5 to 20.5 CFU/g. The finding in the present study, combined with those reported previously, suggests that under certain growth conditions the average lag phage duration of cells that are sparsely distributed in foods could be significantly longer than that indicated from foods inoculated with high levels of inoculum.

No studies have been performed on the behavior of L. monocytogenes in milkshakes made from ice cream, but a few studies analyzed artificially inoculated dairy products. Alavi et al. (1999) reported that the generation time of L. monocytogenes with an initial inoculum level of 1,000 CFU/mL in liquid whole milk stored at 22.5◦C was 1.40 h. Xanthiakos et al. (2006) reported that with an initial inoculum level of 10<sup>3</sup> to 10<sup>4</sup> CFU/mL in pasteurized milk stored at 16◦C, the lag phase duration of L. monocytogenes was 7.12 h and the generation time was 1.27 h. Panagou and Nychas (2008) found that with an initial inoculum level of 100 CFU/g, L. monocytogenes in a vanilla cream dessert-type product held at 15◦C had an average lag phase duration of 7.85 h and an average generation time of 1.45 h. Gougouli et al. (2008) studied the behavior of L. monocytogenes in artificially inoculated ice cream extensively under multiple chilling and freezing conditions. A pro-longed (500 h) chilling-freezing experiment monitoring the recovery and growth of L. monocytogenes every 20 to 40 h demonstrated that L. monocytogenes cells in ice cream did not suffer significant injury in freezing conditions up to −30◦C, and during consecutive freezing and thawing cycles. This suggested that the relatively long lag phase and generation time determined in the present study were not direct results of cell injury. Due to the overall low prevalence of L. monocytogenes in foods, most studies that used naturally contaminated samples could only yield a small number of positive samples and thus not enough biological replicates could be generated (Jorgensen and Huss, 1998; Encinas et al., 1999; Lappi et al., 2004; Beaufort et al., 2007). The 100% prevalence rate of L. monocytogenes (Chen, 2015) in the ice cream samples used in the present study allowed the use of a large number of naturally contaminated biological replicates to generate the unique and highly confident research data.

#### REFERENCES


#### CONCLUSION

This is the first report on the characterization of the behavior of L. monocytogenes in milkshakes prepared from the ice cream products naturally contaminated with low levels of L. monocytogenes. This provides relevant information for future risk assessment using the 2015 U.S. ice cream listeriosis outbreak as a case study, especially since the ice cream scoops used in the present study were produced from the production line that was implicated in the outbreak. The results demonstrated the value of using samples contaminated with very low levels of L. monocytogenes to perform recovery and growth evaluations of L. monocytogenes in foods. It is important to keep in mind that the rather long lag phase and generation time observed only pertained to the strains/samples in the present study, and our conclusions by no means undermine the critical importance of stringent temperature control by food handlers.

### AUTHOR CONTRIBUTIONS

YC is the principle investigator (PI) of the project and wrote the manuscript. TH and DM are co-PIs of the project. EA, AW, MH, IS, and AL are research scientists working on the experiments.

#### ACKNOWLEDGMENTS

This research was funded in part by the University of Maryland Joint Institute for Food Safety and Applied Nutrition through a cooperative agreement with the U.S. Food and Drug Administration (FDA), #FDU001418, and in part by an appointment to the Research Participation Program at the FDA Center for Food Safety and Applied Nutrition, administered by the Oak Ridge Institute for Science and Education through an interagency agreement between the U.S. Department of Energy and the U.S. FDA.


growth in structured media. Int. J. Food Microbiol. 110, 43–51. doi: 10.1016/j.ijfoodmicro.2006.03.002


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the U.S. FDA.

Copyright © 2016 Chen, Allard, Wooten, Hur, Sheth, Laasri, Hammack and Macarisin. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Biofilm-Forming Capacity in Biogenic Amine-Producing Bacteria Isolated from Dairy Products

Maria Diaz, Victor Ladero\*, Beatriz del Rio, Begoña Redruello, María Fernández, M. Cruz Martin and Miguel A. Alvarez

Department of Technology and Biotechnology of Dairy Products, Instituto de Productos Lácteos de Asturias (IPLA-CSIC), Villaviciosa, Spain

Biofilms on the surface of food industry equipment are reservoirs of potentially food-contaminating bacteria—both spoilage and pathogenic. However, the capacity of biogenic amine (BA)-producers to form biofilms has remained largely unexamined. BAs are low molecular weight, biologically active compounds that in food can reach concentrations high enough to be a toxicological hazard. Fermented foods, especially some types of cheese, accumulate the highest BA concentrations of all. The present work examines the biofilm-forming capacity of 56 BA-producing strains belonging to three genera and 10 species (12 Enterococcus faecalis, 6 Enterococcus faecium, 6 Enterococcus durans, 1 Enterococcus hirae, 12 Lactococcus lactis, 7 Lactobacillus vaginalis, 2 Lactobacillus curvatus, 2 Lactobacillus brevis, 1 Lactobacillus reuteri, and 7 Lactobacillus parabuchneri), all isolated from dairy products. Strains of all the tested species - except for L. vaginalis—were able to produce biofilms on polystyrene and adhered to stainless steel. However, the biomass produced in biofilms was strain-dependent. These results suggest that biofilms may provide a route via which fermented foods can become contaminated by BA-producing microorganisms.

#### Edited by:

Giovanna Suzzi, University of Teramo, Italy

#### Reviewed by:

Giorgia Perpetuini, University of Teramo, Italy Fausto Gardini, University of Bologna, Italy

> \*Correspondence: Victor Ladero ladero@ipla.csic.es

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 24 February 2016 Accepted: 11 April 2016 Published: 12 May 2016

#### Citation:

Diaz M, Ladero V, del Rio B, Redruello B, Fernández M, Martin MC and Alvarez MA (2016) Biofilm-Forming Capacity in Biogenic Amine-Producing Bacteria Isolated from Dairy Products. Front. Microbiol. 7:591. doi: 10.3389/fmicb.2016.00591 Keywords: biofilm, biogenic amines, histamine, tyramine, putrescine, lactic acid bacteria, dairy industry

# INTRODUCTION

Food safety is a major priority in today's food industry. Bacterial biofilms on industrial surfaces are a cause for concern since they may act as reservoirs of contaminating microorganisms (Winkelströter et al., 2014). Dairy products in particular are susceptible to such contamination (Srey et al., 2013), with equipment surfaces one of its major sources (Kumar and Anand, 1998). Stainless steel type 304 is the most common material in contact with food in the dairy industry (Zottola and Sasahara, 1994). While inert, easy to clean and highly resistant to corrosion, it can, however, develop small cracks and crevices where biofilm formation is facilitated (Winkelströter et al., 2014). In addition, some parts of food processing equipment may have inaccessible areas where bacteria can evade cleaning treatments. The main biofilm-related risk is the growth of pathogens and spoilage microorganisms.

The capacity of biogenic amine (BA)-producers to form biofilms has not been investigated. BAs are low-molecular weight organic compounds derived from their corresponding amino acids via enzymatic decarboxylation. Although, BAs play an important role in human physiology, the ingestion of food containing them in large quantities can have toxicological effects on the digestive, circulatory, and respiratory systems (ten Brink et al., 1990; Shalaby, 1996; Ladero et al., 2010a). The most important BAs (both qualitatively and quantitatively) in foods and beverages are histamine, tyramine, and putrescine. Together with fish and wine, dairy products—especially cheese can develop BA concentrations that may exceed 1000 mg kg−<sup>1</sup> (Linares et al., 2012).

BAs form in food via the activity of bacteria with aminoacyl decarboxylase activity (Halasz et al., 1994). Their appearance in dairy products has mainly been attributed to Gram positive bacteria of the lactic acid bacteria (LAB) group. These can be present in the microbiota of milk, as part of the starter culture, or be introduced by contamination during manufacturing (Linares et al., 2011), with equipment surfaces a potentially important source of contamination (Novella-Rodríguez et al., 2004). The post-ripening processing of cheese, particularly grating, extends the contact of food with equipment surfaces, increasing the number of histamine-producing bacteria present in the final product (Ladero et al., 2009) and therefore the histamine concentration that may be reached.

The dairy histamine-producing species Lactobacillus parabuchneri has been reported to produce biofilms (Diaz et al., 2016b), but this capacity has not been studied in other BA producers. The aim of the present work was to test the ability of BA-producing bacteria isolated from different cheeses to form biofilms on polystyrene and adhere to stainless steel surfaces.

#### MATERIALS AND METHODS

#### Bacterial Strains and Culture Conditions

In this work we have examined the biofilm-forming capacity of 56 BA-producing strains belonging to three genera and 10 species (12 Enterococcus faecalis, 6 Enterococcus faecium, 6 Enterococcus durans, 1 Enterococcus hirae, 12 Lactococcus lactis, 7 Lactobacillus vaginalis, 2 Lactobacillus curvatus, 2 Lactobacillus brevis, 1 Lactobacillus reuteri, and 7 Lactobacillus parabuchneri), all isolated from dairy products. The ability of all (except of the Lactobacillus parabuchneri strains) to produce BAs was known from previous work (Ladero et al., 2010b,c, 2011a,b, 2012a,b; Diaz et al., 2015; del Rio et al., 2015). Given the reported ability of L. parabuchneri to form biofilms (Diaz et al., 2016b), and the apparent importance of the species in the accumulation of histamine in cheese (Diaz et al., 2016a), seven new strains isolated from cheese (following the protocol of Diaz et al., 2015) were included among those examined. These were identified at the species level by 16S rRNA sequencing (Diaz et al., 2015) and their ability to produce histamine checked by UHPLC (Redruello et al., 2013).

Lactobacilli were grown at 37◦C in MRS (Oxoid, Basingstoke, UK), while enterococci and lactococci were grown at 30◦C in M17 (Oxoid) supplemented with 0.5% (w/v) glucose (GM17). To test the individual capacity of the L. parabuchneri strains to produce histamine, the culture medium was supplemented with 5 mM histidine.

# Analysis of Biofilm Formation on Polystyrene

The ability of the test strains to produce a biofilm on polystyrene was performed as described by Diaz et al. (2016b). Briefly, MRS or GM17 overnight cultures were diluted to approximately 10<sup>6</sup> cfu mL−<sup>1</sup> with fresh medium and used to fill polystyrene 96-well microtitre plates (NuncTM MicroWellTM Plates with a NunclonTM Delta Surface; Thermo Fisher Scientific, Waltham, MA, USA). Negative controls consisted of wells filled with the corresponding uninoculated culture medium. All plates were then incubated at 30 or 37◦C depending on the species. Biofilm biomass was determined using the crystal violet staining method (CV assay) (Kubota et al., 2008). After 24, 30, or 48 h of incubation, the supernatant was removed and the wells rinsed with PBS buffer to eliminate non-adhered cells. The potential biofilm present was then stained with 0.5% (w/v) CV in distilled water (dH2O); the excess dye was removed with dH2O. The bound dye was then extracted using acetone/ethanol (80:20, v:v) and quantified by absorbance at 595 nm using a Benchmark Plus microplate spectrophotometer (BioRad, Hercules, CA, USA). The mean ± SD of the optical density (OD) of three replicates was calculated for each strain. ANOVA with post-hoc Bonferroni correction was used to analyse all data. Significance was set at p < 0.05. All statistical calculations were undertaken using SPSS v.15.0 software (SPSS Inc., IL, USA). Biofilm production capacity was expressed using cut-off values (Extremina et al., 2011). The cutoff value between biofilm-producers and non-producers (ODc) was defined as the mean of the negative controls (ODnc) plus three SDs. The strains were then classified as belonging to one of the following categories: ODc < OD≤ 2 × ODc = weak biofilm producer, 2 × ODc < OD ≤ 4 × ODc = moderate biofilm producer, and OD > 4 × ODc = strong biofilm producer.

#### Analysis of Bacterial Adherence to Stainless Steel

The test surfaces used were 1 cm<sup>2</sup> stainless steel (type AISI 304) coupons. These were washed with soap and dH2O, rinsed with dH2O, and then immersed in acetone for 30 min to remove any grease or fingerprints. They were then rinsed once again in dH2O, autoclaved, and immersed singly in tubes containing MRS or GM17 broth inoculated with 10<sup>6</sup> cfu mL−<sup>1</sup> of the assayed strain (performed in triplicate). Each coupon was then incubated at 30 or 37◦C for 24 or 48 h before removal using sterile forceps. Non-adhered cells were removed by rinsing the coupon three times in PBS buffer. The coupons were then reimmersed in 5 mL PBS buffer, and the adhered cells detached from the coupon by sonication in an ultrasonic bath (Ultrasons-H, Selecta, Spain) for 15 min. The bacterial suspension produced was serially diluted in PBS, and 100 µL of 10<sup>0</sup> , 10−<sup>1</sup> , 10−<sup>2</sup> , 10−<sup>3</sup> , and 10−<sup>4</sup> dilutions were plated on MRS or GM17 and incubated for 48 h (Kruszewski et al., 2013). Three replicates were performed for each strain using independent bacterial cultures. Bacterial counts were expressed as log<sup>10</sup> cfu cm−<sup>2</sup> (mean ± SD of three replicates). To confirm the tolerance of the cells to sonication, bacterial suspensions of all the examined strains were sonicated for 15 min. Pre- and post-sonication suspensions were serially diluted in PBS, plated, incubated for 48 h, and the cells enumerated; no significant differences were seen between preand post-sonication cell counts confirming that no cells were killed by this procedure.

# Scanning Electron Microscopy of Cells Adhered to Stainless Steel

The method of Kubota et al. (2008) was followed, with some modifications, to observe by scanning electron microscopy (SEM) the cells adhered to the stainless steel coupons. Briefly, the latter were rinsed twice in PBS and then fixed in 2.5% glutaraldehyde (Sigma-Aldrich) in PBS for 16 h at room temperature. The fixed bacteria were then dehydrated using a graded series of acetone solutions (50–100% v/v), and the coupons dried with CO<sup>2</sup> using a CPD-030 critical point dryer (Bal-Tec AG, Balzers, Liechtenstein). They were then coated with gold using a SCD 004 Sputtering Coater (Bal-Tec AG, Balzers, Liechtenstein) and observed using a JSM-6610LV SEM (JEOL USA, Inc, Peabody, MA, USA).

# RESULTS

#### Biofilm Formation on Polystyrene

Maximum biofilm biomass values were obtained at different times of incubation depending on the species. For all the Enterococcus strains assayed, biofilm biomass was maximal at 30 h; for all the L. lactis, L. curvatus, and L. brevis strains, maximum values were reached at 24 h; and for the L. reuteri and L. parabuchneri strains, maxima were recorded at 48 h. All incubation time results shown are those at which maximum biomass was reached.

All the E. faecalis strains were able to form biofilms on polystyrene. Six were classified as strong biofilm producers, and six as moderate biofilm producers (**Figure 1**). The only E. hirae strain assayed was a weak biofilm producer (**Figure 1**). All the E. durans strains tested were able to form biofilms; four were strong biofilm producers, and two were weak producers (**Figure 1**). Of the six E. faecium strains analyzed, three were strong biofilm producers, one was a weak producer, and two were unable to form a biofilm (**Figure 1**).

All the L. lactis strains were able to produce biofilms on polystyrene (**Figure 2**). Three L. lactis subsp. cremoris and two L. lactis subsp. lactis were strong biofilm producers. The remaining strains—two L. lactissubsp.cremoris and five L. lactissubsp. lactis strains—were weak producers.

None of the L. vaginalis strains were able to form a biofilm (**Figure 3**). All the L. curvatus, L. brevis, and L. reuteri strains were, however, strong biofilm producers (**Figure 3**). Two out of seven L. parabuchneri strains were strong biofilm producers, while the remaining strains were only weak producers (**Figure 3**).

# Bacterial Adherence to Stainless Steel

The strains selected for this assay were the strongest biofilm producers in the polystyrene surface assay, i.e., E. faecalis 15a, E. hirae 268, E. durans 411, E. faecium AA3, L. lactis subsp. cremoris CECT 8666, L. lactis subsp. lactis 1AA59, L. curvatus VI6, L. brevis 3811, L. parabuchneri IPLA 11150, and L. reuteri IPLA 11078. Although, no L. vaginalis strain was able to form a biofilm on polystyrene, L. vaginalis IPLA 11064 was tested with the steel surface. Two incubation times (24 and 48 h) were tested, but no differences were observed. **Figure 4** shows the number of adhered

cells of each strain after incubation for 48 h (>10<sup>4</sup> cfu cm−<sup>2</sup> for all strains assayed).

## Scanning Electron Microscopy Imaging of Cells

SEM photomicrographs of cells adhered to the stainless steel coupons were captured for the same strains as used in the previous assay. Although, no differences were seen in the counts at different incubation times (24 and 48,h), differences in aggregation type and bacterial ultrastructure were observed.

E. faecalis 15a (**Figure 5A1**), E. faecium AA3 (**Figure 5B1**) and E. durans 411 returned clearer images after 24 h of incubation (data not shown) and appeared uniformly spread on the coupons. No adhering E. hirae 268 cells were observed at either 24 or 48 h. The E. faecalis cells were observed embedded in an extracellular matrix (**Figure 5A2**); this was not observed for the other two species. Structures that might be involved in anchoring to the surface were observed on E. faecium cells (**Figure 5B2**).

L. lactis subsp. cremoris CECT 8666 and L. lactis subsp. lactis 1AA59 returned clearer images after 24 h of incubation. The strain 8666 was uniformly spread across the coupon surface (**Figure 6A1**), while 1AA59 formed more compact aggregates (**Figure 6B1**). In both cases, an extracellular matrix was observed, but with a different appearance (see **Figures 6A2,A3,B2**). Similar structures to those observed in E. faecium, and that might be involved in anchoring to the surface, were also observed for both L. lactis strains (**Figures 6A3,B3**).

L. parabuchneri IPLA 11150, L. reuteri IPLA 11078, and L. brevis 3811 returned clearer images after 48 h of incubation. L. vaginalis IPLA 11064 and L. curvatus VI6 showed no adhered cells at either 24 or 48 h. L. parabuchneri was distributed across the coupon, showing aggregations with an extracellular matrix (**Figures 7A1,A2**). L. reuteri was distributed across the coupons in small aggregations (**Figure 7B1**) in a clear extracellular matrix (**Figure 7B2**). L. brevis was seen only in the fissures of the coupon (**Figures 7C1,C2**).

# DISCUSSION

The formation of BA in fermented foods by bacteria, especially cheese, is a cause of much concern. The industrial equipment used in cheesemaking and processing is a source of contamination, and the ability of microorganisms to adhere to the surfaces of such equipment increases their contamination potential. In this work, the ability of 56 BA-producing dairy LAB strains to adhere to a polystyrene surface was examined using the CV assay, a technique that allows easy quantification of total biofilm biomass. The strongest biofilm producers of each species were then examined for their ability to adhere to stainless steel coupons, and the adhered cells observed by SEM.

The formation of biofilms by clinical Enterococcus isolates has been much studied since it affects pathogenic potential (Langsrud, 2009). However, enterococci are also a cause of concern for the food industry, and the biofilm-forming ability of several food-related Enterococcus species has also been studied (Gomes et al., 2008; Jahan and Holley, 2014; da Silva Fernandes et al., 2015). In the present study, the biofilm-forming capacity of tyramine- or tyramine and putrescine-producing Enterococcus species was examined. The results of the CV assay showed that all the strains analyzed, except for two E. faecium strains, were able to form biofilms, although the total biomass produced differed. These results agree with those of previous studies that showed several foodborne E. faecalis and E. faecium isolates to be either weak, moderate or strong biofilm producers, while some isolates formed no biofilm at all (Gomes et al., 2008; Jahan and Holley, 2014). E. durans has also been described as a strong or moderate biofilm producer (Amel et al., 2015; Pieniz et al., 2015). The only strain of E. hirae analyzed in the present work was a weak biofilm producer; to our knowledge, the capacity of this species to form biofilms has not been previously studied.

All the Enterococcus strains selected to see whether they could adhere to stainless steel did so, with counts reaching 10<sup>4</sup> cfu cm−<sup>2</sup> . The capacity of E. faecalis, E. faecium, and E. durans to adhere to stainless steel has been previously reported (Amel et al., 2015; da Silva Fernandes et al., 2015). The Enterococcus strains that adhered to the stainless steel coupons were observed by SEM, except for E. hirae 268; - this strain attached only weakly to the metal surface and was unable to resist the treatment required prior to observation. Of the strains that covered the surface of the steel coupons (**Figure 5**), none formed the complex three-dimensional structures reported

FIGURE 3 | Biofilm-producing capacity on polystyrene of the biogenic amine-producing Lactobacillus strains incubated at 37◦C. The L. vaginalis, L. curvatus and L. brevis strains were incubated for 24 h. The L. parabuchneri and L. reuteri strains were incubated for 48 h. Data represent means ± SD (error bars) of three experiments. Values marked with the same letter do not differ significantly (p > 0.05 according to the Bonferroni post-hoc test). (\_\_\_) Cut-off line (ODc); (\_\_ \_\_ \_\_) 2 × ODc; (.......) 4 × ODc.

by da Silva Fernandes et al. (2015). Some aggregations of E. faecalis 15a cells were observed (**Figure 5A2**), while E. faecium AA3 cells appeared separated from one another, but all cells were sufficiently well adhered not to be detached by the PBS washes or the treatment required prior to SEM observations. To our knowledge, this is the first time that images of E. durans cells adhered to stainless steel have been captured.

The ability of L. lactis to form a biofilm on a surface has been related to the latter's physicochemical properties (Giaouris et al., 2009; Oxaran et al., 2012). However, few studies have ever been performed on L. lactis biofilm formation. In the present work, 12 putrescine-producing L. lactis strains were shown by the CV assay to be either weak or strong biofilm producers. The two strains that most strongly formed biofim on polystyrene—L. lactis subsp. cremoris CECT 8666 and L. lactis subsp. lactis 1AA59—also adhered to stainless steel, reaching counts of 10<sup>4</sup> cfu cm−<sup>2</sup> . SEM images showed cells of both strains to be embedded in an extracellular matrix and to be clearly adhered to the coupons. This is the first time that L. lactis adhered on stainless steel have been observed by SEM. Unlike that seen for enterococci, the L. lactis CECT 8666 cells showed tridimensional structures including filamentous ones that might help them adhere to the surface (**Figures 6A2,A3**). The extracellular material formed a pod-like covering over the cells (**Figures 6A2,A3**). L. lactis 1AA59 cell clusters also appeared to be attached via the extracellular matrix (**Figure 6B2**). For both L. lactis strains (**Figures 6A3,B3**) examined, and E. faecium AA3 (**Figure 5B2**), small protuberances of the cell surface were observed, which may have helped anchor the cells to the coupons. Certainly, aggregation and biofilm formation in L. lactis has been associated with the production of functional pili (Oxaran et al., 2012). Although, the adhesion of the putrescine-producing strains analyzed in this work would be undesirable for the food industry, allowing the development of L. lactis biofilms has been proposed as a means of preventing the growth of pathogens on food industry surfaces (Leriche et al., 1999; Zhao et al., 2004).

The presence of several Lactobacillus species (L. curvatus, Lactobacillus fermentum, Lactobacillus delbruekii, Lactobacillus paracasei, Lactobacillus plantarum, and L. reuteri) on the surfaces of dairy equipment has been reported (Somers et al., 2001; Scatassa et al., 2015). In the present work, two tyramine and putrescine-producing L. brevis strains were strong biofilm producers on polystyrene. This agrees with previous results for L. brevis strains isolated from onions (Kubota et al., 2008). The two tyramine-producing L. curvatus strains tested were also strong biofilm producers. Biofilm formation by L. curvatus has been previously described, although it was found to be a weaker biofilm producer than in the present work (Pérez Ibarreche et al., 2014). Fifteen histamine-producing strains, seven belonging to L. vaginalis, seven to L. parabuchneri and one to L. reuteri, were examined by the CV assay. The seven histamine-producing L. vaginalis strains were unable to form biofilms. All the L. parabuchneri strains were able to form biofilms, and were either weak or strong biofilm producers, depending on the strain. This agrees with the results of previous reports (Diaz et al., 2016b). The L. reuteri strain examined was a strong biofilm producer. Numerous studies have described the formation of biofilms by L. reuteri, some strains of which are considered probiotics (Leccese Terraf et al., 2014). A system regulating biofilm formation in L. reuteri was recently characterized (Su and Ganzle, 2014). In the present work, L. reuteri IPLA 11078, along with L. vaginalis IPLA 11064, L. curvatus VI6, L. brevis 3811 and L. parabuchneri IPLA 11150 were all able to adhere to the stainless steel coupons, and reached counts of over 10<sup>4</sup> cfu cm−<sup>2</sup> . However, when the coupons were observed under the SEM, only L. parabuchneri

IPLA 11150 (**Figure 7A**) and L. reuteri IPLA 11078 (**Figure 7B**) cells were seen attached. For L. parabuchneri IPLA 11150, the cells aggregated into clumps composed of long chains of undivided cells. A previous study on Pseudomonas aeruginosa showed that elongated cells are inclined to form cohesive clumps (Yoon et al., 2011). According to the CV assay, the present L. parabuchneri and L. reuteristrains were strong biofilm producers. However, while the same assay suggested L. curvatus VI6 and L. brevis 3811 to be strong biofilm producers, the absorbance measured was very close to the lower limit for such classification. The few cells seen attached to the coupons might be the result of them not being able to resist the treatment required prior to SEM observation. This would appear to be supported by the fact that L. brevis 3811 was not homogenously attached to the surface of the coupons, but seen in their fissures. This is an important finding for the food industry since steel surfaces do develop cracks and these appear able to protect bacteria from cleaning procedures.

The present results show all the species examined, except L. vaginalis, had BA-producing strains able to form biofilms and that they could adhere to stainless steel, a material commonly used to make equipment in the food industry. Although, the cells attached to the stainless steel coupons commonly showed none of the three-dimensional structures reported by da Silva Fernandes et al. (2015), counts of over 10<sup>4</sup> cfu cm−<sup>2</sup> were recorded in all cases, with E. faecalis, E. faecium, E. durans, and L. parabuchneri exceeding 10<sup>5</sup> cfu cm−<sup>2</sup> . Given the large surface area of industrial equipment, BA-producing bacteria that can adhere to steel clearly pose a food contamination threat. This problem is of particular concern in the dairy industry since post-ripening treatments such as cheese grating bring food into close and prolonged contact with equipment surfaces. In fact, it has already been shown that cheese that has undergone post-ripening processing (cutting, slicing, or grating) has higher histamine levels than non-processed cheese (Ladero et al., 2009).

It is therefore important that our knowledge of the adhesion and biofilm forming capacities of BA-producers be improved, to prevent food contamination by these spoilage bacteria and eventually the accumulation of biogenic amines in food.

# AUTHOR CONTRIBUTIONS

MD carried out the experiments and drafted the manuscript; BR, VL collaborated in conducting some experiments; VL, Bd, and MF participated in the study design and writing of the manuscript; MC, MA provided the general concept, designed the experiments, and supervised the experimental work and the manuscript. All authors contributed to the discussion of the research and approved the final manuscript.

# ACKNOWLEDGMENTS

The authors are very grateful to Prof. Christian Magni for providing some of the strains used in this work, the Unidad de Microscopía Electrónica y Microanálisis of the University of Oviedo for assistance with the SEM techniques, and Adrian Burton for language assistance. This work was funded by the Spanish Ministry of the Economy and Competitiveness (AGL2013-45431-R) and the Plan for Science, Technology and Innovation 2013–2017 financed by the European Regional Development Fund and the Principality of Asturias (GRUPIN14-137). MD is a beneficiary of an FPI fellowship from the Spanish Ministry of the Economy and Competitiveness.

# REFERENCES


dairy environment. J. Dairy Sci. 84, 1926–1936. doi: 10.3168/jds.S0022- 0302(01)74634-6


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Diaz, Ladero, del Rio, Redruello, Fernández, Martin and Alvarez. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Prevalence, Molecular Characterization, and Antibiotic Susceptibility of *Vibrio parahaemolyticus* from Ready-to-Eat Foods in China

#### Tengfei Xie1, 2, Xiaoke Xu<sup>2</sup> , Qingping Wu<sup>2</sup> \*, Jumei Zhang<sup>2</sup> and Jianheng Cheng<sup>2</sup>

*<sup>1</sup> School of Bioscience and Bioengineering, South China University of Technology, Guangzhou, China, <sup>2</sup> Guangdong Institute of Microbiology, State Key Laboratory of Applied Microbiology Southern China: Guangdong Provincial Key Laboratory of Microbial Culture Collection and Application, Guangdong Open Laboratory of Applied Microbiology, Guangzhou, China*

#### *Edited by:*

*Maria Schirone, University of Teramo, Italy*

#### *Reviewed by:*

*Aoife Boyd, National University of Ireland Galway, Ireland Learn-Han Lee, Monash University Malaysia, Malaysia*

> *\*Correspondence: Qingping Wu wuqp203@163.com*

#### *Specialty section:*

*This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology*

*Received: 16 January 2016 Accepted: 04 April 2016 Published: 21 April 2016*

#### *Citation:*

*Xie T, Xu X, Wu Q, Zhang J and Cheng J (2016) Prevalence, Molecular Characterization, and Antibiotic Susceptibility of Vibrio parahaemolyticus from Ready-to-Eat Foods in China. Front. Microbiol. 7:549. doi: 10.3389/fmicb.2016.00549* *Vibrio parahaemolyticus* is the leading cause of foodborne outbreaks, particularly outbreaks associated with consumption of fish and shellfish, and represents a major threat to human health worldwide. This bacterium harbors two main virulence factors: the thermostable direct hemolysin (TDH) and TDH-related hemolysin (TRH). Additionally, various serotypes have been identified. The extensive use of antibiotics is a contributing factor to the increasing incidence of antimicrobial-resistant *V. parahaemolyticus.* In the current study, we aimed to determine the incidence and features of *V. parahaemolyticus* in ready-to-eat (RTE) foods in China. We found 39 *V. parahaemolyticus* strains on Chinese RTE foods through investigation of 511 RTE foods samples from 24 cities in China. All isolates were analyzed for the presence of *tdh* and *trh* gene by PCR, serotyping was performed using multiplex PCR, antibiotic susceptibility analysis was carried out using the disk diffusion method, and molecular typing was performed using enterobacterial repetitive intergenic consensus sequence PCR (ERIC-PCR) typing and multilocus sequence typing (MLST). The results showed that none of the isolates were positive for *tdh* and *trh.* Most of the isolates (33.3%) were serotype O2. Antimicrobial susceptibility results indicated that most strains were resistant to streptomycin (89.7%), cefazolin (51.3%), and ampicillin (51.3%). The isolates were grouped into five clusters by ERIC-PCR and four clusters by MLST. We updated 10 novel loci and 33 sequence types (STs) in the MLST database. Thus, our findings demonstrated the presence of *V. parahaemolyticus* in Chinese RTE foods, provided insights into the dissemination of antibiotic-resistant strains, and improved our knowledge of methods of microbiological risk assessment in RTE foods.

Keywords: *Vibrio parahaemolyticus,* ready-to-eat foods, antibiotic resistance, ERIC-PCR, MLST, serotype, virulence gene

# INTRODUCTION

Vibrio parahaemolyticus is a gram-negative halophilic bacterium that naturally occurs worldwide in estuarine environments. This microorganism is recognized as one of the leading causes of foodborne illness worldwide and has been shown to cause acute gastroenteritis in humans. Complications such as septicemia can sometimes lead to death in patients with V. parahaemolyticus infection (Qadri et al., 2005; Lopatek et al., 2015). Previous studies have focused on the prevalence of V. parahaemolyticus in shellfish, oysters, water, seafood, and shrimp (Khouadja et al., 2013; Lopatek et al., 2015). However, no reports have identified isolates of V. parahaemolyticus on ready-to-eat (RTE) foods such as cooked meat, roasted poultry, and cold vegetable dishes in sauce, that are highly popular in China. Various foodborne pathogens may be present in RTE foods and may cause illnesses in consumers because RTE foods do not require heat treatment or other forms of curing before eating (Shi et al., 2015). Moreover, the high genetic diversity on RTE foods can facilitate identification of strain relatedness and epidemiological investigations.

Traditionally, V. parahaemolyticus is considered susceptible to antimicrobials. However, during the past few decades, antimicrobial resistance has emerged and evolved in many bacterial genera owing the excessive use of antimicrobials in human, agriculture, and aquaculture systems (Cabello, 2006; Kang et al., 2016). For example, tetracyclines are recommended as antibiotics in the treatment of severe Vibrio infections, and thirdgeneration cephalosporins with doxycycline or fluoroquinolone alone are also used on occasion. Moreover, antibiotics are commonly used to treat fish (Han et al., 2007; Devi et al., 2009). Some V. parahaemolyticus isolates from seafood and other environments are commonly resistant to ampicillin, aminoglycosides (streptomycin and gentamicin), ciprofloxacin, chloramphenicol, and other antibiotics (Oh et al., 2011; Raissy et al., 2012; Shaw et al., 2014; Lopatek et al., 2015). Therefore, the potential presence of antibiotic-resistant V. parahaemolyticus on RTE foods may be an important public health problem related to disease management and control.

The virulence of V. parahaemolyticus is mainly attributed to the production of two major factors: thermo-stable direct hemolysin (TDH) encoded by the tdh gene and TDH-related hemolysin encoded by the trh gene (Honda and Iida, 1993). Clinical strains commonly contain either these genes, and the presence of these genes is associated with the pathogenicity of the strain in humans (Su and Liu, 2007; Jones et al., 2012; Pazhani et al., 2014). Detection of V. parahaemolyticus virulence factors is typically based on molecular biological analysis and amplification of V. parahaemolyticus-specific sequences (Shirai et al., 1990; West et al., 2013). To date, there are 13 O-serogroups and over 70 K-serogroups, differentiated on the basis of somatic (O) and capsular (K) antigens in V. parahaemolyticus (Ishibashi et al., 2000; Jones et al., 2012). The emergence of the first pandemic strain belonging to serovar O3:K6 (Okuda et al., 1997) supported the view that the serotype of V. parahaemolyticus is correlated with virulence. A multiplex polymerase chain reaction (PCR)-based O-antigen serotyping method has been adopted because other antiserum-based approaches are time consuming, expensive and can be associated with the risk of cross-reactions (Chen M. et al., 2012; Xu et al., 2014).

Molecular typing of V. parahaemolyticus, including pulsed field gel electrophoresis (PFGE) (Marshall et al., 1999), repetitive extragenic palindromic sequence PCR (REP-PCR) (Wong and Lin, 2001), and ribotyping (Bag et al., 1999), has been shown to be a useful tool for providing genetic relatedness information (Olive and Bean, 1999). Enterobacterial repetitive intergenic consensus sequence PCR (ERIC-PCR) has been proven useful for subtyping of V. parahaemolyticus strains with highly conserved repetitive intergenic consensus sequences (De Bruijn, 1992; Chen W. et al., 2012). Additionally, multilocus sequence typing (MLST), which is based on sequence analysis of selected housekeeping genes (e.g., recA, gyrB, dnaE, dtdS, pntA, pyrC, and tnaA), is becoming an important method for investigation of the evolution and epidemiology of V. parahaemolyticus owing to its high repeatability (González-Escalona et al., 2008; Banerjee et al., 2014).

V. parahaemolyticus is not frequently found in RTE foods, but is still considered hazardous in humans. Therefore, the objectives of the current study were as follows: (i) to determine the prevalence and contamination level of V. parahaemolyticus in RTE foods in China and (ii) to determine the genetic variation and phenotypic characteristics of V. parahaemolyticus isolates from RTE foods. The information generated in this study will provide insights into the distribution and population of V. parahaemolyticus across Chinese RTE foods and differentiation of V. parahaemolyticus strains on different RTE foods.

# MATERIALS AND METHODS

### Sample Collection

From November 2011 to May 2014, a total of 511 RTE food samples were collected from retail markets in 24 cities, covering most provincial capitals of China (**Figure 1**). The samples consisted of 371 deli meat samples, 97 cold vegetable dishes or noodles in sauce, and 43 fried rice or noodle samples. The samples were placed in sterile sealed plastic bags, transported to the laboratory in a cold box below 4◦C, and analyzed immediately.

#### Qualitative Detection

The bacteriological media used herein, unless indicated, were purchased from Guangdong Huankai Co. Ltd. (Guangzhou, China). For qualitative detection, the V. parahaemolyticus was isolated according to GB 4789.7-2013 for food microbiological examination of V. parahaemolyticus (National Food Safety Standards of China) with minor modifications. In brief, 25 g of each sample was homogenized for 60 s in a stomacher bag (Huankai Co. Ltd., Guangzhou, China) with 225 mL of alkaline peptone water (APW) containing 3% NaCl. Homogenates were incubated at 37◦C for 16−18 h. After incubation, a loopful from the top 1 cm was streaked onto thiosulfate-citrate-bile saltssucrose (TCBS) agar plates and incubated at 37◦C for 18−24 h. Presumptive colonies (green or blue green colonies, 2−3 mm in diameter) were streaked onto Chromogenic Vibrio Medium and

incubated at 37◦C for 24 h. The mauve colonies were selected for further V. parahaemolyticus identification by analysis of oxidase activity, Gram staining, 3.5% NaCl triple sugar iron (TSI) tests, halophilism tests, and API 20E diagnostic strips (Biomerieux Company, France).

#### MPN Enumeration (MPN) of *V. parahaemolyticus*

Enumeration of V. parahaemolyticus in RTE food samples was performed using the MPN method according to the Bacteriological Analytical Manual standard method (Kaysner and Depaola, 2004) and a previous study (Xu et al., 2014).

#### Detection of tdh and *trh* Genes

Detection of the V. parahaemolyticus tdh and trh genes was performed by PCR as described previously (Gutierrez West et al., 2013). The oligonucleotide primers were synthesized by Sangon Biotech (Shanghai, China) (Tdh-F: CTGTCCCTT TTCCTGCCCCCG, Tdh-R: AGCCAGACACCGCTGCCATTG; Trh-F: ACCTTTTCCTTCTCCWGGKTCSG, Trh-F: CCGCTC TCATATGCYTCGACAKT). Each reaction mixture included the following (total volume, 25 µL): 2 × PCR Mix (Qiagen), 12.5 µL; 0.5 µM each primer, dd H2O, 9.5 µL; and DNA template, 1 µL. Both genes were amplified using the following thermalcycling program: denaturation at 95◦C for 5 min; 40 cycles of 95◦C for 1 min, 62◦C for 1 min, and 72◦C for 1 min; and a final extension of 72◦C for 2 min. PCR was conducted in a Bio-Rad PTC-200 Thermal Cycler (Bio-Rad, Hercules, CA, USA). The amplified products were then analyzed electrophoretically on a 2% agarose gel containing Gold View. The images were captured digitally and analyzed using a Gel Image system (Bio-Rad). V. parahaemolyticus strains ATCC 33847 (tdh+) and ATCC 17802 (trh+) were used as positive controls, and distilled water was used as the negative control.

# Multiplex Serotyping PCR

The serotypes of V. parahaemolyticus isolates were identified using the PCR-based O-antigen serotyping technique. The primer concentrations and amplification conditions used were as described previously (Chen M. et al., 2012). The primers used for this assay were synthesized by Sangon Biotech (Shanghai, China). The 12 primer pairs were divided into two groups to amplify target DNA; PCR group 1 was used to detect serogroups O1, O2, O4, O5, O6, and O10, whereas PCR group 2 was used to detect serogroups O3/O13, O7, O8, O9, O11, and O12. The PCR was performed in a 25-µL reaction mixture containing the following: 2 × PCR mix (Dongshen, Guangzhou, China), 12.5 µL; 0.5 µM each primer, dd H2O, 9.5 µL; and DNA template, 1 µL. All amplifications were carried out with the following protocol: 30 cycles of 95◦C for 30 s, 60◦C for 45 s, and 72◦C for 1 min. The thermal cycler was prewarmed to 80◦C before all the reaction tubes were added in order to reduce nonspecific amplification. PCR was conducted in a Bio-Rad PTC-200 Thermal Cycler (Bio-Rad, California, USA). The amplified products were then analyzed electrophoretically on a 2% agarose gel containing Gold View. The images were captured digitally and analyzed using the Gel Image system (Bio-Rad, California, USA). V. parahaemolyticus ATCC 17802 and ATCC 33847 were used as control strains.

#### Antimicrobial Susceptibility

The susceptibility of the V. parahaemolyticus isolates to antibiotics was examined by the disk diffusion method according to the guidelines of the Clinical and Laboratory Standards Institute (CLSI, 2012) and a previous study (Xie et al., 2015). Briefly, Muller-Hinton agar and a panel of 12 antibiotics disks were selected for resistance tests. These 12 antibiotic disks (Oxoid, Hampshire, UK) were ampicillin (10 µg), azithromycin (15 µg), cefazolin (30 µg), cephalothin (30 µg), chloramphenicol (30 µg), ciprofloxacin (5 µg), gentamicin (10 µg), kanamycin (30 µg), nalidixic acid (30 µg), streptomycin (10 µg), trimethoprimsulfamethoxazole (25 µg), and tetracycline (30 µg). The results were expressed as sensitive (S), intermediate (I), or resistant (R) following the methods of the CLSI. Escherichia coli ATCC 25922 and V. parahaemolyticus ATCC 17802 were used as quality control organisms.

#### ERIC-PCR Analysis

ERIC-PCR analysis was performed on the V. parahaemolyticus isolates using a previously described protocol with some modifications (Chen W. et al., 2012; Xie et al., 2015). The reaction mixture (25 µL per reaction) consisted of 12.5 µL 2 × Long Taq Mix (Dongsheng Biotech, Guangzhou, China), 0.6 µM of each primer (5′ -ATGTAAGCTCCTGGGGATTCAC-3′ and 5′ - AAGTAAGTGACTGGGGTGAGCG-3′ ), and 100 ng template DNA. PCR was performed in a DNA thermocycler (Applied Biosystems, CA, USA) using the following procedure: one cycle of denaturation at 95◦C for 5 min; 35 cycles of 94◦C for 45 s, 52◦C for 1 min, and 72◦C for 3 min; and a final extension at 72◦C for 10 min. The PCR products were separated by electrophoresis on 2.0% agarose gels followed by Goldview staining (0.005%, v/v; SBS Genetech, Beijing) and imaging with a UV Imaging System (GE Healthcare, WI, USA). The images were captured in TIFF file format for further analysis.

The size of each band in the ERIC patterns was determined, and the data were coded as 0 (absence) or 1 (presence). Cluster analysis was performed with NTSYS-pc (Version 2.10), a numerical taxonomy and multivariate analysis software package (Rohlf, 2000), based on Dice's similarity coefficient (SD), with a 1% position tolerance and the unweighted pair group method using arithmetic averages (UPGMA).

# MLST

MLST analysis was conducted via the V. parahaemolyticus MLST website and database (http://pubmlst.org/vparahaemolyticus/) (Jolley et al., 2004). PCR conditions were denaturation at 96◦C for 1 min; primer (Table S1; synthesized by Sangon Biotech, Shanghai, China) annealing at 58◦C for 1 min; and extension at 72◦C for 1 min, for 30 cycles; with a final extension step at 72◦C for 10 min. PCR was performed with a Bio- Rad PTC-200 Thermal Cycler (Bio-Rad, California, USA) according to the manufacturer's directions. PCR products were sequenced on a BGI instrument (Shenzhen, China). The alignments of these sequences were determined using BioEdit. The sequences were analyzed online (http://pubmlst.org/vparahaemolyticus/) to assign allele numbers and define STs. New sequences for alleles and new ST profiles were submitted to the V. parahaemolyticus MLST database.

The evolution tree of the concatenated sequences of the seven loci was built based on the method of the Kimura-2-parameter in Mega 6.0 (Tamura et al., 2013). The ratio between the number of synonymous and nonsynonymous substitutions, showing the type of selection at each locus, was calculated using the method of Nei and Gojobori in Mega 6. The hypotheses of neutrality (dS = dN), purifying selection (dS/dN >1), and positive selection (dS/dN < 1) were tested using DNAsp 5.10 (Lüdeke et al., 2015).

# RESULTS

### *V. parahaemolyticus* in RTE Food Samples

Of the 511 samples tested, eight were positive by both qualitative and MPN methods; 12 showed positive results by the qualitative method only, while 10 were positive with the MPN method only. Thirty (5.9%) samples positive for V. parahaemolyticus were detected among 511 samples after qualitative and MPN analyses, including 22 (5.9%) of the 371 deli meat samples, seven (7.2%) of the 97 samples of cold vegetable dishes or noodles in sauce, and one (2.3%) of the 43 fried rice or noodle samples. In 18 positive samples detected by the MPN method, V. parahaemolyticus densities ranged between 3 and 100 MPN/g in all of the samples. The densities of rest positive samples were below 3.0 MPN/g. According to the National Food Safety Standards of China (GB 2727-2005, GB2726-2005), the pathogenic bacteria should not be detected in deli meat such as cooked meat or roasted meat.

#### Detection of *tdh* and *trh* Genes in *V. parahaemolyticus* Isolates

A total of 39 V. parahaemolyticus isolates were confirmed and tested for the presence of the trh and tdh genes. None of the isolates possessed the tdh and trh genes.

# Serotyping by Multiplex PCR

With the exception of serotypes O9, O10, and O11, all other serotypes were detected among the isolates. Serotype O2 was the most prevalent (13 isolates), followed by serotype O4 (eight isolates). The results of the O-antigen serotyping for all 39 isolates are shown in **Table 1**. The serotypes of V. parahaemolyticus ATCC 17802 and ATCC 33847 were O1 and O4, respectively (Figure S2).

#### Antimicrobial Susceptibility

The susceptible, intermediate, and resistance rates of the 39 examined V. parahaemolyticus isolates with respect to 12 antibiotics are shown in **Table 2**. The resistance to streptomycin (S), ampicillin (AMP), cefazolin (KZ), cephalothin (KF), kanamycin (K), gentamicin (CN), and trimethoprimsulfamethoxazole (SXT) reached 89.7, 51.3, 51.3, 41.0, 41.0, 2.6, and 2.6%, respectively. None of the strains showed resistance to azitromycin (AZM), chloramphenicol (C), ciprofloxacin (CIP), nalidixic acid (NA), or tetracycline (TE). Most of the isolates

#### TABLE 1 | Results of the PCR-based O-antigen serotyping of 39 *V. parahaemolyticus* isolates.


TABLE 2 | Antimicrobial resistance profiles of 39 *Vibrio parahaemolyticus* isolates.


\**R, resistant; I, intermediate resistance; S, susceptibility.*

were resistant to streptomycin, with resistance and intermediate rates of 89.3 and 10.3%, respectively, which was consistent with a previous study (Xie et al., 2015). The next-highest susceptible rates were observed for trimethoprim-sulfamethoxazole (87.1%) and gentamicin (51.3%). In addition, there were four multidrugresistant isolates showing resistance to five antibiotics, of which one isolate was collected from deli meat in Xiamen, two isolates were obtained from deli meat in Chengdu, and one isolate was obtained from deli meat in Jinan (Table S2).

#### ERIC-PCR

The results of ERIC-PCR analysis of the 39 isolates are shown in **Figure 2**. ERIC-PCR resulted in 3–8 amplification bands, with sizes ranging from 130 to about 6000 bp. The ERIC image shows that bands with molecular sizes of 500, 1500, and 2500 bp were common (Figure S1). All the isolates were classified into five clusters (designated as A, B, C, D, and E). Most isolates were distributed on cluster A. Only one strain (no. 34) from Zhanjiang divided into cluster D. The standard strain ATCC 17802 was in cluster E alone. The other standard strain (ATCC 33847) was

belonged on cluster A and on the same sub-cluster along with isolates nos. 46 and 47.

## MLST

All V. parahaemolyticus isolates were analyzed by MLST using the sequences generated from internal fragments of seven HK genes. Numbers for alleles and sequence types (STs) were assigned according to the database created for V. parahaemolyticus on submitting the sequence results (**Table 3**). The potential novel loci would appear some mismatching base. Seven gene locus mismatching the ST types on database maybe the novel one. The administrator of the website was notified when we obtained novel loci and ST types; after confirmation, new numbers will be given. The numbers of alleles observed for each MLST locus in our study were distributed as follows: 22 (dna E), 22 (gyr B), 27 (rec A), 20 (dtd S), 18 (pnt A), 24 (pyr C), and 18 (tna A). Thirty-three STs were observed among the 39 isolates. There were ten novel loci: dna E 302, 303; gyr B 428; rec A 306, 307; dtd S 371; pnt A 213; pyr C 339, 340; and tna A 238. Most of the isolates (26) were novel STs, namely 1228–1251, 1301, and 1302. Among all 3682 sites, the number of variable sites was 230; the total number of mutations was 245; the haplotype



*"*\**"The novel locis and STs.*

number was 33; and the haplotype diversity was 0.989. All loci showed ratios of synonymous and nonsynonymous substitutions (dN/dS) below 1. A minimum evolution tree was constructed using the concatenated sequences of each allele (**Figure 3**).

#### DISCUSSION

V. parahaemolyticus is a major seafood-borne gastroenteritiscausing bacterium that is frequently isolated from aquatic products (Letchumanan et al., 2015b). In our study, we analyzed 511 samples from Chinese RTE foods and isolated 39 V. parahaemolyticus strains. In our study, some samples only detected positive in the qualitative method. It may be related to the variation in culture concentration. Enrichment culture was streaked onto TCBS agar plates with undiluted liquid in qualitative method, while culture represent 1/10, 1/100, 1/1000 dilution were streaked onto TCBS agar plates in MPN method. For some samples were only positive by the MPN method, the reason maybe attribute to the incomplete homogeneous of the sample culture solution. When a loopful (10 µL) of bacterial suspension was streaked onto the TCBS

sequences. \* indicates the novel STs.

agar plate could not certain to form colonies on the agar. A combination of qualitative and MPN methods avoid missing detection of V. parahaemolyticus effectively. Unlike Listeria monocytogenes (Shi et al., 2015) and Salmonella (Yang et al., 2016), V. parahaemolyticus was not frequently detected on RTE foods. Previous study reported that no V. parahaemolyticus was positive in 145 samples of RTE food in Korea (Chung et al., 2010), similar result was found in Iran (Zarei et al., 2012). However, a report indicated that the prevalence of V. parahaemolyticus in RTE foods was an important cause of food poisoning in Shanghai, China (Tian et al., 2008). As we know, RTE foods do not need further processing before consumption. Thus, identification of contamination in RTE foods is critical for assuring food safety. Our research can provide insights into food safety in RTE foods.

TDH and TRH are considered major virulence factors in V. parahaemolyticus (Ceccarelli et al., 2013; Letchumanan et al., 2014; Raghunath, 2014). Moreover, the presence of tdh- and/or trh-positive V. parahaemolyticusstrainsrepresents a major public health risk. In our study, strains identified in Chinese RTE foods strains were negative for both trh and tdh genes. These results are consistent with the findings of a previous study reported in India (Raghunath et al., 2008), but this was contradictory to most findings from other previous studies (Zhao et al., 2011; Letchumanan et al., 2015a). The overall mechanism of V. parahaemolyticus pathogenesis remains unclear (Ceccarelli et al., 2013); although TDH and TRH have been shown to be correlated with pathogenic strains, they do not fully explain the pathogenicity of V. parahaemolyticus (Lynch et al., 2005). Several studies have reported the presence of clinical strains without tdh and trh (Shirai et al., 1990). Thus, even in the absence of tdh or trh, V. parahaemolyticus still remains pathogenic, and some environmental isolates lacking tdh and/or trh can produce putative virulence factors. For example, some oyster isolates will contain T3SS1 genes without tdh and/or trh (Mahoney et al., 2010; Jones et al., 2012).

In 1996, pandemic O3:K6 serovars were shown to be responsible for V. parahaemolyticus outbreaks. The serotype O3:K6 V. parahaemolyticus emerged from India and spread throughout the world, including to China and the USA (Okuda et al., 1997; Honda et al., 2008; Tan et al., 2010). Our study indicated that serovar O2 was the predominant serotype among the strains isolated from RTE foods. These findings are in contrast with the results of a previous study, in which the O3 serotype was identified as the predominant serotype in China (Zhang et al., 2006; Zhao et al., 2011). O3:K6, O1:Kut, O4:K8, and O2:K3 were also the dominant serovars identified in outbreaks of V. parahaemolyticus in China (Zhang et al., 2013; Ma et al., 2014). The O3 isolate was the same as the most frequent serotype among our clinical isolates (source: Shenzhen Centres for Disease Control). We also showed the presence of other O-type serovars of V. parahaemolyticus from RTE foods; the results demonstrated the diverse distributions in different RTE food types and locations in China. The changes in pandemic serogroups of V. parahaemolyticus have been reported to occur over time, an increasing number of nonpandemic serogroups have been shown to carry pandemic marker genes (Matsumoto et al., 2000). Therefore, these O2 strains from Chinese RTE foods may have the potential for pathogenesis in humans.

With the steady expansion of the Asian aquaculture industry, in order to increase production, aquaculture farmers are using different antibiotics to prevent (prophylactic use) and treat (therapeutic use) pathogenic bacterial infections in aquatic products (Cabello et al., 2013; Huang et al., 2015; Tan et al., 2016). Furthermore, the continuous and extensive use of antibiotics in humans, has led to the emergence of antimicrobial-resistant V. parahaemolyticus strains worldwide (Sani et al., 2013; Yano et al., 2014). In our study, the highest resistance rate was observed for streptomycin. Similarly, previous study have demonstrated the occurrence of streptomycin- and ampicillin-resistant V. parahaemolyticus isolates (Pazhani et al., 2014). We also found a small number of isolates showing resistance to gentamicin and trimethoprim-sulfamethoxazole, which are first-line drugs used in clinic treatment. Moreover, we found that most of the strains (21/39 isolates) were multidrug resistant. As RTE foods are eaten without cooking, the presence of these strains will increase the health risks of consuming such foods in humans. Thus, it may be important to evaluate variations in antimicrobial susceptibility profiles in V. parahaemolyticus strains.

Molecular subtyping is widely used for the analysis of genetic diversity. ERIC-PCR provides discriminatory values and can be used for rapid V. parahaemolyticus typing (Khan et al., 2002). Compared to V. parahaemolyticus isolates from the 1997 Canadian outbreak using three method, they found that ERIC-PCR and ribotyping were the most informative typing methods (Marshall et al., 1999). Using this approach, the isolates were classified into five clusters at 0.65 similarity. Most of the strains were on cluster A; which showed they were may be genetically related. The reference strain ATCC 17804 harbors trh grouped into a single cluster E, exhibiting differences with other strains. Strain (no. 34), from Zhanjiang, was grouped into cluster D alone and was genetically diverse from other isolates. Clustering based on ERIC-PCR did not coincide with the isolation sources or patterns of antibiotic resistance. The result was similar to other studies, showing the high genetic diversity in V. parahaemolyticus strains. MLST is a good method for typing owing to its reproducibility, as shown in the sequencing of seven housekeeping genes; this method has been widely used for analysis of the V. parahaemolyticus sequence (Lüdeke et al., 2015). In our study, all the isolates could be grouped into four main clusters (I, II, III, and IV). ST 291, ST 396, and ST 300 were identified in a public database as environmental isolates from China; ST1013 was separated from environmental samples from USA. Thus, some V. parahaemolyticus isolates from RTE foods were similar to environment strains. Additionally, one ST162 strain was reported as a clinical isolate from the USA. With the identification of 10 novel loci and 22 novel STs, this study substantially contributed to the diversity in the MLST database. As most strains identified in RTE foods V. parahaemolyticus stains were novel STs, these results suggested that V. parahaemolyticus strains found in RTE foods may be distinct from other stains. In our study, both ERIC-PCR and MLST confirmed the genetic diversity within strains.

In summary, this is the first comprehensive study that described the prevalence, serotype, virulence, antibiotic resistance phenotype, and molecular subtyping of V. parahaemolyticus from RTE foods in China. This study showed that none of the isolates possessed tdh and trh, and serotype O2 was found to be prevalent. The antimicrobialresistance patterns revealed that the streptomycin-resistant were widespread (89.7%) and the isolates resistance to some clinical antibiotics such as cephalothin, gentamicin. ERIC-PCR and MLST typing showed genetic diversity. The novel loci and STs indicated genetically diverse on RTE foods isolates. As RTE foods are common and popular food choices in China, therefore the continuous monitoring of food-borne pathogens such as V. parahaemolyticus are vital to ensure the safety of these food products.

# AUTHOR CONTRIBUTIONS

TX, XX are the common first authors, finsh the article experiment and write the article together. QW (Corresponding Author) give the idea and experiments support. JZ, JC help to finish the experiment on article.

#### ACKNOWLEDGMENTS

XX, TX contributed equally to this work. This work was supported by research grants from the Key Projects in the National Science & Technology Pillar Program during the Twelfth 5-year Plan Period (No. 2013BAD16B05), Guangdong province science and technology plan projects

#### REFERENCES


(No.2012A020100010), Natural Science Foundation of Guangdong Province (No. S2012030006235).We would like to thank Editage for providing editorial assistance.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.00549

isolates from Louisiana Gulf and retail raw oysters. Appl. Environ. Microbiol. 73, 7096–7098. doi: 10.1128/AEM.01116-07


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Xie, Xu, Wu, Zhang and Cheng. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Use of *Lactobacillus plantarum* Strains as a Bio-Control Strategy against Food-Borne Pathogenic Microorganisms

Mattia Pia Arena<sup>1</sup> , Amandine Silvain<sup>2</sup> , Giovanni Normanno<sup>1</sup> , Francesco Grieco<sup>3</sup> , Djamel Drider <sup>2</sup> , Giuseppe Spano<sup>1</sup> \* and Daniela Fiocco<sup>4</sup>

<sup>1</sup> Department of Science of Agriculture, Food and Environment, University of Foggia, Foggia, Italy, <sup>2</sup> Laboratoire Régional de Recherche en Agroalimentaire et Biotechnologies, Institut Charles Viollette-Université Lille 1, Université de Lille, Lille, France, 3 Institute of Sciences of Food Production (ISPA), Lecce Section, Consiglio Nazionale delle Ricerche, Lecce, Italy, <sup>4</sup> Department of Clinical and Experimental Medicine, University of Foggia, Foggia, Italy

#### *Edited by:*

Rosanna Tofalo, University of Teramo, Italy

#### *Reviewed by:*

Ebrahim Hadavi, Islamic Azad University, Iran Learn-Han Lee, Monash University Malaysia, Malaysia

> *\*Correspondence:* Giuseppe Spano giuseppe.spano@unifg.it

#### *Specialty section:*

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

*Received:* 27 December 2015 *Accepted:* 21 March 2016 *Published:* 13 April 2016

#### *Citation:*

Arena MP, Silvain A, Normanno G, Grieco F, Drider D, Spano G and Fiocco D (2016) Use of Lactobacillus plantarum Strains as a Bio-Control Strategy against Food-Borne Pathogenic Microorganisms. Front. Microbiol. 7:464. doi: 10.3389/fmicb.2016.00464 Lactobacillus plantarum is one of the most versatile species extensively used in the food industry both as microbial starters and probiotic microorganisms. Several L. plantarum strains have been shown to produce different antimicrobial compounds such as organic acids, hydrogen peroxide, diacetyl, and also bacteriocins and antimicrobial peptides, both denoted by a variable spectrum of action. In recent decades, the selection of microbial molecules and/or bacterial strains able to produce antagonistic molecules to be used as antimicrobials and preservatives has been attracting scientific interest, in order to eliminate or reduce chemical additives, because of the growing attention of consumers for healthy and natural food products. The aim of this work was to investigate the antimicrobial activity of several food-isolated L. plantarum strains, analyzed against the pathogenic bacteria Listeria monocytogenes, Salmonella Enteritidis, Escherichia coli O157:H7 and Staphylococcus aureus. Antagonistic activity was assayed by agar spot test and revealed that strain L. plantarum 105 had the strongest ability to contrast the growth of L. monocytogenes, while strains L. plantarum 106 and 107 were the most active microorganisms against E. coli O157:H7. The antimicrobial ability was also screened by well diffusion assay and broth micro-dilution method using cell-free supernatants (CFS) from each Lactobacillus strain. Moreover, the chemical nature of the molecules released in the CFS, and possibly underlying the antagonistic activity, was preliminary characterized by exposure to different constraints such as pH neutralization, heating, catalase, and proteinase treatments. Our data suggest that the ability of L. plantarum cultures to contrast pathogens growth in vitro depends, at least in part, on a pH-lowering effect of supernatants and/or on the presence of organic acids. Cluster analysis was performed in order to group L. plantarum strains according to their antimicrobial effect. This study emphasizes the tempting use of the tested L. plantarum strains and/or their CFS as antimicrobial agents against food-borne pathogens.

Keywords: *Lactobacillus plantarum*, antimicrobial compounds, inhibiting activity, cell-free supernatant (CFS), organic acid, pathogens

# INTRODUCTION

Lactobacilli are widespread microorganisms which are extensively used in the food field both as technological starters in the fermented products and as probiotics due to their strainspecific healthy properties (Cebeci and Gürakan, 2003; Georgieva et al., 2009; Altay et al., 2013). Among Lactobacilli, Lactobacillus plantarum is one of the most versatile species, including strains with valuable technological skills and recognized probiotic features (da Silva Sabo et al., 2014; Guidone et al., 2014). Moreover, a number of probiotic L. plantarum strains hold multipurpose features as they can both carry out appreciable fermentative and metabolic processes, e.g., increasing the amount of specific beneficial compounds such as vitamins in the fermented food product, and promote the maintenance of consumers' health, since their capacity to modulate the host immune response and to de novo produce vitamins in the human gut (Arena et al., 2014, 2015). Concurrently, the increasing attention of consumers for healthy and natural food prompts food industry and scientific research to investigate the application of natural compounds for the processing of food products, in order to eliminate or reduce chemical additives used as antimicrobial agents. Thus, in recent decades, several lines of research have tried to find "the natural solution" to "the chemical problem." Among these, the selection of microbial molecules, and/or bacterial strains able to produce such compounds, to be used as antimicrobials and preservatives, proved that Lactic Acid Bacteria (LAB) could be suitable candidates for such "natural purpose" (Šuškovic et al., 2010; da Silva Sabo et al., 2014 ´ ).

LABs, including several L. plantarum strains, have been shown to produce different antimicrobial agents such as organic acids, hydrogen peroxide, diacetyl, bacteriocins, and antimicrobial peptides, with a variable spectrum of action (Herreros et al., 2005; Tharmaraj and Shah, 2009; Cortés-Zavaleta et al., 2014). Several lactobacilli, including L. plantarum, exhibit antagonistic activity against pathogenic and spoilage microorganisms. Such antimicrobial effect has been often ascribed to the production of organic acids, including lactic and phenyllactic acids (Tharmaraj and Shah, 2009; Neal-McKinney et al., 2012; Tejero-Sariñena et al., 2012; Rodríguez-Pazo et al., 2013). However, also the synthesis of bacteriocins and/or bacteriocin like substances has been reported to account for the antagonistic activity exerted by probiotic lactobacilli (Kos et al., 2011; Al Kassaa et al., 2014).

The antagonistic activity of selected microorganisms and/or their extracellular antibacterial agents included in the cell free supernatants (CFS) offer valuable opportunities for food preservation (Kecerová et al., 2004) as well as feed supplements or in veterinary medicine (Nousiainen et al., 2004; Bilkova et al., 2011; Cortés-Zavaleta et al., 2014). Because of their widespread association with foods and their generally recognized as safe (GRAS) status, the use of LAB and/or their metabolites as natural drugs has attracted considerable interest in recent years (Reis et al., 2012). In the food industry, the use of the bacteriocins nisin and pediocin has allowed to reduce the addition of chemical preservatives and the intense thermal treatments, thus enhancing sensory and nutritional properties without impairing safety (De Vuyst and Leroy, 2007; Sobrino-Lopez and Martin-Belloso, 2008). Moreover, several other antimicrobial peptides produced by probiotic LABs have been characterized and suggested for potential and relevant applications in food preservation and safety (Reis et al., 2012; Gupta and Srivastava, 2014).

As the antimicrobial activity of LAB bacteriocins is usually restricted to Gram-positive bacteria, organic acids and organic acid-producing LAB could have even wider applications in food safety (De Vuyst and Leroy, 2007; Mu et al., 2012). In this regard, the use of probiotics which produce antimicrobial metabolites, including organic acids, has been proposed as part of effective bio-control strategies to contrast the contamination of animal feed by spoilage and pathogenic microorganisms, and to reduce pathogen loads in livestock (Gerbaldo et al., 2012; Neal-McKinney et al., 2012), consequently decreasing food-borne illness in humans. Recently, culture supernatants from probiotic LAB, with in vitro inhibitory action on Clostridium difficile (CD), were suggested as a basis for alternative therapies to treat CD infections in humans (Joong-Su et al., 2013). Accordingly, cellfree probiotic extracts were proposed as alternative ingredients to probiotic live cells for nutritional and medicinal applications (Saadatzadeh et al., 2013).

Our main objective was to understand whether Lactobacillus spp. could represent a natural alternative to the chemical antimicrobials commonly used in the food preparation. Therefore, this study evaluated the antimicrobial activity of 79 wine-derived L. plantarum strains against seven pathogenic bacteria, generally involved in foodborne poisoning and infections. The pathogens used in this work were Listeria monocytogenes, which can cause abortions and/or gastrointestinal diseases leading to death (Sip et al., 2012), Escherichia coli O157:H7, which provokes haemorrhagic colitis and haemolytic uremic syndrome (Mead and Griffin, 1998), Salmonella Enteritidis, which determines abdominal pain, nausea, vomiting, and diarrhea (Liu et al., 2011), and methicillinresistant and methicillin-sensitive strains of Staphylococcus aureus, which is involved in harmful gastroenteritis (Gutiérrez-Larraínzar et al., 2012). Additionally, we investigated on the chemical nature of the molecules possibly accounting for the observed antimicrobial activity.

# MATERIALS AND METHODS

#### Bacterial Strains and Growth Conditions

This study was carried out on 79 L. plantarum strains deposited into the culture collection of Foggia University (Italy; UNIFG) and previously isolated from wine and must (**Table 1**). All L. plantarum strains were growth on de Man–Rogosa–Sharpe (MRS; Sigma-Aldrich, St. Louis, MO, USA) at 30◦C.

The pathogenic bacteria used were: L. monocytogenes CECT 4032; S. Enteritidis CECT 409, E. coli O157:H7 CECT 4267, two methicillin-resistant strains of S. aureus MSSA1220, and S. aureus MRSA1209, two methicillin-susceptible strains of S. aureus MRSA1208 and S. aureus MRSA1070. All pathogens were grown in tryptone soy broth (TBS, Oxoid) and incubated at 37◦C, with the exception of S. aureus strains that were grown in Brain Heart Infusion broth (BHI, Oxoid).



#### Antimicrobial Activity

The antimicrobial activity was evaluated by (i) agar spot test, (ii) well-diffusion method, and (iii) broth microdilution method. The agar spot test was carried out according to Gaudana et al. (2010). Briefly, overnight cultures of lactobacilli were spotted (5 µL) on MRS agar and incubated for 24 h at 37◦C. Pathogen overnight cultures were mixed 1:100 with TSB or BHI soft agar (containing 0.6% agar, w/v) and poured over MRS agar plates containing the developed colonies of lactobacilli. Plates were incubated for 24 h and the radii of the inhibition zones were measured.

For the well diffusion assays, cultures of lactobacilli were grown in MRS broth (pH 6.5) for 18 h and, then, centrifuged (8000 × g for 20 min, 4◦C). The cell-free supernatant (CFS) was recovered and sterilized by filtration through Millex-GV 0.22 µm hydrophilic Durapore PVDF membrane (Millipore, Billerica, MA, USA). To investigate on the chemical nature of the potentially inhibitory substances secreted by each L. plantarum strain, showing antagonistic effects, filtered CFSs (CFS-A) were submitted to different treatments. An aliquot of filtered CFS was sequentially treated according to Herreros et al. (2005). First of all, CFSs (CFS-A) were heated at 80◦C for 10 min (CFS-B) and neutralized with 2 M NaOH (to pH 6.5; CFS-C), in order to rule out inhibiting effects due to organic acids. The neutralized CFSs (CFS-C) were subjected to the following treatments: (i) catalase digestion (1 mg/ml; Sigma-Aldrich Corporation, USA) at 37◦C for 1 h, in order to eliminate the possible inhibitory action of hydrogen peroxide (CFS-D); (ii) separate digestion at 37◦C for 2 h with different proteases, i.e., proteinase K (CFS-E; 1 mg/ml), trypsin (CFS-F, 1 mg/ml), α-chemotrypsin (CFS-G, 1 mg/ml), and papain (CFS-H, 1 mg/ml; all purchased from from Sigma, USA); (iii) heating at 80◦C for 60 (CFS-I) and 90 min (CFS-L), 100◦C for 60 (CFS-M) and 90 min (CFS-N), and at 121◦C for 15 min (CFS-O).

Another aliquot of filtered CFS was separately treated according to Cortés-Zavaleta et al. (2014). A part of this aliquot (CFS-A<sup>∗</sup> ) was heated at 80◦C for 10 min (CFS-B<sup>∗</sup> ) and then neutralized with 2 M NaOH (to pH 6.5) (CFS-C<sup>∗</sup> ), as mentioned above. Another portion of CFS-A<sup>∗</sup> was exposed to catalase (1 mg/ml; 37◦C for 1 h; CFS-D<sup>∗</sup> ), or to different proteases, i.e., proteinase K (CFS-E<sup>∗</sup> ; 1 mg/ml, 37◦C for 2 h), trypsin (CFS-F ∗ , 1 mg/ml, 37◦C for 2 h), α-chemotrypsin (CFS-G<sup>∗</sup> , 1 mg/ml, 37◦C for 2 h), and papain (CFS-H<sup>∗</sup> , 1 mg/ml, 37◦C for 2 h), or to different thermal treatments, i.e., 80◦C for 60 (CFS-I<sup>∗</sup> ) and 90 min (CFS-L<sup>∗</sup> ), 100◦C for 60 (CFS-M<sup>∗</sup> ) and 90 min (CFS-N<sup>∗</sup> ), and 121◦C for 15 min (CFS-O<sup>∗</sup> ).

All treated CFSs were collected and 100 µl of each were used to fill 6 mm diameter wells previously punched on MRS agar plates. The plates were incubated for 2 h at 4◦C in order to permit CFSs diffusion onto MRS agar. Overnight cultures of pathogenic bacteria were inoculated (1% v/v) into fresh TSB or BHI soft agar (0.6% agar, w/v) and poured over MRS agar plate containing CFSs. All plates were incubated at 37◦C for 24 h and, then, the inhibition zones around the wells were measured.

Broth microdilution assays were assessed as described by Mayrhofer et al. (2008). Overnight cultures of pathogenic bacteria were inoculated (1% v/v) into fresh medium, i.e., TSB or BHI, and seeded in 96-well plates (Costar-Corning Incorporated, Corning, NY, USA). 200 µl of test solution consisting of 100 µl of pathogenic culture and 100 µl of CFS were mixed into the wells. Untreated CFSs (CFS-A) were diluted in physiological solution (NaCl 8.5 g/L) and used at different percentages, i.e., 5, 10, 15, 20, 25, 30, 35, 40, 45, 50% in the final volume (200 µl), in order to determine the minimum percentage of CFS able to inhibit the growth of target pathogens. Plates were incubated at 37◦C for 24 h and growth of pathogenic bacteria was monitored by measuring optical density (OD<sup>600</sup> nm). Furthermore, after identifying the minimum inhibiting percentage, treated CFSs were also used against pathogenic bacteria. The antimicrobial activity was expressed as inhibition (%) of pathogen growth relative to the control (i.e., pathogen grown in optimal conditions, without CFS).

#### Lyophilization of Cell-Free Supernatant

The supernatants of lactobacilli were collected by centrifugation and 10-fold concentrated by lyophilization as reported by Bermudez-Brito et al. (2013).

# Statistical Analysis

Three independent experiments were conducted for all trials. Cluster analysis was used to determine the grouping of lactobacilli according to their antimicrobial activity against target pathogens. Statistical comparisons were performed by oneway ANOVA test (p < 0.005 was considered as statistically significant). All statistical study was performed using IBM SPSS Statistics 21.0 software program (IBM, Armonk, NY, USA).

# RESULTS

#### Agar Spot Test

In this study, 79 L. plantarum strains were investigated for their possible antimicrobial activity against seven pathogenic bacteria, i.e., L. monocytogenes, S. Enteritidis, E. coli O157:H7, and four strains of S. aureus. The preliminary screening of all Lactobacillus strains, carried out by agar spot test, revealed a different range of antimicrobial activity, depending both on L. plantarum strain tested and on pathogen considered. **Figure 1** reports data obtained with the best inhibiting L. plantarum strains, i.e. which determined overall inhibition halos of more than 5 radius mm, according to the classification proposed by Gaudana et al. (2010). As shown in **Figure 1**, some strains exhibited a very strong ability to inhibit the growth of food pathogens. In particular, L. plantarum 105 exhibited the major ability to inhibit L. monocytogenes, while both L. plantarum 106 and L. plantarum 107 presented the highest antagonistic effect on growth of E. coli O157:H7. S. Enteritidis, S. aureus R1070, R1208, S1209, and S1220 were mainly inhibited by L. plantarum 119, L. plantarum 32, L. plantarum 106, and L. plantarum 108, respectively. Contrariwise, L. plantarum 118 and 119 did not show any inhibition effects on the growth of S. aureus R1208. Similarly, L. plantarum 30 was not able to affect the development of S. aureus S1209.

#### Well-Diffusion Assays

The antagonistic effect of those L. plantarum strains which exhibited appreciable antimicrobial activity, as determined by agar spot test (**Figure 1**), was further assessed by well diffusion assay using CSFs. In contrast to the results obtained with the agar spot assays, no CFS was able to contrast the growth of pathogenic bacteria (data not shown). In order to ascertain if the concentration of the inhibiting substance could be not adequate to sustain antagonistic action, the CFSs were concentrated by lyophilization, prior to their use in well diffusion tests. The 10x concentrated CFSs exhibited inhibition activities that were similar to that previously obtained by agar spot test (**Table 2**). The concentrated and differently treated CFSs were also tested, however, the inhibiting effect was lost after pH neutralization, while enzymatic and heat treatment had no impact on the inhibitory effect (data not shown).

#### Broth Microdilution Assays

CFSs were further tested against pathogens by broth microdilution method. Interestingly, untreated CFSs from all selected L. plantarum strains determined significant inhibition of pathogen growth when used at ≥25% (25:75,

CFS:growth medium), i.e., 30, 40, 50, 60, 70, 80, 90% (data not shown). **Figure 2** shows the antagonism resulted by untreated CFSs of Lactobacillus strains using 25% of CFSs-A. E. coli O157:H7 growth was reduced by around 70 and 93% by CFS of L. plantarum 108 and 104, respectively. L. monocytogenes was inhibited up to 90% by L. plantarum 116, while the growth of S. Enteritidis was reduced by 96 % in presence of L. plantarum 30 CFS. The growth of all strains of S. aureus was significantly contrasted by all CFSs and, the highest reductions were around 90% (S. aureus R1070), 99% (S. aureus R1208), 85% (S. aureus S1209), and 86% (S. aureus S1220). Based on these findings, an aliquot consisting of 100 µl of untreated, 2-fold diluted CFSs was determined as the amount of CFSs showing more than 50% of inhibition ability for 98% of cases analyzed, and chosen for the further assays. The only two-fold diluted CFS causing an inhibition lower than 50% was that from L. plantarum 104 against L. monocytogens and S. Enteritidis.

In order to investigate on the nature of the inhibitory substances secreted by each L. plantarum strain showing antagonistic effects, CFS-As were submitted to different treatments. As a result, the potential inhibiting molecules lost their antagonistic ability after pH neutralization (data not shown).

# Statistical Relationship among Lactobacilli for their Antimicrobial Activity

Cluster analysis was performed on the inhibition halos data obtained by agar spot test, in order to group Lactobacillus strains showing similar antimicrobial activity. Euclidean distance was used to measure the proximity between two data and average linkage clustering was used as linkage criteria. As a result, four clusters of lactobacilli were distinguished (**Figure 3**). Group A contained 10 strains of L. plantarum (103, 104, 105, 108, 109, 117, 119, 120, 121, 122); group B was constituted of strain 118, group C consisted of L. plantarum 106 and 107; group D comprised four strains of L. plantarum (30, 31, 32, 33). Furthermore, ANOVA was elected as a method to study the statistical differences (p < 0.005) among the four groups (data not shown). Overall, the results indicated that the strains included in group A were able to contrast the growth of L. monocytogenes significantly better than strains in groups B, C, and D. L. plantarum 118, within group B, showed no or very low activity against S. aureus R1208 and S. aureus S1220, while L. plantarum 106 and L. plantarum 107, belonging to group C, were mostly active against E. coli O157:H7. Group D strains, including L. plantarum 30, 31, 32, and 33, were able to inhibit the growth of S. aureus R1208 significantly better than lactobacilli of group A, B, and C.


TABLE 2 | Antimicrobial activity of CFS of *L. plantarum* strains determined by well-diffusion assay and expressed as the size of inhibition zones around the wells (mm).

CFS-A, untreated and 10x concentrated cell free supernatant; CFS-B, 10x concentrated CFS heated at 80◦C for 10 min.

#### DISCUSSION

All analyzed Lactobacillus strains were shown to inhibit the growth of pathogens in a lactobacillus strain- and pathogen strain-depending manner. Using the agar spot method, 17 L. plantarum strains were identified as very strong inhibitors, according to the classification made by Gaudana et al. (2010), as they showed inhibition halos of more than 5 mm against the majority of the food pathogens tested. Cluster analysis was useful to group the Lactobacillus strains in four clusters, each of them denoted by a different antimicrobial activity. Within each group, peculiar abilities to contrast the growth of target pathogens were underlined such as a higher inhibitory activity against Gram-positive bacteria L. monocytogenes and S. aureus R1208 and Gram-negative bacteria E. coli O157:H7 by group A, group D, and group C, respectively.

The antimicrobial activity of the tested L. plantarum strains was mostly observed when they were grown on solid media and then brought into contact with pathogenic bacteria. This could be a considerable feature to be sought in the choice of starter or probiotic microorganisms. Indeed, live microorganisms carry out antimicrobial and preservative activity in the food when used as starters. Moreover, as probiotics, they can provide a protective benefit for the consumer when, following ingestion, can activate their metabolism in the intestine (De Vuyst and Leroy, 2007; Tejero-Sariñena et al., 2012; Arena et al., 2014).

The antimicrobial capability was also confirmed when 10x concentrated CFSs from L. plantarum strains were used in agar well diffusion assay, thus indicating that a minimal concentration of antimicrobial compounds is necessary to sustain that antagonism. CSF may include also other molecules, besides those effectively secreted by bacteria (i.e., medium components and/or intracellular compounds which may be accidentally released during CFS preparation). However CFS are routinely used to preliminarily screen the antimicrobial capacity of lactobacilli by well diffusion method (Herreros et al., 2005; Guo et al., 2010; Al Kassaa et al., 2014; Wang et al., 2014). For most of the tested L. plantarum strains, we found a good correspondence between the antimicrobial activities as assessed by either agar spot test or by well diffusion, using CFS. This indicates that the inhibitory effects mainly depend on exudates (which are included in the CFS) and only in part, if any, on other antagonistic mechanisms which require a more direct interaction, possibly occurring during co-culture on solid media (e.g., metabolic competition).

In addition, 1x CFSs were tested by broth microdilution method in order to understand whether the CFSs components could reduce the growth of pathogens in liquid-medium respect to agar-medium. The results suggested a greater capability of CFSs to contrast pathogenic bacteria in liquid-medium than in agar plates. Minimum amounts of CFS with inhibitory effect on pathogens were determined, indicating that two-fold diluted CFSs could exhibit more than 50% of inhibition for 98% of the cases analyzed. Absent or low activity by not concentrated CFSs in agar well diffusion tests was also previously reported (Saadatzadeh et al., 2013). It was also demonstrated that the antimicrobial effect was improved by CFS lyophililization and, more interestingly, such procedure was praised as an innovative strategy in bacterial products preparation (Saadatzadeh et al., 2013).

In order to investigate on the nature of the inhibitory substances secreted by each L. plantarum strain, both 1x CFSs and 10x CFSs were submitted to different treatments and

tested against pathogens by broth method and well diffusion assay, respectively. Since pH neutralization eliminated the antimicrobial feature of CSF, while neither protease nor heat treatment had any impact on the inhibitory effect (both in sequential and in separate treatments, see the experimental section), we hypothesize that acidic pH and/or to the presence of organic acids could account for most of the observed antimicrobial activity. Indeed, although CFSs from bacterial cultures may contain many cellular metabolites, organic acids have been indicated as the principal antimicrobial agents when the antimicrobial activity is reduced or eliminated by alkaline neutralization (Bilkova et al., 2011; Zhang et al., 2011; Tejero-Sariñena et al., 2012). However, data obtained by CFS neutralization provide only a preliminary indication of the active compounds. Further experiments, including monitoring the pH of CFS and the acidification of growing cultures, as well as HPLC analysis, could help to better substantiate the potential role of organic acids. Moreover, organic acids, if any, could have enhanced the activity of other antimicrobial metabolites, which might require acidification and/or acid-mediated cell membrane disruption to exert an apparent antagonistic effect (Alakomi et al., 2000).

The antimicrobial effect of organic acids has been observed for several lactobacilli. The growth-inhibiting activity of different LAB, i.e., strains belonging to Lactobacillus, Bifidobacterium, Lactococcus, Streptococcus, and Bacillus genera, against pathogens such as Salmonella typhimurium, E. coli, Enterococcus faecalis, S. aureus, and Clostridium difficile, was attributed to a pH reduction and/or to the production of organic acids, including lactate and acetic acid (Tejero-Sariñena et al., 2012). Moreover, De Keersmaecker et al. (2006) also found a strong antimicrobial activity of L. rhamnosus strains against S. typhimurium, which was ascribed to the accumulation of lactic acid.

The organic acids secreted by LAB determine an environmental pH reduction that can be adverse for those microorganisms sharing the same niche (Tharmaraj and Shah, 2009). In their undissociated form, organic acids can penetrate the cytoplasmic membrane of target microorganisms, thus resulting in intracellular acidification and in the collapse of the transmembrane proton motive force. Such mode of action is pH-dependent, because the undissociated forms are prevalent when the pH value is below the pKa of the organic acid (Batish et al., 1997; Dalié et al., 2010; Schillinger and Villareal, 2010; Cortés-Zavaleta et al., 2014). Accordingly, the production of undissociated organic acids was indicated as the main mechanism through which several intestinal lactobacilli contrast the growth of a range of both Gram-negative and Gram-positive pathogenic bacteria in liquid media (Annuk et al., 2003; Topisirovic et al., 2007; Toy et al., 2015). Despite, the neutral pH conditions of the large intestine, probiotic bacteria could produce locally high concentrations of organic acids, thus establishing chemical microenvironments where the antagonistic action can be carried out (Alakomi et al., 2000).

Besides their pH lowering properties, sometimes, the antimicrobial effect of organic acids reflects a specific mode of action which may be relatively pH-independent. For instance, lactate, i.e., the main acid produced by LAB fermentation, was proved to specifically permeabilize the outer membrane of Gram negative species, causing structural alterations in the phospholipid component (Alakomi et al., 2000). Likewise, probiotic lactobacilli inhibited Campylobacter jejuni growth by the secretion of lactic acid, which was shown to disrupt

membrane integrity through a mechanism which is not solely pH-dependent (Neal-McKinney et al., 2012).

In the last decades, the main molecules which have been extensively studied as antimicrobial agents have been bacteriocins (Adebayo et al., 2014; Gupta and Srivastava, 2014). Bacteriocins are ribosomally-synthesized peptides that can act against bacteria of the same species (narrow spectrum) or of the same genera (broad spectrum). These compounds can be produced directly in fermented food either by bacteriocinproducing starter cultures (fermentative and bioconservative actions) or by protective culture strains (only bioconservative action). Additionally, isolated and purified bacteriocins can be used as food additives or included in the packaging materials (Kos et al., 2011; Fan and Song, 2013). One of the major drawbacks in the use of bacteriocins as natural antimicrobials is that their proteinaceous structure can be easily altered by diverse proteases possibly occurring in the food, being secreted by different bacteria and/or occurring in the human digestive tract (Saavedra et al., 2004). Moreover, the efficacy of bacteriocins in food can be decreased by their adsorption to food components, poor solubility and uneven distribution within the food matrix (Hartmann et al., 2011). Compared to LAB bacteriocins, which are mainly active against Gram-positive bacteria, organic acids exhibit a broader spectrum of antimicrobial action. Besides, organic acids are not sensitive to proteases and may be better solubilized. Therefore, the bioprotective potential of organic acid-producing LAB is high and suited to wide applications in food safety and nutritional medicine (Mu et al., 2012; Pawlowska et al., 2012).

The use of CFS as antimicrobial ingredients could be an interesting strategy in food preparation. CFS produced by selected bacteria could be effective in inhibiting pathogens, especially when the inoculation of live inhibiting microorganisms may not be feasible, e.g., in food subjected to refrigeration in which the psychrotrophic L. monocytogenes, but not lactobacilli, could easily grow. Furthermore, the use of CFS rather than purified antimicrobials could determine the advantage to have different biologically active substances, with possible synergistic effects, in one product (Hartmann et al., 2011). It is worthwhile mentioning that some recent in vitro studies have suggested potential and intriguing biomedical applications of CFSs. For instance, CFS from LAB, with antimicrobial activity against C. difficile, was proposed as a plausible alternative to the therapies for the treatment of CD-associated gut disorders (Joong-Su et al., 2013). Moreover, CFSs from Lactobacillus strains were ascribed health beneficial effects, including inhibition of cancer metastatis (Escamilla et al., 2012), positive modulation of the intestinal immune response (Bermudez-Brito et al., 2013) and cholesterolreducing properties (Kim et al., 2008).

To sum up, this study provides evidence that several of the screened L. plantarum strains possess a significant ability to contrast various pathogenic bacteria, including both Gram negative and Gram positive species, which can contaminate food and are responsible for diseases in humans. The biosynthesis of organic acids is proposed as one of the main mechanism through which the antimicrobial activity is exerted. The antagonistic feature could be a distinctive trait to take into account for the selection of starters and or probiotics that could also function as bio-control agents against potentially harmful microorganisms during food processing and storage. In a future perspective of reducing or eliminating the use of chemical compounds, this work contributes to the existing knowledge in a context in which the consumer's attention is increasingly aimed at healthy and natural food products.

#### REFERENCES


#### AUTHOR CONTRIBUTIONS

GS designed the experimental plan, analyzed the results and read the final paper. MA made the experimental trials and wrote up the paper. AS supported the technical part. GN provided some biological materials, participated to the experimental plan and read the final paper. FG and DD participated to the experimental plan and read the final paper. DF made the experimental trials and read the final paper.

#### ACKNOWLEDGMENTS

This work was partially supported by a grant from the project PON02\_00186\_3417512 in the framework of the project named "S.I.Mi.S.A.," (PON02\_00186\_2937475).


pathogens and spoilage bacteria. Food Control 26, 555–563. doi: 10.1016/j.foodcont.2012.02.025


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Arena, Silvain, Normanno, Grieco, Drider, Spano and Fiocco. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Structural Elucidation and Toxicity Assessment of Degraded Products of Aflatoxin B1 and B2 by Aqueous Extracts of *Trachyspermum ammi*

Wajiha Iram<sup>1</sup> , Tehmina Anjum<sup>1</sup> \*, Mazhar Iqbal <sup>2</sup> , Abdul Ghaffar <sup>3</sup> and Mateen Abbas <sup>4</sup>

1 Institute of Agricultural Sciences, University of the Punjab, Lahore, Pakistan, <sup>2</sup> Health Biotechnology Division, National Institute for Biotechnology and Genetic Engineering, Faisalabad, Pakistan, <sup>3</sup> Department of Chemistry, University of Engineering and Technology, Lahore, Pakistan, <sup>4</sup> Department of Toxicology, Quality Operating Laboratory, University of Veterinary and Animal Sciences, Lahore, Pakistan

In this study aqueous extract of seeds and leaves of Trachyspermum ammi were evaluated for their ability to detoxify aflatoxin B1 and B2 (AFB1; 100µg L−<sup>1</sup> and AFB2; 50µg L−<sup>1</sup> ) by in vitro and in vivo assays. Results indicated that T. ammi seeds extract was found to be significant (P < 0.05) in degrading AFB1 and AFB2 i.e., 92.8 and 91.9% respectively. However, T. ammi leaves extract proved to be less efficient in degrading these aflatoxins, under optimized conditions i.e., pH 8, temperature 30◦C and incubation period of 72 h. The structural elucidation of degraded toxin products by LCMS/MS analysis showed that eight degraded products of AFB1 and AFB2 were formed. MS/MS spectra showed that most of the products were formed by the removal of double bond in the terminal furan ring and modification of lactone group indicating less toxicity as compared to parent compounds. Brine shrimps bioassay further confirmed the low toxicity of degraded products, showing that T. ammi seeds extract can be used as an effective tool for the detoxification of aflatoxins.

Keywords: aflatoxin, degradation, plant extract, LCMS/MS, toxicity

# INTRODUCTION

Mycotoxins are chemically and biologically active secondary metabolites produced by fungi in cereals, nuts, fruits and vegetables (Sinha and Sinha, 1991; Aly, 2002). About 25% of the world cereals are contaminated with known mycotoxins produced by variety of toxigenic fungi. The food and agriculture organization (FAO) estimates that about 1000 million metric tons of foodstuffs could be contaminated with mycotoxins each year (Bhat et al., 2010). Currently, more than 400 mycotoxins are identified, among them, aflatoxins are the most serious carcinogenic, hepatotoxic, teratogenic, and mutagenic secondary metabolites which adversely affect humans and animal health. They are classified as group-1 carcinogens by International Agency for Research on cancer (IARC, 2002). Aflatoxin contamination can occur at any stage of food production from pre-harvest to storage (Wilson and Payne, 1994). Factors that affect aflatoxin contamination include the climate, genotype of the crop planted, soil type, temperature fluxes, intercropping with infected grains, early and delayed harvesting, improper drying and meager storage conditions (Folayan, 2013).

Aflatoxins are biologically active polyketide derived secondary metabolites which consist of a group of closely related highly oxygenated bisfurano-coumarin heterocyclic compounds, mostly

#### *Edited by:*

Giovanna Suzzi, Università degli Studi di Teramo, Italy

#### *Reviewed by:*

Issmat Kassem, The Ohio State University, USA Robert Lawrence Brown, Agricultural Research Service, USA

*\*Correspondence:*

Tehmina Anjum tehminaanjum@yahoo.com

#### *Specialty section:*

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

*Received:* 24 November 2015 *Accepted:* 04 March 2016 *Published:* 30 March 2016

#### *Citation:*

Iram W, Anjum T, Iqbal M, Ghaffar A and Abbas M (2016) Structural Elucidation and Toxicity Assessment of Degraded Products of Aflatoxin B1 and B2 by Aqueous Extracts of Trachyspermum ammi. Front. Microbiol. 7:346. doi: 10.3389/fmicb.2016.00346

produced by Aspergillus flavus and Aspergillus parasiticus (Ellis et al., 1991; Bhatnagar et al., 1992). The G series of aflatoxins differs chemically from B series by the presence of a β lactone ring, instead of cyclopentenone ring. In AFB1 and AFG1 a double bond undergoes reduction forming vinyl ether at the terminal furan ring but not in AFB2 and AFG2 (Samarajeewa et al., 1990). AFB1 and AFG1 are carcinogenic and considerably more toxic than AFB2 and AFG2 probably due to these small difference in structure (Jaimez et al., 2000).

Various physical, chemical and biological methods have been described for detoxification of mycotoxins. Routinely, physical and chemical methods like roasting, flaking, canning, alkalization, oxidation, reduction and acidification are being frequently used. But these methods have not yet proven to be as effective and desirable because they are considered to be potentially unsafe as some form toxic residues or even alter the nutritional contents and flavor of treated commodity. In addition these methods require sophisticated equipment rendering for their high cost and environmental pollution (Joseph et al., 2005; Shukla et al., 2012). So, health hazards from exposure to such methods and economic consideration make biological control and natural plants extracts ideal alternatives to protect food and feed from fungal and mycotoxin contamination as they are biologically safe, eco-friendly, cheap, easily available, lack residual effects and are easily degradable (Reddy et al., 2007).

For the last decade, the use of herbal food additives has been encouraged (Mirzaei-Aghsaghali, 2012). The intensive efforts have been made by various researches for the clarification of biochemical structures and physiological functions of various food and feed additives like prebiotics, probiotics, organic acids and plant extracts. Numerous aromatic plants have been found to inhibit the microbial growth and thereby traditionally used to extend the shelf life of food (Ahmed et al., 2014). Similarly, several medicinal herbs and spices have been reported to counteract deleterious effects of mycotoxins either by chemical modification or by inclusion into the plant matrix (Wallnofer et al., 1996). Powder and extract of many medicinal herbs and higher plants have been shown inhibitory effect on growth of toxigenic fungi and production of toxins (Solis et al., 1993; Momoh et al., 2012). Based on the literature following plants were found to be effective in detoxifying aflatoxins and growth inhibition of toxigenic fungi: Withania somnifera (Linn.), Camellia sinensis (Linn.), Citrus medica (Linn.), Syzigium aromaticum (L.) Merr. Et Perry, Curcuma longa (L.), Allium sativum L. and Ocimum sanctum (Linn.), Trachyspermum ammi (L.), Eucalyptus globolus (Linn.), Olea europaea (Linn.), Thymus vulgaris, Hibiscus sabdariffa (Linn.), Boswellia sacra, Adhatoda vasica Nees and Barleria lupulina Lindl (Krishnamurthy and Shashikala, 2006; Reddy et al., 2009; Velazhahan et al., 2010; Al-Rahmah et al., 2011; El-Nagerabi et al., 2012, 2013; Kannan and Velazhahan, 2014; Vijayanandraj et al., 2014).

In the present investigation, Trachyspermum ammi (Family: Apiaceae, Common name: Ajwain) was used to evaluate its aflatoxin detoxifying potential. T. ammi has been known

to possess known antimicrobial, antispasmodic, antiflatulent, antioxidant and antirheumatic effect due to the presence of several active compounds (Bairwa et al., 2012). Phytochemical studies on T. ammi revealed the presence of alkaloids, phenolics, steroids, fixed oils, glycosides, tannins, saponin and flavonoids, cumene, thymene, amino acids and thymol (Asifa et al., 2014). Literature showed that phenols, thymol and carvacol, are responsible for its antimicrobial, anti mycotoxigenic, antiseptic, antitussive and expectorant properties (Pathak et al., 2010). Ajwain is generally regarded as safe when taken in the recommended doses, however, in rare cases, it can cause nausea and headache (Grossberg and Fox, 2008). Although previous

studies have been conducted with T. ammi seeds extract to detoxify aflatoxins but their resulting degradation products are not described in detail. However, in the present manuscript structural identification and fragmentation patterns of proposed degradation products were included along with their toxicity assessment. Various parameters were optimized for in vitro and in vivo detoxification of aflatoxins.

# MATERIALS AND METHODS

#### Extraction and Purification of Aflatoxins

Aflatoxin B1 and B2 were extracted from a toxigenic isolate of Aspergillus flavus (isolated from stored maize samples) grown on coconut cream media by solvent extraction method as described by Yazdani et al. (2010) with some modifications. For extraction, colony margins were scraped together with surrounding zones into 250 mL Erlenmeyer flask containing 10 mL chloroform: acetone (85:15 v/v) and shacked for 30 min at 200 rpm. The crude extracts were filtered through gauze, and then through Whatman No.1 filter paper. Then, the filtrate was passed through the immunoaffinity column (Aflatest column, VICAM, Waters, USA) in solid phase extraction assembly for the separation of aflatoxins, according to the method described by Stroka et al. (2000) with some modifications. The column was rinsed with HPLC grade water and 10 ml of the sample was passed through it. Based on the highly specific antibody antigen reaction, the aflatoxins present in the sample form a conjugate with the antibody and the remaining impurities are separated out. Finally aflatoxins bound in the column are cleaved from their respective antibodies using methanol and compared with standard AFB1 and AFB2 purchased from (Sigma-Aldrich, St. Louis, MO, USA) through High Performance Liquid Chromatography. Stock solutions of AFB1 (1000µg L−<sup>1</sup> ) and AFB2 (500µg L−<sup>1</sup> ) were prepared in methanol and stored at 4◦C. The working solutions of AFB1 (100µg L−) and AFB2 (50µg L−<sup>1</sup> ) were prepared by diluting the stock solution.

# Preparation of Plant Extract

Plant extracts were prepared according to the method described by Velazhahan et al. (2010) with some modifications. Samples were surface-sterilized using 1% sodium hypochlorite for 10 min and washed several times with sterile distilled water. After that aqueous extract of Trachyspermum ammi leaves and seeds was prepared by homogenizing 10 g of leaves/seeds with 10 mL of sterile distilled water. Homogenate was filtered through muslin cloth and centrifuged at 14,000 rpm for 20 min. Supernatant was sterilized using syringe filter assembly and used for further detoxification studies.

#### *In vitro* Toxin Inactivation Assay

For detoxification studies, 50µL of working solution containing (100µg L−<sup>1</sup> ) AFB1 and (50µg L−<sup>1</sup> ) AFB2 was mixed with 250µL of T. ammi plant extracts and incubated for various intervals of time. After incubation, 250µL of Chloroform was added to above mixture and mixed well by vortexing. After that, the mixture was centrifuged at 13,000 rpm for 10 min in a centrifuge (Eppendorf, 5424C) to separate the chloroform fraction. After centrifugation, the chloroform fraction was

transferred to another glass tube, evaporated to dryness under gentle stream of nitrogen and re-dissolved in methanol. Control consisted of 50µL of toxin in 250µL of water and was incubated under same conditions. All experiments were conducted in triplicate.

#### *In vitro* Optimization of Parameters for Toxin Detoxification pH

The optimal pH was determined by modifying the original pH of the T. ammi aqueous extracts in the range of 2.0–10.0 (adjusted using either 1 N HCl or 1 N NaOH) and then assayed for toxin detoxification activity. Distilled water with same pH range as well as untreated extract was used as control.

#### Temperature and Incubation Time

For assessing optimum temperature and incubation period, T. ammi extracts were incubated with toxins at 25◦C, 30◦C, 35◦C, 40◦C, 45◦C, 50◦C, 55◦C, and 60◦C for 3, 6, 12, 24, 48, and 72 h respectively. After incubation, the toxin content in the reaction mixture was determined as described above.

# Effect of Boiling on Toxin Detoxification Properties of Plants Extracts

In order to study the effect of boiling on toxin detoxification properties of plants extracts, 1mL of aqueous plant extract was added in 1.5 mL eppendorf tube and placed in a boiling water bath for 5–10 min, cooled to room temperature and then tested for toxin detoxification activity.

# Detoxification of Maize Samples Using Plant Extracts (*In vivo* Studies)

In In vivo studies, ten grams of maize seeds were kept in each 250 ml Erlenmeyer flask and spiked with 3ml of aflatoxins (with concentration B1 100µg L−<sup>1</sup> and B2 50µg L−<sup>1</sup> ) according to the method described by Das and Mishra (2000) with some modification. These samples were then incubated with 10 ml of T. ammi seeds and leaves aqueous extract at 30◦C for 72 h.

After incubation, aflatoxins extraction was performed according to the method described by Stroka et al. (2000) with some modifications. Maize samples were extracted with water–acetonitrile (15: 85 v/v) and incubated on shaking water bath for 2 h. After incubation, the extracts were filtered through filter paper (Whatman, Inc., Clifton, NJ, USA). Immunoaffinity columns were conditioned with double distilled water. Then, the filtrate was passed through the column in a solid phase extraction assembly. Toxins were slowly eluted from the column with 1mL methanol in a glass vial. The residual toxin was qualitatively and quantitatively analyzed by TLC and HPLC respectively. Controls consist of untreated maize sample, sample with toxin without T. ammi extract, sample with T. ammi extract without toxin. Experiments were done in triplicate.

# Detection and Quantification of Treated Toxin

The detection and qualification of residual toxin was determined by thin layer chromatography (TLC) according to the method described by Ramesh et al. (2013) with some modifications.

Twenty microliters of chloroform methanol fraction of treated and control samples were spotted on 0.25 mm silica gel 60F<sup>254</sup> (20 × 20 cm, Merck) TLC plate and developed in chloroform: acetone (92:8 v/v). The developed plates were viewed under UV light at 365 nm.

Quantitative analysis of treated and untreated toxin was done by using High Performance Liquid Chromatography (HPLC) after derivatization. Derivatization was carried out as described by Hernandez-Hierro et al. (2008) with some modifications. For this purpose, elute was evaporated to dryness with gentle stream of nitrogen, redissolved in 200µL of n-Haxane, vortexed for 30 s. Next, 50µL of trifluroacetic acid (TFA) was added to it. Finally, 950µL of acetonitrile-water (1:9) was added to above solution and filtered by using syringe filter assembly. The filtrate was analyzed by HPLC.

A HPLC system (Agilent 1100 series, Agilent Technologies, Santa Clara, CA, USA) with a reversed- phase C18 column (Merck, Darmstadt, Germany) and a fluorescence detector was used for quantification. Mobile phase consisting of water: methanol: acetonitrile in the volume ratio 60:20:20 at a flow rate of 1 mL/min was applied and aflatoxin was detected at excitation and emission wavelengths of 360 and 440 nm respectively. For HPLC method validation, calibration curves were drawn using a series of calibration solutions in methanol. Each standard solution was chromatographed in duplicate. Further, identification of degraded toxin metabolites was carried out by mass spectral studies.

#### LCMS Analysis of Degraded Toxin

Toxin products were analyzed by using surveyor LC system equipped with mass spectrophotometer and PDA plus detectors (Thermo Fisher Scientific). The system was validated with known standards individually and in mixture form. All analysis were performed in triplicate using luna phenomenex C<sup>18</sup> column (150 × 4.6 mm, 3µm), in isocratic mode. Following are the LC-MS conditions for Aflatoxins. Injection volume was 10µL. The mobile phase consisted of Methanol: Acetonitrile: Water (22.5: 22.5: 55.0 v/v). Column temperature was maintained at 30◦C. The total operation time was 25 min with the flow rate of 0.5 mL min−<sup>1</sup> . MS conditions were as follows: capillary temperature was 335◦C, sheath gas flow and Auxiliary gas flow was 20 L min−<sup>1</sup> and 4 L min−<sup>1</sup> respectively. Source voltage, capillary voltage and tube lens voltage was 5 KV, 49 V, and 120 V respectively. Toxins incubated with water instead of plant extracts, under optimized conditions of pH (8.0) and temperature (30◦C) were run as control in LCMS analysis.

#### ESI—MS/MS Conditions for Aflatoxins through Direct Insertion Pump

Samples were further analyzed by mass spectrometer with electrospray ionization (ESI) to predict the molecular formulae as well as elemental composition of degraded products of AFB1 and AFB2. Mass spectrometery/Mass spectrometery was performed on a Thermo Scientific LTQ XL System fitted with electrospray ionization (ESI) source operating in positive ionization mode

with optimum conditions set as follows: capillary voltage to 49.0 V, source voltage to 5.0 KV, Tube lens voltage to 110 V, and capillary temperature to 275◦C. Sheath and auxiliary gas flow were adjusted to get stable spray i.e., 3 L min−<sup>1</sup> and 0.4 L min−<sup>1</sup> respectively. Data were collected in positive mode within the range of 100–500 m/z. The final identification of an unknown compound was based on the accurate mass measurement of parent and fragments ions, as well as other useful MS/MS spectrum information (Wang et al., 2011). Untreated toxins (AFB1 and AFB2) and water treated toxins were run as control in MS/MS experiments.

#### Testing Biological Toxicity of Degraded Products

The biological toxicity of degraded toxin products was tested using brine shrimps (Artemia salina) bioassay. The procedure for the bioassay generally followed the method developed by Solis et al. (1993) with some modifications. Brine shrimps dry eggs were procured from local market. 100–200 mg of shrimps eggs were hatched in artificial sea water (34 g sea salt/L of deionized water) by incubation under 60 W lamp, providing direct light and warmth (26◦C). After an incubation period, the hatched nauplii were separated from shells and transferred to fresh sea water.

300µl of treated and untreated AFB1 (100µg L−<sup>1</sup> ) and AFB2 (50µg L−<sup>1</sup> ) solution was added to 96 well plate separately and dried overnight. After complete evaporation of solvent, toxins were re-dissolved in 200µL of sea water. 200µL of sea water containing 40–45 organisms were pipetted into each well, resulting in a final volume of 400µL and incubated for 24–96 h at 26◦C. Mortality was determined by counting the immobile (dead) larvae under stereoscope microscope. Toxicity of each solution was evaluated in triplicate.

#### Statistical Analysis

Results obtained in various experiments were subjected to statistical analysis by using DSSTAT software. Data were analyzed by analysis of Variance (ANOVA) and differences among the means were determined for significance at P ≤ 0.05 using Tukey's multiple range test.

# RESULTS

#### Effect of Temperature and Incubation Period on Toxin Detoxification by *T. ammi* Extracts (*In vitro*)

Aqueous extracts of Trachyspermum ammi leaves and seeds were evaluated for their ability to detoxify aflatoxin B1 and B2 at different temperatures and incubation time. The extent of detoxification was compared with that of control under same conditions. Time course study of toxin degradation showed that detoxification of AFB1 and AFB2 started within 3 h of incubation and percentage of degradation increased with increase in incubation time (**Table 1**). Results indicated that at lowest tested temperature of 25◦C, aqueous extract of T. ammi seeds showed higher detoxification of AFB1 and AFB2 after 3 h of incubation i.e., 52.4 and 69.3% respectively as compared to T. ammi leaves extract. This percentage of detoxification increased with increase in incubation time to 72 h that detoxified AFB1 to 79.8% and AFB2 up to 78.9%.

It is evidenced from the results that percentage detoxification of AFB1 and AFB2 by T. ammi extracts was progressively increased with the consequent increase in temperature from 30 to 55◦C. The highest inactivation was observed at 60◦C. At this temperature, respective control (water) showed 10.4 and 8.6% detoxification of AFB1 and AFB2 after 72 h of incubation. However, toxin treated with T. ammi leaves and seeds extract showed 79.46 and 94.7% detoxification of aflatoxin B1 while detoxification of aflatoxin B2 was 66.6 and 90.1% respectively, under same conditions. This detoxification may be due to synergistic effect of heat and moisture (**Table 1**).

In the present investigation, it was found that T. ammi seeds extract was effective in detoxifying aflatoxin B1 and B2 at all tested temperatures and incubation periods. However, for further studies, 30◦C was selected as it is more or less near to room temperature and moreover was found close to be the existing temperature of storehouses in Punjab especially in summer. Therefore, by selecting this temperature cost of maintaining temperature in storehouses can be greatly reduced.

## Effect of pH on Toxin Detoxification by *T. ammi* Extracts (In vitro)

The comparative assessment of T. ammi leaves and seeds extract to detoxify AFB1 and AFB2 at different pH values revealed that least significant detoxification occurred at pH 2. Results indicated that at pH 2, percentage reduction of AFB1 and AFB2 was 81.2 and 74.9% respectively in samples treated with T. ammi seeds extract after 72 h of incubation at 30◦C.The efficacy of T. ammi seeds extract to detoxify AFB1 and AFB2 was significantly (P < 0.05) increased with increase in pH from 4 to 10 (**Figure 1**). Distilled water with pH adjusted to 2, 4, 6, 8, and 10 was used as a control. Control data showed that at pH 10, 19.94% of AFB1 and 17.45% of AFB2 was degraded after 72 h of incubation at 30◦C while 15.88 and 13.05% degradation of AFB1 and AFB2 was observed at pH 8 under same conditions. The percentage of degradation decreases as the pH decreases to neutral or acidic range.

Maximum degradation of AFB1 and AFB2 was observed at pH 10 after treatment with T. ammi seeds extract i.e., 94.9 and 94.1% as compared to T. ammi leaves extract with degradation percentage of 76.2 and 61.9% respectively.

However, at high basic pH conditions aflatoxins are known to become unstable and sensitive, therefore to avoid this in further experimentations pH 8 was selected which is 100 times less alkaline than pH 10. It was also evidenced from the results that T. ammi seeds extract significantly (P < 0.05) detoxified AFB1 and AFB2 at slightly alkaline pH 8 and their results were closely comparable with the results obtained at pH 10 (**Figure 1**).

# Effect of Boiling on Toxin Detoxification Properties of *T. ammi* Extracts

The aflatoxins detoxification efficacy of T. ammi leaves and seeds extract was significantly (P < 0.05) decreased upon boiling



Capital alphabetic letters with different values indicate significant differences (p < 0.05) among controls and tested plant extracts at different temperature

parenthesis.

 and incubation period in each column. Standard deviation values are shown in

at 100◦C for 10 min. The pH of treated (boiled) and untreated (Unboiled) extracts was adjusted to 8. Unboiled T. ammi leaves extract showed 73.0 and 60.6% detoxification of AFB1 and AFB2 as compared to boiled extracts i.e., 54.4 and 45.7% respectively. Similarly 92.8 and 91.9% detoxification of AFB1 and AFB2 was recorded after treatment with unboiled T. ammi seeds extract in comparison with boiled seeds extract with 69.9 and 74.4% detoxification. Results indicated that upon boiling, AFB1 detoxifying activity of T. ammi seeds and leaves extract were decreased up to 23 and 18% respectively. While in case of AFB2, 15 and 17.4% decrease in detoxification was recorded after treatment with boiled extracts of T. ammi leaves and seeds respectively. So, these results clearly depicted that unboiled plants extracts are more efficient in degrading aflatoxins as compared to boiled extracts.

# *In vivo* Detoxification of Aflatoxins in Maize Samples

In vivo analysis followed a similar trend as that was recorded in in vitro studies. These studies were carried out under conditions optimized in previous in vitro assays i.e., pH 8, temperature 30◦C and incubation time 72 h. Data obtained from in vivo studies showed that maximum detoxification of AFB1 and AFB2 in spiked maize samples was carried out by T. ammi seeds extract after 72 h of incubation i.e., 89.6 and 86.5% respectively. As compared to T. ammi seeds extract, in T. ammi leaves extract 68.8 and 53.7% detoxification of AFB1 and AFB2 was observed in spiked samples (**Table 2**).

HPLC chromatograms confirmed that after T. ammi seeds extract treatment, trace amount of aflatoxin was present along with other peaks whose footprints were not found in chromatogram of parent compounds which may be attributed to toxins degradation products (**Figure 2**).

### Structural Characterization of AFB1, AFB2, and Their Degradation Products

Both aflatoxin B1 and B2 exhibited good ESI ionization efficiency in the positive ion mode with molecular base ion at m/z 313.17 and m/z 315.17 for protonated adduct [M+ H]<sup>+</sup> while m/z 335 and m/z 337 for sodium adduct [M+ Na]<sup>+</sup> respectively. Identity of parent compound was validated by its fragmentation into daughter ions. The protonated molecule was chosen as the precursor ion for aflatoxins in the product ion scan mode because the sodium adduct did not exhibit specific fragmentation for any compound.

#### MS/MS Analysis of AFB1 and AFB2

MS/MS spectrum of AFB1 showed that continuous loss of carbon monoxide (CO) was the main fragmentation pathway. Methyl and methanol losses occurred on methoxy group located on side chain of benzene. The double bond equivalence (DBE) of AFB1 was 12 (**Figure 3A**). However, MS/MS fragmentation pathway of AFB2 revealed that daughter ions were formed by loss of carbon monoxide, oxygen, hydrogen and methyl group (**Figure 3B**). The DBE of AFB2 was 11. The identification of degradation products

were based on accurate mass measurement of ions and similar fragmentation pathways with AFB1 and AFB2.

Results showed that in most of the degraded products obtained after treatment with T. ammi seeds extract additional reactions occurred which leads to the loss of double bond in terminal furan ring responsible for toxicity. Structural formulas of possible hypothesized degraded products of AFB1 and AFB2 are shown in **Figures 4A,B**.

#### MS/MS Analysis for Confirmation of Degraded Products of AFB1

The degradation product at m/z 303.08 corresponded to molecular formulae C16H14O<sup>6</sup> was formed due to the elimination of CH<sup>2</sup> and addition of hydrogen atoms. The DBE of C16H14O<sup>6</sup> was less than AFB1 i.e., 10. Loss of H2O, CO2, CO and O was the main fragmentation pathway (**Figure 5**).

The degradation product C16H16O<sup>7</sup> (with 321.25 m/z) was formed due to the addition of hydroxyl group on the terminal furan ring and replacement of methoxy group on the side chain of benzene ring with hydroxyl group. The DBE of C16H16O<sup>7</sup> content was lower than AFB1 i.e., 9. Fragments demonstrated that the loss of oxygen and carbon monoxide was the main fragmentation pathway which was different from that of AFB1 (**Figure 6**).

The degradation product C17H20O<sup>9</sup> (with 369.33 m/z) had more H8O<sup>3</sup> molecules than AFB1. The DBE of C17H20O<sup>9</sup> was 7, which was lower than AFB1 implying that additional reactions occurred on the furan rings. Fragmentation pathway TABLE 2 | *In vivo* detoxification of AFB1 and AFB2 by aqueous extracts of *T. ammi*.


Values are mean of three replicates.

Data were analyzed by analysis of Variance (ANOVA).

Values with different letters show significant difference (P = 0.05) as determined by Tukey's Multiple Range test.

was different from that of AFB1. The precursor ion yielded a series of product ions which were represented by 351.17 [M-H2O]+, 337.08 [M-O2] <sup>+</sup>, 325.25 [M-CO2] <sup>+</sup>, 309.08 [M-CO3] +, 301.33 [M-CH8O3] <sup>+</sup> and 285.17 [M-CH8O4] <sup>+</sup> (**Figure 7**).

The degradation product 345.92 corresponded to molecular formula C17H12O<sup>8</sup> was formed due to the addition of hydroxyl groups on the terminal furan ring. The DBE of C17H12O<sup>8</sup> was same as that of AFB1 i.e., 12. The fragments of C17H12O<sup>8</sup> showed losses of CH4O, H2O, CO<sup>2</sup> and CO. More details on the fragmentation pathway are shown in **Figure 8**.

The degradation product C17H8O<sup>8</sup> (with 341.08 m/z) had two more oxygen and four less hydrogen atoms. Addition of oxygen occurred on the terminal furan ring and benzene ring. The DBE of C17H8O<sup>8</sup> was one more than AFB1 i.e., 13. Series of product ions formed by the precursor ion were represented by 323.17 [M-H2O]+, 295.08 [M-CH2O2] <sup>+</sup>, 291.17 [M-H2O3] <sup>+</sup>, 279.25 [M-CH2O3] <sup>+</sup>, 267.0 [M-C2H2O3] <sup>+</sup>, 263.08 [M-CH2O4] <sup>+</sup> (**Figure 9**).

The degradation product C17H18O<sup>8</sup> (with 351.25 m/z) was formed by addition of two hydroxyl groups on the double bond of terminal furan ring. The DBE of C17H18O<sup>8</sup> was less than AFB1 i.e., 9 with different fragmentation pathway from that of AFB1. Product ions formed from parent ion C17H18O<sup>8</sup> was represented by 337.17 [M-CH2] <sup>+</sup>, 321.33[M-CH2O]+, 301.33[M-CH6O2] +, 281.33[M-CH10O3] <sup>+</sup> and 241.33[M-C2H6O5] <sup>+</sup> (**Figure 10**).

#### MS/MS Analysis for Confirmation of Degraded Products of AFB2

The degradation product C17H14O<sup>5</sup> (with m/z 299.17) was formed by the loss of oxygen atom from the side chain of benzene ring. The DBE of C17H14O<sup>5</sup> was same as that of AFB2. Fragments of C17H14O<sup>5</sup> showed losses of CH3, C2O<sup>2</sup> and CO. More detail on fragmentation pathway are shown in **Figure 11**.

However, the degradation product C17H16O<sup>6</sup> at m/z 317.25 had two more hydrogen atoms and one less DBE than AFB2. Loss of CO2, CO, CH3O and O was the main fragmentation pathway (**Figure 12**).

#### Assessment of Biological Toxicity of Degraded Products

Biological toxicity of degraded toxin products were tested using brine shrimps (Artemia salina) bioassay. The brine shrimps assay actually proved to be a convenient system for monitoring biological activity (Hartl and Humpf, 2000). In this study, the degraded toxin products were incubated with brine shrimp larvae at 26◦C for 24–96 h. The percentage of mortality was compared with that of control (**Table 3**).

Results indicated that only 11.7–23.3% mortality in brine shrimps larvae was observed after treatment with degraded toxin products. However, mortality level was 83.0–91.7% when larvae were incubated with untreated AFB1 (100µg L−<sup>1</sup> ) and AFB2 (50µg L−<sup>1</sup> ), under same conditions. Percentage of mortality was increased with increase in incubation period.

#### DISCUSSION

Various food and feed additives like phenolic compounds and plant extracts can be used to reduce toxic effects of mycotoxins (Nahm, 1995; Dvorska et al., 2007). According to the literature, essential oils and extracts of various spices and herbs like cinnamon, peppermint, basil, lemongrass may be TABLE 3 | Percent mortality of brine shrimps (*Artemia salina*) larvae at 26◦C after treatment with toxin (AFB1 and AFB2) detoxified with *T. ammi* seeds extract at various incubation periods.


Sea water was taken as a control. Other Controls consist of Methanol and untreated toxin AFB1 (100 ng/ml) and AFB2 (50 ng/ml) dried and redissolved in sea water.

Values are mean of three replicates.

Data were analyzed by analysis of Variance (ANOVA).

recommended as a plant based safe food additive in protecting the food from deteriorating fungi as well as from aflatoxin contamination (Montes-Belmont and Carvajall, 1998; Burt, 2004; Yang et al., 2007). Similarly, there are several mycotoxins binding commercial products. Some of them are developed and approved in North America and Western Europe, such as Mycofix <sup>R</sup> and Mycosorb <sup>R</sup> . These binders are the combination of various things including herbal and yeast cell wall component extracts (Marroquin-Cardona et al., 2009). They have been used worldwide to neutralize or detoxify the mycotoxins in poultry, pig, ruminant feed as well as fish and shrimp diets. The aforementioned products are being extensively used in Pakistan.

In this present study, both in vitro and in vivo assays were performed with aqueous extracts of T. ammi leaves and seeds to check their aflatoxin B1 and B2 detoxification potential under optimized conditions of temperature, pH and incubation period. The results of in vitro assays showed that the percentage of detoxification by plants extracts increased with increase in temperature to 60◦C but this detoxification could be due to synergistic action of heat and moisture (Basappa and Shantha, 1996; Rustom, 1997). Similarly, Hajare et al. (2005) worked on aflatoxin inactivation by using Ajwain seeds extract under optimized conditions. According to his findings, highest inactivation was observed at 60◦C but further studies were conducted on 45◦C to reduce the effect of heat and moisture on toxin inactivation.

The pH of reaction mixture plays an important role in the process of detoxification by using plant extracts. The maximum detoxification was observed at pH 10 followed by pH 8. The percentage of detoxification decreased as the pH changed to neutral or acidic range. Subsequently, Mendez-Albores et al. (2004) also found that aflatoxin florescence, attributed to the coumarin moiety, diminish or even disappear in alkaline treatment. In addition, the similar results were in accordance with the findings of Kannan and Velazhahan (2014) who explored the potential of Barleria lupulina leaf extract on detoxification of aflatoxins. Results of the present study showed that in vivo decontamination of maize samples followed a similar trend as that was recorded in in vitro studies. These studies were carried out under conditions optimized in previousin vitro assays i.e., pH 8, temperature 30◦C and incubation time of 72 h.

Furthermore, T. ammi plant extracts showed varied degree of reduction in aflatoxin B1 and B2 detoxification upon boiling. Similar findings were recorded in a study conducted by Velazhahan et al. (2010). The reason behind is that the activity of certain plant phytochemicals like phenolics and alkaloids highly reduces upon boiling as described by Momoh et al. (2012), which is might be responsible for the alteration and breakage of the molecular structure of phytochemicals.

After detoxification, structural changes in aflatoxin molecule have been observed in several studies conducted with microorganisms, physical and chemical agents, ultraviolet (UV) rays, Gamma rays and plant products (Alberts et al., 2006; Albores et al., 2008; Guan et al., 2010; Velazhahan et al., 2010; Wang et al., 2011; Farzaneh et al., 2012; Inoue et al., 2013; Luo et al., 2013; Samuel et al., 2014; Vijayanandraj et al., 2014). Aflatoxins have been widely researched for their toxicity by various scientists (Guengerich, 2001, 2008; Hussein and Brasel, 2001). Their toxicity data showed that aflatoxins have cyclopentene ring and furan moiety in their chemical structure. In AFB1 the presence of double bond in the terminal furan ring is key factor for its toxic and carcinogenic activities (Wang et al., 2011). In contrast, aflatoxin B2 which have a saturated furan ring is hundreds times less carcinogenic (Dvorackova, 1990). The degraded products of AFB2 may be active but were less potent than that of the parent compound. Thus, removing the double bond of terminal furan ring is a major aim of detoxification. In this present study, among six degraded products of AFB1 acquired after T. ammi seeds extract treatment, 50% products (with m/z 303, 341, 351) showed removal of double bond in the terminal furan ring while in products with m/z 369 and 321 both lactone group modification and double bond removal in the terminal furan ring occurred. Therefore, toxicity of most of the degraded products compared with that of aflatoxin was reduced to a much lower level.

Biological toxicity of degraded toxin products were tested using brine shrimps (Artemia salina) bioassay. There are several studies on the effects of aflatoxin on the brine shrimp (Artemia salina) eggs and larvae (Harwing and Scott, 1971; Schmidt, 1989; Logrieco et al., 1996; Durakovic et al., 2002; Moretti et al., 2007). According to previous findings (Brown, 1969; Hartl and Humpf, 2000; Favilla et al., 2006) the Artemia salina larvae appears to be as susceptible as biological indicator of toxicity of some mycotoxins in foods and feeds. The results of present investigation showed significant reduction in larval mortality after incubation with treated toxins as compared to untreated toxins. Similar findings were also observed by Samuel et al. (2014), who worked on detoxification of aflatoxin B1 by Pseudomonas putida. He compared the toxicity of treated and untreated AFB1 toward HeLa cells and concluded that degraded products are nontoxic (D1) or much less toxic (D2 and D3) than AFB1 to the cells at the tested concentrations.

From the findings of present investigation, T. ammi seeds extracts can be used for development of biologically safe herbal additives to food and feed products processed for human consumption to avoid the toxic effects of aflatoxins. Based on previous literature, use of ajwain in food showed no safety issues (Gemeda et al., 2014). Direct spray of aqueous plant extract is convenient for the farmers because these can be easily prepared and its application does not require any technical knowledge. However, there may be some limitations regarding formulations and shelf life of extract and research in this direction is needed.

### AUTHOR CONTRIBUTIONS

WI: PhD student who performed all the experimental work. TA: PhD supervisor, who guided and planned this project. MI:

#### REFERENCES


Provided expertise and equipment for mass spectroscopy. AG: Helped in data analyses. MA: provided lab facilities for high performance liquid chromatography.

#### FUNDING

This project did not received any funding from any agency, but was run on indigenous resources of Fungal biotechnology lab, Institute of agricultural sciences, University of the Punjab, Pakistan.

#### ACKNOWLEDGMENTS

Authors are extremely thankful to Dr. MI (National Institute of Biotechnology and Genetic Engineering), Dr. AG (University of Engineering and Technology) and Dr. MA (University of Veterinary and Animal Sciences) for the provision of specialized apparatus and technical guidance for successful accomplishment of this study.


Artemia salina by Fusarium species Gibberella fujikuroi species complex. Inter. J. Food Microbiol. 118, 158–163. doi: 10.1016/j.ijfoodmicro.2007. 07.004


Yazdani, D., Zainal-Abidin, M. A., Tan, Y. H., and Kamaruzaman, S. (2010). Evaluation of the detection techniques of toxigenic Aspergillus isolates. Afr. J. Biotechnol. 9, 7654–7659. doi: 10.5897/AJB10.1128

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Iram, Anjum, Iqbal, Ghaffar and Abbas. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Analysis of Multilocus Sequence Typing and Virulence Characterization of *Listeria monocytogenes* Isolates from Chinese Retail Ready-to-Eat Food

Shi Wu1, 2, Qingping Wu<sup>1</sup> \*, Jumei Zhang<sup>1</sup> , Moutong Chen<sup>1</sup> and Weipeng Guo<sup>1</sup>

<sup>1</sup> Guangdong Institute of Microbiology, State Key Laboratory of Applied Microbiology Southern China, Guangdong Provincial Key Laboratory of Microbial Culture Collection and Application, Guangdong Open Laboratory of Applied Microbiology, Guangzhou, China, <sup>2</sup> School of Bioscience and Bioengineering, South China University of Technology, Guangzhou, China

#### *Edited by:*

Giovanna Suzzi, Università degli Studi di Teramo, Italy

#### *Reviewed by:*

Luca Settanni, Università degli Studi di Palermo, Italy Fatih Ozogul, Çukurova Üniversity, Turkey Rosanna Tofalo, Università degli Studi di Teramo, Italy

> *\*Correspondence:* Qingping Wu wuqp203@163.com

#### *Specialty section:*

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

*Received:* 20 October 2015 *Accepted:* 01 February 2016 *Published:* 16 February 2016

#### *Citation:*

Wu S, Wu Q, Zhang J, Chen M and Guo W (2016) Analysis of Multilocus Sequence Typing and Virulence Characterization of Listeria monocytogenes Isolates from Chinese Retail Ready-to-Eat Food. Front. Microbiol. 7:168. doi: 10.3389/fmicb.2016.00168 Eighty Listeria monocytogenes isolates were obtained from Chinese retail ready-to-eat (RTE) food and were previously characterized with serotyping and antibiotic susceptibility tests. The aim of this study was to characterize the subtype and virulence potential of these L. monocytogenes isolates by multilocus sequence typing (MLST), virulence-associate genes, epidemic clones (ECs), and sequence analysis of the important virulence factor: internalin A (inlA). The result of MLST revealed that these L. monocytogenes isolates belonged to 14 different sequence types (STs). With the exception of four new STs (ST804, ST805, ST806, and ST807), all other STs observed in this study have been associated with human listeriosis and outbreaks to varying extents. Six virulence-associate genes (inlA, inlB, inlC, inlJ, hly, and llsX) were selected and their presence was investigated using PCR. All strains carried inlA, inlB, inlC, inlJ, and hly, whereas 38.8% (31/80) of strains harbored the listeriolysin S genes (llsX). A multiplex PCR assay was used to evaluate the presence of markers specific to epidemic clones of L. monocytogenes and identified 26.3% (21/80) of ECI in the 4b-4d-4e strains. Further study of inlA sequencing revealed that most strains contained the full-length InlA required for host cell invasion, whereas three mutations lead to premature stop codons (PMSC) within a novel PMSCs at position 326 (GAA→TAA). MLST and inlA sequence analysis results were concordant, and different virulence potentials within isolates were observed. These findings suggest that L. monocytogenes isolates from RTE food in China could be virulent and be capable of causing human illness. Furthermore, the STs and virulence profiles of L. monocytogenes isolates have significant implications for epidemiological and public health studies of this pathogen.

Keywords: *Listeria monocytogenes*, MLST, *inlA*, PMSC, virulence genes, epidemic clones

# INTRODUCTION

Listeria monocytogenes is a gram-positive, facultative intracellular bacterium that is responsible for listeriosis. Listeriosis can cause meningitis, newborn septicemia, encephalomyelitis, or even death in humans, especially in the elderly, pregnant women, or newborn. Every year, ∼1591 cases of listeriosis in humans are report, with a 19% case-fatality rate in the United States (Scallan et al., 2011). In the European Union, a total of 1763 confirmed human cases of listeriosis (notification rate of 0.44 cases per 100,000 population) were reported in 2013 (EFSA, 2015). As an important foodborne pathogen, it is widespread in nature and lives naturally in plants and soil environments. Its ability to survive and grow over a wide range of environmental conditions, including refrigeration temperatures, high salt concentration and low pH, makes it a potential hazard in foods (Ryser and Marth, 2007).

Differences in virulence between L. monocytogenes strains may also influence infection and clinical outcome. Some strains are highly pathogenic and sometimes deadly, whereas others are less virulent or even avirulent and produce little harm in the host (Olier et al., 2002). Several methods have been used to differentiate L. monocytogenes strains. Based on somatic (O) and flagellar (H) antigens represents a conventional approach to understanding L. monocytogenes isolates ecological distribution and epidemiology. However, there are 13 serotypes and only four serotype (1/2a, 1/2b, 1/2c, and 4b) cause almost all cases of listeriosis in human. To further discriminate these strains, numerous molecular methods have been developed for epidemiological investigations and of help for the surveillance and control of listeriosis. Multilocus sequence typing (MLST) method is one of the most robust tools for investigating the global epidemiology of microbial populations (Sullivan et al., 2005). Based on the sequencing of seven housekeeping genes, it is highly discriminatory and provides unambiguous results that can be comparable directly among laboratories via the internet. In recent years, MLST has evolved to the reference method for global epidemiology and population biology (Maiden, 2006; Ragon et al., 2008).

Certain virulence and virulence-associated genes also play very important roles in intracellular survival, cell-to-cell spread, and virulence of L. monocytogenes. The presence of a number of virulence factors such as surface-associated internalins, listeriolysin O, and listeriolysin S (LLS) in L. monocytogenes significantly regulates its pathogenicity (Cotter et al., 2008; Shen et al., 2013). Of which, internalin A (InlA), encode by inlA, is responsible for facilitating the entry of L. monocytogenes into nonprofessional phagocytic cells expressing the human isoform of E-cadherin (Lecuit et al., 2001). It has been shown that mutations in inlA leading to a premature stop codon (PMSC) significantly reduce the invasion of the strain to human epithelial cells (Nightingale et al., 2005a). In addition, a small number of closely related L. monocytogenes clones have caused multiple outbreaks worldwide. These epidemic clones (EC) are divided into seven groups: ECI, ECII and ECIV within serotype 4b, ECIII, ECV and ECVII in serotype 1/2a, and ECVI within serotype 1/2b (Kathariou, 2003; Knabel et al., 2012; Lomonaco et al., 2013).

Since the main food linked to listeriosis outbreaks was readyto-eat (RTE) food, the consumer had limit opportunities to destroy the pathogen before the food consumed. Therefore, the risk associated with strains isolated from RTE food may be more severe. In our previous study, 80 RTE L. monocytogenes isolates were examined (Wu et al., 2015a). In order to determine the potential pathogenic profile and relative risk of these RTE L. monocytogenes isolates, this study provide a phylogenetic framework based on MLST analysis of L. monocytogenes isolates and evaluate their virulence-genes and EC-specific markers. In addition, we analyzed the full-length sequences of important virulence factor (inlA) to investigate correlations among the MLST types, amino acid sequence of inlA, and virulence potential.

# MATERIALS AND METHODS

# Bacterial Isolates

A total of 80 L. monocytogenes isolates were collected from retail ready-to-eat food samples in 24 Chinese cities were analyzed, comprising 27 isolates from cold vegetable dishes in sauce, 7 isolates from cold noodles dishes in sauce, 17 isolates from roast chicken, 12 isolates from roast duck, 11 isolates from roast pork, and 5 isolates from pasteurized milk (**Table 1**). The isolates were obtained between September 2012 and January 2014 according to the GB 4789.30-2010 of food microbiological examination of L. monocytogenes (National Food Safety Standards of China) with slight modifications and the most probable number (MPN) method (Gombas et al., 2003). They were identified by Gram stain, catalase, and oxidase tests and Micro ID Listeria identification system (Microgen, Camberley, UK). Serovars and antibiotic susceptibility were confirmed in previous study (Wu et al., 2015a). Additionally, 35 isolates were selected for the sequencing analysis of inlA, based on MLST profile, molecular serogroup, food origin, and antimicrobial resistance. Each isolate was incubated at 37◦C overnight on TSB-YE (Tryptic soy agar with yeast extract). Genomic DNA was extracted using a Genomic DNA Extraction kit (Dongsheng Biotech, Guangzhou, China) according to the manufacturer's instructions. The concentration of genomic DNA was determined at 260 nm using a NanoDrop-ND-1000 UV-Vis Spectrophotometer (Thermo Fisher Scientific, MA, USA).

#### Multilocus Sequence Typing

The MLST scheme used to characterize L. monocytogenes isolates is based on the sequence analysis of the following seven housekeeping genes: abcZ (ABC transporter), bglA (betaglucosidase), cat (catalase), dapE (succinyl diaminopimelate desuccinylase), dat (D-amino acid aminotransferase), ldh (Llactate dehydrogenase), and lhkA (histidine kinase) (Salcedo et al., 2003). The PCR amplification conditions were as follow: an initial cycle of 94◦C for 4 min; 35 cycles of 94◦C for 30 s, 52◦C for 30 s (45◦C for bglA), 72◦C for 2 min, and a final extension at 72◦C for 10 min. The DNA fragments were purified using a PCR purification kit (Qiagen, Genman) and were sequenced in each direction with Big Dye fluorescent terminators on an ABI3730XL sequencer (Applied BioSystems).

#### TABLE 1 | Characteristics of 80 *L. monocytogenes* strains isolated from retail ready-to-eat food in this study.


(Continued)

#### TABLE 1 | Continued


<sup>a</sup>SUS, susceptibility; P, penicillin G (5U); AMP, ampicillin (10µg); KF, Cephalothin (30µg); C, chloramphenicol (30µg); CIP, Ciprofloxacin (5µg); DA, Clindamycin (2µg); E, Erythromycin (15µg); CN, gentamicin (10µg); K, Kanamycin (30µg); MEZ, Mezlocillin (30µg); RD, Rifampicin (5µg); S, Streptomycin (25µg); TE, tetracycline (30µg); VA, Vancomycin (30µg). The serovars and antibiotic susceptibility of L. monocytogenes isolates were detected in previous study by Wu et al. (2015a).

# Determination of Virulence Genes and Epidemic Clone

All primers and PCR conditions are presented in Supplementary Table 1.

The isolates were identified using duplex PCR detection containing hly (707 bp) and inlB (367 bp) genes, which are specific to L. monocytogenes, as previously described (Xu et al., 2009). Multiplex PCR (Liu et al., 2007) was used to determine the presence of the virulence genes inlA, inlC, and inlJ. The llsX gene was detected by PCR assays to identify the LLS-positive L. monocytogenes strains (Clayton et al., 2011). Presumptive major ECs (ECI, ECII, and ECIII) were found in the isolates as described previously (Chen and Knabel, 2007).

#### *inlA* Gene Sequencing

The 2400 bp long inlA gene was sequenced in 35 isolates representing the clonal diversity of L. monocytogenes. External primers were used to amplification covering the whole inlA ORF and internal primers for sequencing (Supplementary Table 2). The inlA sequences were assembled using Seqman (DNASTAR, Lasergene). Mutation types were determined according to the site of mutation that leads to PMSC in inlA (Nightingale et al., 2005a) and by comparing the obtained inlA sequence data to that of the L. monocytogenes EGDe reference strain (Glaser et al., 2001).

### Data Analysis

For each MLST locus, an allele number was given to each distinct sequence variant, and a distinct sequence type (ST) number was attributed to each distinct combination of alleles among the seven genes. Sequence types (STs) were determined by using the L. monocytogenes MLST database (http://bigsdb. web.pasteur.fr/listeria/listeria.html). Sequence Type Analysis and Recombinational Tests software (S.T.A.R.T. ver.2; http:// pubmlst.org/software/analysis/start2) was used to analyze the data of MLST.

Nucleotide diversity (π, average pairwise nucleotide difference/site; and k, average pairwise nucleotide difference/sequence), number of polymorphic sites, number of mutations, number of alleles, G+C content, Tajima's D test for neutrality, number of synonymous mutations (Ks), number of nonsynonymous mutations (Ka), and the Ks/Ka ratios [the number of nonsynonymous substitutions/nonsynonymous site (Ks) to the number of synonymous substitutions/synonymous site (Ka)] with a Jukes and Cantor correction were calculated using DnaSP version 5.10.01. Sequence analysis of inlA was performed using the ClustalX algorithm (version 1.83) (Thompson et al., 1997), which was followed by phylogenetic analysis using the maximum likelihood algorithm in MEGA 6 (version 6.05) (Tamura et al., 2011).

# RESULTS

## Multilocus Sequence Typing of *L. monocytogenes* Isolates

MLST detected a total of 14 different sequence types in the 80 isolates, including four new STs (ST804, ST805, ST806, and ST807) (**Table 2**). The most common allelic profile was ST8 (24/80, 30% of isolates) followed by ST1 (16/80, 20% of isolates) independently of the isolates source. With the exception of ST5 and ST805, the remainder of the STs included more than one isolate. Of the STs, ST-87, ST5, ST224, ST804, ST805, and ST3 belonged to serovar II.2 (1/2b-3b-7), ST8, ST7, ST122, ST155, and ST806 belonged to serovar I.1 (1/2a-3a), ST1 and ST807 belonged to serovars II.1 (4b-4d-4e), and ST9 belonged to serovar I.2 (1/2c-3c). Based on each ST matches at least one other ST at ≥6 of the 7 loci, three clonal complexes (CCs) were identified, including CC1 (ST1, ST807), CC3 (ST3, ST804), and CC155 (ST155, ST806). A phylogenetic tree based on the seven concatenated MLST sequences (**Figure 1**) shows the relatedness between the RTE strains. STs correlated well with serotype and lineage. Most isolates recovered from same city were clustered into one type. Furthermore, it should be noted that the multidrug resistant strains 1194-4LM, 1330-2LM, and 1342-1LM were belonged to new STs (ST805, ST806) indicating that these isolates were genetically diverse from other isolates.

The evolutionary characteristics of the seven housekeeping genes among the isolates were analyzed using the software dnasp5, which can distinguish between randomly (i.e., neutrally)

TABLE 2 | Allelic profile (STs) of RTE *L. monocytogenes* isolates for MLST.


and non-randomly evolving DNA sequences. Tajima's D test indicated that abcZ, cat, dapE, dat, ldh, and lhkA evolved randomly, whereas bglA evolved non-significantly (0.9151, P > 0.10) (**Table 3**). However, seven gene portions harbored a total of 195 polymorphisms (5.93%; range 3.51–12.1% per gene). The average nucleotide diversity π was 2.6%, ranging from 0.9 to 6.2% per gene. The GC% observed in all alleles ranged from 36.4 to 42.9%, which was consistent with the 39% value observed across the entire L. monocytogenes EGDs genome (Glaser et al., 2001).

#### Prevalence of Virulence Associated Genes and EC Markers

Eighty isolates of L. monocytogenes from retail RTE food were detected for the presence of virulence genes. The result showed that all isolates harbored listeriolysin O genes (hly) and internalin genes (inlA, inlB, inlC, and inlJ), and 38.8% (31/80) of isolates harbored listeriolysin S genes (llsX) (**Table 4**). The llsX-positive isolates including twenty-one 4b-4d-4e isolates, eleven 1/2b-3b-7 isolates. EC markers were identified in 26.3% (21/80) of isolates by multiplex PCR. ECI was the only EC marker identified and was observed in 4b-4d-4e isolates from roast meat (roast chicken/duck/pork) in this study. These EC isolates were also found to be positive for llsX.

# Analysis of Selected Isolates for *inlA* Sequence

Sequencing the full 2400 bp inlA ORF in 35 L. monocytogenes isolates yielded 14 different inlA allelic types, indicating 88.6% of gene diversity. A total of 126 (5.2%) sites were polymorphic, including 85 sites with synonymous substitutions and 41 sites with non-synonymous substitutions. The average number of nucleotide differences per site between two sequences (π) was 0.01907; the average number of nucleotide differences (k) was 45.834. The GC% observed in inlA was 37.1%.

The phylogenetic tree generated using the inlA sequences of 35 RTE isolates and the L. monocytogenes EGDe reference strain

(NC\_003210.1; serotype 1/2a, ST35) revealed a similar grouping based on MLST analysis (**Figure 2**). The same lineage exhibited close relationships. Truncated forms of InlA have been described and are associated with reduced virulence (Nightingale et al., 2005a; Van Stelten et al., 2010), three distinct inlA alleles were found to harbor PMSC at position 492 (PMSC type 6), carried by ST 121 (1309-1LM, 1/2a-3a), the mutation at position 685 (PMSC type 11) harbored by the serotype 1/2c-3c isolate (2311-1LM, ST9) and a nonsense mutation at position 326 (GAA→TAA) of the inlA gene where a change of glutamic acid codon to a stop codon occurs showed a novel mutation type of PMSC (1159-3LM, 1/2c-3c, ST9).

Comparison of the amino-acid sequences of inlA between 35 L. monocytogenes isolates and the homologous sequence type strain EGDs, 13 internalin A variants were grouped. Each STs showed same variants, excluding CC1 (ST1, ST807), CC3 (ST3, ST804), and ST5 and ST805. Isolates from ST1 divided into two variants. In total, 40 amino acids were substituted (40/800, 5%).



Ks, number of synonymous mutations; Ka, number of non-synonymous mutations; π, Nucleotide diversity.

\*Indicate statistical differences of P < 0.05.

\*\*Indicate statistical differences of P < 0.01.

\*\*\*Statistical differences of P < 0.001.

#### TABLE 4 | Prevalence of virulence markers in 80 RTE *Listeria monocytogenes* strains.


Most of the substitutions (31/40, 77.5%) occurred in the leucinerich repeats (LLR), the inter-repeat (IR) domain and the B-repeat domain (**Figure 3**). The most conserved region of InlA among the 35 RTE L. monocytogenes isolates was the LRR domain. In parallel, a previous study (Ragon et al., 2008) reported the high constraint of the LRR-region and the moderate constraint of the IR- and B-repeat regions.

#### DISCUSSION

As an important foodborne pathogen, L. monocytogenes remains a signification public health and food safety threat worldwide. In China, many studies of the prevalence of L. monocytogenes in food have been reported (Chen et al., 2009, 2014), but the outbreaks and infection of listeriosis in food are very limited. However, not all L. monocytogenes isolates have an equal capacity to cause disease. Therefore, it is important to investigate the molecular characteristics and virulence potential of L. monocytogenes isolates for designing and implementing more effective prevention strategies. The RTE isolates analyzed in this study were isolated from most of provincial capitals of China (Wu et al., 2015a), which could be better understanding the characterization of this important pathogen in China.

To enable increased characterization, MLST can be complemented by the nucleotide sequence determination of one or more highly diverse genes, such as those encoding antigens or antibiotic resistance determinants (Sullivan et al., 2005). In this study, 80 L. monocytogenes RTE isolates belonged to limited number of major clones. Except four new STs (ST804, 805, 806, 807), the other detected STs were already described in the L. monocytogenes Institute Pasteur MLST database (http://bigsdb.web.pasteur.fr/perl/bigsdb/bigsdb.pl? db=pubmlst\_listeria\_isolates\_public&page=profiles): 327 strains of ST1 (51% from human isolates), 253 strains of ST3 (37% from human and 55% from food and environment), 143 strains of ST7 (63% from animal and feed), 37 strains of ST8 (including 15 human and 10 food), 130 strains of ST9 (35% from human and 21% from food), 52 strains of ST5 (50% from human and 23% from food), 22 strains of ST87 (55% from human and 23% from food), 70 strains of ST155 (50% from human), 6 strains of ST224 (2 isolates from human and 4 from environment), and 74 strains of ST121 (54% from food). Strains of all STs observed in our study have been associated to various extents with human listeriosis and outbreaks, indicating that L. monocytogenes strains of these STs have at least a theoretical pathogenic potential. It is worth nothing that ST1, comprising 20% of the isolates in this study, have the same sequences as F2365 (a serotype 4b isolates from a soft cheese outbreak in California in 1985), CLIP12848 (a serotype 4b isolates from outbreak in France in 1989), and CLIP68868 (a serotype 4b isolates from outbreak in Sweden 1995) which should the worthy of the attention of the food hygiene supervision department. Interestingly, novel STs (ST805, ST806) were found in multidrug resistant strains (**Figure 1**), showing some allele numbers had exchanged. Multiple resistance in L. monocytogenes was linked to the presence of a self-transferable plasmid that was proposed to originate in Enterococcus-Streptococcus (Charpentier and Courvalin, 1999). Therefore, the self-transferable plasmid may also impact the homologous recombination of housekeeping genes. However, further research is needed to determine the reason underlying this correlation.

The presence of hly, inlA, inlB, inlC, inlJ, and llsX was evaluated in L. monocytogenes isolates recovered from patients, food and the environment (Kathariou, 2003; Wieczorek et al., 2012; Wu et al., 2015b). In this study, PCR based analysis of the 80 RTE isolates showed that most of these isolates possessed virulence genes similar to those of clinical isolates (Mammina et al., 2009). The hly gene is encoding the listeriolysin O (LLO), the determinant that is required for the disruption of the phagocytic vacuole and the release of bacteria into the cytoplasm, a prerequisite for their intracellular proliferation. Therefore, LLO is an essential virulence factor and its absence leads to total avirulence (Portnoy et al., 1992). L. monocytogenes adheres to and actively through actions of internalins, a complex family of LRR-containing proteins (Bierne et al., 2007). In this study,

we selected inlA, inlB, inlC, and inlJ for tested internalins in L. monocytogenes isolates. These genes are claimed to play a role in pathogenesis of human listeriosis. Besides, the llsX gene (encoding LLS) was employed as a genetic marker to detect L. monocytogenes with Listeria pathogenicity island 3 (LIPI-3) (Cotter et al., 2008), and strains from the present study exhibited 38.8% (31/80) positivity. The LLS, which is present in a subset of strains of lineage I, is induced only under oxidative stress conditions and contributes to murine virulence and in survival in polymorphonuclear neutrophils (Cotter et al., 2008). Overall, our result showed that the positive-llsX strains were found exclusively in lineage I (shown **Table 3** and **Figure 1**), the evolutionary lineage of L. monocytogenes that contributes to the majority of spontaneous and epidemic outbreaks of listeriosis (Jeffers et al., 2001). Besides, most of positive-llsX strains involving ECI (21/32, 65.6%) have been related to several major outbreaks in the US and Europe in the past (Kathariou, 2003; Liu, 2008). In general, ECs is defined as a small number of isolates of a presumably common ancestor (Chen and Knabel, 2007). Specific ECs continue to be associated with sporadic cases around the world (Neves et al., 2008; Mammina et al., 2009; Lomonaco et al., 2013). Recently, Cantinelli et al. (2013) conclude that the "epidemic clone" denominations represent a redundant but largely incomplete nomenclature system for MLST-defined clones, which must be regarded as successful genetic groups that are widely distributed across time and space. In our study, L. monocytogenes isolates of MLST-CC1, including ST1 and ST807 were both ECI which was consistent with this conclusion. However, the high occurrence of multiple virulence-associated genes and ECI markers in L. monocytogenes isolates from Chinese RTE food could pose a significant health risk as these isolates can be pathogenic and potentially capable of causing an epidemic.

Certainly, the presence of these virulence-associated genes does not mean a particular strain is virulent, as these genes are normally present in L. monocytogenes but if absence of these genes lead to avirulence. Over the past decade, previous studies identified naturally occurring PMSC mutations in inlA and have demonstrated that these mutations are responsible for virulence attenuation (Van Stelten and Nightingale, 2008). L. monocytogenes strains with PMSCs were frequently isolated from RTE foods in the US, although inlA PMSCs were markedly underrepresented among ECs (Van Stelten et al., 2010). In our study, the sequencing of inlA genes from isolates revealed that three mutations resulted in PMSCs: type 6, 11 and a novel type at position 326 (GAA→TAA), which means most of strains (32/35, 91.4%) are capable of producing full-length InlA required for host cell invasion. Generally, PMSCs have been reported 30%-45% of prevalence in food but 5% of prevalence in clinical isolates (Jacquet et al., 2004; Van Stelten et al., 2010; Chen et al., 2011). Three mutations were both harbored by the lineage II strains (2 strains with ST9, one strain with ST121) (**Figure 2**), which further confirmed that 1/2a and 1/2c (lineage II) more frequently possess PMSCs than 1/2b or 3b serotypes, with 4b strains rarely having PMSCs (Orsi et al., 2011). PMSC seems to be more frequently observed in ST9 and ST121 isolates as previous reports (Ragon et al., 2008; Ciolacu et al., 2014). ST9 and ST121 include isolates from several different countries and from several different environmental, clinical and food source (Ragon et al., 2008; Parisi et al., 2010; Holch et al., 2013). Some L. monocytogenes strains can persist for a longtime in food processing. The repeatedly lost by convergent evolution in the genetically homogeneous of these genotypes, may be attribute to the selective advantage by the loss of a functional InlA protein or a relaxed selective constraint on maintaining InlA function (Ragon et al., 2008). It is worth mentioning that the novel PMSCs type at position 326 (GAA→TAA), indicated that variation occurred in the LLR domain were located in repeats 10 (**Figure 3**). This domain promotes interaction with human surface receptor, E-cadherin, and has been reported as highly conserved, especially from repeats 7 to 15 (Ragon et al., 2008). To date, at least 18 distinct mutations in inlA leading to PMSCs have been observed in L. monocytogenes isolates worldwide, only four of them (type 5, 16, 17, and 18) were found in LLR (Van Stelten et al., 2010). Compared with the inlA variants, L. monocytogenes retained the ability for localized recombination is clearly provided by the inlA gene coding InlA, in agreement with previous reports (Nightingale et al., 2005b; Orsi et al., 2007). The substitution of some amino acid were consistent between lineage I strains and II, such as lysine replaced into arginine at position 3 in all lineage I strains (**Figure 3**). Additionally, the close relationship between inlA sequencing and MLST typing (**Figure 2**) suggested that housekeeping genes and virulenceassociated genes had some similar genetic characterization, representing monophyletic origin. In other words, our data suggest that L. monocytogenes evolved and generated different STs related to some degree of evolutionary consistent.

In summary, the research findings suggest that L. monocytogenes isolated from Chinese RTE food showed genetic relatedness and virulence attribute. Serotype, lineage, MLST types, even antibiotic resistance showed some degree of genetically relationship. Most MLST sequence types (10/14,

#### REFERENCES


71.4%) found in this study have been linked to human listeriosis around the world. All isolates carried several important virulence genes (inlA, inlB, inlC, inlJ, and hly), whereas 38.8% (31/80) of strains harbored listeriolysin S genes (llsX). Besides, all 4b-4d-4e isolates belonged to ECI showed a high potential to cause human diseases. Further study of inlA sequencing showed most of selected strains (32/35, 91.4%) contained PMSC-lacking inlA gene sequences required for the encoding of InlA factor and bacterial invasion of the host cell. Based on the genetic properties observed in L. monocytogenes isolates, it is reasonable to assume L. monocytogenes found in Chinese retail RTE food have potential to cause listeriosis, and is concerning. These data have significant implications for the epidemiological and public health studies of this pathogen.

#### AUTHOR CONTRIBUTIONS

Conceived and designed the experiments: QW, JZ, SW. Performed the experiments: SW. Analyzed the data: SW, MC. Contributed reagents/materials/analysis tools: WG. Contributed to the writing of the manuscript: SW, QW.

#### FUNDING

We would like to acknowledge the financing support of National Natural Science Foundation of China (No. 31471664), the Science and Technology Projects of Guangdong (2013B050800026), and the Science and Technology Projects of Guangzhou (201504010036).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.00168


Liu, D. (2008). Handbook of Listeria monocytogenes. Boca Raton, FL: CRC Press.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The reviewer RT and handling Editor declared a current collaboration and the handling Editor states that the process nevertheless met the standards of a fair and objective review.

Copyright © 2016 Wu, Wu, Zhang, Chen and Guo. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Biosurfactant Produced by *Salmonella* Enteritidis SE86 Can Increase Adherence and Resistance to Sanitizers on Lettuce Leaves (*Lactuca sativa* L., *cichoraceae*)

*Eliandra M. Rossi1,2\*, Luniele Beilke1, Marília Kochhann1, Diana H. Sarzi1 and Eduardo C. Tondo2*

*<sup>1</sup> Laboratório de Microbiologia, Departamento de Ciências Biológicas e da Saúde, Universidade do Oeste de Santa Catarina, São Miguel do Oeste, Brazil, <sup>2</sup> Laboratório de Microbiologia e Controle de Alimentos, Instituto de Ciência e Tecnologia de Alimentos, Universidade Federal do Rio Grande do Sul, Porto Alegre, Brazil*

#### *Edited by:*

*Rosanna Tofalo, University of Teramo, Italy*

#### *Reviewed by:*

*Mehrdad Mark Tajkarimi, University of North Carolina at Greensboro, USA Ebrahim Hadavi, Islamic Azad University, Iran*

*\*Correspondence: Eliandra M. Rossi eliandra\_bio@yahoo.com.br*

#### *Specialty section:*

*This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology*

*Received: 27 October 2015 Accepted: 07 January 2016 Published: 22 January 2016*

#### *Citation:*

*Rossi EM, Beilke L, Kochhann M, Sarzi DH and Tondo EC (2016) Biosurfactant Produced by Salmonella Enteritidis SE86 Can Increase Adherence and Resistance to Sanitizers on Lettuce Leaves (Lactuca sativa L., cichoraceae). Front. Microbiol. 7:9. doi: 10.3389/fmicb.2016.00009*

*Salmonella* Enteritidis SE86 is an important foodborne pathogen in Southern Brazil and it is able to produce a biosurfactant. However, the importance of this compound for the microorganism is still unknown. This study aimed to investigate the influence of the biosurfactant produced by *S*. Enteritidis SE86 on adherence to slices of lettuce leaves and on resistance to sanitizers. First, lettuce leaves were inoculated with *S*. Enteritidis SE86 in order to determine the amount of biosurfactant produced. Subsequently, lettuce leaves were inoculated with *S*. Enteritidis SE86 with and without the biosurfactant, and the adherence and bacterial resistance to different sanitization methods were evaluated. *S*. Enteritidis SE86 produced biosurfactant after 16 h (emulsification index of 11 to 52.15 percent, *P <* 0.05) and showed greater adherence capability and resistance to sanitization methods when the compound was present. The scanning electron microscopy demonstrated that *S*. Enteritidis was able to adhere, form lumps, and invade the lettuce leaves' stomata in the presence of the biosurfactant. Results indicated that the biosurfactant produced by *S*. Enteritidis SE86 contributed to adherence and increased resistance to sanitizers when the microorganism was present on lettuce leaves.

Keywords: lettuce, microbial adherence and resistance, *Salmonella* Enteritidis SE86, biosurfactant, disinfection

# INTRODUCTION

*Salmonella* Enteritidis SE86 is a recognized food pathogen responsible for several foodborne disease (FBD) outbreaks in Southern Brazil (Geimba et al., 2004; Oliveira et al., 2009; Tondo and Ritter, 2012; Capalonga et al., 2014; Tondo et al., 2015). Several studies have been carried out taking into account the importance of this pathogen with the aim of understanding the reasons that it continues to be an important foodborne pathogen in this region since 1999 (Geimba et al., 2004; Capalonga et al., 2014; Tondo et al., 2015). Among all the characteristics that may contribute to that, we may highlight its great acid adaptation capability when the pathogen is exposed to acidic environments and, as a consequence, an increase in virulence (Perez et al., 2012) and ability to survive in simulated gastric fluid, (pH 1.5; 2), besides its resistance to sodium hypochlorite at 200 and 400 ppm (Machado et al., 2010). A previous study (Machado, 2007) demonstrated that *S*. Enteritidis SE86 was able to produce expressive amounts of biosurfactant during its growth in BHI broth. Nevertheless, the characteristics and functions of this compound have not been studied yet.

The term biosurfactant is described as a "surface active agent" produced by microorganisms (Marchant and Banat, 2012). These are amphiphilic compounds used as detergents or wetting, emulsifying, dispersing, and foaming agents in many industrial formulations (Nitschke and Pastore, 2002). Even though they have been highly used, the physiological function of biosurfactants for microbial cells is still not completely understood, and the way these compounds rule food microorganisms is practically unknown. One of the few studies concerning biosurfactants in foods was published by Mellor et al. (2011), who reported that a biosurfactant produced by *Pseudomonas fluorescens* was able to alter the characteristics of chilled chicken meat (increased decomposition) and the compound facilitated the survival of the bacterium.

The production of biosurfactants is usually associated with the presence of large amounts of microorganisms (Ron and Rosenberg, 2001) and this factor can contribute to increased pathogenicity. Also, several researchers have reported that biosurfactants can contribute to the adherence of pathogens to surfaces and the formation of biofilms (Ron and Rosenberg, 1999; Ron and Rosenberg, 2001; Nitschke and Pastore, 2002). Furthermore, the ability of microorganisms to produce biosurfactants can also be linked with their resistance to sanitizers, because generally they present organic compounds that can protect bacterial cells.

Recently, it was stated that bacteria such as *S*. Enteritidis have a natural tendency to stick to surfaces, which includes lettuce leaves (Lima et al., 2013).

Lettuce (*Lactuca sativa* L., *cichoraceae*) is the most consumed green leaf in the world; this is a plant of easy acquisition, standing out due to its nutritional quality and because it is considered a low cost leafy vegetable (Abreu et al., 2010; Lima et al., 2013). During their growing cycle, lettuces can be contaminated by *Salmonella* and, as a consequence, several cases of salmonellosis have been related to the consumption of lettuces (Horby et al., 2003; Sagoo et al., 2003; Takkinen et al., 2005; Nygard et al., 2008; Irvine et al., 2009). In order to avoid contamination, lettuce leaves must be washed and sanitized before going to the table. However, if a biosurfactant is produced by *Salmonella*, microbial cells can easily adhere to the leaves and be protected against inactivation.

The aim of this study was to investigate the influence of the biosurfactant on adherence and resistance of *S*. Enteritidis SE86 to sanitizers on lettuce leaves.

#### MATERIALS AND METHODS

#### Lettuce Samples

All lettuces used in this study were purchased in a supermarket in Porto Alegre, Capital of Rio Grande do Sul, Southern state of Brazil. Before the experiments started, lettuces were transported to the laboratory, inside thermal boxes, at 4◦C for a maximum period of one hour. Before experiments, injured leaves were removed and the remaining ones were washed with potable water. Whole lettuce leaves were used for the experiments on resistance to sanitizers described in "Influence of Biosurfactant on the Efficiency of Sanitation Methods Used for Disinfection of Whole Lettuce Leaves Contaminated with *S.* Enteritidis SE86." Slices of lettuce leaves with sizes of 10 cm × 10 cm were used for the experiments of adherence, according to Sagong et al. (2011). This was done in order to express results as number of CFU/cm2. All sliced leaves were cut similarly, aiming to avoid interference in the results.

Before experiments, whole lettuce leaves and sliced leaves were washed and sanitized with potable water with 200 ppm sodium hypochlorite added, for 15 min (Antunes, 2009). After that, leaves were rinsed with sterile distilled water with 0.5% sodium thiosulfate added (Synth, Diadema-SP).

#### Microorganism

In this study, we used the *S*. Enteritidis SE86 strain, which was isolated from a cabbage involved with a salmonellosis outbreak in the State of Rio Grande do Sul, Brazil, in 1999. This strain was characterized by Geimba et al. (2004) and presents the same profile and genotypic characteristics of *S.* Enteritidis responsible for several cases of salmonellosis that occurred from 1999 to 2012 in Rio Grande do Sul (Capalonga et al., 2014; Tondo et al., 2015). For the tests, the strain was cultivated in BHI broth (Oxoid, Basingstoke, England) at 36 ± 1◦C, for approximately 18 h.

### Biosurfactant Production on Lettuce Leaves

Four whole lettuce leaves were submerged in 100 ml of minimal medium containing 4.4 log CFU/ml of *S*. Enteritidis SE86 and incubated at 36 ± 1◦C for 120 h.

Aliquots of 6 ml were removed every two hours for up to 60 and, after each time period, aliquots were withdrawn every 24 h for up to 120 h of culture in order to determine the emulsification index (IE24), pH, and bacterial count. Bacterial counts were performed in triplicate by seeding the samples onto plates containing xylose lysine deoxycholate agar (XLD: Merck, Darmstadt, Germany) and incubated at 36 ± 1◦C for 24 h. The pH was evaluated by aliquots (10 ml) of the samples and then analyzed with a pH meter (PHTECK). The emulsification index (IE24) was assessed using the method described by Cooper and Goldenberg (1987).

All experiments were repeated three times and the averages were subsequently expressed as the final result.

### Preparation of the Inoculum of *S*. Enteritidis SE86 With and Without Biosurfactant

The inoculum of *S*. Enteritidis SE86 without biosurfactant was prepared using 40 ml of BHI broth (Merck, Darmstadt, Germany) incubated at 36 ± 1◦C for 72 h. After incubation, the culture was centrifuged at 3500 rpm for 15 min and washed with phosphate buffered saline (PBS) three times. Then, the washed cells were inoculated in 100 ml minimal medium until they reached a concentration of approximately 8.0 log CFU/ml.

In order to prepare the inoculum of *S*. Enteritidis SE86 with biosurfactant, the compound was partially purified, according to the following procedures. The biosurfactant recovery was prepared by centrifuging at 3500 rpm for 15 min inoculum *S*. Enteritidis SE86 in BHI broth incubated at 36 ± 1◦C for 72 h. The supernatant was homogenized with ethanol (–4◦C) at 95% concentration 4:1 and stored at 4◦C for 24 h. Subsequently, the precipitate (biosurfactant) was recovered by centrifugation at 3500 rpm for 15 min and the supernatant was discarded. After the alcohol had evaporated completely, the pellet was resuspended in sterile distilled water and dialyzed. The dialysis was done using a membrane tube (SIGMA) submitted to constant agitation in distilled water for 24 h (Kumar et al., 2004; Ciapina, 2008; Pacheco et al., 2010). Hundred milliliter of solution with partially purified biosurfactant and 8.0 log CFU/ml of *S.* Enteritidis SE86 were used.

#### Resistance to Sanitizers of *S.* Enteritidis SE86 With and Without Biosurfactant *In Vitro*

The susceptibility of *S.* Enteritidis SE86 with and without biosurfactant to sanitizers *in vitro* was evaluated using sodium hypochlorite (50 and 200 ppm) and vinegar (2 and 20%). The test was performed according to the methodology recommended by Ordinance 101/93 published by the Brazilian Ministry of Agriculture and Food Supply (Brasil, 1993).

Initially, the concentrations of sodium hypochlorite (Q. Boa<sup>R</sup> ) and vinegar (fermented acetic acid from red wine and alcohol koller<sup>R</sup> ) were prepared in sterile distilled water. Nine milliliter of each sanitizer were aseptically placed into sterile vials, to which was added 1 ml of bovine serum albumin solution (1%). After that, 0.1 ml inoculum (*S*. Enteritidis SE86 with and without the biosurfactant) was added separately to each tube containing sanitizers and the exposure time was measured. After 5, 10, 15, 20, and 30 min of exposure, an aliquot of 0.01 ml of suspension was transferred into tubes containing BHI broth. The tubes were incubated for 96 h at 36 ± 1◦C, and the bacterial growth was checked every 24 h. In the case of bacterial growth, the test was considered positive (resistant). The negative confirmation of results (tubes without growth) was performed through inoculation on trypticase soy agar (TSA agar, Merck, Darmstadt, Germany) incubated at 36 ± 1◦C for 24 h.

Each experiment was performed in triplicate on different days.

# Adherence of *S*. Enteritidis SE86 to Slices of Lettuce Leaves (*Lactuca sativa L., cichoraceae*)

Adherence of *S*. Enteritidis SE86 to slices of lettuce leaves was assessed using the methods proposed by Lima et al. (2013) with the following adaptation: the slices of lettuce leaves were cleaned, as described in Section "Lettuce Samples".

Before each treatment, three slices of lettuce leaf were immersed in 100 ml of minimal medium containing *S*. Enteritidis SE86 at a concentration of approximately 8.0 log CFU/ml, with and without the biosurfactant, for 15, 30, and 60 min at room temperature (25◦C). The preparation of the inoculum of *S.* Enteritidis SE86 with and without biosurfactant is described in Section "Preparation of the Inoculum of *S*. Enteritidis SE86 With and Without Biosurfactant". After that, slices of lettuce leaf were submerged in 100 ml of PBS and immediately sonicated for five minutes, using ultrasonic equipment (LF Equipamentos, Anhangaba SP) with intensity of 40 kHz. Sonication was used in order to remove adhered cells following the methods described by Sinde and Carballo (2000). This method was used because it does not damage cells and is considered very efficient in removing bacteria from biomaterials, especially from rough or irregular surfaces (An and Skowronski, 2000).

The counting of *S*. Enteritidis SE86 was performed on XLD agar incubated at 36 ± 1◦C for 24 h. Counts were done in triplicate and each experiment was repeated five times.

### Influence of Biosurfactant on the Efficiency of Sanitation Methods Used for Disinfection of Whole Lettuce Leaves Contaminated with *S*. Enteritidis SE86

First of all, 250 g of whole lettuce leaves were immersed into 500 ml of the *S.* Enteritidis SE86 inoculum with and without biosurfactant (prepared as described in Section "Preparation of the Inoculum of *S*. Enteritidis SE86 With and Without Biosurfactant") for 60 min.

Sanitation treatments were performed by immersing artificially contaminated lettuce leaves (25 g) in 500 ml of each treatment solution (i.e., potable water for 30 min; 50 and 200 ppm sodium hypochlorite for 15 and 30 min; 2 and 20% vinegar aqueous solution for 15 min). At the end of the contact time, each treatment solution was drained off, and leaves were rinsed with 200 ml of neutralizing buffer solution (0.5% thiosulfate sodium, Synth, Diadema SP) for 30 s, as recommended by Abadias et al. (2008), and then rinsed with potable water.

The negative and positive controls were non-contaminated lettuce leaves and lettuce leaves artificially contaminated with *S*. Enteritidis SE86, respectively. Washing was carried out only with potable water in order to evaluate of bacterial removal.

After treatments, lettuce leaves (25 g) were blended in a Stomacher bag containing 225 ml 0.1% peptone water (Merck, Darmstadt, Germany) for 60 s. *S*. Enteritidis SE86 counting was carried out on XLD agar after incubation at 36 ± 1◦C for 24 h. Typical colonies (black) were counted in triplicate and the identity of the microorganism confirmed by biochemical tests.

All treatments were performed ten times on different days and the measurement of free chlorine in solutions was done using a Spectroquant<sup>R</sup> Kit (Merck).

# Scanning Electron Microscopy of *S*. Enteritidis SE86 on Surface of Lettuce Leaves With and Without Biosurfactant

Lettuce slices (1 cm × 1 cm) were prepared using the central region of lettuce leaves (washed and disinfected as described in Section "Lettuce Samples").

Three artificially contaminated lettuce slices with and without biosurfactant were let for 1 h at room temperature (25◦C). After that, leaves were gently washed twice using 0.1% peptone water and fixed with 3.0% glutaraldehyde and 0.05 M phosphate buffer, pH 7.0, for 1 h. The slices were washed four times (15 min each) with phosphate-buffered saline. After that, samples were dehydrated by increasing concentrations of ethanol solution (30, 50, 70, 80, 95, and 100%), with 15 min of contact each, and finally acetone PA for 30 min. The slices were dried with CO2 in a critical-point drier (CPD 030; Bal-Tec), coated with gold (BAL-TEC SCD 050), and taken for observation on a JSM 5800 scanning electron microscope (SEM). Three lettuce slices submerged only in sterile distilled water were used as negative controls.

#### Statistical Analysis

The ANOVA test was applied (Assistat 7.7 Beta) with *P <* 0.05, in order to assess significant differences in the adherence of *S*. Enteritidis SE86 and its resistance to washing and disinfecting methods on lettuce leaves.

#### RESULTS

#### Production of Biosurfactant on Lettuce Leaves

The results showed that *S*. Enteritidis SE86 produced biosurfactant when in contact with the lettuce leaves for more than 16 h, presenting an emulsification index (IE24) of 11% when bacterial population reached 7.11 log CFU/ml (**Figure 1**). The greater emulsification index was 52.15%, after 120 h of contact with lettuce leaves and when the population was 9.8 log CFU/ml and the pH remained at 7.0 during all experiments. During the preparation of the inoculum in minimal medium, SE86 also produced emulsifier (EI24 46%); however, after 120 h, the IE24 decreased to 3% (results not shown), probably because the energy sources were depleted.

# *In Vitro* Resistance to Sanitizers

The *in vitro* testing of the susceptibility to sanitizers revealed that the biosurfactant was responsible for increasing the survival of *S*. Enteritidis SE86 in 50ppm sodium hypochlorite and 2 and 20% vinegar solution. As an example, SE86 without biosurfactant was completely inactivated by 50 ppm sodium hypochlorite in 15 min, while with biosurfactant, the microorganism survived for 30 min. Similarly, SE86 without surfactant was eliminated by 2 and 20% vinegar after 15 and 0 min of exposure, respectively. Nevertheless, the presence of surfactant made SE86 survive for 20 and 5 min, respectively (**Table 1**). The biosurfactant did not influence the survival of SE86 exposed to 200 ppm sodium hypochlorite.

#### Influence of Surfactant on the Adherence of *S*. Enteritidis SE86 to Slices of Lettuce Leaf

The results of this study showed that there were significant differences (*P <* 0.05) in the adherence of *S*. Enteritidis SE86 to slices of lettuce leaf when biosurfactant was present (**Table 2**). The highest counts of adhered SE86 were observed after 60 min of contact with slices of lettuce leaf. At that time, average counts of 7.3 log CFU/cm<sup>2</sup> and 4.1 log CFU/cm<sup>2</sup> were obtained on lettuce leaves with and without biosurfactant, respectively.


TABLE 1 | *In vitro* susceptibility testing to disinfectants (200 and 50 ppm sodium hypochlorite and 2 and 20% vinegar solution) of *S*. Enteritidis SE86 with and without biosurfactant.

*R: resistant; S: sensitive.*

TABLE 2 | Median and standard deviation about the adherence of *S*. Enteritidis SE86 with and without biosurfactant on slices of lettuce leaves at different times.


*Values represent the means of five replicates. Different letters represent significant differences (P < 0.05).*

Scanning electron microscopy demonstrated that *S*. Enteritidis SE86 was able to adhere to the slices of lettuce leaf, forms lumps, and enter the stomata when the biosurfactant was present (**Figure 2**).

# Resistance to Sanitation Methods of Lettuce Leaves

It was observed that all treatments reduced the amount of *S*. Enteritidis SE86 on lettuces, but lettuces contaminated with *S*. Enteritidis SE86 and with biosurfactant demonstrated higher numbers of survival cells (significant difference *P <* 0.05) than lettuces contaminated with *S*. Enteritidis SE86 without the surfactant (**Figure 3**). Reductions in counts of *<sup>S</sup>*. Enteritidis SE86 with biosurfactant ranged from 1.0 to 2.8 log CFU/g, whereas the reductions of *S*. Enteritidis SE86 without biosurfactant ranged from 1.3 to 3.3 log CFU/g (*<sup>P</sup> <sup>&</sup>lt;* 0.05) (**Table 3**). It was observed that the most effective treatment of lettuce contaminated with *S*. Enteritidis SE86 and biosurfactant was washing it with potable water and submerging it in 200 ppm of sodium hypochlorite for 15 min. This showed a reduction of 2.8 log CFU/g. However, when lettuce leaves were contaminated only with *S*. Enteritidis SE86, the most effective reduction (3.3 log CFU/g) was obtained by the treatment that washed leaves with potable water and sanitized them with 50 ppm sodium hypochlorite for 30 min. This result showed that *S*. Enteritidis SE86 without biosurfactant was inactivated by lower concentrations of sodium hypochlorite.

*Salmonella* Enteritidis SE86 with biosurfactant was more resistant on lettuce leaves than in *in vitro* tests (**Table 1**; **Figure 3**).

# DISCUSSION

Microorganisms develop survival abilities in different environments and biosurfactant production can be an advantage to survive in foods (Mellor et al., 2011). However, the exact physiological function of biosurfactants is not yet completely elucidated (Nitschke and Pastore, 2002; Hamme et al., 2006; Abdel-Mawgoud et al., 2009; Jirku et al., 2015).

Results of the present study demonstrated that *S*. Enteritidis SE86 was able to produce biosurfactant on lettuce leaves and this ability may have facilitated the access to cutin on lettuce leaves, one functional component of the cuticle deposited on the surfaces and within the epidermal walls of aerial parts of plants. Cutin is composed of three dimensional polyesters of long fatty acid chains (Bacic et al., 1988) and the amphipathic property of biosurfactant may facilitate access to nutrients present on lettuce leaves, supplying energy for bacterial growth.

Several research groups have reported that environmental microorganisms are able to produce biosurfactants (Chen et al., 2012; Jain et al., 2013; Ayed et al., 2014; Maoa et al., 2015; Rosa et al., 2015); however, there are no scientific reports showing the production of surfactants by foodborne pathogens. To our knowledge, the present study is the first that demonstrates biosurfactant production by *Salmonella*. Other reports have demonstrated the production of biosurfactants by degradative microorganisms on foods. For example, according to Mellor et al. (2011), the biosurfactant produced by *Pseudomonas fluorescens* contributed to increasing the total bacterial count on chicken stored aerobically for three days, suggesting that the biosurfactant contributed to the bioavailability of nutrients for the bacteria. These researchers suggested that the biosurfactant becomes a competitive advantage for the microorganism to maintain their survival, thereby enhancing the decomposition of chicken meat. Shaheen et al. (2010) have reported that a type of biosurfactant called surfactin may have contributed to the formation of biofilms by *Bacillus cereus* inside milk tanks.

The ability of *Salmonella* to adhere to lettuce leaves was reported by several studies (Wei et al., 2006; Patel and Sharma, 2010; Kroupitski et al., 2011; Lima et al., 2013). Results similar to the ones obtained in this study were found by Kroupitski et al. (2011), who found 7.0 log CFU of *S.* Typhimurium on the central region of lettuce leaves. Also, Lima et al. (2013) demonstrated that the count of *S.* Enteritidis cells that adhered to lettuce leaves

differed *(P <* 0.05*)* between the hydroponic and conventional systems, reaching 5.2 ± 0.56 and 4.6 ± 0.26, respectively.

The influence of biosurfactants on bacterial adherence to surfaces has been quite well studied, and the results are variable. Hassan and Frank (2003) stated that Tween 85 surfactant reduced the adherence of *Escherichia coli* O157:H7 to lettuce leaves. Other researchers (Sotirova and Vasileva-Tonkova, 2009) reported that *Pseudomonas aeruginosa* NBIMCC 1390 with rhamnolipid biosurfactant increased cell hydrophobicity to 31% adherence and that these compounds caused changes in the bacterial cell surface.

The results of our study suggest that the biosurfactant contributed to increase the survival of *S*. Enteritidis SE86 on lettuce leaves. According to Wei et al. (2006), high surface adherence of bacterial populations is a competitive tool against other microorganisms.

The results of *in vitro* resistance to sanitizers and resistance to sanitation methods of lettuce leaves showed that *S*. Enteritidis SE86 with biosurfactant is more resistant to antimicrobial activity of the compounds tested.

The bactericidal action of sodium hypochlorite is the result of microbial cell oxidation, after contact of sanitizer and cells

(Watters et al., 2002; Møretrø et al., 2012; Bermúdez-Aguirre and Barbosa-Cánovas, 2013). According to Møretrø et al. (2012), pH and the presence of organic matter can affect the antimicrobial action of sodium hypochlorite. In our study, it was observed that biosurfactant decreased the antimicrobial action of sodium hypochlorite, probably because this organic compound linked to the sanitizer or avoided the contact of cells with the sanitizer.

TABLE 3 | Reduction (log CFU/g) in counts of *S*. Enteritidis SE86 with and without biosurfactant on whole lettuce leaves after the treatments were performed.


*Log CFU/g: colony forming unit/g converted to log*10*.* <sup>∗</sup>*The period of time the solutions spent submerged is shown in parenthesis next to each treatment used. Statistical analysis between treatments. Different letters indicate statistically significant differences between treatments (P < 0.05).*

Some studies have shown that acetic acid (vinegar) can reduce the amount of bacteria on foods and surfaces, including whole lettuce leaves (Karapinar and Gonul, 1992; Oliveira et al., 2012). Our study showed that *S*. Enteritidis SE86 in the presence of biosurfactant was more resistant to both vinegar concentrations (solution 2 and 20%). The counts of *S*. Enteritidis SE86 with biosurfactant on whole lettuce leaves sanitized with vinegar solution showed a reduction of 1.8 and 1.9 log CFU/g, whereas the reductions of *S*. Enteritidis SE86 without biosurfactant were 2.5 and 2.6 log CFU/g (**Table 3**).

Vinegars are able to decrease the external and internal pH of cells, inactivating microbial enzymes, and damaging membrane function and metabolic activities such as the transport of nutrients (Chang and Fang, 2007; Ölmez and Kretzschmar, 2009). The less effective antimicrobial action of vinegar solution on lettuce contaminated with *S*. Enteritidis SE86 in the presence of biosurfactant suggested that biosurfactant protected SE86 from contact with the vinegar solution, or this compound was able to neutralize pH action.

The greater resistance of *S*. Enteritidis SE86 in the presence of biosurfactant on lettuce leaves suggests that the biosurfactant production may be a mechanism used by the bacterium to maintain its survival in different environments.

Thus this study demonstrated that *S*. Enteritidis SE86 can use the biosurfactant to increase its adhesion to the surface of lettuce leaves, form lumps, and also to penetrate the stomata of lettuce leaves. These effects may influence the increase of resistance to vinegar and sodium hypochlorite during lettuce sanitization. Furthermore, the surfactant production by adhered cells may protect them, avoiding contact with sanitizers.

#### CONCLUSION

Based on the results found during this study, it can be concluded that high counts of *S.* Enteritidis SE86 were able to produce biosurfactant on lettuce leaves. The presence of biosurfactant S. Enteritidis SE86 increased the adherence to slices of lettuce leaf and decreased the antimicrobial action of sanitizers (vinegar and sodium hypochlorite) used to sanitize whole lettuce leaves. In addition, when SE86 was added with biosurfactant and was analyzed by scanning electron microscopy, lumps of cells were observed and the bacterium was able to enter the stomata. The same results were not observed in the absence of biosurfactant.

New studies are necessary to investigate other probable functions of biosurfactant produced by SE86. As a perspective of the present study, we suggest investigating the influence of this biosurfactant on the microbial ecology of lettuce leaves, and on the multiplication and survival of SE86 in other foods.

#### REFERENCES


#### AUTHOR CONTRIBUTIONS

ER: contributes to the development of experimental research, data analysis, and preparation of the article. LB: contributes to the development of experimental research. MK: contributes to the development of experimental research. DS: contributes to the development of experimental research. ET: contributes to the data analysis and preparation of the article.

#### ACKNOWLEDGMENTS

This research received support from Universidade do Oeste de Santa Catarina, São Miguel do Oeste, Santa Catarina, Brazil, Universidade Federal do Rio Grande do Sul, Brazil and Support Fund for Maintenance and Development of Higher Education (FUMDES) of Santa Catarina, Brazil.


*Dessas Superfícies*. Ph.D. dissertations, Universidade Federal do Rio Grande do Sul (UFRGS), Porto Alegre, 143.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2016 Rossi, Beilke, Kochhann, Sarzi and Tondo. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Engineering Bacterial Surface Displayed Human Norovirus Capsid Proteins: A Novel System to Explore Interaction Between Norovirus and Ligands

#### Mengya Niu<sup>1</sup> , Qianqian Yu<sup>1</sup> , Peng Tian<sup>2</sup> , Zhiyong Gao<sup>3</sup> , Dapeng Wang<sup>1</sup> \* and Xianming Shi <sup>1</sup>

*<sup>1</sup> Department of Food Science and Technology, MOST-USDA Joint Research Center for Food Safety, School of Agriculture and Biology, Shanghai Jiao Tong University, Shanghai, China, <sup>2</sup> Produce Safety and Microbiology Research Unit, Western Regional Research Center, Agricultural Research Service, United States Department of Agriculture, Albany, CA, USA, <sup>3</sup> Beijing Center for Diseases Prevention and Control, Beijing, China*

#### Edited by:

*Maria Schirone, University of Teramo, Italy*

#### Reviewed by:

*Tineke H. Jones, Agriculture and Agri-Food Canada, Canada Miriam Berti, Istituto Zooprofilattico Sperimentale Abruzzo e Molise, Italy*

> \*Correspondence: *Dapeng Wang norovirus@163.com*

#### Specialty section:

*This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology*

Received: *10 October 2015* Accepted: *04 December 2015* Published: *22 December 2015*

#### Citation:

*Niu M, Yu Q, Tian P, Gao Z, Wang D and Shi X (2015) Engineering Bacterial Surface Displayed Human Norovirus Capsid Proteins: A Novel System to Explore Interaction Between Norovirus and Ligands. Front. Microbiol. 6:1448. doi: 10.3389/fmicb.2015.01448* Human noroviruses (HuNoVs) are major contributors to acute nonbacterial gastroenteritis outbreaks. Many aspects of HuNoVs are poorly understood due to both the current inability to culture HuNoVs, and the lack of efficient small animal models. Surrogates for HuNoVs, such as recombinant viral like particles (VLPs) expressed in eukaryotic system or P particles expressed in prokaryotic system, have been used for studies in immunology and interaction between the virus and its receptors. However, it is difficult to use VLPs or P particles to collect or isolate potential ligands binding to these recombinant capsid proteins. In this study, a new strategy was used to collect HuNoVs binding ligands through the use of ice nucleation protein (INP) to display recombinant capsid proteins of HuNoVs on bacterial surfaces. The viral protein-ligand complex could be easily separated by a low speed centrifugation step. This system was also used to explore interaction between recombinant capsid proteins of HuNoVs and their receptors. In this system, the VP1 capsid encoding gene (ORF2) and the protruding domain (P domain) encoding gene (3′ terminal fragment of ORF2) of HuNoVs GI.1 and GII.4 were fused with 5′ terminal fragment of INP encoding gene (*inaQn*). The results demonstrated that the recombinant VP1 and P domains of HuNoVs were expressed and anchored on the surface of *Escherichia coli* BL21 cells after the bacteria were transformed with the corresponding plasmids. Both cell surface displayed VP1 and P domains could be recognized by HuNoVs specific antibodies and interact with the viral histo-blood group antigens receptors. In both cases, displayed P domains had better binding abilities than VP1. This new strategy of using displayed HuNoVs capsid proteins on the bacterial surface could be utilized to separate HuNoVs binding components from complex samples, to investigate interaction between the virus and its receptors, as well as to develop an oral vaccine for HuNoVs.

Keywords: human noroviruses, cell surface display, receptor, ice nucleation protein, histo-blood group antigens

# INTRODUCTION

Human noroviruses (HuNoVs) are one of major nonbacterial pathogens for foodborne gastroenteritis (Atmar and Estes, 2006). HuNoVs belong to the family of Caliciviridae. The singlestranded, positive-sensed RNA viral genome can be divided into three open reading frames (ORFs) (Jiang et al., 1993). The ORF2 encodes the major structural protein (VP1) which have two domains: the shell (S) and the protruding (P) domain (Zheng et al., 2006). The viral capsid is composed of 180 capsid protein monomers organized into 90 dimers (Tan et al., 2008).

Due to a lack of cultivable system or animal models, surrogate viruses such as feline calicivirus (FCV), murine norovirus (MNV), and Tulane virus (TV) are often used to study the fundamental biology of the viruses such as viral replication pattern and mechanism of infection (Farkas, 2015). Recombinant human norovirus capsid proteins expressed either by eukaryotic or prokaryotic systems are often used for studies for immunogenicity, diagnosis assays, and hostreceptor interaction (Gray et al., 1993; Green et al., 1993; Hutson et al., 2003; Huang et al., 2005). The virus-like particles (VLPs), spontaneously formed in a recombinant baculovirus system, have morphological and antigenical similarities to viral particles (Jiang et al., 1992). However, large production of recombinant proteins is still difficult owing to low protein yield and lack of VLPs. Tan et al. applied E. coli expression system to produce recombinant norovirus capsid proteins (Tan et al., 2004). They demonstrated that the E. coli-expressed capsid proteins maintained the same antigenicity and receptor binding specificity as those of the baculovirus-expressed VLPs, although the E. coli-expressed capsid proteins did not form VLPs. Tan et al. further demonstrated that a smaller particle (P particle) could form, which was expressed in E. coli (Tan et al., 2004). The P particles expressed in vitro is an octahedral nanoparticle formed by 24 copies of P monomers, most likely organized into 12 P dimers. These P particles are easily produced in E. coli, extremely stable, and highly immunogenic (Tan and Jiang, 2005b). However, one disadvantage of the bacterial system for expression of recombinant HuNoVs capsid proteins is that the expressed capsid proteins cannot be purified easily. After the fusion tag is removed, gel filtration and anion-exchange chromatography are required for purification of recombinant HuNoVs capsid proteins (Tan et al., 2004). Another disadvantage is that VP1 and P particles expressed by bacteria are either soluble proteins (VP1) or small particles. Therefore, it will be hard to either collect or isolate the viral capsid protein-ligand complex.

Bacterial ice nucleation proteins (INPs) are a family of proteins that enable Gram-negative bacteria to promote crystal formation at relatively high temperatures (Kawahara, 2002). Different INP coding genes from Pseudomonas syringae, Erwinia herbicola, and Xanthomonas campestris were well characterized (Wolber et al., 1986; Schmid et al., 1997; Jung et al., 1998a; Li et al., 2012). INP composes three distinct structural domains: an N-terminal domain, a C-terminal domain and a highly repetitive central domain (Shimazu et al., 2001). So far, INPs have been applied in various perfect bacterial cell surface display systems, including host cells of E. coli (Jung et al., 1998b; Kwak et al., 1999; Li et al., 2009), Salmonella typhi (Lee et al., 2000), Vibrio anguillarum (Xu et al., 2008), P. syringae (Shimazu et al., 2003). By transformation of bacteria with the gene encoding a fusion target protein with the anchoring motifs of INP, the target protein could be directly displayed on the surface of the bacteria (Kim and Yoo, 1999; Kwak et al., 1999; Cochet and Widehem, 2000). It was reported that the N-terminal domain of InaQ (named as InaQN) is responsible for the transmembrane transport and membrane-binding activity of INP (Li et al., 2012).

In order to solve the problem of collecting ligands binding to viral capsid proteins, recombinant HuNoVs capsid proteins were displayed on the surface of bacteria with the help of InaQN. It was reported that histo-blood group antigens (HBGAs) have been recognized as receptors for HuNoVs (Hutson et al., 2003; Tan and Jiang, 2005a). Therefore, in this study, Type III porcine gastric mucin (PGM) containing HBGAs (Tian et al., 2008) was used to evaluate the binding efficacy between the viral receptors and displayed HuNoVs VP1s and P domains.

#### MATERIALS AND METHODS

#### Bacterial Strains and Plasmids

E. coli DH5α and BL21 (ThermoFisher, Shanghai, China) were used as competent cells for recombinant plasmid construction and protein expression. Plasmid pMD19-T (TaKaRa, Dalian, China) inserted with different gene fragments was used for subcloning into the prokaryotic expression plasmid pET-28a (ThermoFisher, Shanghai, China). pCR-TORO/GI.1-ORF2+3 plasmid with inserted gene of HuNoV GI.1 ORF2 was kindly provided by Dr. Peng Tian (PSMRU, WRRC, USDA, CA, USA). pTrc-HisC-inaQ plasmid was kindly provided by Prof. Lin Li (State Key Laboratory of Agricultural Microbiology, Huazhong Agricultural University, Wuhan, China).

#### Cloning of inaQn, HuNoVs GI.1/GII.4 ORF2 and P Domain Coding Fragments

The coding sequence of InaQN domain (inaQn) was subcloned from plasmid pTrc-HisC-inaQ according to the previous report (Li et al., 2012).

HuNoV GI.1 ORF2 and its 3′ terminal fragments (named P) were amplified from recombinant plasmid pCR-TORO/GI.1- ORF2+3. A 1584 bp fragment for HuNoV GI.1 ORF2 (GenBank No. M87661) and a 909 bp fragment for HuNoV GI.1 ORF2 3 ′ terminal were amplified by PCR. Each 20.0µL PCR reaction including 2.0µL 10 × PCR Buffer, 1.5µL 25 mmol/L Mg2+, 1 U Taq E (ThemoFisher, Shanghai, China), 1.0µL 10 mmol/L dNTPs, 1.0µL 10 mmol/L of upstream and downstream primers respectively, 2.0µL plasmid and double distilled water (dd H2O). Thermal cycling condition consists of initial denaturation at 95◦C for 5 min, 35 cycles of template denaturation at 95◦C for 30 s, primer annealing at 52◦C for 30 s, primer extension at 72◦C for 1 min 40 s, and final extension at 72◦C for 10 min.

Genomic RNA of HuNoV GII.4 strain was extracted by Trizol Kit (ThemoFisher, Shanghai, China) from clinical gastroenteritis samples provided by Beijing Center for Disease Control and Prevention, China. A 1623 bp fragment (GenBank No.

Niu et al. Construction of Mock Human Norovirus


TABLE 1 | Sequences of (RT-) PCR primers.

TABLE 2 | Recombinant plasmids with restriction enzyme sites.


KM114291) for HuNoV GII.4 ORF2 was amplified by twostep RT-PCR kit (TaKaRa, Dalian, China) in accordance with manufacturer's protocol. The 3′ terminal fragment (948 bp) of ORF2 was amplified according to previous reports (Wang et al., 2008).

All primers used in this study were listed in **Table 1**. PCR products were analyzed by electrophoresis on 1.5% agarose gels and target bands were recycled using AxyPrep DNA gel extraction kit (Corning, Shanghai, China). The recovered nucleic acid fragments were inserted into plasmid pMD19- T, named pMD19-ORF2(GI.1), pMD19-ORF2(GII.4), pMD19- P(GI.1), and pMD19-P(GII.4), respectively. The recombinant plasmids were sequenced by Shanghai Majorbio Bio-pharm Technology Co., Ltd.

After restriction enzymes digestion (**Table 2**), inaQn fragment and ORF2 or ORF2 3′ terminal fragments from GI.1 or GII.4 were inserted into pET-28a to create recombinant plasmids pET28a-inaQn-ORF2(GI.1), pET28a-inaQn-ORF2(GII.4), pET28a-inaQn-P(GI.1), and pET28a-inaQn-P(GII.4), respectively. In addition, pET28a-ORF2(GI.1), pET28a-ORF2(GII.4), pET28a-P(GI.1), and pET28a-P(GII.4) without the inaQn gene were also constructed as negative controls. All recombinant plasmids were confirmed by sequencing by Shanghai Majorbio Bio-pharm Technology Co., Ltd.

#### Expression of Recombinant Plasmids

Recombinant E. coli BL21 strains were transformed with different recombinant plasmids, including pET28a-inaQn-ORF2(GI.1), pET28a-inaQn-ORF2(GII.4), pET28a-inaQn-P(GI.1), and pET28a-inaQn-P(GII.4), then cultured in Luria-Bertani (LB) (0.5% yeast extract, 1% trypton, and 1% NaCl) liquid medium containing 100µg/mL kanamycin with shaking (150 rpm) at 37◦C overnight. The transformed BL21 cells (50µL) were transferred to 5 mL fresh LB medium (100µg/mL kanamycin) with shaking (150 rpm) at 37◦C. When the culture reached to optical density in 600 nm (OD600) of around 0.6, 0.4 mmol/L (final concentration) isopropyl β-D-1-thiogalactopyranoside (IPTG) was added and the cells were incubated at 16◦C for 24 h with shaking (120 rpm). The cells were kept at 4◦C for further use.

# Preparation of Antibodies for VP1 of HuNoVs GI.1 and GII.4

Expression of Recombinant Capsid Proteins in E. coli E. coli BL21 with recombinant plasmids pET28a-ORF2 (GI.1) and pET28a-ORF2 (GII.4) were grown in LB medium with 100µg/mL kanamycin with shaking (150 rpm) at 37◦C overnight. The next day, the cells (2 mL) were subcultured in fresh LB (200 mL) with 100µg/mL kanamycin with shaking (150 rpm) at 37◦C. The recombinant plasmids were expressed in E. coli BL21 at 37◦C for 4 h after induction by 0.4 mmol/L IPTG. All induced cells were collected for further use.

#### Purification of Recombinant Protein and Preparation of Specific Antibodies

Recombinant capsid protein was purified according to previous report with minor modifications (Wang et al., 2008). Briefly, the cells expressing viral capsid proteins at the condition described in Section Expression of Recombinant Capsid Proteins in E. coli were washed 3 times using PBS (pH 7.2, NaCl 137 mmol/L, KCl 2.7 mmol/L, Na2HPO<sup>4</sup> 10 mmol/L, KH2PO<sup>4</sup> 2 mmol/L). The cells were resuspended in 20 mL Buffer A [pH 8.0, 50 mmol/L Tris-HCl, 1 mmol/L EDTA, 50 mmol/L NaCl, 5% glycerin (v/v), 0.5% Triton X-100 (v/v)] with 100µg/mL lysozyme (Sigma, St. Louis, USA). After incubation at 37◦C for 30 min, cells were subjected to ultrasonic treatment (BILON92-II, Shanghai, China) under the condition of 6 s with 4 s interval for 90 cycles on the ice. Then cell lysate was centrifuged at 10,000 × g for 15 min at 4◦C. The precipitate (crude inclusion body) was washed by 10 mL Buffer A. Twenty milliliters lysis buffer (pH 9.0, 50 mmol/L Tris-HCl, 1 mmol/L EDTA, 50 mmol/L NaCl, 5% glycerin (v/v), 0.5% Triton X-100 (v/v), and 8 mol/L urea) was added to resuspend the precipitate and then resolved slowly on the ice for 2 h. After centrifugation at 10,000 × g at 4◦C for 30 min, the supernatant was dialyzed in TE buffer (pH 8.0, 10 mmol/L Tris-HCl, 1 mmol/L EDTA) at 4◦C for 12 h. The TE buffer was then replaced and continued second dialysis. After dialysis, the supernatant was centrifuged at 10,000 × g at 4◦C for 30 min. The recombinant protein in supernatant was analyzed by SDS-PAGE according to our previous report (Wang et al., 2008).

#### Preparation of Primary Antibody Against VP1

Ten female Balb/c mice (Hubei Provincial Center for Disease Control and Prevention, China), 5–6 weeks old, were immunized subcutaneously with 0.1 mg recombinant VP1 of HuNoV GI.1 or GII.4 every 2 weeks respectively. The sera were collected and used as primary antibodies against the recombinant viral capsid proteins (VP1 and P particles) after 3 months of immunization. The sera were stored at −80◦C for further use.

#### Ethics Statement

All animal studies were carried out in strict accordance with the recommendations in the Guide for the Institutional Animal Care and Use Committee. The protocol was approved by Laboratory Animal Feeding Standard Operation Procedure (SOP-Ani-019- 2012), the Institutional Animal Care and Use Committee, and the Laboratory Animal Center of Shanghai Jiao Tong University (Permit Number: A2015020).

#### Whole Cell Enzyme Immunoassay (EIA)

A modified sandwich EIA was developed to evaluate the recombinant VP1 or P domains display efficiency on the surface of host cells with or without InaQN. Briefly, the transformed cells with recombinant plasmids pET28a-ORF2 (GI.1), pET28a-ORF2 (GII.4), pET28a-P (GI.1), pET28a-P (GII.4) and pET28a-inaQn-ORF2(GI.1), pET28a-inaQn-P(GI.1), pET28a-inaQn-ORF2(GII.4), pET28a-inaQn-P(GII.4), and plasmid pET28a as control were collected after induction of viral protein expression as described in previous Section Expression of Recombinant Plasmids by centrifugation at 5000 × g for 5 min. The pellet was washed with PBS (pH 7.2) twice and adjusted OD<sup>600</sup> to around 1.0 in 500µL PBS in a 1.5 mL Eppendorf tube. After centrifugation at 5000 × g for 5 min, the pellet was resuspended with 500µL primary antibodies against VP1 obtained from Section Preparation of Primary Antibody Against VP1 (1:10,000 diluted in Tris-Buffered Saline and Tween-20 with 1% bovine serum albumin) and incubated for 30 min at 37◦C. After washing with Tris-Buffered Saline and Tween-20 (TBST) for three times, the bacterial pellet was resuspended with 500µL Peroxidase-Conjugated Goat Anti-Mouse IgG (H+L) (Yeasen, Shanghai, China) at a dilution of 1:5000 and incubated for 30 min at 37◦C. After three washes with TBST, the bacterial pellet was resuspended in 500µL PBS. Finally, a 50µL of suspension was taken and mixed with 50µL 3,3′ ,5,5′ -tetramethylbenzidine [Friendbio Science and Technology (Wuhan) Co., Ltd., Hubei, China]. After keeping in the dark for 10 min, 50µL of mixture was transferred to 96 well module (Lantian Biological Equipment Factory, Jiangsu, China), and 50µL 2 mol/L H2SO<sup>4</sup> was added to stop the reaction. OD<sup>450</sup> value was measured using a Sunrise Microplate Reader (Tecan Sunrise, Switzerland). In addition, bacteria transformed with recombinant plasmids without inaQn gene including pET28a-ORF2 (GI.1), pET28a-ORF2 (GII.4), pET28a-P (GI.1), and pET28a-P (GII.4) were also tested.

### Measuring the HBGAs Binding Abilities

Type III PGM purchased from Sigma (St, Louis, MI, USA) was used as viral receptor. Each well of the Nunc Immuno Module (VWR, CA, USA) was coated with 100µL of PGM (1 mg/mL in 0.05 mol/L carbonate-bicarbonate buffer, pH 9.6) at 4 ◦C overnight. After being washed with PBS (pH 7.2) for 3 times, the wells were blocked with 120µL of 1% bovine serum albumin (BSA) in PBS at 37◦C for 1 h. The wells were washed with PBS (pH 7.2) for 3 times and used immediately.

The transformed bacteria were cultured as described in Section Expression of Recombinant Plasmids. The cells with recombinant plasmids pET28a-inaQn-ORF2(GI.1), pET28a-inaQn-P(GI.1), pET28a-inaQn-ORF2(GII.4), pET28ainaQn-P(GII.4) were collected by centrifugation at 3000 × g for 5 min and diluted to 10<sup>2</sup> or 10<sup>3</sup> CFU/mL in PBS. The diluted cells (100µL per well) were added into 48 wells, half of which were coated with PGM, and the remainder was not coated with PGM. After incubation at 37◦C for 30 min, 10µL suspension from each well was taken onto the LB agar plate containing 100µg/mL kanamycin for overnight at 37◦C. After incubation, the colonies were counted. In addition, recombinant plasmids without inaQn gene were constructed as a control including pET28a-ORF2 (GI.1), pET28a-ORF2 (GII.4), pET28a-P (GI.1), and pET28a-P (GII.4). The binding ratio was calculated by comparing the difference of colony forming units (CFU) before and after binding with or without PGM.

$$\begin{aligned} \text{Binding ratio } &= \, n/N \times 100\%\\ n &= \text{ (CFU in well without PGM) }\\ &- \text{(CFU in well with PGM) }\\ N &= \text{CFU in well without PGM} \end{aligned}$$

#### Statistical Analysis

IBM SPSS Statistics Software (version 19) was used for analyzing statistics. Each experiment was repeated at least three times (N = 3) as independent replicates in triplicates within each experiment (n = 3). One-way ANOVA was utilized for data analysis. Differences in means were considered significant when the p < 0.05.

#### RESULTS

#### The Fusion Viral Proteins Expressed on the Surface of Transformed Cells

A modified sandwich EIA was developed to evaluate the recombinant VP1 or P domains display efficiency on the surface of host cells with or without InaQN (**Figure 1**). The results indicated the fusion protein could display on the surface of bacterial cells and maintain the antigenicity of HuNoVs (**Figure 1**). The absorbance was significantly higher in bacteria

transformed with plasmids of inaQn gene fused with ORF2 or P domain coding gene than that of plasmids of ORF2 or P domain coding gene without inaQn gene (p < 0.01). The P/N (Positive result/Negative result) ratio of pET28a-inaQn-ORF2 (GI.1), pET28a-inaQn-P (GI.1), pET28a-inaQn-ORF2 (GII.4), and pET28a-inaQn-P (GII.4) were 9.45, 8.93, 6.53, and 8.98 respectively. Without the inaQn, the P/N ratio of pET28a-ORF2 (GI.1), pET28a-P (GI.1), pET28a-ORF2 (GII.4), and pET28a-P (GII.4) were 3.81, 4.25, 2.55, and 2.77 respectively. The presence of inaQn gene significantly increased display of HuNoVs capsid proteins on the surface of the transformed bacterial cells. There was no significant difference in signals of displayed VP1 and P domains of GI.1 transformed bacteria (p > 0.05). However, the signals of displayed P domains were significantly higher than VP1 in GII.4 transformed bacteria (p < 0.01). No significant difference was found among pET28ainaQn-ORF2 (GI.1) (OD<sup>450</sup> = 0.803 ± 0.029), pET28a-inaQn-P (GI.1) (OD<sup>450</sup> = 0.759 ± 0.008), and pET28a-inaQn-P (GII.4) (OD<sup>450</sup> = 0.763 ± 0.005) (p > 0.05). However, absorbance signal of pET28a-inaQn-ORF2 (GII.4) (OD<sup>450</sup> = 0.555 ± 0.012) was significantly lower than the residual groups (p < 0.01).

#### Binding of HBGAs with Surface Displayed VP1 and P Domains

To identify the HBGAs receptor-binding capacity, E. coli BL21 cells were transformed with plasmids pET-28a, pET28a-ORF2 (GI.1), pET28a-inaQn-ORF2 (GI.1), pET28a-P (GI.1), pET28ainaQn-P (GI.1), pET28a-ORF2 (GII.4), pET28a-inaQn-ORF2 (GII.4), pET28a-P (GII.4), and pET28a-inaQn-P (GII.4). The HBGAs binding capacity was evaluated by colony counting assay. When bacteria transformed with HuNoVs capsid protein encoding genes were added to PGM coated wells, the binding capacity between the HBGAs in PGM and displayed VP1 or

P domains was determined by comparing the binding ratio of CFU in wells with or without PGM (**Figure 2**). The HBGAs binding capacity was 10.5 ± 0.4%, 12.5 ± 0.3%, 24.0 ± 0.3%, and 27.3 ± 0.9% in bacteria transformed with pET28a-inaQn-ORF2 (GI.1), pET28a-inaQn-P (GI.1), pET28a-inaQn-ORF2 (GII.4), and pET28a-inaQn-P (GII.4), respectively. The HBGAs binding capacity was 5.5 ± 0.4%, 3.3 ± 0.1%, 12.3 ± 0.2%, and 7.4 ± 0.3% in bacteria transformed with pET28a-ORF2 (GI.1), pET28a-P (GI.1), pET28a-ORF2 (GII.4), and pET28a-P (GII.4) without inaQn, respectively. Therefore, the HBGAs binding capacity of InaQN fused VP1s and P particles was significantly higher than groups without InaQN (p < 0.01). The HBGAs binding capacities of displayed P domains from both genotypes are slightly higher than displayed VP1. However, the HBGAs binding capacities of both displayed VP1 and P domains of GI.1 were significantly lower than that of GII.4.

#### DISCUSSION

In this study, cell surface display systems for HuNoVs capsid protein VP1 and P domains were constructed using InaQN as a carrier protein. To our knowledge, it is the first report on HuNoVs displayed the surface of transformed bacteria. Our results indicated that displayed VP1 or P domains of HuNoVs were successfully anchored on the surface of bacterial cells. The surface displayed VP1 or P domains maintained the antigenicity (**Figure 1**). The cell surface display using INP has been used in vaccine development. For example, a recombinant oral vaccine for hepatitis C has been tested (Lee et al., 2000). Therefore, this system could be a good candidate for vaccine development. Currently, we are in the process of using these cell surface displayed VP1 and P domains for vaccine candidate for HuNoVs.

The HBGAs have been recognized as receptors or co-receptors for HuNoVs (Huang et al., 2003). In this study, we demonstrated that the displayed viral proteins could interact with the HBGAs viral receptors (**Figure 2**). The HBGAs binding capacity of the displayed VP1 and P domains from GII.4 were significantly higher than displayed VP1 and P domains from GI.1, indicating GII.4 P domains or VP1 may fold more perfectly and be easily recognized by viral receptors than GI.1 P domains or VP1. Recently, sialic acids have been identified as additional cellular receptors/co-receptor for Tulane virus (Tan et al., 2015) and for HuNoVs (Rydell et al., 2009). It is possible that other receptors or ligands for HuNoVs have not been identified. VLP and P particles are limited for valuing in collection or isolation of HuNoVs receptors or ligands. VLP and P particles could not be used to get viral particle-ligand complex unless ultracentrifugation is applied. It is not practical to collect or isolate a viral particle-ligand complex by ultracentrifugation due to the small sample size. In this study, the results demonstrated that the viral capsid protein-receptor complex could be easily collected by low speed centrifugation. The cell surface display system we described in this paper could be an alternative method to replace the ultracentrifugation to collect the viral capsid binding ligands.

There was some background binding in bacteria expressing the viral capsid proteins (VP1 and P domains) without the inaQn gene. The background noise could be a result of non-specific binding of the viral capsid proteins on bacteria or nonspecific binding of the antibodies to bacteria as the antibodies were made from viral capsid proteins, which expressed in E. coli BL21. The background should be reduced significantly when the targets are viral receptors or ligands. Currently, we are in the process of isolating candidate ligands for HuNoVs using this approach.

#### REFERENCES


This new strategy of using cell surface displayed HuNoVs capsid proteins will provide a new approach to isolate HuNoVs ligands, a new approach to characterize interaction between HuNoVs and receptors or ligands, a faster way to construct VP1 or P domains from a new strain or mutated strains. Since a couple of viral antigens displayed on the bacterial surface using INP have been proved to be an attractive platform for production of vaccine (Kim and Yoo, 1999; Kwak et al., 1999; Lee et al., 2000) and E. coli-expressed norovirus P particle could be a platform for antibody production (Tan et al., 2011), we believe that cell surface displayed HuNoVs capsid proteins can also be a good candidate for vaccine development.

#### AUTHOR CONTRIBUTIONS

DW and MN designed the experiments. MN, QY, and DW carried out experiments. MN, DW, and XS analyzed sequencing data and experimental results. ZG provided HuNoV GII.4 clinical sample and analyzed the viral RNA sequence. PT provided plasmid pCR-TORO/GI.1-ORF2+3. MN, DW, and PT wrote and modified the manuscript. XS provided laboratory equipment and place.

#### ACKNOWLEDGMENTS

This work was jointly supported by the grant No. 2012AA101601 from the Ministry of Science and Technology of China and the grant No. 31301475 from the National Natural Science Foundation of China. We want to thank Prof. Lin Li kindly provided plasmid pTrc-HisC-inaQ (State Key Laboratory of Agricultural Microbiology, Huazhong Agricultural University, Wuhan, China).


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2015 Niu, Yu, Tian, Gao, Wang and Shi. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# **Classification and Taxonomy of Vegetable Macergens**

#### *Bukola R. Aremu <sup>1</sup> and Olubukola O. Babalola 1,2 \**

*<sup>1</sup> Department of Biological Sciences, Faculty of Agriculture, Science and Technology, North-West University, Mmabatho, South Africa, <sup>2</sup> Food Security and Safety Niche Area, Faculty of Agriculture, Science and Technology, North-West University, Mmabatho, South Africa*

Macergens are bacteria capable of releasing pectic enzymes (pectolytic bacteria). These enzymatic actions result in the separation of plant tissues leading to total plant destruction. This can be attributed to soft rot diseases in vegetables. These macergens primarily belong to the genus *Erwinia* and to a range of opportunistic pathogens namely: the *Xanthomonas* spp., *Pseudomonas* spp., *Clostridium* spp., *Cytophaga* spp., and *Bacillus* spp. They consist of taxa that displayed considerable heterogeneity and intermingled with members of other genera belonging to the *Enterobacteriaceae*. They have been classified based on phenotypic, chemotaxonomic and genotypic which obviously not necessary in the taxonomy of all bacterial genera for defining bacterial species and describing new ones These taxonomic markers have been used traditionally as a simple technique for identification of bacterial isolates. The most important fields of taxonomy are supposed to be based on clear, reliable and worldwide applicable criteria. Hence, this review clarifies the taxonomy of the macergens to the species level and revealed that their taxonomy is beyond complete. For discovery of additional species, further research with the use modern molecular methods like phylogenomics need to be done. This can precisely define classification of macergens resulting in occasional, but significant changes in previous taxonomic schemes of these macergens.

#### **Keywords: classification, macergens, pectolytic, proteolytic, taxonomy, species**

# **INTRODUCTION**

Marcergens are soft rot causing bacteria, responsible for plant tissue maceration resulting in total tissue collapse (Beattie, 2006; Bhai et al., 2012). Soft rot diseases of vegetables are the most characteristic symptom of tissue maceration in a plant. These begin as small water soaked lesion, expands and intensifies until the tissue turns soft and watery (Reddy, 2015). Apparently, the outer surface of the diseased plant might stay unbroken, while tanning and depressed, or enclosed in an exuding bacterial mucus layer (Heyman et al., 2013). Foul smells are common owing to the discharge of explosive complexes through tissue degradation. Best bacterial growth follows plant cell lysis in these diseases (Rich, 2013). Soft-rotting bacteria are distinguished for the speed at which they stimulate soft rot. Warehoused crop may turn to liquid in only a few hours (Reddy, 2015). These pathogens usually enter through wound spots or natural openings such as lenticels and persist

#### *Edited by:*

*Giovanna Suzzi, Università degli Studi di Teramo, Italy*

#### *Reviewed by:*

*Ivonne Delgadillo, University of Aveiro, Portugal Rosanna Tofalo, Università degli Studi di Teramo, Italy*

#### *\*Correspondence:*

*Olubukola O. Babalola olubukola.babalola@nwu.ac.za*

#### *Specialty section:*

*This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology*

*Received: 21 August 2015 Accepted: 16 November 2015 Published: 27 November 2015*

#### *Citation:*

*Aremu BR and Babalola OO (2015) Classification and Taxonomy of Vegetable Macergens. Front. Microbiol. 6:1361. doi: 10.3389/fmicb.2015.01361*

**Abbreviation:** DNA: Deoxy Ribonucleic Acid; ITS: Internal transcribed spacer; MLSA: Multilocus Sequence Analysis; NADH: Nicotinamide Adenine Dinucleotide (Reduced); RNA: Ribonucleic Acid; rDNA: Ribosomal Deoxy Ribonucleic Acid; rRNA: Ribosomal Ribonucleic Acid

in the intercellular spaces and vascular tissues till the environmental conditions become fit for disease development.

University, North Carolina Cooperative Extension Sevice (Lan et al., 2013).

Parenchymatous tissues are macerated by massive quantities of pectic exoenzymes exudates produced during this period. These enzymes comprise of cellulolytic enzymes, pectate lyases, and pectin methylesterases, which are responsible for the total tissue destruction (Parthiban et al., 2012).

Soft rot can be found worldwide, anywhere ample storage tissues of vegetables and ornamentals are found (Golkhandan et al., 2013; Elbanna et al., 2014). Potatoes, carrots, and onions are among the most affected vegetables, along with tomato and cucumber (Mir et al., 2010) (**Figure 1**). Soft rot of fleshy vegetables and ornamental plants can be caused by more than six genera of pectolytic bacteria comprising; *Erwinia*, *Pseudomonas*, *Clostridium*, *Bacillus*, *Cytophaga*, and *Xanthomonas* (Elbanna et al., 2014). The estimated rate of infection of macergens on harvested crop ranges from 15 to 30%. *Erwinia* are the major macergens causing tissue degradation in vegetables (Choi and Kim, 2013; Waleron et al., 2014). Although, *Erwinia* are the primary macergens, it is not a single taxon. It is reclassified into genera such as *Pectobacterium* and *Dickeya* (Brady et al., 2012; Nabhan et al., 2012; Czajkowski et al., 2013). Macergens comprise of multiple groups ranging from the very complex *Pseudomonas*, a gamma-*proteobacteria* to as diverse as *Bacillus* and *Clostridium* which are firmicutes. *Bacillus* spp., *Clostridium* spp., *Pseudomonas marginalis*, and *Pantoea agglomerans* only cause soft rot when conditions are favorable to do so, thus are secondary invader called opportunistic pathogens (da Silva, 2013). Among all these pectolytic bacteria, soft rot *Erwinias* are the most important primary macergens that can macerate both the growing and harvested crop (Baz et al., 2012). All other bacteria are referred to as secondary because they can only destroy the parenchymatous tissues of plant under extreme environmental conditions or secondary invaders after *Erwinias* or other pathogens have infected the plant.

These macergens infect and destroy plant tissues both pre- and post-harvest and this species causes the greatest damage to harvested vegetables (Lee et al., 2012). There is need to ensure a continuous cold chain from immediately after post-harvest, to retail for successful management of these ubiquitous spoilage bacteria that only thrive well at temperatures of 20°C and above (Tournas, 2005). The fluorescent *Pseudomonads* (*P. fluorescens* and *P. viridiflava*) can macerate plant parenchymatous tissues at a temperature below 4°C. This cause higher occurrence of these bacteria on decayed vegetables both at wholesale and retail markets. These softrotting fluorescent *Pseudomonads* and *Erwinia* therefore become the major threat to commercial fresh product operations and fresh vegetables precisely, from the farm to retail and wholesale outlets (Saranraj et al., 2012). There are currently no commercial agents available specifically for controlling soft rot (Yaganza et al., 2014).

Despite advances in vegetable production and disease management, many challenges face growers of vegetables, out of which the major one is the damage caused by macergens (Wu et al., 2012). Macergens damage the tissues of vegetable thereby reducing the quality, yield, shelf-life and consumer satisfaction of these plants (Akhtar, 2015). They usually cause great economic losses due to their ability to infect and macerate vegetable tissues at any point in time, be it, the field, transit, storage or marketing period (Lee et al., 2012). In the nature of

#### **TABLE 1 | List of interesting** *Erwinia* **species.**


today's worldwide market, there are extremely high expectations for growers to provide ample supplies of high-quality, disease-free produce that have extended shelf-life (Kewa, 2012; Cheverton, 2015). The traditional methods to identify these macergens are extremely slow, more complex and obsolete (Hawks, 2005). Also, resistance genes active against macergens have been found in multiple host species, but their sequences and mechanisms remain unknown (Nykyri et al., 2012). Hence, means of quick identification of these bacteria are very essential. But the understanding of the taxonomy of these macergens will go a long way in shedding light to understand their biology and ultimately to the best method of controlling them. At present, there is very few knowledge available on the biology, ecology and epidemiology of macergens affecting vegetables in lowland and highland tropics. In order to increase crop production an assessment of biology, ecology and epidemiology of these bacteria need to be successfully implemented. Thus, this review focuses on the classification and taxonomy of the macergens to the species level. This is very important for more exploration in biotechnology.

# **TYPES OF MICROORGANISMS ON VEGETABLES**

The majority of Gram negative rods identified from raw vegetables were fluorescent *Pseudomonads* spp., *Klebsiella* spp., *Serratia* spp., *Flavobacterium* spp., *Xanthomonads* spp., *Chronobacterium* spp., and *Alcaligenes* (Elbanna et al., 2014). In vegetables like broccoli, cabbage, mungbean sprouts and carrot, Gram positive rods were predominantly isolated. *Coryneform* bacteria and catalase negative *cocci* were also predominantly isolated from broccoli, raw peas and raw sweet corn. In India, the mesophilic microflora of potatoes mainly comprised Gram positive bacteria, *Bacillus* spp., and *Micrococcus* spp. as fluorescent *Pseudomonads*, *Cytophaga* spp., *Flavobacterium* spp., *Xanthomonas*. spp., and *Erwinia* spp. *Leuconostoc mesenteroides* was the most common and abundant species found in vegetables amongst lactic acid bacteria (Andrews and Harris, 2000).

# **TAXONOMY OF MACERGENS**

#### **Genus** *Erwinia*

*Erwinia* belongs to the phylum *Proteobacteria*, class *Gammaproteobacteria*, order *Enterobacteriales* and family *Enterobacteriaceae*. For the past several decades, *Enterobacteria* that macerate and decay plants tissues, often referred to as the pectolytic Erwinias, were placed in genus *Erwinia*. Named after the eminent plant pathologist, Erwinin F. Smith. They are nonspore forming, facultative Gram negative rod-shaped anaerobes of approximately 0.7 *×* 1.5 µm in size with peritrichous flagella. This genus contains diverse set of group of organisms represented in **Table 1**. Since its establishment many new genera have been generated from *Erwinia*.

#### Nomenclature of *Erwinia*

Traditionally two species (*Erwinia carotovora* and *Erwinia chrysanthemi*) are circumscribed as the important plant pathogenic strains, but has been reclassification into a new genus, *Pectobacterium*, with multiple species being proposed (Gardan et al., 2003). *Pectobacterium* spp. (Waldee, 1945; formerly *Erwinia carotovora*) and *Dickeya* spp. (formerly *Erwinia chrysanthemi*) species are related to soft rot *Enterobacteria* pathogens with broad host ranges. These species formerly were known as the soft rot *Erwinia* spp., but several studies have shown that the soft rot *Enterobacteria* and *E. amylovora*, the type strain of the *Erwinia* genus, are too divergent to be included in one clade. Therefore, the soft rot *Erwinia* spp. were moved to two new genera as *Pectobacterium* and *Dickeya* (Nabhan et al., 2013). *Pectobacterium* and *Dickeya* spp. are considered broadhost range pathogens in part because, they have been isolated from so many plant species and in part because, single strains are pathogens of numerous plant species under experimental conditions. Within the genus *Pectobacterium*, there are five major clades designated I, II, III, IV, and V, which differs from previous studies. These comprise five subspecies or species-level clades of *Pectobacterium* namely; *Pectobacterium carotovorum* subsp. *carotovorum* (syn. *Erwinia carotovorum* subsp. *carotovorum*), *Pectobacterium atrosepticum* (syn. *Erwinia carotovorum* subsp. *atrosepticum*) *Pectobacterium wasabiae* (syn. *Erwinia carotovorum* subsp. *wasabiae*), *Pectobacterium betavasculorum* (syn. *Erwinia carotovorum* subsp. *betavasculorum*), and *Pectobacterium carotovorum* subsp. *brasiliense* (Hauben et al., 2005; Nabhan et al., 2012). The reconstructed phylogenies agree that *P. atrosepticum*, *P. betavasculorum*, and *P. wasabiae* do form individual clades and place the *brasiliensis* strains in an individual clade.

Previous suggestions to separate the pectolytic *Enterobacteria* into the genus *Pectobacterium* has not found favor among phytobacteriologists. Initially the suggestion was made by Waldee (1945), who recommended the segregation on the basis of the unique pectolytic activity of the bacteria. Consequently, Hauben et al. (1998) revived the suggestion to support the proposal by adding evidence from the 16S ribosomal DNA sequence analysis of various plant-associated members of the *Enterobacteriaceae*. Although the phenotypic characterization and analysis of a single DNA fragment might have been considered insufficient for the subdivision at the generic level, the DNA-DNA hybridization study conducted by Gardan et al. (2003) provides further stimulation to change in favor of the new nomenclature. Samson et al. (2005), have proposed several new species from new genus, *Dickeya* for *E. chrysanthemi*, comprising of six genomic species namely: *Dickeya dianthicola*, *D. dadantii*, *D. zeae*, *D. chrysanthemi*, *D. dieffenbachiae*, *D. paradisiaca*.

A recently initiated multi-locus sequencing project, as well as DNA hybridization data from the 1970s, supports the transfer of *E. carotovora* and *E. chrysanthemi* to two separate genera as well as the elevation of some soft rot *Erwinia* subgroups to the species level (Brady et al., 2012).

All the phylogenetic analyses completed to date have suffered from the small number of strains available for some *Enterobacteria* species, which makes it difficult to determine the relatedness of these taxa. Unfortunately, the naming and re-naming of species has caused considerable confusion in the literature, resulting in manuscripts being published with names that were used for only a few years. Since *Erwinia* has remained the preferred name used in the literature, the comprehensive phylogenetic study of the entire group of soft rot *Enterobacteria* remains uncompleted (Charkowski, 2006; Elbanna et al., 2014). The pectolytic *Erwinia* are ubiquitous in environments that support plant growth, and because they may be found in association with asymptomatic plants, they have been viewed as opportunistic pathogens analogous to medical bacteria that infect only immunologically compromised individuals. *Pectobacterium carotovorum*, in the family *Enterobacteriaceae*, is a highly diverse species consisting of at least two valid names, *P. carotovorum* subsp. *carotovorum* and *P. carotovorum* subsp. *odoriferum* and a suggested third taxon, *P*. *carotovorum* subsp. *brasiliense* (De Boer et al., 2012). Despite the lack of valid *carotovorum* publication, the *P. carotovorum* subsp. *brasiliense* name has been used in more than 10 publications since first published in 2004 as *Erwinia carotovora* subsp. *brasiliense* (Ma et al., 2007). Assigning strains to this taxon was based mainly on the genetic information of the 16S-23S intergenic spacer region of the rRNA operon, partial sequence of 16S rRNA gene and multilocus sequence analysis (MLSA) of housekeeping genes and MALDI-TOF characterization (Wensing et al., 2012). **Table 2** depicts the molecular method employed in the characterization of *Pectobacterium* and *Dickeya* species. *Pectobacterium carotovorum* subsp. *brasiliense* was first described as causing blackleg disease on potatoes (*Solanum tuberosum* L.) in Brazil and has since been described as also causing soft rot in *Capsicum annum* L., *Ornithogalum* spp., and *Daucus carota* subsp. *Sativus*. Strains of this taxon were isolated in the USA, Canada, South Africa, Peru, Germany, Japan, Israel, and Syria (Ngadze et al., 2012; Moleleki et al., 2013).

Genus *Pseudomonas* was first described in 1894 as one of the most diverse and ubiquitous bacterial genera whose species have been isolated worldwide from soil, decayed plant materials and rhizopheric region, quite a numerous plant species (Migula, 1894). They comprise a heterogeneous group of species which were grouped into five groups based on RNA homology (Saranraj et al., 2012). The RNA-homology group I belong to the fluorescent group because of their ability of producing pyoverdines. Pectolytic *Pseudomonas* belongs to this rRNA group I organism of gamma *Proteobacteria*. They are non-sporulating, Gram-negative, strict aerobic, rod-shape with polar flagella (Özen and Ussery, 2012). The strains of these bacteria called *P. marginalis* or *P. fluorescens* can be attributed to soft rot diseases in vegetables. The very complex groups of fluorescent, oxidase positive soft rot *Pseudomonas* are opportunistic macergens. **Table 3** represents the molecular methods for the description of *Pseudomonas* species belonging to macergens.

#### Nomenclature of *Pseudomonas*

The nomenclature of bacteria in the genus *Pseudomonas* has changed considerably during the last decennia. *P. marginalis* or *P. fluorescens* are pectinolytic that cause strains soft rot on a wide range of hosts. The taxonomic and phytopathogenic status of *P. marginalis* is not well known. However, these are biochemically and phenotypically indistinguishable from saprophytic strains of *P. fluorescens* biovars II, *P. putida*, and *P. chlororaphis* (now includes *P. aureofaciens*). Based on their ability to degrade pectin and macerate the plant parenchymatous tissues they are referred to as *P. marginalis*. Recently, based on 16S rRNA analysis Anzai et al. (2000) came up with 57 strains of *Pseudomonas* sensu stricto with seven subclusters: *P. syringae* group, *P. chlororaphis* group, *P. fluorescens* group, *P. putida* group, *P. stutzeri* group, *P. aeruginosa* group, and *P. pertucinogena* group (Novik et al., 2015). Also, in the same genus *Pseudomonas*, some species have been found to be misclassified for instance *P. aureofaciens* and *P. aurantiaca*, which were reclassified into *P. chlororaphis* (Peix et al., 2007).

Ever since the discovery of genus *Pseudomonas*, it has undergone several taxonomic changes not only as far as the number of species included, but also as far as the criteria used for their definition and delineation. In Bergey's Manual of Systematic Bacteriology's current edition, an extensive list of methods used in *Pseudomonas* taxonomy was integrated (Palleroni, 2005). These methods, which consist of cell morphology and structure, cell wall composition, pigment types, nutritional and metabolic characteristics, susceptibility to different compounds, antibiotic production, pathogenicity of other organisms, antigenic structure and genetic and ecological studies

#### **TABLE 2 | Molecular methods of identifying macergens.**


*PFGE: Pulse-field gel electrophoresis; 16S-23S intergenic transcribed region of the rRNA operon; MLSA: multilocus sequence analysis of housekeeping genes; MALDI-TOF MS: matrixassisted laser desorption/ionization time-of-flight mass spectrometry; AFLP: amplified fragment length polymorphism; MLST: multilocus sequence tagging; RFLP: restriction fragment length polymorphism; RAPD: random amplification of polymorphic DNA; rep-PCR: repetitive sequence-based PCR 3.2 Genus* Pseudomonas*.*

revealed the efforts for characterizing *Pseudomonas* species. The phenotypic taxonomic markers comprise a set of tests, namely: cell shape, flagella type, consumption of carbon sources such as organic acids, polyalcohols and amino acids, ability to grow in different culture conditions, antibiotic resistance, production of antibiotic substances and exocellular enzymes (Palleroni, 2005).

In *Pseudomonas* taxonomy, the effectiveness of chemotaxonomic studies has been confirmed, such as quinone systems, fatty acid, protein, polar lipid or polyamine profiles, which are usually useful in the taxonomy of most bacterial groups. Generally, *Pseudomonas* species were reclassified by chemotaxonomic markers into other genera such as *P. mephitica* into *Janthinobacterium lividum* (Kämpfer et al., 2008). Janse et al. (1992), used whole fatty acid analysis in the study of a broad collection opportunistic phytopathogenic to clarify the taxonomic position of some *P. marginalis* strains included in the *P. fluorescens* group. Also, Janse et al. (1992) reported that other bacteria (*P. putida*, *P. aureofaciens*, and *P. tolaasii*) within the fluorescent oxidase positive pseudomonads group also exhibit pectinolytic ability. Hence, they are referred to as *P. fluorescens* supercluster. The study of polyamine composition in Proteobacteria revealed putrescine as the main polyamine present in the *P. fluorescens* complex, thus help in the delineation of species from this group. Recently, the polar lipid patterns of representative species of genus *Pseudomonas* were analyzed which showed the presence of phosphatidylglycerol, diphosphatidylglycerol and phosphatidylethanolamine as major polar lipids (Cámara et al., 2007).

Siderotyping an interesting taxonomic tool was used in characterizing fluorescent and then to non-fluorescent based on the isoelectrophoretic. Characterization of the major siderophores and pyoverdines and determination of strains pyoverdine mediated iron uptake specificity led to characterization of several *Pseudomonas* strains at species level, through species-specific pyoverdines (Novik et al., 2015). Mass spectrometry for the determination of molecular mass of pyoverdines has helped recently to improve siderotyping resolution power and accuracy (Meyer et al., 2008).

Currently fluorescent spectroscopy fingerprinting, the most modern techniques for biomolecules analysis are being applied to *Pseudomonas* taxonomy, by emission spectra of three intrinsic fluorophores (NADH, tryptophan, and the complex of aromatic amino acids and nucleic acid), which have been able to differentiate *Pseudomonas* at genus level from

#### **TABLE 3 | Molecular methods for the description of** *Pseudomonas* **species belonging to macergens.**


*16S-23S intergenic transcribed region of the rRNA operon; MLSA: multilocus sequence analysis of housekeeping genes; MALDI-TOF MS: matrix-assisted laser desorption/ionization time-of-flight mass spectrometry; AFLP: amplified fragment length polymorphism; MLST: multilocus sequence tagging; RFLP: restriction fragment length polymorphism; rep-PCR: repetitive sequence-based PCR.*

#### **TABLE 4 | Macergens host pathogenicity.**


*Burkholderia*, *Xanthomonas*, or *Stenotrophomonas* with very high sensitivity, and moreover at species level *P. chlororaphis*, *P. lundensis*, *P. fragi*, *P. taetrolens* and *P. stutzeri* grouped separately from *P. putida*, *P. pseudoalcaligenes*, and *P. fluorescens*, which correlate with the phylogenetic clusters earlier obtained by Anzai et al. (2000); Peix et al. (2007), and Tourkya et al. (2009).

Hence, other gene sequences like housekeeping genes have been used in the last decade as phylogenetic molecular markers in taxonomic studies such as the *rec*A, *atp*D, *car*A, *gyr*B, and *rpo*B, whose effectiveness has been demonstrated in genus *Pseudomonas* for species differentiation (Hilario et al., 2004). For instance, the effectiveness of *rpo*B has been reported in discriminating closely related *Pseudomonas*, with a phylogenetic resolution of the *rpo*B tree roughly three times higher than that of the 16S rRNA gene tree (Tayeb et al., 2005). These genes also enhanced differentiation of subspecies within *P. chlororaphis* (Hilario et al., 2004; Peix et al., 2007). Nevertheless, the analysis of housekeeping genes is not frequently used so far in *Pseudomonas* species description, but only *gyr*B, *rpo*B and *rpoD* have been integrated in the current description of *P. xiamenensis* (Lai and Shao, 2008).

16S-23S rRNA intergenic spacer is another phylogenetic marker used increasingly in taxonomic studies for discrimination of very closely related bacteria, at species and intraspecific levels, even at the strain level because of its high variability both in size and sequence (Sakamoto et al., 2001). This region can be amplified by using universal primers, and specific protocols (Locatelli et al., 2002). The efficacy of this phylogenetic marker has been reported in the differentiation of *Pseudomonas* species (Guasp et al., 2000). The selection of the minimal principles necessary for species delineation and description is selected for each bacterial genus by a committee created by experts in the given genus. The methods used in the taxonomy of the genus *Pseudomonas* and its related genera have been standardized by the subcommittee on the taxonomy. However, the minimal standards for genus *Pseudomonas* species description are yet to be cleared after the 2002 meeting of this subcommittee (De Vos and Yabuuchi, 2002). Hence, the new species description of this genus must be based on the general minimal standards for bacterial species characterization (Stackebrandt et al., 2002). These general minimal standards needed for the classification of new species and/or subspecies must comprise 16S rRNA sequencing, DNA-DNA hybridization, fatty acid analysis and phenotypic classification.

#### **Genus** *Xanthomonas*

The genus *Xanthomonas* belong to the family *Xanthomonadaceae*. This family composed of 10 genera that dwell in an extreme environment. The genus *Xanthomonas* belongs to the gamma proteolytic subdivision (Mbega et al., 2014). They are Gramnegative, aerobic, rod-shape, motile, non-spore forming with a single polar flagellum, comprises of 27 species infecting more than 400 dicots and monocots plant species (Rodriguez et al., 2012).

#### Nomenclature of *Xanthomonas*

Traditionally, genus *Xanthomonas* is referred to as a taxon of pathogenic plant bacteria (Dye et al., 1974; Bradbury, 1984). *Xanthomonas* usually produce some extracellular polysaccharide namely: xanthan and xanthomonadin, a membrane-bound, brominated, aryl-polyene, yellow pigment (Adriko et al., 2014). This yellow pigment is responsible for their pathogenicity and virulence (Subramoni et al., 2006). However, the yellow-pigmented *X.* spp. (*X. campestris*) are the only one associated with tissue maceration of the postharvest vegetables and fruits (Liao and Wells, 1987). They are opportunistic macergens because they are entering through natural openings or after infection of the plant by *Erwinia* spp.

Genetically, it can be differentiated into over 141 pathovars (pv.) based on specificity range (Swings and Civerolo, 1993). But *Xanthomonas* classification of *X*. *campestris* pathovar was based on the host pathogenicity system (**Table 4**)

Initially, this genus undergone diverse taxonomic and phylogenetic studies based on their phenotype and host specificity. Until Vauterin et al. (1995) revised the reclassification of *Xanthomonas* by DNA-DNA hybridization into 20 species based on their genomic relatedness. Phenotypic fingerprinting techniques such as 50S-polyacrylamide gel electrophoresis (50S-PAGE) of cellular proteins and gas chromatographic analysis of fatty acid methyl esters (FAME) reasonable supported these genomic groups to an extent. Hence, both techniques are useful tools in specific and interspecific differentiation of *Xanthomonas* levels (Rademaker et al., 2000).

Other analyses like Multi-Locus Sequence Analysis (MLSA), Amplified Fragment Length Polymorphism (AFLP) were also used in characterisation of this genus, revealing the complexity and diversity of the genus previously described by DNA-DNA hybridization (Ferreira-Tonin et al., 2012; Hamza et al., 2012). Not quite long, the phylogeny of species representing the principal lineages of the genus *Xanthomonas* were reported based on their genome (Rodriguez et al., 2012). The 16S ribosomal DNA sequences and MLSA classified *Xanthomonas* species into two major groups (Vicente and Holub, 2013). Group I comprising: *X. albilineans*, *X. hyacinth*, *X. theicola*, *X. sacchari* and *X. translucens*, and Group II made up of *X. arboricola*, *X. axonopodis*, *X. bromi*, *X. campestris*, *X. cassavae*, *X. codiaei*, *X. cucurbitae*, *X. fragariae*, *X. hortorum*, *X. melonis*, *X. oryzae*, *X. pisi*, *X. populi*, *X. vasicola*, and *X. vesicatoria* (Rodriguez et al., 2012). Thus, taxonomy of this genus are still subjected to debate since the last decade (Rodriguez et al., 2012; Vandroemme et al., 2013; Lamichhane, 2014).

# **CONCLUSION**

The taxonomy of all these macergens is far from being complete because of the controversial issues arising from their classification which were based on host pathogenicity (**Table 1**). This may be affected by the sudden change in the ecosystem. This classification is not based on scientific research perspective for defined taxa and the consequences brought about by these marcergens may become difficult to understand. It is majorly based on symptoms that is similar in all the macergens, and this is unreliable according to (Sławiak et al., 2013). Although, some scientific method like MLSA were used for the classification they have limitation of single locus analysis. Thus, a proper classification is imperative, in order to reflect an understanding of their existing natural diversity and relationships among them. This will help plant breeders, farmers, and legislators to ensure quick and effective disease diagnosis and management, in order to avoid unnecessary destruction of economically valuable crops. The knowledge of genomic diversity within the macergens pathovars is necessary for host resistance disease based management strategies for the plant breeders.

As a concluding comment, we would like to stress that we applaud further developments in molecular methods of analyzing macergens for a better classification of these macergens. It is our belief, however, that any future progress in taxonomy as a scientific discipline will depend only on the availability of new experimental data that will broaden and refined the view on bacterial diversity.

# **AUTHOR CONTRIBUTIONS**

BR involved in data collection from internet, drafting of the manuscript or revising it critically for important intellectual content; have given final approval of the version to be published; and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. OO involved in collection of data, drafting of the manuscript, revising it critically, responsible for any aspect of the article and also help in the general supervision of the article.

# **ACKNOWLEDGMENTS**

The authors would like to thank the North West University and National Research Foundation, South Africa for funds (Grant no. UID81192 OO Babalola) that have supported research in their laboratory.

# **REFERENCES**


*Pectobacterium wasabiae* sp. nov. *Int. J. Syst. Evol. Microbiol.* 53, 381–391. doi: 10.1099/ijs.0.02423-0


Swings, J. G., and Civerolo, E. L. (1993). *Xanthomonas*. London: Chapman & Hall.


threat to brassica crops. *Mol. Plant Pathol.* 14, 2–18. doi: 10.1111/j.1364- 3703.2012.00833.x


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2015 Aremu and Babalola. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# The Membrane Proteins Involved in Virulence of *Cronobacter sakazakii* Virulent G362 and Attenuated L3101 Isolates

*YingWang Ye1,2,3,4\* †, Jina Gao1†, Rui Jiao1†, Hui Li1,2,3,4 †, Qingping Wu2,3,4\*, Jumei Zhang2,3,4 and Xian Zhong2,3,4*

*<sup>1</sup> School of Biotechnology and Food Engineering, Hefei University of Technology, Hefei, China, <sup>2</sup> State Key Laboratory of Applied Microbiology Southern China, Guangzhou, China, <sup>3</sup> Guangdong Provincial Key Laboratory of Microbiology Culture Collection and Application, Guangzhou, China, <sup>4</sup> Guangdong Institute of Microbiology, Guangzhou, China*

#### *Edited by:*

*Giovanna Suzzi, Università degli Studi di Teramo, Italy*

#### *Reviewed by:*

*Learn-Han Lee, Monash University Malaysia, Malaysia Maria Schirone, Università degli Studi di Teramo, Italy*

#### *\*Correspondence:*

*Yingwang Ye yeyw04@mails.gucas.ac.cn; Qingping Wu wuqp203@163.com*

*†These authors have contributed equally to this work.*

#### *Specialty section:*

*This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology*

*Received: 23 August 2015 Accepted: 26 October 2015 Published: 09 November 2015*

#### *Citation:*

*Ye Y, Gao J, Jiao R, Li H, Wu Q, Zhang J and Zhong X (2015) The Membrane Proteins Involved in Virulence of Cronobacter sakazakii Virulent G362 and Attenuated L3101 Isolates. Front. Microbiol. 6:1238. doi: 10.3389/fmicb.2015.01238* *Cronobacter sakazakii* is an opportunistic foodborne pathogen and the virulence differences were previously documented. However, information about membranous proteins involved in virulence differences was not available. In this study, virulent characterization such as biofilm formation and flagella motility between virulent *C. sakazakii* isolate G362 and attenuated L3101 were determined. Then, twodimensional gel electrophoresis (2-DE) technology was used to preliminarily reveal differential expression of membranous proteins between G362 and L3101. On the mass spectrometry (MS) analysis and MASCOT research results, fourteen proteins with differential expression were successfully identified. At the threshold of twofold changes, five out of eight membranous proteins were up-regulated in G362. Using RT-PCR, the expression abundance of the protein (*enzV, ompX, lptE, pstB,* and *OsmY*) genes at mRNA levels was consistent with the results by 2-DE method. The findings presented here provided novel information and valuable knowledge for revealing pathogenic mechanism of *C. sakazakii*.

Keywords: membranous proteins, *Cronobacter sakazakii*, virulence factors, two-dimensional gel electrophoresis (2-DE)

## INTRODUCTION

The genus of *Cronobacter* is a group of opportunistic pathogens linked with life-threatening infections in neonates (Holý and Forsythe, 2014). *Cronobacter* sp. was isolated from various kinds of samples including food, soil, environments, clinical samples (Gurtler et al., 2005). In addition, these infections often occurred among poor immunity or low-birth weight newborn through consumption of contaminated powdered infant formula (Muytjens et al., 1983; Tall et al., 2014). The *Cronobacter* genus consists of *Cronobacter sakazakii*, *C. malonaticus*, *C. turicensis*, *C. muytjensii*, *C. condiment*, *C. universalis*, and *C. dublinensis* (Joseph et al., 2012a,b). Furthermore, *C. sakazakii* is the predominant species within *Cronobacter* isolated from powdered milk (Xu et al., 2014).

Membranous proteins were required for adhesion, invasion, and the release of toxin of pathogens, and further played important roles on host-pathogen interaction or host infections in foodborne pathogens (Allaoui et al., 1993; Monteville et al., 2003; Ellis and Kuehn, 2010). Transmembrane signal proteins of two-component regulatory system in gram-negative pathogens contributed to immune escape or biofilm formation (Krachler et al., 2011). In addition, outer membrane proteins A (OmpA) and X (OmpX) in *C. sakazakii* played critical roles in adhesion to and invading human cells (Mohan Nair and Venkitanarayanan, 2007; Mohan Nair et al., 2009; Kim et al., 2010). Kothary et al. (2007) reported that *zpx* gene encoding the cellbound zinc-containing metalloprotease might play important roles in dissemination of *Cronobacter* into the systemic circulation. However, little information about membranous proteins involved in virulence differences among *C. sakazakii* isolates is available.

To determine the potential virulence factors of pathogenic bacteria, the use of global techniques such as proteomics was widely applied in exploring different factors between virulent and attenuated bacteria (Srinivasa Rao et al., 2004; Hecker et al., 2010; Niemann et al., 2011).

In this study, the virulent characterization including biofilm formation, flagella motility, and O serotypes between G362 and L3101 was determined. Then, the 2-DE technology coupled with mass spectrometry was applied to identify global molecular information about differentially expressed membranous proteins. Finally, Real-time PCR was used to further validate the expression abundance of these novel membranous protein genes between G362 and L3101.

# MATERIALS AND METHODS

# *Cronobacter sakazakii* Isolates

The virulence differences between *C. sakazakii* G362 and L3101 from food samples were previously determined through intraperitoneal injection and histopathologic analysis (small intestine, kidney, and liver; Ye et al., 2015). The significant

morphological change in small intestine, liver, and kidney further reveals the virulence differences between G362 and L3101.

# Different Phenotypic Characterization between G362 and L3101

The biofilm formation between G362 and L3101 was determined using crystal violet staining (CVS) method as described by Hochbaum et al. (2011). The O serotypes were detected as described by Sun et al. (2012). In addition, the primers, PCR conditions, and PCR mixtures also referred to the description by Sun et al. (2012).

In addition, the motility of two isolates (G362 and L3101) was also determined. Media consisting of 10 g/L tryptose, 5 g/L NaCl, and 1.5 g/L agarose was used to test the swimming motility. Swimming plate was inoculated with *C. sakazakii* isolates from an overnight culture of TSA agar using a sterile toothpick and incubated at 37◦C for 8 h. Media (8 g/L nutrient broth, 5 g/L D-glucose, and 5 g/L Difco bacto-agar) was made to determine the swarming motility. Swarming plate was inoculated with *C. sakazakii* isolates from swimming plate using a sterile toothpick and incubated at 37◦C for 12 h. The diameter of motility at the agar dish interface was measured in triplicate.

# Membranous Proteins Extraction and Quantification

Two *C. sakazakii* isolates (G362 and L3101) were inoculated into tryptic soy broth (TSB, Huankai, Guangzhou) for incubation at 37◦C for 16 h. Two grams (wet weight) of G362 and L3101 were suspended in 5 ml of phosphate buffer saline (PBS, pH7.4) at 4◦C. Then the cell suspension was sonicated in an ice bath (SONICS VC-505). Following the sonication process, Triton X-114 was applied for membranous proteins extraction using the method described by González de la Vara and Alfaro (2009). Then, two different phase collections and the insoluble pellet were performed to analyze SOD activity (Huang et al., 2006). Acetone precipitation was employed to remove the excess salts in the aqueous phase collection and detergent phase collection respectively. After lysis buffer (5 M urea, 2 M thiourea, 2% SB3-10, 2% CHAPS, 65 mM DTT, 40 mM Tris; Bio-Rad, USA) dissolution and centrifuged at 18, 000 × *g* for 1 h at 4◦C, membrane proteins supernatant was stored at −80◦C for use. Proteins qualification was measured by the bicinchoninic acid (BCA) assay as described by Krieg et al. (2005).

# 2-DE Analysis

For each sample, 100 µl (8 µg/µl) of membrane proteins extract, diluted with 200 µl of rehydration buffer (7 M urea, 2 M thiourea, 1% ASB-14, 65 mM DTT, 0.2% Bio-Lyte (w/v), 40 mM Tris, 0.001% Bromophenol blue; Bio-Rad, USA), was applied to a reswelling cassette with 17 cm immobiline dry strip (pH 3- 10 NL, Bio-Rad, USA). Rehydration was allowed to proceed at 20◦C to 16 h under silicone oil. The isoelectric focusing (IEF) was performed using the PROTEAN IEF system (Bio-Rad, USA). Each IPG strip current limit during IEF was set 50 µA at 20◦C. The focusing was conducted in next five steps: (1) constant voltage at 250 V for 1 h; (2) constant voltage at 500 V for 30 min; (3) constant voltage at 1000 V for 2 h; (4) under linear ramping mode voltage increased from 1000 to 10, 000 V in 5 h; (5) at last constant voltage at 10, 000 V until the total volt-hours reached 68,000 V-h. The following steps including equilibration, SDS-PAGE electrophoresis, coomassie brilliant blue stain (CBB-stained), gel analysis, and subsequent MS analysis were performed as described by Cordwell et al. (2008) with modification. In brief, spots in gels between isolates corresponding to the same protein identifications were detected using PD-Quest (Bio-Rad) and the relative spot intensities calculated. Significantly changed spots were selected by rate increased/decreased twofold. Then Protein spots of interest were subjected to in-gel digestion, and then cut from the gels for mass spectrometric analysis. Protein identities were based on a combination of peptide fingerprint and MS/MS spectra. MS


c*Accession numbers in uniprot protein database (available at* 

d*pI, the predicted isoelectric point calculated from the protein sequence.*

e*The predicted molecular weight calculated from the protein sequence.*

f*Number of experimental*

 *peptides matched to the predicted protein peptides.*

*http://www*.*uniprot*.*org).*

TABLE 2 | Primers used in the study by real-time PCR.


and MS/MS data were searched using MASCOT version 1.9.05 (Matrix Science) as search engine against the NCBI database. *C. sakazakii* species database was defined as a matching species. For each isolate sample, the 2-D gels in triplicate were analyzed.

#### Expression Abundance of Membranous Protein Genes by Real Time PCR

The overnight culture (1.0 ml) of *C. sakazakii* isolates was subject to extraction of RNA using bacterial RNA extraction kit (BIOMIGA, USA). Then, cDNA was obtained using firststrand cDNA synthesis kit (BIOMIGA, USA) according to the manufacturer' instructions. The RT-PCR mixture consisted of 2 µl cDNA, 10 µl 2 × SYBR Green qPCR mix (Giagen, Beijing), 0.2 µM for each primer, and RNase-free water in a final volume of 20 µl. The PCR program was initiated at 95◦C for 5 min, followed by 40 cycles at 95◦C for 10 s, 59◦C for 20 s, 72◦C for 25 s using ABI Stepone plus (Applied Biosystems). The expression abundance of targeted genes between G362 and L3101 was determined in triplicate using --Ct method.

#### RESULTS AND DISCUSSION

*Cronobacter sakazakii* is an important foodborne pathogen associated with severe infections in poor immunity or lowbirth weight newborn. The virulence differences of *C. sakazakii* G362 and L3101were previously documented by Ye et al. (2015). In this study, the virulent characterization including biofilm formation, O serotypes, and motility was further determined. In **Figure 1A**, both G362 and L3101 were O5 serotype by PCR assay (Sun et al., 2012). Xu et al. (2014) found that O2 serotype in *C. sakazakii* is the predominant serotype among isolates from powdered milk. However, the correlation between O serotypes and virulence was not revealed. Additionally, the significant differences of biofilm formation between G362 and L3101 were observed in **Figure 1B**, which might be involved in their adhesion to surfaces. The biofilm formation was associated with virulence in *Candida* (Hasan et al., 2009), *Escherichia coli* (Naves et al., 2008), and *Pseudomonas* (Ghadaksaz et al., 2015). Interestingly, the differences of motility (swimming and swarming) in **Figure 1C** were also founded between G362 and L3101 at 37◦C for 8 h (swimming motility) and 12 h (swarming motility). Swimming ability of L3101 is stronger than that of G362, while the weakest swarming motility was also observed in L3101. This is the first report to determine the relationship between flagella motility and biofilm formation in *C. sakazakii*. Our result preliminarily indicated that swimming motility had negative effects on the biofilm formation, while swarming motility contributed to the biofilm formation. The biofilm formation required flagella motility in *Pseudomonas aeruginosa* (Barken et al., 2008), and motility and adhesion to surface was also connected in *E. coli* (Wood et al., 2006). However, flagella motility was not necessarily required for biofilm formation in *P. aeruginosa* (Caiazza et al., 2007) and *Burkholderia pseudomallei* (Lee et al., 2010). In *C. sakazakii*, the regulation of biofilm formation and flagella motility remains to be revealed.

Thereafter, to reveal the membranous proteins involved in virulence differences among *C. sakazakii* isolates, the membranous proteomic profiles of G362 and L3101 were firstly constructed using the 2-DE technology (**Figure 2**). Changes of membranous proteins in intensity (>twofold) were considered as significant differences. The expression abundance of 14 proteins was successfully identified between G362 and L3101. In **Table 1**, 8 out of 14 differentially expressed proteins was confirmed as membrane-associated proteins and expression abundance of seven membranous proteins was increased in G362. The different expression abundance of genes of membranous proteins including enzV, ompX, lptE, pstB, and OsmY were detected using real time PCR. The primers were listed in **Table 2** and 16S rRNA gene was used as internal control. From **Figure 3**, the expression of five membranous protein genes at mRNA levels using real-time PCR was consistent to results by 2-D method. The relative expression abundance of *osmF, ompX, enzV,* and *lptE* genes in G362 were 6.88, 4.49, 1.48, and 3.82 fold changes than those in L3101, respectively, while the expression of *pstB* in L3101 was 61.38 fold changes than that in G362. The relative expression of five factors by real-time PCR was consistent with results using 2-DE method.

Based on proteomic analysis, the outer membrane proteins (ompA and ompX) were increased in G362 compared to those in L3101. The ompA and ompX proteins were reported to enhance adherence and invasion of human cells (Mohan Nair and Venkitanarayanan, 2007; Mohan Nair et al., 2009; Liu et al., 2012). Outer membrane A of *E. coli* has been reported enhanced the invasion of brain microvascular endothelial (BMEC) by assisting through the blood-brain barrier (Prasadarao et al., 1996). Our findings also indicated that ompA and ompX might be involved in virulence differences of *C. sakazakii* at membranous protein levels.

Besides, the expression abundance of novel proteins of LPSassembly lipoprotein LptE, Osmolarity sensory histidine kinase EnvZ, and Osmotically inducible protein OsmY in G362 were increased compared to those in L3101. The LPS-assembly lipoprotein LptE encoded by the gene *lptE* in *E. coli* is essential for the biosynthesis of lipopolysaccharide, resulting in strong immunogenicity with host cells (Chng et al., 2010). Inner membrane osmolarity sensory histidine kinase EnvZ-OmpR pair in *C. sakazakii* would be activated in response to environmental osmoic stress (Amalaradjou and Venkitanarayanan, 2011). The receptor ompR could influence bacterial biofilm formation and flagella motility by regulating genes *ompF* and *ompC*. Further findings indicated that virulence of *ompR* mutants decreased in *Salmonella typhimurium* (Bernardini et al., 1990; Raczkowska et al., 2011). Additionally, Osmotically inducible protein OsmY also played important roles in pathogenic bacteria. In *S. typhimurium* and *E. coli*, OsmY was indirectly associated with virulence behavior (Bader et al., 2003; Dong and Schellhorn, 2009). In addition, putative multidrug resistance protein MdtD and putative ABC transporter permease were also up-regulated

#### REFERENCES


in G362. Some authors reported mdtABCD together with BaeSR two-component system contributed to resisting to host defenses through membrane drug-efflux pump (Baranova and Nikaido, 2002; Zoetendal et al., 2008; Leblanc et al., 2011). ABC transporter permease, a multi-pass membrane protein, was responsible for oligogalacturonide transport, but its detailed mechanism involved in virulence of bacteria is not documented.

Interestingly, the expression abundance of some proteins like phosphate import ATP-binding protein PstB, GDP-L-fucose synthase and methylthioribose kinase were increased in L3101. To date, the correlation of these proteins and bacterial virulence was not documented.

In summary, the relationship between flagella motility and biofilm formation was preliminarily determined, which indicated that motility was not necessarily required for biofilm formation. Then, membranous proteomic profiles between virulent and attenuated *C. sakazakii* isolate were firstly constructed. In spite of ompA and ompX, some novel factors (envZ, LptE, MdtD, and OsmY) were successfully identified by proteomics which were potentially involved in virulence differences. The findings here provided novel insights and better knowledge for revealing pathogenic mechanisms of *C. sakazakii*. However, the detailed roles and regulation of these potential virulence factors of *C. sakazakii*, and study on interaction between host and bacteria are also urgent for exploration of pathogenic mechanism.

#### ACKNOWLEDGMENTS

We gratefully acknowledge the financial support of the National Natural Science Foundation of China (31201292) and Guangdong Province, Chinese Academy of comprehensive strategic cooperation project (2012B090400017).


and contributes to invasion of human brain microvascular endothelial cells. *Foodborne Pathog. Dis.* 6, 495–501. doi: 10.1089/fpd.2008.0228


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2015 Ye, Gao, Jiao, Li, Wu, Zhang and Zhong. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# An insight into the isolation, enumeration, and molecular detection of *Listeria monocytogenes* in food

*Jodi Woan-Fei Law1†, Nurul-Syakima Ab Mutalib2†, Kok-Gan Chan3 and Learn-Han Lee1\**

*<sup>1</sup> Biomedical Research Laboratory, Jeffrey Cheah School of Medicine and Health Sciences, Monash University, Bandar Sunway, Malaysia, <sup>2</sup> UKM Medical Molecular Biology Institute, UKM Medical Centre, Kuala Lumpur, Malaysia, <sup>3</sup> Division of Genetics and Molecular Biology, Institute of Biological Sciences, Faculty of Science, University of Malaya, Kuala Lumpur, Malaysia*

#### *Edited by:*

*Maria Schirone, University of Teramo, Italy*

#### *Reviewed by:*

*Qiaobin Xiao, Cornell University, USA Juan Aguirre, Food Safety Adviser Chile, Chile Paula Teixeira, Universidade Católica Portuguesa, Portugal*

#### *\*Correspondence:*

*Learn-Han Lee lee.learn.han@monash.edu; leelearnhan@yahoo.com*

*†These authors have contributed equally to this work.*

#### *Specialty section:*

*This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology*

*Received: 20 August 2015 Accepted: 20 October 2015 Published: 03 November 2015*

#### *Citation:*

*Law JW-F, Ab Mutalib N-S, Chan K-G and Lee L-H (2015) An insight into the isolation, enumeration, and molecular detection of Listeria monocytogenes in food. Front. Microbiol. 6:1227. doi: 10.3389/fmicb.2015.01227* *Listeria monocytogenes*, a foodborne pathogen that can cause listeriosis through the consumption of food contaminated with this pathogen. The ability of *L. monocytogenes* to survive in extreme conditions and cause food contaminations have become a major concern. Hence, routine microbiological food testing is necessary to prevent food contamination and outbreaks of foodborne illness. This review provides insight into the methods for cultural detection, enumeration, and molecular identification of *L. monocytogenes* in various food samples. There are a number of enrichment and plating media that can be used for the isolation of *L. monocytogenes* from food samples. Enrichment media such as buffered *Listeria* enrichment broth, Fraser broth, and University of Vermont Medium (UVM) *Listeria* enrichment broth are recommended by regulatory agencies such as Food and Drug Administration-bacteriological and analytical method (FDA-BAM), US Department of Agriculture-Food and Safety (USDA-FSIS), and International Organization for Standardization (ISO). Many plating media are available for the isolation of *L. monocytogenes*, for instance, polymyxin acriflavin lithiumchloride ceftazidime aesculin mannitol, Oxford, and other chromogenic media. Besides, reference methods like FDA-BAM, ISO 11290 method, and USDA-FSIS method are usually applied for the cultural detection or enumeration of *L. monocytogenes*. most probable number technique is applied for the enumeration of *L. monocytogenes* in the case of low level contamination. Molecular methods including polymerase chain reaction, multiplex polymerase chain reaction, real-time/quantitative polymerase chain reaction, nucleic acid sequence-based amplification, loop-mediated isothermal amplification, DNA microarray, and next generation sequencing technology for the detection and identification of *L. monocytogenes* are discussed in this review. Overall, molecular methods are rapid, sensitive, specific, time- and labor-saving. In future, there are chances for the development of new techniques for the detection and identification of foodborne with improved features.

Keywords: *Listeria* species, isolation, enumeration, molecular detection, foodborne pathogens

# INTRODUCTION

*Listeria monocytogenes* has become an important foodborne pathogen and it can be found in a variety of foods which include raw foods and processed foods (Gasanov et al., 2005; Janzten et al., 2006; Liu, 2006; Jeyaletchumi et al., 2010a; Välimaa et al., 2015). *L. monocytogenes* has been a serious threat to the food industry due to its ability to survive the most common food processing conditions such as extreme pH, high salt concentration, low water activity, and refrigeration temperatures (Liu, 2006; Zunabovic et al., 2011; Jadhav et al., 2012). It is known that *L. monocytogenes* can be eliminated or reduced by pasteurization process because it cannot survive pasteurization temperatures in food processing (Jadhav et al., 2012). For instance, a study conducted by Murphy et al. (2003) showed that hot water pasteurization and steam pasteurization resulted in a 7 log10 (CFU/g) reduction of *L. monocytogenes* in inoculated single packaged fully cooked chicken breast fillets, 227 g packaged fully cooked chicken strips and 454 g packaged fully cooked chicken strips when pasteurized at 90◦C for 5, 25, and 35 min, respectively. However, post-processing or post-pasteurization contamination by *L. monocytogenes* may occur due to crosscontamination or biofilms (Jadhav et al., 2012). *L. monocytogenes* can attach to food contact surfaces such as stainless steel and polystyrene during food processing and form biofilms which is important for their survival in hostile environments (Jadhav et al., 2012; Da Silva and De Martinis, 2013; Välimaa et al., 2015). Biofilms may persist for a long period of time and they can tolerate high concentrations of disinfectants, sanitizers, and antimicrobials (Välimaa et al., 2015). Hence, this may result in the contamination of food contact surfaces which then lead to higher risk of food contamination during and/or after processing. Food contaminated with *L. monocytogenes* has posed a great concern to the food industry as it can cause serious infection known as listeriosis when ingested and it is also one of the causes of recalls which may result in large economic losses (Jemmi and Stephan, 2006). Thus, microbiological food testing is important to ensure the safety of food products (Dwivedi and Jaykus, 2011; Välimaa et al., 2015).

#### DETECTION AND IDENTIFICATION OF *Listeria monocytogenes*

#### Enrichment Media and Selectivity

There are various selective enrichment and plating media that have been developed and utilized for the isolation and detection of *L. monocytogenes* in food and environmental samples. As required by majority of the regulatory agencies, the isolation methods must be able to detect one *Listeria* organism per 25 g of food. In order to achieve this sensitivity, enrichment methods are required to allow the organism to grow and reach a detectable level of <sup>∼</sup>104–105 CFU ml−1, before plating onto selective media and confirmation of cultures. Antimicrobial agents are employed in enrichment and plating media to suppress competing microflora as *Listeria* cells are slow growing and can be rapidly out-grown by competitors. The most common selective agents are acriflavine, nalidixic acid, and cycloheximide (Beumer and Hazeleger, 2003; Gasanov et al., 2005; Janzten et al., 2006; Jeyaletchumi et al., 2010a). The function of acriflavine is to inhibit the growth of other Gram-positive bacteria and it is often used in combination with other selective agents, for instance, polymyxin B-sulfate, cycloheximide, potassium thiocyanate, and nalidixic acid. Nalidixic acid is used for the inhibition of Gram-negative bacteria while cycloheximide is used for the inhibition of fungi (Beumer and Hazeleger, 2003; Janzten et al., 2006; Jeyaletchumi et al., 2010a). Besides, there are other antimicrobials that are usually added into the media such as broad-spectrum ceftazidime and moxalactam as well as lithium chloride (Janzten et al., 2006).

In addition, esculin is an important carbohydrate that is usually incorporated in *Listeria* enrichment and plating media (Bush and Donnelly, 1992; Janzten et al., 2006; Jeyaletchumi et al., 2010a). All *Listeria* sp. are capable of esculin hydrolysis and this process will result in the formation of an intense black color in the media. This is due to the presence of esculin and ferric iron in the media, in which the ferric iron forms complex with 6,7-dihidroxycoumarin, the product of esculin cleavage by β-Dglucosidase, resulting in a black precipitate (Fraser and Sperber, 1988; Janzten et al., 2006; Jeyaletchumi et al., 2010a). Hence, cultures that produce an intense black color indicate the presence of *Listeria*.

The regulatory agencies have recommended several selective enrichment media for *L. monocytogenes* such as buffered *Listeria* enrichment broth (BLEB), Fraser broth, and University of Vermont Medium (UVM) *Listeria* enrichment broth. BLEB is recommended in the US Food and Drug Administration (FDA) bacteriological and analytical method (BAM) for the isolation and identification of *L. monocytogenes*. BLEB is a modification of the formula by Lovett et al. (1987), in which disodium phosphate is added into the medium to increase the buffering capacity of the medium, thus, resulting in the improvement of enrichment properties. Selective agents which include acriflavine, cycloheximide, and nalidixic acid are added into the medium after an initial 4 h of non-selective pre-enrichment (Magalhães et al., 2014).

University of Vermont Medium *Listeria* enrichment broth is recommended in the US Department of Agriculture-Food and Safety (USDA-FSIS) method for the isolation and detection of *L. monocytogenes*. UVM is based on the formulation of Donnelly and Baigent (1986). This enrichment broth is suggested as the primary enrichment broth for the recovery of heatinjured *L. monocytogenes*. Nalidixic acid and acriflavine are the selective agents added into the medium (Magalhães et al., 2014).

International Organization for Standardization (ISO) 11290 method has suggested the use of Fraser broth for the selective enrichment of *L. monocytogenes* in food and environmental samples. Fraser broth is based on the formulation of Fraser and Sperber (1988) whereby it is a modification of USDA secondary enrichment broth by the addition of ferric ammonium citrate and lithium chloride. The selective agents added into Fraser broth are nalidixic acid and acriflavine (Magalhães et al., 2014). Presumptive *L. monocytogenes* can be detected within 48 h by using Fraser broth (Fraser and Sperber, 1988).

The frequently recommended selective differential plating media by FDA-BAM, ISO, and USDA for the isolation of *Listeria* sp. are PALCAM (polymyxin acriflavin lithium-chloride ceftazidime aesculin mannitol) and Oxford (Zunabovic et al., 2011). PALCAM and Oxford are both useful for the isolation of *Listeria* sp. from food samples with injured *Listeria* cells and/or rich in competitive microflora (El Marrakchi et al., 2005). PALCAM agar was initially developed by Van Netten et al. (1989) for the detection and enumeration of *L. monocytogenes* and other *Listeria* sp. from food samples. PALCAM agar consists of Columbia Blood Agar with 23.0 g/L protease peptones, 0.5 g/L glucose, 1.0 g/L starch, 3.0 g/L yeast extract, and 5.0 g/L sodium chloride. The selectivity of this medium is achieved by the addition of 15 g/L lithium chloride, 0.01 g/L polymyxin B, 0.005 g/L acriflavine, and 0.02 g/L ceftazidime. The differentiation on PALCAM agar is based on (1) esculin hydrolysis, addition of 0.8 g/L esculin and 0.5 g/L ferric salt and (2) mannitol fermentation, addition of 10 g/L mannitol and 0.08 g/L phenol red (Van Netten et al., 1989; Magalhães et al., 2014). PALCAM agar plated with bacteria is usually incubated for 24–48 h at 37◦C (Jamali et al., 2013; Ajay Kumar et al., 2014; Osman et al., 2014). Since all *Listeria* sp. can hydrolyze esculin, they are visually confirmed by a blackening of the medium and their colonies are about 2 mm diameter, graygreen in color with a black sunken center and a black halo. Occasionally, *Enterococcus* sp. or *Staphylococcus* sp. may grow on PALCAM agar. However, they can be distinguished from *Listeria* sp. via mannitol fermentation. Mannitol fermentation causes a color change in the colony and/or surrounding medium from gray or red to yellow due to the production of acids. Colonies of these mannitol fermenting organisms are yellow with a yellow halo or gray with a brown-green halo (Van Netten et al., 1989; Ajay Kumar et al., 2014; Osman et al., 2014).

Oxford agar was initially developed by Curtis et al. (1989) for the isolation of *L. monocytogenes* from clinical specimens. Oxford agar has been extensively used in many studies for the isolation and detection of *L. monocytogenes* from various food samples (Pinto et al., 2001; Rudol and Scherer, 2001; Gudbjörnsdóttir et al., 2004; Mena et al., 2004; Alessandria et al., 2010). Oxford consists of Columbia Blood Agar with 23 g/L protease peptones, 5.0 g/L sodium chloride, and 1.0 g/L starch. The selectivity of Oxford is achieved by the addition of 15 g/L lithium chloride, 0.005 g/L acriflavine, 0.02 g/L colistin sulfate, 0.4 g/L cycloheximide, 0.002 g/L cefotetan, and 0.01 g/L fosfomycin. The differentiation of *Listeria* sp. on Oxford agar is based on esculin hydrolysis which is aided by the addition of 1 g/L esculin and 0.5 g/L ferric ammonium citrate into the agar (Curtis et al., 1989; Janzten et al., 2006; Magalhães et al., 2014). Oxford agar is incubated for 24–48 h at 37◦C after plating (Curtis et al., 1989; Alessandria et al., 2010). After 24 h of incubation, *L. monocytogenes* colonies are olive-green with a black halo. After 48 h of incubation, *L. monocytogenes* colonies are about 2–3 mm in diameter, the color turns darker with a black sunken center and surrounded by black zones. Other *Listeria*

sp. colonies have a similar appearance to *L. monocytogenes* colonies. Colonies of other *Listeria* sp. are black with a black halo after 24 h of incubation and they remain the same after 48 h of incubation but with a sunken center (Curtis et al., 1989; Magalhães et al., 2014). *Staphylococcus* sp. may grow on Oxford agar occasionally and their colonies are yellow in color, irregular in size as well as shape (Curtis et al., 1989). A variation of Oxford agar has been developed and it is known as Modified Oxford Agar (MOX). MOX is recommended for the isolation and identification of *L. monocytogenes* from processed meat and poultry products whereas Oxford agar is recommended for the isolation of *L. monocytogenes* from enrichment broth cultures (Magalhães et al., 2014).

The main limitation of PALCAM and Oxford is the inability to distinguish between *L. monocytogenes* from nonpathogenic *Listeria* sp. (El Marrakchi et al., 2005; Zunabovic et al., 2011). Hence, these plating media are not able to provide a rapid detection of *L. monocytogenes* from foods. This has led to the development of chromogenic media which can improve the isolation of *L. monocytogenes* as they are able to differentiate *L. monocytogenes* and/or pathogenic *Listeria* sp. from other non-pathogenic *Listeria* sp. (Beumer and Hazeleger, 2003). Chromogenic media detect essential determinants of pathogenicity of *Listeria* sp. and majority of these media are commercially available as ready-to-use plates (Janzten et al., 2006; Zunabovic et al., 2011). Besides, presumptive *L. monocytogenes* can be identified after 24 h by using chromogenic media (Jeyaletchumi et al., 2010a). Several studies have demonstrated that chromogenic media such as Agar *Listeria* according to Ottaviani and Agosti (ALOA) and CHROMagarTM *Listeria* are more sensitive, specific, time and cost saving in *L. monocytogenes* detection compared to nonchromogenic media such as PALCAM and Oxford (Vlaemynck et al., 2000; Hegde et al., 2007; Jamali et al., 2013).

Agar *Listeria* according to Ottaviani and Agosti is a chromogenic medium developed by Ottaviani et al. (1997) for the isolation of *Listeria* sp. and specific detection of *L. monocytogenes*. The selectivity of ALOA is achieved by the addition of lithium chloride and antimicrobials such as ceftazidime, polymyxin B, nalidixic acid, and cycloheximide (Beumer and Hazeleger, 2003; Magalhães et al., 2014). The differentiation of *Listeria* sp. on ALOA is achieved by the incorporation of chromogenic substrate (5-bromo-4-chloro-3 indolyl-β-D-glucopyranoside, X-glucoside) in the medium for the detection of β-D-glucosidase activity, common to all *Listeria* sp. The differentiation of *L. monocytogenes* from other *Listeria* sp. is based on the production of phosphatidylinositol-specific phospholipase C (PI-PLC) which is encoded by the virulence gene *plc*A present in *L. monocytogenes*. ALOA detects PI-PLC that is present in *L. monocytogenes* and in some strains of *L. ivanovii* through the hydrolysis of L-α-phosphatidylinositol in the medium by PI-PLC. This results in the production of water insoluble fatty acids and the formation of an opaque halo around the colonies. In ALOA, all *Listeria* sp. produce blue-green colonies and pathogenic *Listeria* sp. such as *L. monocytogenes* and *L. ivanovii* produce blue-green colonies with an opaque halo (Ottaviani et al., 1997; Beumer and Hazeleger, 2003; Janzten et al., 2006; Jeyaletchumi et al., 2010a; Zunabovic et al., 2011; Magalhães et al., 2014; Park et al., 2014).

CHROMagarTM *Listeria* (Becton Dickson Diagnostics) is one of the variations of ALOA that have been developed for the isolation and detection of *L. monocytogenes*. On CHROMagarTM *Listeria*, colonies of *L. monocytogenes* are blue with a white halo and colonies of other *Listeria* sp. are blue without halo. Some strains of *L. ivanovii* may grow on CHROMagarTM *Listeria* and they also produce blue colonies with a white halo (Magalhães et al., 2014). Additional variations of ALOA and commercially available media include Biosynth Chromogenic Medium (BCM) *L. monocytogenes* detection system (Biosynth), Compass L. mono (Biokar Diagnostics), BrillianceTM *Listeria* Agar (Oxoid) and chromID Ottaviani Agosti Agar (bioMérieux; Janzten et al., 2006; Zunabovic et al., 2011).

Rapid' *L. mono* agar is a chromogenic medium that operates differently than ALOA. Rapid' *L. mono* agar detects PI-PLC that is present in *L. monocytogenes* and *L. ivanovii* through the hydrolysis of a different substrate by PI-PLC, which is 5-bromo-4-chloro-3-indolyl-myo-inositol-1-phospjhate (X-IP). Cleavage of X-IP by PI-PLC results in the production of blue colonies. Hence, *L. monocytogenes* and *L. ivanovii* appear as blue colonies on Rapid' *L. mono* agar. Furthermore, the addition of xylose into the medium enables the differentiation of *L. monocytogenes* from *L. ivanovii*. The ability of *L. ivanovii* to metabolize xylose results in the production of blue colonies with a yellow halo. As for *L. monocytogenes*, the colonies produced are blue without halo due to the inability to metabolize xylose. Other *Listeria* sp. that grow on Rapid' *L. mono* agar will appear as white colonies with or without a yellow halo (Janzten et al., 2006; Zunabovic et al., 2011; Magalhães et al., 2014).

In general, chromogenic media are able to isolate and distinguish *L. monocytogenes* from other *Listeria* sp. and thus allowing a more rapid detection of *L. monocytogenes*. However, the sensitivity and specificity of the culture media may be affected by the types of food matrices (Andritsos et al., 2013). For instance, the study conducted by Aragon-Alegro et al. (2008) indicated that the sensitivity and specificity of CHROMagarTM *Listeria* for the detection of *L. monocytogenes* in sliced cooked ham (56.2% sensitivity and 73.6% specificity), minced beef meat (92.7% sensitivity and 76.8% specificity) and frankfurters (91.2% sensitivity and 84.2% specificity) were different. Hence, there are no particular medium which is perfect for the isolation of *L. monocytogenes* from various food samples (Churchill et al., 2006; Andritsos et al., 2013).

#### Cultural Detection and Enumeration of *Listeria monocytogenes*

Traditionally, the detection and identification of pathogens in foods involve the use of culture methods followed by phenotypic confirmation based on standard culture (e.g., haemolysis and phospholipase C), biochemical and immunological identification (Gasanov et al., 2005; Janzten et al., 2006). The conventional methods are simple, sensitive, inexpensive, and important when bacterial culture is required as the end result from positive samples (Churchill et al., 2006; Janzten et al., 2006; Law et al., 2015). Generally, the culture methods involve a twostage enrichment process followed by plating on a selective differential agar (Beumer and Hazeleger, 2003; Janzten et al., 2006). The procedures may vary depending on the number of cells expected in a sample and/or the official culture reference methods used. The success of culture methods depends on several factors. For instance, the amount and state of the bacteria in the sample, the selectivity of the media (balance between inhibition of competitors and inhibition of the target bacteria), electivity of the isolation medium (difference between the target bacteria and competitive microflora) and the conditions of incubation (e.g., temperature, time, and oxygen; Beumer and Hazeleger, 2003).

The food samples are homogenized prior to the two-stage enrichment process which is divided into pre-enrichment stage and selective enrichment stage that involve incubation for ∼24– 72 h at 30–37◦C (Churchill et al., 2006; Janzten et al., 2006). Preenrichment is carried out in non- or half-selective enrichment medium in order to revive the injured target pathogen and to increase the amount of the target pathogen. In addition, preenrichment allows the dilution of inhibitory compounds present in foods such as preservatives and rehydration of bacterial cells sampled from dried or processed food matrices (Gasanov et al., 2005; Dwivedi and Jaykus, 2011; Jadhav et al., 2012; Välimaa et al., 2015). As for selective enrichment, it involves the use of selective medium that will increase the amount of target pathogen while suppress the growth of competing background microflora, thus, enabling the isolation and detection of the target pathogen (Dwivedi and Jaykus, 2011; Välimaa et al., 2015). Selective and differential plating is carried out after the two-stage enrichment process. The analysis is completed if there are no typical colonies can be observed on the selective differential agar and the results are reported as negative. If presumptive positive colonies are isolated, further tests are required to confirm the isolated pathogen such is described below (Jeyaletchumi et al., 2010a; Dwivedi and Jaykus, 2011; Välimaa et al., 2015).

The well-known culture reference methods for the isolation and detection of *L. monocytogenes* in foods are the FDA-BAM, ISO 11290 method, and USDA-FSIS method. These methods are recommended for the detection of *L. monocytogenes* from different food matrices and they utilize different enrichment media as well as plating media. Besides, the incubation time and temperature employed by each culture reference method are slightly different (Gasanov et al., 2005; Churchill et al., 2006; Janzten et al., 2006; Jeyaletchumi et al., 2010a; Välimaa et al., 2015). Numerous researchers have employed these culture reference methods for the investigation of *L. monocytogenes* in foods (Jeyaletchumi et al., 2010a; Goh et al., 2012; Lambertz et al., 2012; Jamali et al., 2013; Kramarenko et al., 2013; Wang et al., 2013; Osman et al., 2014). The culture reference methods are summarized in **Table 1**.

The qualitative information of the pathogen is provided by conventional methods. As for the quantitative information of the pathogen, it is required if the pathogen is detected in the food sample. The enumeration of the level of *L. monocytogenes* contamination in food sample can be done according to the ISO 11290-2 method (ISO, 2004b) and the protocols mentioned in FDA-BAM as well as USDA-FSIS method (**Table 1**). Besides,


#### TABLE 1 | Summary of each culture reference method for the isolation and detection of *L. monocytogenes* in foods and the detection limit of each method.

the FDA and USDA have issued Compliance Policy Guides for food industry regarding the appropriate measures required to control *L. monocytogenes* in food and prevent contamination of food with *L. monocytogenes* (Kraiss, 2008). Enumeration of *L. monocytogenes* in food is important because an initial contamination as few as 1 CFU/100 g *L. monocytogenes* can cause the food unsafe in 32 days, while 10 CFU/g *L. monocytogenes* can cause the food unsafe in 8 days (Salvat and Fravalo, 2004). *L. monocytogenes* is able to grow over a wide range of temperatures, from around −0.4 to 45◦C with an optimum temperature of 37◦C (International Commission on Microbiological Specifications for Foods [ICMSF], 1996). Hence, this may cause the prevalence of *L. monocytogenes* in food to increase and reach unsafe levels during storage periods or long holding time before retailing. The infectious dose of *L. monocytogenes* for healthy or susceptible individuals has not been established, however, it is estimated to be <sup>∼</sup>107–109 CFU in healthy individuals and 105–107 CFU in susceptible individuals such as immunocompromised people or pregnant women (Farber et al., 1996; Smith et al., 2003). In order to enumerate the level of food sample contamination by presumptive *L. monocytogenes*, the primary enrichment broth is quantified prior to incubation, by direct spread plate count on chromogenic media (Hitchins and Jinneman, 2013). If the level of contamination is low, the enumeration of *L. monocytogenes* is done by the most probable number (MPN) technique (Janzten et al., 2006; Jeyaletchumi et al., 2010a; Hitchins and Jinneman, 2013). Besides, some samples may contain particulate material that will interfere with plate count enumeration methods. Hence, MPN technique is applied for these types of samples (Sutton, 2010). MPN technique allows the estimation of population density of viable microorganisms in a sample through replicate liquid broth growth in 10-fold dilutions (Sutton, 2010; Letchumanan et al., 2014). The theoretical basis for MPN technique is to dilute the sample to some extent that inocula in the tubes will occasionally contain viable organisms. A reasonably accurate estimation of the most probable number of cells in the sample can thus be achieved by replicates and dilution series (Sutton, 2010). The FDA-BAM has described 10 fold serial dilution of sample in BLEB with the use of three or more tube MPN culture procedure on each dilution. The samples are incubated at 30◦C for 48 h, followed by streaking on selective agar medium (Hitchins and Jinneman, 2013). *L. monocytogenes* can be directly enumerated if chromogenic media is used after MPN enrichment (Janzten et al., 2006; Jeyaletchumi et al., 2010a).

Most probable number technique is more sensitive as compared to direct plating, however, it is more labor intensive and it requires ∼7 days to complete the identification (Janzten et al., 2006; Jeyaletchumi et al., 2010a; Dwivedi and Jaykus, 2011). In MPN technique, the use of selective agar media or chromogenic media may not be selective enough as they may allow the growth of other competitive background microflora, thereby causing difficulties in determining presumptive *L. monocytogenes* (Jeyaletchumi et al., 2010a). MPN technique combined with polymerase chain reaction (PCR) technique is developed in order to overcome these limitations. MPN-PCR technique involves the detection of a particular gene in the target bacteria by PCR instead of isolation of the target bacteria for the enumeration of the bacteria in a sample (Letchumanan et al., 2014). Hence, this technique allows the direct enumeration of *L. monocytogenes* in food without interference of background microflora. The enumeration of *L. monocytogenes* by MPN-PCR technique can be completed in 2 days and this method has higher sensitivity than the standard MPN technique (Jeyaletchumi et al., 2010a). Several researchers have reported the success of MPN-PCR technique for the enumeration of *L. monocytogenes* in various food samples such as fermented sausages (Martín et al., 2004), salad vegetables (Jeyaletchumi et al., 2010b), and raw chicken (Goh et al., 2012).

# MOLECULAR DETECTION OF *Listeria monocytogenes*

The detection of *L. monocytogenes* in food samples by conventional methods is simple, sensitive, and inexpensive if compared with molecular methods (Janzten et al., 2006; Law et al., 2015). However, conventional methods are laborious and time consuming as they require more than a week for the detection and confirmation of pathogen (Dwivedi and Jaykus, 2011; Law et al., 2015; Letchumanan et al., 2015b). Due to the recent advances in molecular technology, molecular methods have been used as an alternative to culture and serological methods for food testing (Gasanov et al., 2005). The detection of a pathogen present in food by nucleic-acid based molecular methods is based on the detection of specific DNA or RNA sequences in the target pathogen. Hence, these genetic methods can provide highly accurate and reliable results as compared to phenotypic methods. Nevertheless, molecular methods require specialized instruments and highly trained personnel (Gasanov et al., 2005; Jadhav et al., 2012; Law et al., 2015). There are various molecular methods available for the detection and identification of *L. monocytogenes*, for instance, PCR, multiplex polymerase chain reaction (mPCR), real-time/quantitative polymerase chain reaction (qPCR), nucleic acid sequence-based amplification (NASBA), loop-mediated isothermal amplification (LAMP), DNA microarray as well as next generation sequencing (NGS) technology.

Polymerase chain reaction has been widely used for the detection of various foodborne bacterial pathogens. This method requires two single-stranded synthetic oligonucleotides or specific primers for the amplification of a specific target DNA sequence in a cyclic three steps process involving the use of a thermal cycler. The PCR amplification products are separated by agarose gel electrophoresis and visualized on the gel as bands with a DNA stain. The specific detection of the genus *Listeria* by PCR involves PCR primers based on the highly conserved 16S rRNA sequence present in all *Listeria* sp. with a resulting 938 bp amplification product (Levin, 2003; Burbano et al., 2006; Goh et al., 2012; Jamali et al., 2013). *L. monocytogenes* can be differentiated from other *Listeria* sp. by exploiting the molecular differences within the PCR amplified 16S rRNA gene, 23S rRNA gene and 16S–23S rRNA intergenic spacer regions (Wang et al., 1992; Graham et al., 1996, 1997; Sallen et al., 1996). In addition, PCR method also detects *L. monocytogenes* at the species level by targeting the virulence genes of the organism (Levin, 2003). Several virulence genes have been identified in *L. monocytogenes* and targeted for the PCR detection of the organism, for example, *hly* (*hlyA*) gene codes for listeriolysin O (LLO; Deneer and Boychuk, 1991; Johnson et al., 1992; Agersborg et al., 1997; Aznar and Alarcón, 2003; Amagliani et al., 2004; Burbano et al., 2006), *iap* gene codes for an invasion-associated protein known as p60 (Agersborg et al., 1997; Aznar and Alarcón, 2003; Swetha et al., 2012), *actA* gene codes for a surface protein known as ActA which is required for intracellular bacterial propulsion and cell to cell invasion (Moriishi et al., 1998; Levin, 2003), *lma*A gene codes for *L. monocytogenes* antigen (lmaA), which also known as *Dth*-18 gene codes for delayed-type hypersensitivity protein (DTH-18 factor; Wernars et al., 1991; Johnson et al., 1992; Levin, 2003), *inlA* gene codes for internalin A (Almeida and Almeida, 2000; Ingianni et al., 2001; Jung et al., 2003), *inlB* gene codes for internalin B (Pangallo et al., 2001; Jung et al., 2003), *prfA* gene codes for positive regulator factor A (PrFA; Simon et al., 1996), *pepC* codes for aminopeptidase C (Winters et al., 1999), *fbp* gene codes for fibronectin-binding protein (Gilot and Content, 2002) and *plcB* Phospholipase C protein (Volokhov et al., 2002). Among these targeted genes, the *hly*A gene is the most frequently chosen target gene for the PCR detection of *L. monocytogenes* (Aznar and Alarcón, 2003; Jadhav et al., 2012). The *hlyA* gene codes for a protein with pore forming activity, which is known as listeriolysin O. This protein is found to be essential for the virulence of *L. monocytogenes* as it is responsible for the lysis of phagocyte vacuole and followed by the escape of *L. monocytogenes* from the vacuole (Kathariou et al., 1987; Cossart et al., 1989; Levin, 2003; Liu, 2006). Besides, it has been discovered that all clinical isolates of *L. monocytogenes* have hemolytic activity due to listeriolysin O and thus the *hlyA* gene is a relevant marker for the identification of *L. monocytogenes* (Groves and Welshimer, 1977; Golsteyn-Thomas et al., 1990).

Other than simple PCR, multiplex PCR (mPCR) is available for a more rapid detection of *L. monocytogenes*. Multiplex PCR is a variant of simple PCR in which multiple gene targets are simultaneously amplified by using several sets of specific primers in a single reaction (Liu et al., 2007). The primer design, concentration of primers, PCR buffer concentrations, quantities of DNA template, Taq DNA polymerase, balance between magnesium chloride and deoxynucleotide concentrations and cycling temperatures are very important for a successful mPCR assay (Markoulatos et al., 2002; Zhao et al., 2014; Law et al., 2015). Multiplex PCR is capable of detecting multiple virulence-associated genes of *L. monocytogenes* in a single PCR mixture. Hence, the possible failure in the detection of virulent *L. monocytogenes* can be prevented (Cooray et al., 1994). In the study conducted by Cooray et al. (1994), *L. monocytogenes* in milk samples was successfully detected by mPCR with primers targeting three virulence-associated genes, *prfA, hlyA,* and *plcB*. Liu et al. (2007) had developed an mPCR assay targeting *inlA*, *inlC,* and *inlJ* genes for the rapid species-specific and virulence-specific determination of *L. monocytogenes*. Besides, mPCR is employed for simultaneous detection of *L. monocytogenes*, *Listeria* sp. and other foodborne pathogens such as *Salmonella* sp., *Escherichia coli* O175:H7, *Vibrio parahaemolyticus*, *Staphylococcus aureus*, *Bacillus cereus* as well as *Campylobacter jejuni* in various food samples (Lawrence and Gilmour, 1994; Gilbert et al., 2003; Jofré et al., 2005; Germini et al., 2009; Kumar et al., 2009; Yuan et al., 2009; Zhang et al., 2009; Zarei et al., 2012; Yang et al., 2013). A novel mPCR assay that can simultaneously detect and discriminate six *Listeria* species including *L. monocytogenes, L. grayi, L. ivanovii, L. innocua, L. welshimeri,* and *L. seeligeri* was first developed by Ryu et al. (2013). A rapid mPCR assay for simultaneous detection of *L. monocytogenes*, *S. aureus*, *E. coli* O157:H7, *Salmonella* Enteritidis, and *Shigella flexneri* in meat samples was developed by Chen et al. (2012). Lee et al. (2014) had carried out mPCR assay that can simultaneously detect *L. monocytogenes*, *B. cereus*, *E. coli* O157:H7, *V. parahaemolyticus*, *Salmonella* sp., and *S. aureus* in ready-to-eat food samples. Furthermore, the major *L. monocytogenes* serovars such as 1/2a, 1/2b, 1/2c, and 4b can be differentiated by mPCR targeting marker genes *Imo0730*, *Imo1118*, ORF2819, and ORF2110 (Doumith et al., 2004; Hamdi et al., 2007; Erol and Ayaz, 2011).

The development of real-time or quantitative PCR (qPCR) provides high-throughput analysis and low risk of crosscontamination since post-PCR processing for the detection of PCR products is not required (Fricker et al., 2007). Fluorescent dye such as SYBR green, hydrolysis probe such as TaqMan assays and oligonucleotide hybridization probes such as molecular beacons are used to monitor the PCR products in qPCR (Law et al., 2015). Recently, qPCR is widely used for the detection of foodborne pathogens and multiplex qPCR is also developed for this purpose. This method offers rapid and specific identification as well as quantification of *L. monocytogenes* in a variety of food samples such as soft cheese, fruit juice, fish, vegetables, salads, milk, meat, and crustaceans (Berrada et al., 2006; O'Grady et al., 2008; Kim and Cho, 2010; Garrido et al., 2013; Gianfranceschi et al., 2014). Oravcová et al. (2005) had developed a realtime 5 -nuclease PCR targeting a sequence of the gene *actA* for the identification and quantification of *L. monocytogenes*. In this study, TaqMan probe was used for the detection and quantification of qPCR products. Besides, Barbau-Piednoir et al. (2013) developed a combination of four qualitative SYBRgreen qPCR assays for the detection and discrimination of *Listeria* sp. and *L. monocytogenes* with high accuracy. In these assays, the *iap* and *prs* genes were targeted for detection of *Listeria* sp. and *hlyA* gene was targeted for detection of *L. monocytogenes*. The successful detection of *L. monocytogenes* in fresh produce using molecular beacon-qPCR targeting the *hlyA* gene was first reported by Liming et al. (2004). Furthermore, a novel 5 exonuclease multiplex qPCR assay for the identification of six *Listeria* sp. including *L. monocytogenes*, *L. seeligeri*, *L. ivanovii*, *L. welshimeri*, *L. grayi,* and *L. innocua* was developed by Hage et al. (2014). In this study, two sets of triplex PCR were designed with one set identifying *L. seeligeri*, *L. welshimeri,* and *L. monocytogenes* and another set identifying *L. ivanovii*, *L. grayi,* and *L. innocua*. The *Listeria* species were differentiated by targeting their respective species-specific target genes and TaqMan probe was used to monitor the multiplex qPCR products.

Additionally, commercial qPCR kits for the detection of *L. monocytogenes* are available and this allows laboratories in food industry to adapt qPCR testing easily (Janzten et al., 2006). The examples of these commercial qPCR kits include BAX<sup>R</sup> System Real-time PCR Assay *Listeria monocytogenes* (DuPont-Qualicon), Probelia<sup>R</sup> *Listeria monocytogenes* PCR System (Bio-Rad), LightCycler<sup>R</sup> *Listeria monocytogenes* Detection Kit (Roche/Biotecon), TaqMan<sup>R</sup> *Listeria monocytogenes* Detection Kit (Applied Biosystems), GeneVision<sup>R</sup> Rapid Pathogen Detection System for *Listeria monocytogenes* (Warnex), ADIAFOOD rapid pathogen detection system for *Listeria monocytogenes* (AES Chemunex), CycleavePCR<sup>R</sup> *Listeria monocytogenes* (*inlA* gene) Detection Kit (TaKaRa Bio, Inc.) and iQ-Check *L. monocytogenes* kit (Bio-Rad Laboratories; Liming et al., 2004; Rodríguez-Lázaro et al., 2004; Becker et al., 2005; Janzten et al., 2006; Liu et al., 2012).

There is no doubt that PCR-based detection methods are rapid, highly sensitive, and specific. However, these methods require thermocycling system. Alternative methods have been developed for the amplification of nucleic acids under isothermal conditions. Two of the most commonly used isothermal nucleic acid amplification methods for the detection of foodborne pathogens are LAMP and NASBA.

Several types of LAMP assays such as multiplex LAMP, realtime LAMP, *in situ* LAMP and reverse-transcription LAMP have been developed and utilized for the detection of foodborne pathogens (Ye et al., 2011; Law et al., 2015). Studies have shown that LAMP assay has high specificity and it exhibits higher sensitivity than PCR assays in the detection of *L. monocytogenes*. For example, Tang et al. (2011) conducted a sensitive and specific LAMP assay for the detection of *L. monocytogenes* with primers that target the *hlyA* gene region. In this study, the LAMP assay was evaluated against conventional PCR method for the detection of *L. monocytogenes* in food. The results indicated that LAMP assay was 100 times more sensitive than the conventional PCR assay. Besides, a real-time quantitative LAMP that amplifies the *hlyA* gene of *L. monocytogenes* was designed by Shan et al. (2012). This LAMP assay was then used to detect *L. monocytogenes* in four different types of retail food samples such as raw meat, vegetables, deli, and seafood. The study also proven that LAMP assay was more sensitive than PCR in the detection of *L. monocytogenes*. A double LAMP (dLAMP) assay was first conducted by Wu et al. (2014) for the detection of *L. monocytogenes* in food samples including pork, beef, chicken, mutton, shrimp, fish, and quick-frozen rice flour products. LAMP primers targeting the *hlyA* and *iap* genes of *L. monocytogenes* were used to ensure the dLAMP assay is more rapid, sensitive and specific. The results of this study showed that dLAMP assay was more sensitive and less time consuming as compared to normal LAMP assay. Recently, LAMP has been commercialized as kits for the detection of *L. monocytogenes*, for instance, Loopamp<sup>R</sup> *Listeria monocytogenes* Detection Kit (Eiken Chemical, Co., Ltd.) and Isothermal Master Mix (OptiGene; Wang et al., 2015).

In general, NASBA often involves in the amplification of mRNA targets under isothermal conditions (Leone et al., 1998). NASBA selectively amplifies the mRNA targets even in the presence of genomic DNA and it has been used to detect various foodborne pathogens (Simpkins et al., 2000). The main advantage of NASBA over other molecular detection methods is its ability to detect viable bacterial cells that are present in environmental samples and food samples (Simpkins et al., 2000; Cook, 2003). A highly specific NASBA system was developed by Blais et al. (1997) for the detection of *L. monocytogenes* with primers targeting the *hlyA* mRNA sequences. The NASBA system was capable of detecting low numbers of *L. monocytogenes* (*<*10 CFU/g) in artificially contaminated dairy and egg products after 48 h enrichment period. Nevertheless, false-positive results were reported and the researchers suggested that the reason for this could be due to cross-contamination of NASBA reactions with amplicons from previous amplifications performed at the same site. The post-NASBA product detection steps involving agarose gel electrophoresis, enzyme-linked gel assay, enzymatic bead-based detection and numerous probing and/or blotting techniques can be laborious. Hence, homogenous real-time NASBA that

TABLE 2 | Application of molecular methods for the detection and identification of *L. monocytogenes* in various food samples.


utilizes fluorescently labeled probes (e.g., molecular beacon) to monitor the amplicons is developed in order to overcome this problem (Leone et al., 1998). A molecular beacon-based real time NASBA assay for the detection of *L. monocytogenes* in cooked ham and smoked salmon slices was first described by Nadal et al. (2007). Sequence from the mRNA transcript of *hly* gene was used as a target for this assay and the detection limit of this assay for *L. monocytogenes* was 400 CFU/mL. This study also involved the use of a commercial NASBA kit, which was NucliSens<sup>R</sup> Basic Kit (bioMérieux) for the detection of *L. monocytogenes*.

DNA microarrays, which were initially being applied for the study of gene expression, could be used for the investigation of microbial evolution and epidemiology as well as for the detection of foodborne pathogens (Gasanov et al., 2005; Severgnini et al., 2011). DNA microarrays comprise multiple specific oligonucleotide probes (with sequence length ranges from 25 to 80 bp) or PCR probes which are coated on to glass slides or chips. The target nucleic acid which can be either DNA, mRNA or cDNA is labeled with fluorescent dye and then applied to the DNA microarray. The target nucleic acid will bind to its corresponding oligonucleotide probe and the hybridization is detected by production of fluorescent signal from probe-sample complex. DNA microarrays are capable of detecting multiple foodborne pathogens simultaneously and

thus suitable for high-throughput analysis (De Boer and López, 2012; Law et al., 2015). An oligonucleotide DNA microarray assay that can simultaneously detect and discriminate six *Listeria* sp. including *L. monocytogenes*, *L. ivanovii*, *L. innocua*, *L. seeligeri*, *L. grayi,* and *L. welshimeri* was performed by Volokhov et al. (2002). The microarray assay was based on *iap*, *hly*, *inlB*, *plcA*, *plcB,* and *clpE* genes for the identification of *Listeria* species. Oligonucleotide DNA microarray was also used for the simultaneous detection of different foodborne bacterial pathogens which include *L. monocytogenes*, *E. coli* O157:H7, *Salmonella enterica* and *C. jejuni* in food samples (Suo et al., 2010). DNA microarrays provide high-throughput analysis but the shortcoming is that large amounts of target DNA or RNA are needed for these methods (Gasanov et al., 2005).

The majority of bacterial genome sequences available today have been generated using the Sanger chain termination sequencing chemistries. Despite being very instrumental in the rise of the field of genomics, it is time consuming as well as resource intensive (Sanger et al., 1977; Medini et al., 2008). The post-Sanger era sequencing technologies, the NGS technologies, have been developed since 2005 to permit extremely rapid highthroughput whole genome sequencing (WGS) hence providing a broader application of comparative genomics (Medini et al., 2008; Shendure and Ji, 2008; Letchumanan et al., 2015a). Due


TABLE 3 | Advantages and limitations of molecular methods for the detection and identification of *L. monocytogenes.*

to rapid decreasing costs for sequencers and reagents, a bacterial genome sequence can be obtained within a few days for less than US\$500 (Didelot et al., 2012), and more than 36,000 bacterial genome sequences are available in public databases (Reddy et al., 2015). Other than serving as a detection tool, WGS is also a feasible tool for retrospective epidemiological analyses and is frequently used for the latter purpose. Genome sequencing of several *L. monocytogenes* strains have revealed serotype- and strain-specific characteristics of *L. monocytogenes* (Orsi et al., 2008; Fretz et al., 2010; Gilmour et al., 2010) and provided novel insights into the genomic causes underlying pathogenicity and survival in food and food processing settings (Buchrieser and Glaser, 2011).

A listeriosis outbreak in Oklahoma, USA in the year 1988 was linked to the consumption of turkey franks contaminated with *L. monocytogenes* produced in a food processing facility in Texas, USA (Centers for Disease Control, and Prevention [CDC], 1989). In 2000, 11 states in the US faced listeriosis outbreak affecting 29 individuals including four fatalities, and it was linked to consumption of deli turkey meat produced in the same facility (Stone and Shoenberger, 2001; Olsen et al., 2005). Using NGS, Orsi et al. (2008) revealed that the human listeriosis outbreak in 2000 in the USA was caused by a *L. monocytogenes* strain that persisted in that food processing plant for over 12 years in which the same strain has also been responsible for a sporadic case in 1988.

In 2008, *L. monocytogenes* serotype 1/2a caused an outbreak of listeriosis associated with ready to eat meat products in Canada, resulting in 22 deaths and at least 57 illnesses (Gilmour et al., 2010). The authors reported the first real-time application of WGS during an active listeriosis outbreak investigation using the high-throughput to characterize two outbreak-associated isolates of *L. monocytogenes*. In 2009 and 2010, another large listeriosis outbreak occurred in Austria, Germany, and the Czech Republic due to intake of a traditional Austrian cheese called "Quargel," an acid curd cheese with a red smear made from skimmed pasteurized milk (Fretz et al., 2010). Molecular typing via PFGE revealed that two different *L. monocytogenes* strains, both serotype 1/2a (Pichler et al., 2011). From June 2009 to January 2010 Quargel outbreak clone 1 (QOC1) was the culprit in 14 cases, including five with a lethal outcome (Fretz et al., 2010). Whereas between December 2009 and February 2010, Quargel outbreak clone 2 (QOC2) accounted for 20 cases, which resulted in three deaths (Fretz et al., 2010). Rychli et al. (2014) sequenced and analyzed the genomes of both outbreak strains in order to retrospectively investigate the extent of genetic diversity between the two strains. WGS analysis revealed that these two strains have distinct *in vitro* virulence potential despite originating from similar serovar (Rychli et al., 2014).

The development of benchtop sequencers using NGS technology such as 454 or GS FLXTM (Roche), MiSeq (Illumina) and Ion Torrent Personal Genome Sequencer (PGM; Life Technologies) will enable bacterial WGS even in small research and clinical laboratories (Didelot et al., 2012). WGS has already been actively used for the characterization of bacterial isolates in several large outbreaks in the world (Gilmour et al., 2010; Reuter et al., 2013) and also being used as a tool for retrospective epidemiological analyses (Orsi et al., 2008; Rychli et al., 2014). In the near future this technology is likely to substitute currently used typing methodologies due to its ultimate resolution and sensitivity (Sabat et al., 2013). However, NGS also has its limitation. Until now, it is still too laborious and time-consuming to obtain useful data for routine surveillance (Fournier et al., 2014). Library preparation and sequencing protocol requires adept and skillful technician; however, this limitations are likely to be overcome due to higher level of automation which will lead to a more streamline processed. Bioinformatics analysis and interpretation, as well as computational hardware are also another challenges to be solved especially by small laboratories (Fournier et al., 2014). In addition, a fundamentally unsolved question is how the sequences should be examined for epidemiological characterization (Sabat et al., 2013). More examples of studies that involved the application of molecular methods for the detection and identification of *L. monocytogenes* in food samples are listed in **Table 2**.

There are many advantages of using molecular methods for the detection and identification of *L. monocytogenes*. For instance, the main advantages of molecular detection methods are due to their high sensitivity and specificity. Nonetheless, limitations can be found in these methods such as some molecular methods can be costly, complex and require trained personnel. The advantages and limitations of molecular methods are summarized and listed in **Table 3**.

#### CONCLUSION

Early detection of *L. monocytogenes* contaminated food is crucial as it can prevent the outbreaks of foodborne illness. Till date, the culture reference methods mentioned in this review are still applicable and being used in many studies. Simple PCR and mPCR have been used routinely for rapid, sensitive, and specific screening as well as confirmation of *L. monocytogenes*. The introduction of LAMP and NASBA allow accurate and cost-effective screening of *L. monocytogenes*. For large number of samples, high-throughput assays such as qPCR and DNA microarray are often used for the detection of *L. monocytogenes*. Cultural and molecular techniques are continuously being developed and improved in order to provide higher sensitivity and specificity of *L. monocytogenes* detection. The advancements of molecular methods allow the rapid detection of *L. monocytogenes* in food samples with high sensitivity and specificity whilst substituting the lengthy and laborious conventional detection methods. Molecular methods have provide many advantages, nonetheless, there are still some limitations in these methods such as the need to use highly advance technology that are costly compared to conventional methods. The combined use of two or more detection methods is also possible and may improve the accuracy of detecting *L. monocytogenes*. Still, there are great potential for the development and application of new techniques for foodborne pathogens detection and analysis.

#### ACKNOWLEDGMENTS

This work was supported by a University of Malaya for High Impact Research Grant (UM-MOHE HIR Nature Microbiome

#### REFERENCES


Grant No. H-50001-A000027 and No. A000001-50001) awarded to CK-G and External Industry Grants from Biotek Abadi Sdn Bhd (vote no. GBA-808138 and GBA-808813) awarded to LL-H.


oligonucleotide primers targeting the genes encoding internalin AB. *J. Food Prot.* 66, 237–241.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2015 Law, Ab Mutalib, Chan and Lee. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Prevalence, enumeration, and pheno- and genotypic characteristics of *Listeria monocytogenes* isolated from raw foods in South China

#### *Edited by:*

*Giovanna Suzzi, Università degli Studi di Teramo, Italy*

#### *Reviewed by:*

*Alexandra Lianou, Agricultural University of Athens, Greece Graziella Bourdin, Agence Nationale de Sécurité Sanitaire de l'Alimentation, de l'Environnement et du Travail, France*

#### *\*Correspondence:*

*Qingping Wu, Guangdong Institute of Microbiology, State Key Laboratory of Applied Microbiology Southern China, Guangdong Provincial Key Laboratory of Microbial Culture Collection and Application, Guangdong Open Laboratory of Applied Microbiology, Xianlie Zhong Road 100 #, 58th Building, Guangzhou 510070, China wuqp203@163.com*

#### *Specialty section:*

*This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology*

*Received: 06 July 2015 Accepted: 09 September 2015 Published: 29 September 2015*

#### *Citation:*

*Chen M, Wu Q, Zhang J, Wu S and Guo W (2015) Prevalence, enumeration, and phenoand genotypic characteristics of Listeria monocytogenes isolated from raw foods in South China. Front. Microbiol. 6:1026. doi: 10.3389/fmicb.2015.01026*

#### *Moutong Chen1,2, Qingping Wu1\*, Jumei Zhang1, Shi Wu1,2 and Weipeng Guo1*

*<sup>1</sup> Guangdong Institute of Microbiology, State Key Laboratory of Applied Microbiology Southern China, Guangdong Provincial Key Laboratory of Microbial Culture Collection and Application, Guangdong Open Laboratory of Applied Microbiology, Guangzhou, China, <sup>2</sup> School of Bioscience and Bioengineering, South China University of Technology, Guangzhou, China*

*Listeria monocytogenes* is an important foodborne pathogen that can cause serious illness in immunocompromised individuals, pregnant women, the elderly, and newborns. The aim of this study was to: (i) evaluate the prevalence and contamination level [most probable number (MPN)] of *L. monocytogenes* in 567 retail raw foods (fishery products, *n* = 154; raw/fresh meat, *n* = 123; frozen foods, *n* = 110; edible fungi, *n* = 108; vegetables, *n* = 72) collected from South China and (ii) to gain further knowledge on the phenotype and genotype distributions of this important foodborne pathogen. Approximately 22% of the samples were positive for *L. monocytogenes*. The contamination levels were between 0.3 and 10 MPN/g in 75.0%, between 10 and 100 MPN/g in 11.0% and less than 100 MPN/g in 14.0% of the countable samples. Five serogroups were identified among the 177 foodborne *L. monocytogenes* isolates, with 1/2a-3a (42.4%) and 1/2b-3b (26.0%) serogroups being the most dominant. Serogroups I.1 and II.2 were only found in the edible mushrooms, while serogroup III was dominant in the fishery products, suggesting that specific serogroups of *L. monocytogenes* may have distinct ecological niches. Ten (5.6%) *L. monocytogenes* isolates exhibited multidrug resistance. Genetic relatedness analysis revealed the absence of distinct associations between specific food types, antibiotic resistance, serogroups, and genetic diversity. The present study provided the first baseline data on the prevalence, contamination level, and characteristics of *L. monocytogenes* isolated from raw foods in South China. Some multidrug resistant strains belonged to the epidemiologically important serogroups (I.1 and II.1), implying a potential public health risk. In addition, these findings also provide basic information for the Chinese food safety associated authorities to draft appropriate standards to control *L. monocytogenes* contamination and improve microbiological safety of raw foods.

Keywords: *Listeria monocytogenes*, most probable number, enterobacterial repetitive intergenic consensus PCR, random amplified polymorphic DNA, antimicrobial susceptibility profile

# Introduction

*Listeria monocytogenes*, a facultative intracellular foodborne pathogen, is capable of causing serious disease in humans, especially in immunocompromised individuals, pregnant women, the elderly, and newborns. It has the ability to inhabit a wide range of environments and is commonly found in food, nature, and the food processing environment (Abadias et al., 2008). Listeriosis is a severe invasive disease and the manifestations include neurological infections like encephalitis, meningitis, septicemia, and abortion, with a mortality rate of up to 20–30% (Vazquez-Boland et al., 2001; Kang et al., 2013).

Foodborne outbreaks of listeriosis have been documented in Asia, Europe, and the USA (Makino et al., 2005; de Castro et al., 2012; Lomonaco et al., 2013). In recent years, several sporadic cases of listeriosis have been reported in China (Wu et al., 2008; Hsieh et al., 2009; Feng et al., 2011). *L. monocytogenes* isolates associated with outbreaks or sporadic cases of listeriosis have been detected in different kinds of foods including dairy products, raw meat, vegetables, and fishery products (Schlech, 2000; Bell and Kyriakides, 2005; Yücel et al., 2005). Identifying the source of infection is essential to undertake preventive measures and control the exposure to such infection (Franciosa et al., 1998). Thus, determining the prevalence and contamination level of *L. monocytogenes* in retail foods are of utmost importance to control and track the definite source of *L. monocytogenes*. In China, previous studies have focused on the prevalence of *L. monocytogenes* in retail foods. However, these studies only focused their investigation to limited regions, such as Heilongjiang Province, Guangzhou City, and Gansu Province (Hu et al., 2013; Shi et al., 2013; Wang et al., 2013). In this context, it is notable that South China is located in a subtropical region where the climate is suitable for *L. monocytogenes* growth. Several studies have focused on the occurrence of *L. monocytogenes* in ready-to-eat foods (Huang et al., 2005; Mei et al., 2006; Lianou and Sofos, 2007; Jin et al., 2009; Sauders et al., 2009). Additionally, numerous previous studies have reported the contamination level of *L. monocytogenes* in ready-to-eat products in other countries (Lianou and Sofos, 2007). Wang et al. (2013) reported occurrence and counts of *L. monocytogenes* in retail raw foods in Heilongjiang Province (Northeast of China) from 2008 to 2009. However, there are limited studies that focused on the contamination level of *L. monocytogenes* in retail raw foods in South China, especially the contamination level in edible fungi and vegetables, thus, hampering the potential risk analysis of *L. monocytogenes* in retail raw foods. It is necessary to monitor the occurrence of *L. monocytogenes* in retail raw foods because of possible chances of cross-contamination during food processing and food storage.

Since the first multiresistant *L. monocytogenes* strain was isolated from a patient with meningoencephalitis in France in 1988 (Poyart-Salmeron et al., 1990), antimicrobial resistant strains have been commonly recovered from food, natural environment, and clinical cases of listeriosis (Rodas-Suárez et al., 2006; Zhang et al., 2007; Conter et al., 2009; Morvan et al., 2010; Granier et al., 2011).The resistance profiles are varied and may be influence by overuse of antimicrobials in humans and livestock animals, as well as by geographical differences. In China, the rates of antibiotic resistance in 467 foodborne *L. monocytogenes* isolates were 4.5% in 2005, and the antimicrobial resistance was most frequently observed for ciprofloxacin and tetracycline (Yang et al., 2008); similar results were also reported by Yan et al. (2010). Zhao et al. (2012) reported that the rates of antibiotic resistance in 1069 foodborne *L. monocytogenes* isolates were 6.92% in China. The resistance was most prevalently observed for antibiotics such as tetracycline, doxycycline, erythromycin, chloramphenicol, and ciprofloxacin. Based on this trend of antibiotic resistance, it is advisable to monitor the patterns of antibiotic resistance of *L. monocytogenes* in food sources from different regions in China.

*Listeria monocytogenes* has 13 serotypes that can be divided into five serogroups, such as I.1 (1/2a-3a), I.2 (1/2c-3c), II.1 (4b-4d-4e), II.2 (1/2b-3b-7), and III (4a-4c; Doumith et al., 2004). As 95% of the strains responsible for human listeriosis and found in food samples belong to serotype 1/2a, 1/2b, 1/2c, and 4b (Pontello et al., 2012; Althaus et al., 2014), serotype analysis is not very efficient method to differentiate *L. monocytogenes* strains isolated from foods and clinical samples. Recently, numerous molecular subtyping techniques have been developed for the surveillance or tracing the sources of *L. monocytogenes*; these include pulsed field gel electrophoresis (PFGE; Gerner-Smidt et al., 2006), amplified fragment length polymorphism (AFLP; Parisi et al., 2010), random amplified polymorphic DNA (RAPD; Farber and Addison, 1994), multi-locus sequence typing (MLST; Parisi et al., 2010), and enterobacterial repetitive intergenic consensus polymerase chain reaction (ERIC-PCR; Chen et al., 2010). Of these techniques, ERIC-PCR is a relatively simple, highly reliable, and cost-effective method that has been demonstrated to generate distinct DNA fingerprints within a single bacterial species (Jersek et al., 1999).

The objective of the present study was to: (i) determine the prevalence and contamination level of *L. monocytogenes* in retail raw foods in South China and (ii) to determine the genetic variation and phenotypic characteristics of *L. monocytogenes* isolates.

#### Materials and Methods

#### Sampling Procedure

From July 2011 to August 2012, a total of 567 retail raw food samples were collected from rural markets, open-air markets, and large supermarkets in South China, including five districts of Guangzhou City and 11 cities in Guangdong, Fujian, Guangxi, and Hainan Province. Samples comprised of fishery products (*n* = 154, including prawn, Grass carp, fresh squid, speckled fish, yellow croaker, catfish, oyster, *Tilapia mossambica*), raw/fresh meat (*n* = 123, including fresh pork, fresh beef, chicken, duck, Minced pork, fresh pork ribs, fresh pork ball, mutton), frozen foods (*n* = 110, including Chinese dumpling, frozen mutton, frozen chicken wing, frozen han sausage, frozen drumstick, wonton, frozen salted pork), fresh edible fungi (*n* = 108, including *Flammulina velutipes*, *Pleurotus eryngii*, *Hypsizygus marmoreus*, *Coriandrum sativum*, *Lentinus edodes*, oyster mushroom, *Pleurotus geesteranus*), and vegetables (*n* = 72, including Lettuce, cucumber, coriander). Samples were stored in insulated shipping coolers containing frozen gel packs placed on the sides, middle, and the tops of the samples. All the samples were kept below 4◦C during transportation and test was initiated within 4 h after receipt.

#### Qualitative and Quantitative Analysis for Food Samples

An enrichment method according to the National Food Safety Standard of China – Food microbiological examination: *L. monocytogenes* (GB 4789.30-2010; National Standard of the People's Republic of China, 2010) with minor modifications was used for qualitative detection. In brief, samples were analyzed for the presence of *L. monocytogenes* by homogenizing 25 g samples in 225 mL *Listeria* enrichment broth I (LB1; Huankai, Co. Ltd., China). Homogenates were incubated at 30◦C for 24 h. Thereafter, 0.1 mL LB1 enrichment culture was transferred to 10 mL *Listeria* enrichment broth II (LB2) at 30◦C for 24 h. A loopful of the LB2 enrichment culture was streaked onto Chromagar<sup>R</sup> *Listeria* selective agar plates (CHROM-agar, Paris, France) and incubated at 37◦C for 48 h. Three to five presumptive colonies that were typically blue in color with a white halo were selected for identification of *L. monocytogenes* by Microgen ID *Listeria* identification system (Microgen, Camberley, UK) according to the manufacturer's instructions.

To determine the most probable number (MPN), the method was adapted from a previous study conducted by Gombas et al. (2003). Briefly, a nine-tube MPN method was used. The nine tubes were divided into three sets of three tubes each. The second and third sets of tubes contained 10 ml of Fraser broth medium. Three aliquots (10, 1, and 0.1 mL) of the sample homogenate were dispensed into three sets, representing 1.0, 0.1, and 0.01 g of the original sample, respectively. The tubes were incubated at 30 ± 2◦C for 24 ± 2 h, and subsequently 0.1 mL of content from each tube was transferred to a new tube containing 10 ml of fresh Fraser broth. The tubes were incubated at 30 ± 2◦C for 26 ± 2 h. Darkened Fraser tubes were subjected to confirmation. If a Fraser broth failed to darken, it was examined again after an additional 26 ± 2 h of incubation. The MPN values were determined based on the number of positive tube(s) in each of the three sets and the MPN table (U. S. Department of Agriculture, 1998; Hitchins, 1998).

#### Serogroup Analysis using Multiplex PCR

Genomic DNA was extracted from *L. monocytogenes* using Bacterial Genomic DNA Purification Kit (Dongsheng Biotech. Inc., Guangzhou, China) according to the manufacturer's instruction. DNA concentration was determined at O.D. 260 nm using Nano Drop<sup>R</sup> ND-1000UVeVis Spectrophotometer (Thermo Fisher Scientific, Inc., Waltham, MA, USA). Serogroup analysis of 177 isolates was performed using a multiplex PCR as previously described by Doumith et al. (2004),13 serotypes of *L. monocytogenes* were categorized into five distinct phylogenetic groups, viz. I.1 (1/2a-3a), I.2 (1/2c-3c), II.1 (4b-4d-4e), II.2 (1/2b-3b-7), and III (4a-4c). The PCR mixture (50-μL) contained 1.5 unit GoTaq<sup>R</sup> Hotstart polymerase (Promega, Madison, WI, USA), 1 μM for *lmo0737*, *ORF2819*, and *ORF2110*; 1.5 μM for *lmo1118*; and 0.2 μM for *prs*, 2.5 mM MgCl2, 0.2 mM each dNTP, and 40 ng of template genomic DNA. PCR was performed with the following thermal cycle: initial denaturation step at 94◦C for 3 min; 35 cycles of 94◦C for 35 s, 53◦C for 50 s, and 72◦C for 60 s; and a final cycle of 72◦C for 7 min in a thermocycler (Applied Biosystems, Foster City, CA, USA). Five microliters of the reaction mixture was mixed with 5 μL of loading buffer and separated on a 2% agarose gel in TAE buffer. The PCR product was visualized by Goldview<sup>R</sup> staining (0.005%, v/v).The primers are shown in Supplementary Table S1.

#### Enterobacterial Repetitive Intergenic Consensus Sequence Polymerase Chain Reaction (ERIC-PCR)

Enterobacterial repetitive intergenic consensus sequence polymerase chain reaction (ERIC-PCR) typing was carried out on the *L. monocytogenes* isolates and five reference strains using the protocol described by Chen et al. (2014). The ERIC primers was described by Versalovic et al. (1991), ERIC1R: 5 -ATGTAAGCTCCTGGGGATTCAC-3 , and ERIC2: 5 - AAGTAAGTGACTGGGGTGAGCG-3 . The PCR mixture (25-μL) contained one unit GoTaq<sup>R</sup> Hotstart polymerase (Promega, Madison, WI, USA), 0.6 μM of each primer, 2.5 mM MgCl2, 0.2 mM each dNTP, and 40 ng of template genomic DNA. Amplifications were performed in a DNA thermocycler (Applied Biosystems, Foster City, CA, USA) with the following temperature profile: an initial denaturation at 94◦C for 3 min; 35 cycles each consisting of 30 s at 94◦C, 30 s at 46◦C, 30 s at 49◦C, and 3 min at 72◦C; and a final extension at 72◦C for 10 min. The ERIC-PCR products were separated by electrophoresis on a 1.5% agarose gel with Goldview<sup>R</sup> staining (0.005%, v/v), and photographed using an UV Imaging System (GE Healthcare, Waukesha, WI, USA). The images were captured in TIFF file format for further analysis.

#### Cluster Analysis

The observed bands in the gels were evaluated based on the presence (coded 1) or absence (coded 0) of polymorphic fragments for the ERIC and RAPD primers. Analysis of TIFF images was carried out using Gel Pro Analyzer (Version 4.0) and NTSYS-pc (Version 2.10), a numerical taxonomy and multivariate analysis software package (Rohlf, 2000). Similarity between fingerprints was determined by the Dice's similarity coefficient at 1% band position tolerance and dendrograms were generated by unweighted pair group method using arithmetic average (UPGMA).The Simpson's indexes of discrimination (DI) of ERIC-PCR and RAPD were calculated as described by Hunter and Gaston to determine the ability of each typing method (Hunter and Gaston, 1988).

#### Random Amplified Polymorphic DNA

According to a previous study conducted by Chen et al. (2014), the 10-mer primer UBC155 (5 -CTGGCGGCTG-3 ) was chosen for typing the *L. monocytogenes* isolates in this study (Farber and Addison, 1994). The PCR reaction condition used for the selected *L. monocytogenes* isolates and five reference strains was similar to the protocol described by Chen et al. (2014). The reaction


#### TABLE 1 | Antimicrobial resistance of foodborne *L. monocytogenes* isolates collected in South China.

∗*Breakpoint for Entercococcus spp.*

mixtures were placed in a hot-lid cycler (Applied Biosystems, Foster City, CA, USA) and subjected to the following temperature profile: an initial five cycles at 94◦C for 5 min, 35◦C for 5 min, 72◦C for 5 min; then 30 cycles each consisting of 1 min at 94◦C, 2 min at 35◦C, and 2 min at 72◦C; and a final extension at 72◦C for 10 min. The amplicons were electrophoresed on 1.5% agarose gel and photographed that were saved as TIFF file format for genotype analysis.

#### Strain Library Construction

Strain library construction was performed as described in a previous study with minor modifications (Casarez et al., 2007). In brief, results of ERIC-PCR fingerprinting and multiplex PCRbased serogroups were used to screen isolates from the same sample in order to identify clonal isolates and assure as diverse a known source library as possible. Three to five *L. monocytogenes* isolates from each positive sample were fingerprinted by



ERIC-PCR and serogroup analysis using multiplex PCR, and subsequently compared with each other. Isolates from the same sample with *>*90% similarity were considered as clonal. If the ERIC-PCR fingerprint of one isolate with *>*90% similarity from the same sample but has different serogroup determined by PCR, the isolate was included in the strain library. Only the isolate with *>*90% similarity ERIC-PCR fingerprint and the same serogroup from the same sample source were considered clonal. Clonal isolates from individual sample were excluded. At least one *L. monocytogenes* isolate from each known source sample was included in the library.

#### Antimicrobial (AM) Susceptibility Test

Since no resistance criteria exist for *Listeria* antibacterial susceptibility test in Clinical and Laboratory Standards Institute guidelines for the tested AMs, criteria for *Staphylococcus aureus* were used except for where noted (**Table 1**; Clinical and Laboratory Standards Institute, 2014). A panel of 15 antimicrobials at the specific concentration per disk (Oxoid, Boston, MA, USA) was tested in this study (**Table 1**).


TABLE 3 | Results of serogroup analysis carried out for the foodborne *L. monocytogenes* isolates collected in South China#.

*#The 177 L. monocytogenes isolates included in the strain library were used for statistical analysis.*

Antimicrobial susceptibility test was performed as described in a previous study (Kovacevic et al., 2012). *S. aureus* ATCC 25923 and *Escherichia coli* ATCC 25922 were used as quality control strains for this study. Zones of inhibition were measured with a precision caliper to the nearest 0.01 mm. Isolates exhibiting resistance to three or more classes of antibiotics were considered as multidrug-resistant strains (Magiorakos et al., 2012).

#### Results

#### Prevalence and Quantitative Analysis

In this study, 123 (21.7%) samples were positive for *L. monocytogenes* out of 567 collected samples. Depending on the food category, the prevalence of *L. monocytogenes* in frozen food samples was 50.9% (56/110) and 31.5% in fresh edible fungus (34/108, including 29 *F. velutipes* samples and five other edible fungus samples), 15.4% in fresh meat (19/123), 7.8% in fishery products (12/154), and 2.8% in vegetables (2/72; **Figure 1**). Based on quantitative analysis, 75.0% (75/100) of the positive samples were contaminated at levels ranging between 0.3 and 10 MPN/g, that included 45 frozen foods, 12 fresh meat, 12 edible fungi, five fishery products, and one vegetable sample. Only in 14.0% of the positive samples the contamination level exceeded 100 MPN/g, which included 12 edible fungi samples, one frozen food, and one fishery product. The contamination level of *L. monocytogenes* in frozen food samples was low; in 95.8% (45/47) of the positive samples the levels were less than 10 MPN/g, only in one sample the levels exceeded 100 MPN/g. Surprisingly, 38.7% (12/31) of edible fungus samples were more than 100 MPN/g (**Table 2**).

#### Serogroup Analysis using Multiplex PCR

One hundred and seventy-seven *L. monocytogenes* isolates included in the strain library and five reference strains were used for serogroup analysis using multiplex PCR (Supplementary Table S2). As shown in **Table 3**, 42.4% (75/177) of the 177 *L. monocytogenes* isolates were recognized as serogroup I.1, of which isolates mainly isolated from fresh meat (13 isolates), frozen products (28 isolates), and edible fungi samples (31 isolates); 18.1% (32/177) as serogroup I.2, of which 28 isolates were recovered from frozen products samples; 7.3% as serogroup II.1, 26.0% (46/177) as serogroup II.2, which comprised of 20 isolates from edible fungi samples and 21 isolates from frozen products samples; and 6.2% as serogroup III. Additionally, serogroup I.1 (60.8%) and II.2 (39.2%) were identified in *L. monocytogenes* strains that were isolated from fresh edible fungi samples. Serogroup III (53.3%) was predominant in fishery product samples.

#### ERIC-PCR Typing

Using ERIC-PCR, genomic DNA of *L. monocytogenes* isolates displayed DNA bands in sizes of between 200 and 2200 bp, most of isolates from the same sample generated similar fingerprints. Sixty ERIC-types (DI = 0.960) were classified among 177 foodborne isolates. Based on cluster analysis of genetic profiles obtained from ERIC-PCR typing (**Figure 2**), there were three singletons and eight clusters at a relative similarity coefficient of 80%; isolates belonging to a distinct serogroup clustered together, such as in cluster D that included the strains 283-4, 298-1, 302-4, 547-4, and 648-1. It is noteworthy that 23 isolates, from *F. velutipes* samples, belonging to serogroup I.1 fell into cluster D, which included strains 298-1, 318-1, 518-1, 548-3, 618-1, 633- 2, 733-1, 283-4, 648-2, 668-4, 398-1, 483-1, 418-4, 718-1, 798-1, 283-6, 283-7, 418-1, 243-2, 648-1, 698-4, 698-1, and 748-1. For fishery products, four isolates (556-1, 557-1, 588-1, and 589-1) shared the same ERIC-type under cluster F; three isolates (55-1, 85-1, and 189-1) shared similar genetic fingerprints as in cluster K, and all these seven isolates belonged to serogroup III.

#### RAPD Genotyping Analysis

Using UBC-155 primer, it was determined that at a relative similarity coefficient of 80%, 177 isolates and five reference strains fell into 14 clusters and five singletons (designated as I to XIX; **Figure 3**). The discriminatory index of RAPD typing for the 177 isolates was 0.972 based on Simpson's Index of diversity (Hunter and Gaston, 1988). The strong correlation between RAPD-type and serogroup was observed using RAPD typing (**Figure 3**), cluster I included serogroup II.2 and II.1, which accounted for 78.3% (36/46) and 17.4% (8/46), respectively. Most of the isolates in cluster VIII belonged to serogroup III (87.5%); serogroup I.2 (27/36) dominated in cluster IX and serogroup I.1 dominated in cluster X (77.8%), XI (93.1%), and XII (100%). As shown in **Figure 2**, when comparing the isolates included in the clusters of ERIC-PCR and RAPD, an agreement was observed with ERIC-PCR and RAPD subtyping, i.e., isolates in cluster XI and IX typed using RAPD belonged to cluster I of ERIC-PCR).

#### FIGURE 2 | Continued

The dendrogram was constructed based on ERIC-PCR analysis; a: T (t), tetracycline; P (p), penicillin; CIP (cip), ciprofloxacin; K (k), kanamycin, A (a), ampicillin; C (c), cephalothin; CN (cn), gentamycin, E (e), erythromycin, D (d), doxycycline; SXT (sxt), sulfamethoxazole with trimethoprim; V (v): vancomycin; R (r), rifampin; L (l), levofloxacin; /, no resistance; ND, not detected; Name of antibiotics with capital letters implies resistance; Name of antibiotics with lowercase letters implies intermediate resistance. b: -, no gene absent; genes listed mean absent. c: CMCC, China Medical Culture Collection; ATCC, American Type Culture Collection. d: \, unknown.

#### Antimicrobial Susceptibility Test

Antimicrobial susceptibility of 177 *L. monocytogenes* isolates were evaluated using the disk diffusion method. Based on the breakpoint criteria for *S. aureus* or *Enterococcus* spp., 42 antimicrobial susceptibility profiles were identified in this study (**Figure 2**). Penicillin (23.1%), ampicillin (20.9%), and ciprofloxacin (24.3%) were the three most frequent antimicrobial resistance profiles among the *L. monocytogenes* isolates, while the resistance rate of tetracycline and doxycycline was 7.3 and 3.4%, respectively. For the other 12 antibacterial agents, all 177 isolates of *L. monocytogenes* displayed 89.3– 100% susceptibility (**Table 1**). Considering the fact that an intermediate resistant isolate may become a resistant strain under certain circumstances (Ruiz-Bolivar et al., 2011), all 177 isolates were grouped into 42 antibacterial resistance patterns, including 106 (59.9%) isolates of *L. monocytogenes* that were susceptible to the 15 tested antibacterial agents and 10 (5.6%) isolates that were multidrug resistant (**Figure 2**). Notably, 83.3% of the isolates (10/12) that were resistant to up to four antimicrobials were isolated from the livestock and poultry meat and were associated food products (**Figure 2**).

# Discussion

In this study, we collected 567 food samples including raw meat, frozen foods, edible fungi, fishery products, and vegetables from16 cities or districts in South China. The contamination rate of *L. monocytogenes* (21.7%) in food samples from South China was in consistent with that in Southeastern China as shown by Wang et al. (2013). Over 15% of fresh meat samples were positive for *L. monocytogenes* in the present study, however, raw meat and poultry are not considered at high risk for causing foodborne listeriosis provided adequate cooking precedes consumption and that cross-contamination is avoided. Nevertheless, raw or insufficiently cooked meats may serve as sources of crosscontamination for products that are intended to be consumed without heat treatment, and along with inadequate cleaning and sanitation, these foods have been recognized as the main sources of post-processing contamination of RTE meat products (Lianou and Sofos, 2007; Sauders et al., 2009). Quantitative analysis revealed that over 75% of *L. monocytogenes*-positive samples had counts below 10 MPN/g, which was consistent with the results of the retail raw foods collected in Northeast of China (Wang et al., 2013). In the present study, the contamination rate of frozen foods was up to 50.9%, while the MPN value of positive samples were mostly below 10 MPN/g, these results may associated with the tolerance ability of *L. monocytogenes* in low temperature environments (Vazquez-Boland et al., 2001). Surprisingly, 76.3% (29/38) of *F. velutipes* samples were positive for *L. monocytogenes* and the MPN value of 12 positive samples exceeded 100 MPN/g, while the contamination rates of other edible fungi samples were low (data not shown), these data were significantly higher than that reported in other countries as reviewed by Lianou and Sofos (2007). To date, limited data have been available about the prevalence of *L. monocytogenes* in edible fungi products in China. In addition, there are no specific qualitative and quantitative standards to determine the presence of *L. monocytogenes* in the mushroom products. However, some countries have formulated a zero-tolerance policy for *L. monocytogenes* in mushroom products, such as in the United States and Canada (Canadian Food Inspection Agency, 2012; U. S. Food and Drug Administration, 2012). It is necessary to draft a corresponding microbiological standard to ensure the quality of mushroom products in China.

Previous studies reported over 95% of the isolates in human listeriosis and food samples belonged to serotype 1/2a, 1/2b, 1/2c, and 4b (Pontello et al., 2012; Althaus et al., 2014). In this study, 42.4% of *L. monocytogenes* isolates belonged to serogroup I.1 (serotype 1/2a-3a), followed by serogroup II.2 (26.0%), serogroup I.2 (18.1%), serogroup II.1 (7.3%), and serogroup III (6.2%). Serogroup I.1 (serotype 1/2a-3a) predominant among the foodborne isolates, the serogroup compositions of these foodborne isolates was in agreement with that previously reported for other foodborne isolates (Chen et al., 2009; Korsak et al., 2012). Interestingly, isolates recovered from edible fungi samples only belonged to serogroup I.1 and II.2, while Viswanath et al. (2013) reported only serotype 4a contamination in a small-scale mushroom production facility. Serogroup III was rarely isolated from foods (Orsi et al., 2011), however, in this study 11 isolates belonging to serogroup III were isolated from food samples; 72.7% (8/11) isolates were from fishery products, Korsak et al. (2012) reported that serogroup III was mainly isolated from vegetables in Poland. These results indicated that some specific serogroups of *L. monocytogenes* may have distinct ecological niches; Ochiai et al. (2010) also reported this for *L. monocytogenes* isolated from retail meat in Japan. Continuous and comprehensive investigation should be carried out for better understanding of the ecology of *L. monocytogenes*.

As contaminated foods are considered as the transmission source for human clinical listeriosis, we decided to carry out our investigation using *L. monocytogenes* strains that were isolated from food samples in South China, and examining the diversity of their antimicrobial profiles. In the present study, all of the *L. monocytogenes* isolates were susceptible to sulbactam/ampicillin (**Table 1**) and over 90% of isolates were susceptible to 10 antibacterial agents, namely kanamycin, gentamycin, sulfamethoxazole with trimethoprim, doxycycline, chloramphenicol, erythromycin, cephalothin, rifampin, tetracycline, and vancomycin, which indicated that these antibacterial agents are still effective for the treatment of listeriosis. The tetracycline and ciprofloxacin resistance rates were significantly different than that of the *L. monocytogenes* isolated in Northern China as described by Yan et al. (2010), which is possibly due to the influence of geographic differences. Surprisingly, 20% of *L. monocytogenes* isolates exhibited resistance toward the first-choice drug penicillin and ampicillin (**Table 1**), although no penicillinresistant strain was found in the previous studies conducted in China (Chao et al., 2007; Yang et al., 2008; Zhao et al., 2012). Therefore, more attention is required while monitoring the variation in the trend of resistance toward penicillin and

ampicillin. These findings indicated that *L. monocytogenes* is slowly becoming resistant to antibiotics in South China. This emphasized the need to continuously conduct surveillance for antibacterial susceptibility of foodborne *L. monocytogenes* in China.

In the present study, 5.6% (10/177) of *L. monocytogenes* isolates were determined as multidrug resistant (**Figure 2**) and ten isolates that were resistant to four antibiotics were obtained from animal foods and its associated products. These results support that the use of antibiotics in poultry and livestock as growth promoter has led to the emergence of antimicrobialresistant bacteria in the food and associated environment (Harakeh et al., 2009; Wieczorek et al., 2012). These results also indicated that foods from both poultry and livestock are possibly the reservoir of multidrug resistant *L. monocytogenes* strains. However, the results of multidrug resistance were inconsistent with that previously reported (Yan et al., 2010; Zhao et al., 2012), which can be attributed to the different source-composition of foodborne *L. monocytogenes* as well as different criteria used to determine multidrug resistance. In addition, it should be noted that, in our study 31.4% (16/51) isolates recovered from edible fungi samples exhibited resistance to more than two antibiotics. To the best of our knowledge, the processes involved in the production of edible fungi in scale-level plants would not use any antibiotics to prevent edible fungi contaminating other harmful microbes. This is the first report that *L. monocytogenes* isolates recovered from edible fungi samples demonstrated a high frequency of multidrug resistance, indicating that mushrooms may serve as the potential reservoirs of multidrug-resistant strains of *L. monocytogenes*.

Recently, high discriminatory molecular typing methods, i.e., PFGE, RAPD, and ERIC-PCR, have been developed for the differentiation of pathogenic bacteria. ERIC-PCR and RAPD techniques have been extensively used to determine genetic lineages of *L. monocytogenes* (Zhou and Jiao, 2004; Aurora et al., 2009; Chen et al., 2010). In the present study, ERIC-PCR and RAPD also yielded comparable results for the typing of *L. monocytogenes* strains, the discriminatory ability of ERIC-PCR and RAPD were 0.960 and 0.972, respectively. Genetic relatedness analysis revealed that there were no prominent associations between specific food types, antibiotic resistance, serogroups, and genetic diversity. The clustering results generated were different depending on the typing methods used (**Figures 2** and **3**), because the typing methods possibly target different genetic markers (Marshall et al., 1999). Seven isolates (418-1, 418-4, 518-1, 548-3, 698- 1, 748-1, and 798-1) belonging to serogroup I.1 were grouped into ERIC-D and RAPD-XI; ten isolates (27-1, 398-2, 418- 2, 468-1, 518-2, 548-1, 598-1, 618-1, 633-1, 668-1) belonging to serogroup II.2 were clustered into ERIC-C and RAPD-I. These isolates were obtained from *F. velutipes* samples, indicating that these isolates share similar genetic information and can better tolerate the disinfectants and sanitizers used in the mushroom production facility. Further studies are likely to elucidate the characteristics of these specific subtypes of *L. monocytogenes* colonizing the processing environments that in turn will aid in exploring better cleaning and sanitation measures. Previous studies demonstrated that the antibacterial resistance in *L. monocytogenes* was possibly acquired *via* selftransferable plasmids (Poyart-Salmeron et al., 1990), conjugative mobilization (Charpentier et al., 1999; Toomey et al., 2009), and efflux pumps (Rakic-Martinez et al., 2011; Jiang et al., 2012). Interestingly, two multidrug-resistant isolates (351-3, 715-2) were clustered together as revealed by ERIC-PCR, while isolates 351-3, 351-4 recovered from the fresh pork samples displayed unique fingerprints based on RAPD typing, suggesting that the multidrug-resistant isolates may acquire resistance by vertical and horizontal gene transfer. Further studies are required to elucidate the underlying molecular mechanisms for the acquisition of antimicrobial resistance by *L. monocytogenes*.

# Conclusion

Overall, the findings of this study demonstrated high contamination rate of *L. monocytogenes* in raw foods in South China, while the MPN values were relative low. Serogroup I.1 and II.2 were dominant among the foodborne *L. monocytogenes* strains, indicating that some specific serogroups of *L. monocytogenes* may have distinct ecological niches. Approximately 59.9% of the strains were susceptible to 15 antibiotics. All 177 isolates were grouped into 42 antibacterial susceptibility profiles and 5.6% of the isolates were multidrug resistant. Both ERIC-PCR and RAPD displayed excellent discriminatory ability for typing *L. monocytogenes*, the discriminatory index of ERIC-PCR and RAPD were 0.960 and 0.972, respectively. The present study provided the first baseline data on the prevalence, contamination level, and characteristics of *L. monocytogenes* isolated from raw foods in South China. Some multidrug resistant strains belonged to an epidemiologically important serogroup (I.1 and II.1), implying a potential public health risk. In addition, Chinese food safety authorities should draft appropriate standards to control *L. monocytogenes* contamination for the improvement of microbiological safety of retail raw foods.

# Funding

The authors would like to acknowledge the financial support of National Natural Science Foundation of China (U1031003, 31471664); Science and Technology Program of Guangdong Province (2013B050800026); Science and Technology Program of Guangzhou, China (201504010036); Key Projects in the National Science & Technology Pillar Program during the Twelfth Fiveyear Plan Period (2013BAD16B05).

# Supplementary Material

The Supplementary Material for this article can be found online at: http://journal*.*frontiersin*.*org/article/10*.*3389/fmicb*.* 2015*.*01026

## References


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2015 Chen, Wu, Zhang, Wu and Guo. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Presence of toxic microbial metabolites in table olives

#### *Eduardo Medina-Pradas and Francisco Noé Arroyo-López\**

*Food Biotechnology Department, Instituto de la Grasa, Consejo Superior de Investigaciones Científicas, Seville, Spain*

Table olives have an enormous importance in the diet and culture of many Mediterranean countries. Albeit there are different ways to produce this fermented vegetable, brining/salting, fermentation, and acidification are common practices for all of them. Preservation methods such as pasteurization or sterilization are frequently used to guarantee the stability and safety of fermented olives. However, final products are not always subjected to a heat treatment. Thus, microbiota is not always removed and appropriate levels of acidity and salt must be obtained before commercialization. Despite the physicochemical conditions not being favorable for the growth of foodborne pathogens, some illness outbreaks have been reported in the literature. Street markets, inappropriate manipulation and storage conditions were the origin of many of the samples in which foodborne pathogens or their metabolites were detected. Many authors have also studied the survival of pathogens in different styles of table olive elaboration, finding in general that olive environment is not appropriate for their presence. Inhibitory compounds such as polyphenols, low availability of nutrients, high salt content, low pH levels, bacteriocins, or the addition of preservatives act as hurdles against undesirable microorganisms, which contribute to obtaining a safe and good quality product.

#### Keywords: microbial risk, foodborne pathogens, table olives, mycotoxins, *Clostridium*, biogenic amines

#### Production of Table Olives

The fermentation of olive fruit has many centuries of history, particularly in the Mediterranean basin, where this fermented vegetable has had a great influence on the culture and diet of many countries. According to the last consolidated statistics of the International Olive Council, worldwide production currently exceeds 2.4 million tons per year. Spain, Turkey, Egypt, Syria, Algeria, Greece, and Morocco are among the main producers, albeit Argentina, Peru, and USA are also important contributors (International Olive Council [IOC], 2015). Thus, table olive processing is spread worldwide and represents an important economic source for olive-growing countries.

Olive fruit cannot be consumed directly from the tree due to its peculiar characteristics (presence of the bitter glucoside compound oleuropein, high fat, and low sugar content). For this reason, diverse methods were developed to make them palatable. Although many of them share the general process of brining/salting, fermentation, and acidification, they can differ slightly between areas of production. The Trade Standard Applying to Table Olives (International Olive Council [IOC], 2004) defines table olives as: '*the product obtained from suitable olive cultivars, processed to remove their natural bitterness, and preserved (by natural fermentation, heat treatment or preservatives) with or without brine until consumption*.' Among the most important table olive

#### *Edited by:*

*Maria Schirone, University of Teramo, Italy*

#### *Reviewed by:*

*George-John Nychas, Agricultural University of Athens, Greece Rosalba Lanciotti, University of Bologna, Italy*

#### *\*Correspondence:*

*Francisco Noé Arroyo-López, Food Biotechnology Department, Instituto de la Grasa, Consejo Superior de Investigaciones Científicas – Pablo de Olavide University, Building 46, Ctra. Utrera, Km 1, 41013 Seville, Spain fnarroyo@cica.es*

#### *Specialty section:*

*This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology*

*Received: 03 June 2015 Accepted: 10 August 2015 Published: 28 August 2015*

#### *Citation:*

*Medina-Pradas E and Arroyo-López FN (2015) Presence of toxic microbial metabolites in table olives. Front. Microbiol. 6:873. doi: 10.3389/fmicb.2015.00873* industrial processing methods we can find: (i) the so-called Spanish-style (alkali treated green olives), which represent about 50–60% of production, (ii) the so-called Californian-style (ripening of olives by alkaline oxidation), and (iii) directly brined olives (green, changing color or naturally black fruits) (Garrido-Fernández et al., 1997).

In all table olive processing methods described above, microorganisms have an important role, determining the safety, quality and flavor of the final product. Lactic acid bacteria (LAB) and yeasts are considered beneficial microorganisms, opposite to the role played by *Enterobacteriaceae* and *Propionibacteriaceae* (Garrido-Fernández et al., 1997; Arroyo-López et al., 2012; Hurtado et al., 2012). Traditionally, olive fermentation occurred spontaneously, but the process is not fully predictable and sometimes can lead to product spoilage or sanitary risks (Lanza, 2013). The present mini-review deals with the biological hazards posed by microorganisms in table olives, as well as the diverse hurdles that olive fermentation environments offer against growth of undesirable microorganisms.

### Sanitary Risks Caused by Microorganisms or Their Metabolites in Table Olives

Despite fermented table olives having a long history of microbial safety, diverse biological hazards may be present in the finished product (**Table 1**). Among the most relevant, we can mention:

#### (i) Biogenic Amines

The consumption of foods containing high amounts of toxic biogenic amines may cause food intoxication and intolerance, with diverse associated symptoms such as migraines, headaches, depression, diarrhea, insomnia, etc., indicating the need for a better hygiene process. These compounds can be formed in table olives by spoilage microorganisms with amino acid decarboxylase activity. Hornero-Mendez and Garrido-Fernández (1994) reported the presence of biogenic amines (putrescine, cadaverine, and tyramine) in fermented green table olives with "zapatera" spoilage. The concentration of biogenic amines can increase during olive storage but the levels found in the final products are usually low and should not represent a health concern (García-García et al., 2001). Recently, Tofalo et al. (2012) also detected in naturally fermented olives a low quantity of biogenic amines, as well as the presence by RT-qPCR of biogenic amines producing bacteria.

#### (ii) Mycotoxins

These compounds are secondary toxic metabolites produced by some species of mold (mainly *Aspergillus*, *Penicillium*, and *Fusarium* genera) under aerobic and humidity conditions (El Adlouni et al., 2006). Mycotoxins in foods can be of concern for consumers, causing disease in human and other vertebrates with symptoms such as skin irritation, immunosuppression, neurotoxicity, etc. Contamination of table olives with various types of mycotoxins (Ochratoxin, Aflatoxin B, and Citrinin) have been documented in cracked olives (Franzetti et al., 2011), but Greek-style black olives are the most affected (Gourama and Bullerman, 1988; Ghitakou et al., 2006). Fortunately, mycotoxin levels usually found in table olives are too low to cause disease.

#### (iii) Foodborne Pathogenic Bacteria

Diverse works have reported the presence of *Listeria monocytogenes* (Caggia et al., 2004; RASFF Portal, 2012a), *Staphylococcus aureus* (Asehraou et al., 1992; Pereira et al., 2008), and *Enterobacteriaceae* species such as *Yersinia enterocolitica* and *Escherichia coli* (Asehraou et al., 1992; Franzetti et al., 2011; Lucena-Padrós et al., 2014) in table olives. However, there are no reports of illness outbreaks caused by these microorganisms in table olives. Botulism, associated with *Clostridium botulinum* growth, is certainly the most relevant biohazard in table olives. Diverse outbreaks associated with homemade table olives and recalls of suspected products have been reported (Debord et al., 1920; Fenicia et al., 1992; Cawthorne et al., 2005; Jalava et al., 2011; Pingeon et al., 2011; RASFF Portal, 2012b). It should be emphasized that artisanal productions or inadequate storage (pH <sup>≥</sup> 4.5 units) were often the origin of these outbreaks. **Table 2** shows the epidemiological cases of botulism reported in table olives.

#### (iv) Degradation of Organic Acids

Spoilage microorganisms associated with fermented vegetables such as *Lactobacillus buchneri* are able to produce acetic acid from lactic acid consumption under anaerobic conditions (Johanningsmeier and McFeeters, 2013), whilst *Propionibacterium* and *Pectinatus* species are able to convert lactic acid to propionic acid (Breidt et al., 2013; Lucena-Padrós et al., 2014). Oxidative yeasts can also consume the lactic and acetic acids produced during olive fermentation under aerobic conditions (Ruiz Cruz and González Cancho, 1969). However, they are not able to use these acids in the absence of oxygen. Lactic acid consumption in table olives reduces the preservative power of fermented olives and increases the pH values, which can allow for the growth of others undesirable microorganisms with the consequent loss of product quality and food safety.

#### Table Olives: Hurdles against Biological Hazards

The olive fermentation process led by LAB involves the consumption of sugars to produce a wide range of final products with preservative effects; among the most important is lactic acid. These preservative compounds, together with low pH, protein and vitamin content, as well as reduced water activity (chloride salt is added to brine in a range of 5–11%), provide an acidic and salty environment which is adverse for the growth of undesirable microorganisms.

Other compounds excreted by microorganisms can also act as biopreservative agents. Bacteriocins are bacterial proteins or peptides that show a bactericide effect against closely related species. Jimenez-Díaz et al. (1993) isolated a bacteriocin producer *Lactobacillus plantarum* strain from green olive fermentation. The inhibitory compounds produced by this microorganism

#### TABLE 1 | Summary of the main types of biohazards reported in table olives.


#### TABLE 2 | Major illness outbreaks associated with botulism in table olives.


(plantaricins S and T) were active against bacteria that can cause spoilage in olive fermentations (*Propionibacteriaceae* and *Clostridium*) as well as natural competitors of *Lactobacillus plantarum* in olive fermentation brines (Ruíz-Barba et al., 1994). Likewise, yeasts produce toxic proteins or glycoproteins, also known as killer factors, are able to inhibit the growth of fungi and other non-desirable yeast species acting as biocontrol agents. *Debaryomyces*, *Pichia*, and *Candida* are genera with a considerable number of killer strains isolated from table olives (Hernández et al., 2008).

Table olive fermentations also contain antimicrobial compounds that limit the growth of LAB and others microorganisms, mainly in non-alkali treated olives (Medina et al., 2010). It has been recently demonstrated that some phenolic and oleosidic substances such as the dialdehydic form of decarboxymethyl elenolic acid (EDA), as well as EDA linked to hydroxytyrosol (Hy-EDA) present in olive brines, possess significant bacteriocide activity against foodborne pathogens, even greater than other phenolic compounds isolated from foods or synthetic biocides (Medina et al., 2009; Brenes et al., 2011). Thus, survival studies carried out with *E. coli* O157:H7 in Spanish-style table olive fermentation, show inhibition of the pathogen in all assayed conditions (Spyropoulou et al., 2001). Similar behavior was observed in the survival of *Bacillus cereus* in green olive fermentation, where the population declined steadily during the fermentation process (Panagou et al., 2008). Recently, Grounta et al. (2013) investigated the survival of diverse foodborne pathogens artificially inoculated on natural black table olives. They demonstrated that natural black olives are not a favorable environment to support the growth of the assayed pathogens, and the population of all them showed a rapid decline throughout the first 2 days of storage. Medina et al. (2013) studied the survival of diverse food-borne pathogens (*E. coli*, *Salmonella enterica*, *Listeria monocytogenes*, and *S. aureus*) in industrial olive brines from different cultivars (Manzanilla, Gordal, Hojiblanca, etc.). They found a correlation with the presence of polyphenols, considered inhibitory compounds from olives fruit. 5-log reduction of population inoculated was achieved between 5 min to 17 days in the least deleterious brine. Hence, according to the available data, table olive industrial brines of different olive cultivars and elaboration processes, do not constitute a favorable environment for any of the pathogenic bacteria tested.

# How to Reduce Biological Hazards in Table Olives

The objective of table olive producers should be to achieve zero risk in the case of illness and injury caused by toxic microbial metabolites. This can only be achieved by following practices that ensure that the fruits selected for processing are: produced under Good Agricultural Practices (GAP); processed under the principles of Good Manufacturing Practices (GMP) and produced at premises with equipment and personnel strictly following Good Hygienic Practices (GHP). All these requisites must be considered in the framework of food safety management systems, which include not only the HACCP System, but also other food defense tools to prevent intentional adulterations, i.e., CARVER (Criticality Accesibility Recuperability Vulnerability Effect Recognizability), TACCP (Threat Assessment and Critical Control Points), VACCP (Vulnerability Analysis and Critical Control Points).

In many cases, the fermentation of olive fruit still occurs spontaneously, which can sometimes lead to spoilage of the final product or to sanitary risks. In order to prevent these problems, the processing can be controlled through physicochemical (addition of acids, salt, temperature control, preservatives, or application of modified atmospheres) or microbiological approaches. To improve fermentation and consistently produce high quality, safe, final products; many authors have recommended strict process control of the above parameters, in addition to the use of starter cultures (see Corsetti et al., 2012 for a complete review on this aspect). The search for starters with application in olive fermentation and vegetables in general, has for many years, been focused on the activity of LAB and their technological applications. However, in the last decade, several publications

have emphasized the importance of the role that selected yeasts can play when used as starter cultures during table olive processing (Arroyo-López et al., 2012; Bevilacqua et al., 2012, 2013). Moreover, the selection of microorganisms as starters in olive fermentation and vegetables in general, has been exclusively based on diverse technological criteria (homo-fermentative metabolism; high acidification rate and fast consumption of fermentable substrates; organic acids, polyphenols, high pH and salt tolerance; flavor development or production of bacteriocins) (Duran-Quintana et al., 1999; Sánchez et al., 2001; Corsetti et al., 2012; Hurtado et al., 2012; Di Cagno et al., 2013; Heperkan, 2013). However, in addition to technological characteristics, recent studies on the development of starter cultures for table olives have focused on the study of the probiotic potential of native microorganisms. These studies must include both LAB and yeasts for the development of a mixed-multifunctional starter, in order to improve and expand the form of action of the culture by the use of two complementary microorganisms with different properties.

# Conclusion

The harsh environmental conditions found in the fermentation process (low pH, high salt content, presence of inhibitory compounds, sugar consumption, etc.), and the presence of other additional hurdles (production of bacteriocins, killer factors, addition of preservatives, etc.), make table olives an adverse habitat for the development of foodborne pathogens. If such growth ultimately occurs, the presence of undesirable microorganisms or their metabolites is often linked to the storage or selling conditions, not to the fermentation/production process. For all the above mentioned reasons, this widespread Mediterranean fermented vegetable can be considered quite a safe product, if good hygiene and manufacturing practices are followed and appropriated levels of salt (*>*5%) and pH (*<*4.3) are obtained in the final products.

#### Acknowledgments

This paper has received funding from Junta de Andalucía regional Government through the PrediAlo project (AGR-7755: www*.*predialo*.*science*.*com*.*es) and EU's Seventh Framework Program (FP7/2007-2013) under grant agreement n◦243471 (PROBIOLIVES: www*.*probiolives*.*eu). FA-L wishes to express thanks to the Spanish government for his Ramón y Cajal postdoctoral research contract, while EM-P would like to thank The Junta de Andalucía for his postdoctoral research contract. Both authors also express their gratitude to Antonio de Castro and Antonio Garrido-Fernández from Instituto de la Grasa (CSIC, Seville, Spain) for their invaluable comments and help during the writing of this manuscript.

# References


and brines from different Italian cultivars. *Int. J. General Mol. Microbiol.* 102, 121–131. doi: 10.1007/s10482-012-9719-x

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2015 Medina-Pradas and Arroyo-López. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# From genome to toxicity: a combinatory approach highlights the complexity of enterotoxin production in *Bacillus cereus*

Nadja Jeßberger <sup>1</sup> \* † , Viktoria M. Krey 2 †, Corinna Rademacher <sup>3</sup> , Maria-Elisabeth Böhm<sup>2</sup> , Ann-Katrin Mohr <sup>1</sup> , Monika Ehling-Schulz <sup>3</sup> , Siegfried Scherer <sup>2</sup> and Erwin Märtlbauer <sup>1</sup>

<sup>1</sup> Department of Veterinary Sciences, Faculty of Veterinary Medicine, Ludwig-Maximilians-Universität München, Oberschleißheim, Germany, <sup>2</sup> Lehrstuhl für Mikrobielle Ökologie, Zentralinstitut für Ernährungs- und Lebensmittelforschung, Wissenschaftszentrum Weihenstephan, Technische Universität München, Freising, Germany, <sup>3</sup> Functional Microbiology, Department of Pathobiology, Institute of Microbiology, University of Veterinary Medicine Vienna, Vienna, Austria

#### *Edited by:*

Giovanna Suzzi, Università degli Studi di Teramo, Italy

#### *Reviewed by:*

Mehrdad Mark Tajkarimi, University of North Carolina at Greensboro, USA Maria Schirone, Università degli Studi di Teramo, Italy

#### *\*Correspondence:*

Nadja Jeßberger, Department of Veterinary Sciences, Faculty of Veterinary Medicine, Ludwig-Maximilians-Universität München, Schönleutnerstr. 8, 85764 Oberschleißheim, Germany n.jessberger@mh.vetmed. uni-muenchen.de

† These authors have contributed equally to this work.

#### *Specialty section:*

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

*Received:* 07 May 2015 *Accepted:* 21 May 2015 *Published:* 10 June 2015

#### *Citation:*

Jeßberger N, Krey VM, Rademacher C, Böhm M-E, Mohr A-K, Ehling-Schulz M, Scherer S and Märtlbauer E (2015) From genome to toxicity: a combinatory approach highlights the complexity of enterotoxin production in Bacillus cereus. Front. Microbiol. 6:560. doi: 10.3389/fmicb.2015.00560 In recent years Bacillus cereus has gained increasing importance as a food poisoning pathogen. It is the eponymous member of the B. cereus sensu lato group that consists of eight closely related species showing impressive diversity of their pathogenicity. The high variability of cytotoxicity and the complex regulatory network of enterotoxin expression have complicated efforts to predict the toxic potential of new B. cereus isolates. In this study, comprehensive analyses of enterotoxin gene sequences, transcription, toxin secretion and cytotoxicity were performed. For the first time, these parameters were compared in a whole set of B. cereus strains representing isolates of different origin (food or food poisoning outbreaks) and of different toxic potential (enteropathogenic and apathogenic) to elucidate potential starting points of strain-specific differential toxicity. While toxin gene sequences were highly conserved and did not allow for differentiation between high and low toxicity strains, comparison of nheB and hblD enterotoxin gene transcription and Nhe and Hbl protein titers revealed not only strain-specific differences but also incongruence between toxin gene transcripts and toxin protein levels. With one exception all strains showed comparable capability of protein secretion and so far, no secretion patterns specific for high and low toxicity strains were identified. These results indicate that enterotoxin expression is more complex than expected, possibly involving the orchestrated interplay of different transcriptional regulator proteins, as well as posttranscriptional and posttranslational regulatory mechanisms plus additional influences of environmental conditions.

Keywords: *Bacillus cereus,* enterotoxins, Nhe, Hbl, host cell cytotoxicity, food poisoning

# Introduction

Bacillus cereus has become a hygienic and technological problem of increasing importance in the food industry. It is ubiquitous, produces heat resistant endospores and is able to form biofilms (Wijman et al., 2007; Stenfors Arnesen et al., 2008; Nam et al., 2014). Due to its lipo- and proteolytic properties it plays an important role in food spoilage (Andersson et al., 1995), but the main problem is the production of toxins, which are responsible for food poisoning. In 2011, the number of B. cereus-associated food poisoning outbreaks within the European Union increased by 122% (Anonymous, 2013). Various efforts have been made to predict the toxic potential of newly isolated strains. The B. cereus sensu lato group consists of eight closely related species, i.e., B. cereus, Bacillus anthracis, Bacillus thuringiensis, Bacillus weihenstephanensis, Bacillus mycoides, Bacillus pseudomycoides, Bacillus cytotoxicus, and Bacillus toyonensis. Highly toxic as well as non-toxic strains exist among most of these species. Guinebretière et al. (2008) further divided the B. cereus group into seven phylogenetic groups and subgroups (Guinebretière et al., 2008). For B. cereus sensu lato, the most commonly used method in routine diagnostics is the detection and quantification of colonies on selective culture media such as PEMBA or MYP according to international standards (Ehling-Schulz and Messelhäusser, 2013), which does not allow to distinguish the species of the B. cereus group. Therefore, only presumptive B. cereus can be detected (ISO 7932). While molecular methods for quantification of B. cereus have been established, no differentiation between living and dead cell or between spores and vegetative cells could be achieved (Martinez-Blanch et al., 2009; Ceuppens et al., 2010; Dzieciol et al., 2013). Currently, the molecular detection of toxin genes rather than species differentiation is applied (Ehling-Schulz and Messelhäusser, 2013). Toward this end, multiplex PCR systems for the detection of nhe, hbl, cytK, and ces have been established (Guinebretière et al., 2002; Fricker et al., 2007; Wehrle et al., 2009). However, the presence or absence of toxin genes does not allow to reliably infer the toxic potential, as highly variable amounts of toxins are produced in strains sharing the same toxin genes (Dietrich et al., 2005; Jeßberger et al., 2014).

A peptide synthetase, encoded by ces, produces the emetic toxin cereulide, the causative agent of the emetic syndrome leading to vomiting (Ehling-Schulz et al., 2005b; Stenfors Arnesen et al., 2008). Due to its resistance against heat, acids and proteolysis, it can't be inactivated by heat treatment of contaminated food samples or during stomach passage (Ehling-Schulz et al., 2004). The amount of preformed cereulide present in foods can consequently be used as an indicator for the toxic potential and thus for the consumer's risk (Ehling-Schulz and Messelhäusser, 2013). Currently, an ISO method for quantitative detection of cereulide in food samples (EU-CEN action: TC 257/WG6) is under development.

Enteropathogenic B. cereus cause diarrhea due to the production of enterotoxins in the human intestine. This occurs after viable bacteria or most likely spores are ingested together with contaminated foods (Clavel et al., 2004; Ceuppens et al., 2012). So far, the two three component enterotoxin complexes Nhe (non haemolytic enterotoxin, Lund and Granum, 1996) and Hbl (haemolysin BL, Beecher et al., 1995) have been described, as well as the single protein CytK (cytotoxin K, Lund et al., 2000). Only very few strains bear the highly toxic variant CytK1 and these are classified as a separate species, B. cytotoxicus (Guinebretière et al., 2013). The nhe genes are present in all enteropathogenic B. cereus strains analyzed so far. The hbl operon is present in approximately 50% of the strains, whereas its prevalence seems to be higher in clinical and food isolates (Guinebretière et al., 2002; Ehling-Schulz et al., 2005a; Moravek et al., 2006). Prediction of toxicity is based on the quantification of the enterotoxin components in B. cereus culture supernatants. Currently, three test systems are commercially available, detecting the enterotoxin components Hbl L2, NheA, as well as NheB and Hbl L2, respectively. However, results may often be inappropriate for evaluating the risk of contaminated food samples, as the enterotoxins, unlike the emetic toxin cereulide, are largely produced in the intestine.

According to recent studies, further virulence factors such as sphingomyelinase, haemolysin II or exoproteases contribute to pathogenicity. A role of sphingomyelinase as a virulence factor against insects and murine intestinal epithelial cells as well as its interaction with Nhe have been reported (Doll et al., 2013). HlyII was shown to induce in vitro and in vivo apoptosis to macrophages (Tran et al., 2011). In another study, hlyII was preferably found in pathogenic B. cereus, and inhA1 and nprA expression (both genes encoding metalloproteases) was higher in pathogenic than in non-pathogenic strains (Cadot et al., 2010).

So far, it is still not clear why toxin production and thus toxicity is so highly variable among different B. cereus strains. Thus, we aimed to explore whether distinct bacterial (genetic) factors, mechanisms or regulatory levels can be identified upon which high and low toxicity of enteropathogenic B. cereus strains can be distinguished unequivocally. Therefore, we chose a representative set of B. cereus isolates and characterized it with respect to enterotoxin gene presence, sequence and transcription as well as protein secretion, enterotoxin production and cytotoxicity.

# Materials and Methods

#### Bacterial Strains, Growth Conditions, and Sample Preparation

All B. cereus strains used in this study are listed in **Table 1** and Supplemental Table S1. Strains were grown at 30◦C on CGY plates or in CGY broth. Unless stated otherwise, 30 ml cultures were inoculated from 17 h pre-cultures, and were grown in 300 ml Erlenmeyer flasks with 125 rpm at 30◦C. For determination of enterotoxin production and toxicity, all strains listed in Table S1 were routinely grown for 6 h in CGY medium, inoculated with an optical density at 600 nm (OD600) of 0.2. For growth tests in this study, CGY medium was inoculated with a 17 h pre-culture to an OD<sup>600</sup> of 0.05 and OD<sup>600</sup> was recorded every 30 min. 5 milliliter (at 2 h) and 3 ml (at 6 h) samples were taken and centrifuged for 15 min at RT and 3500 rpm. Cell pellets were immediately frozen at −80◦C and used for preparation of RNA and measurement of intracellular protein concentrations. Supernatants were filtered through 0.2µm filters and separated for measurement of extracellular protein concentrations and enterotoxin production. For the latter, 1 mM EDTA was added to the culture supernatant. Supernatants were frozen and stored at −20◦C. All growth tests and sample preparations were carried out in triplicates.

#### Cell Lines and Culture Conditions

CaCo-2 cells were from DSMZ (German Collection of Microorganisms and Cell Cultures, Braunschweig, Germany)



Genotyping was performed by sequence analyses of the genetic markers spoIIIAB and panC, toxin profiling by PCR analyses using specific primers for ces, enterotoxin genes and virulence factors, toxin titers were determined by sandwich EIAs against NheB and toxicity was analyzed by WST-1-bioassay on Vero cells. Isolates were classified highly (hi), medium (m) or low (lo) toxic according to their cytotoxicity (hi: >500; m: 250-500; lo: <250) and NheB titers (hi: >4000; m: 2000-4000; lo: <2000).

\*, strain showed high toxicity but particularly low NheB titers due to binding failure of mab 2B11 in sandwich EIA (A. Didier, in preparation).

\*\*, sequence analysis revealed a truncated hbl operon; as strain is not able to produce Hbl L2 and Hbl B protein (negative in EIAs), it was allocated to profile F.

and Vero cells from ECACC (European Collection of Cell Cultures). Both cell lines were cultured in media recommended by the supplier and as described elsewhere (Jeßberger et al., 2014).

#### Toxin Gene Profiling and Sequence Typing

Total DNA was isolated using the MasterPure™ Gram Positive DNA Purification Kit (Epicentre). Detection of nhe, hbl, cytK and ces was performed by multiplex PCR (Ehling-Schulz et al., 2006). The sporulation stage IIIAB gene (spoIIIAB) was used as a genetic locus for sequence typing and classification of B. cereus isolates to clades (Ehling-Schulz et al., 2005a; Ehling-Schulz and Messelhäusser, 2013). Therefore, sequences were aligned using CLUSTAL X software (Thompson et al., 2002) and cluster analysis of aligned sequences was performed by TREECON using UPGMA method for inferring tree topology (Van De Peer and De Wachter, 1997). Alternatively sequence typing and clade classification was carried out by using the pantothenate synthetase gene panC (Guinebretière et al., 2008).

#### Comparison of Toxin Operon Nucleotide and Amino Acid Sequences

Nucleotide sequences of nheABC and hblCDA were obtained from a de novo genome sequencing approach (MiSeq <sup>R</sup> Reagent Kit v2, 500 cycles, Illumina) (Böhm et al., in preparation), identified via BLASTN algorithm and compared using CloneManager 7. Multiple amino acid sequence alignments of single Nhe and Hbl enterotoxin components were constructed with Clustal Omega v1.2.0 (Sievers et al., 2011). Signal peptide sequences were predicted using the SignalP 4.1 server (Petersen et al., 2011).

#### RNA Isolation, cDNA Synthesis, and Quantitative Real-time PCR (qRT-PCR)

Isolation of total RNA and DNAse I digestion was performed as previously described (Dommel et al., 2010). Using the random primers of the qScript cDNA Supermix (Quanta Biosciences), first strand synthesis of 1 mg of total RNA was performed. qRT-PCR was done as reported repeatedly and the 2−11C<sup>T</sup> method was used for calculation of relative gene expression (Livak and Schmittgen, 2001; Lücking et al., 2009; Dommel et al., 2010). Primer amplification efficiencies (E) were in the desirable range (E = 1.7 − 2.1) for all primer pairs listed in Table S2 as evaluated according to the equation E = 10(−1/slope) (Pfaffl et al., 2002). 16S rrn transcription levels determined by qRT-PCR (Martineau et al., 1996) served as the reference control for normalization applying the 2−11C<sup>T</sup> method (Livak and Schmittgen, 2001). This method is based on the following formula: amount of target transcript = 2 <sup>−</sup>11C<sup>T</sup> with −11C<sup>T</sup> = −(1CT(sample) − 1CT(calibrator)) = − ((CT(reference gene) − CT(target gene))sample − (CT(reference gene) − CT(target gene))calibrator). C<sup>T</sup> denotes the cycle number of the amplification reaction that exceeds the quantification threshold of the instrument. To determine the relative transcription of a target gene, rrnnormalized transcription level of a sample (1CT(sample)) was set relative to the transcript level of an external calibrator (1CT(calibrator)) and multiplied with 100 to obtain relative transcription in percent [%]. Expression level of the hblD gene of the toxin reference strain B. cereus F837/76 at 6 h served as calibrator that was set to 100% (log-2 = 0), all other transcript levels were compared to this condition using the 2−11C<sup>T</sup> method.

#### Enzyme Immunoassays (EIAs)

The enterotoxin components NheB and Hbl L2, L1 and B were detected in indirect and sandwich enzyme immunoassays as described elsewhere (Dietrich et al., 1999, 2005; Moravek et al., 2006). For NheB detection, the sandwich EIA using mabs 2B11 and 1E11-HRP or the indirect EIA using only mab 1E11 were performed. Titers were defined as the reciprocal of the highest dilutions resulting in an absorbance value of ≥1.0.

#### WST-1 Bioassays

WST-1 bioassays were performed as previously described (Dietrich et al., 1999, 2005; Didier et al., 2012; Jeßberger et al., 2014). Characterization of the toxic activity of the 136 B. cereus strains was carried out on Vero cells, as this assay is used in our routine B. cereus diagnostic. For the 19 strains in this study, the human colon carcinoma cell line CaCo-2 was further used. For detection, the optical density at 450 nm was measured with a Tecan photometer. With Ridawin software dose-response curves and thus 50% inhibitory concentrations were calculated, shown as reciprocal titers.

#### Propidium Iodide (PI) Influx Studies

Enterotoxin-induced pore formation in the membranes of CaCo-2 cells was measured by propidium iodide influx tests as described before (Jeßberger et al., 2014). For each B. cereus strain, a specific curve (increase of fluorescence per time) was obtained, of which the value of the highest linear slope was calculated. Strains were compared according to those values.

#### Quantification of Total Protein Amount

Extracellular proteins were obtained from culture supernatants. For isolation of intracellular proteins cell pellets were resuspended in 50 mM Tris, pH 7,5 and disrupted by two passages through a French Pressure cell press (2 kbar). The soluble proteins were collected by centrifugation (15.000 rpm, 45 min, 4◦C). Intra- and extracellular protein concentrations were quantified by Roti-Nanoquant Kit (Roth) in microtiterplates according to the manufacturer's instructions. Colorimetric reactions were measured with Infinite F200 reader (Tecan) at wavelengths of 610/450 nm. Protein concentrations were determined by quotient of optical densities OD610/<sup>450</sup> referred to internal standard generated by Quick Start Bovine Serum Albumin Standard (Biorad).

#### Isolation of Extracellular Protein Extracts for Gelelectrophoretic Separation

Protein extracts were precipitated by addition of ice-cold trichloroacetic acid solution to a final concentration of 10% (v/v) overnight at 4◦C. Precipitated proteins were centrifuged (1 h, 10.000 rpm, 4◦C) and protein pellet was washed 3 times with acetone. Pellets were dried overnight at 4◦C and finally resuspended in DIGE lysis buffer (7 M Urea, 2 M Thiourea, 4% (w/v) CHAPS, 30 mM Tris, pH 8.5).

#### DIGE Labeling and SDS PAGE

For differential comparison of secretion patterns 5µg of protein extracts per sample were labeled with CyDye Fluor minimal dyes (GE Healthcare) according to the manufacturer's instructions. Protein extracts of low toxin producing strains were labeled with Cy5, extracts of high toxin producing strains were labeled with Cy3. Afterwards samples of corresponding low and high toxic strains were pooled, reduced with Laemmli sample buffer containing 0,5 M DTT for 10 min on ice, and alkylated with 14% (w/v) iodoacetamide for 10 min at room temperature. Finally labeled protein extracts were separated by SDS PAGE using SE600 vertical electrophoresis system (Hoefer). Fluorescence images were scanned with a Typhoon 9400 imager (GE Healthcare).

#### Statistical Analyses

Mean values and standard deviations (SD) were calculated from at least three independent experiments each. For statistical analysis, 2-tailed Student's t test was used, where applicable, to determine statistically significant differences by Microsoft Excel. Graphs of raw data were plotted using Microsoft Excel or GraphPad Prism (version 6.0). Spearman correlation was calculated using SigmaStat 3.5.

# Results

#### Selection of the *B. cereus* Strain Set

For selection of the B. cereus strain set to be used in this study, 136 isolates from the strain collections of our institutes were characterized according to their genetic toxin profile. Additionally, they were classified to be of high, medium or low toxicity according to their cytotoxicity in routinely performed WST-1 bioassays on Vero cells and their production of enterotoxin component NheB (Table S1). 19 of these strains were selected for genome sequencing (Böhm et al., in preparation) and growth, enterotoxin gene transcription, enterotoxin production, protein secretion, and cytotoxicity were studied. The genetic relationship of the strains is depicted in Figure S1. The strains were of different origins (food or food poisoning outbreaks), of different genetic toxin profiles and of either high or low toxic activity (**Table 1**). Strains involved in food poisoning outbreaks were isolated either from the contaminated food or from patient feces. Also, the presence of hlyII, inhA1, and nprA was considered. Originally, equal numbers of strains per toxin profile were chosen, later two isolates from recent food poisoning outbreaks in Austria and England were added to profile A.

#### Strong Genetic Conservation of Enterotoxin Components

Whole genome sequences of all 19 B. cereus strains were obtained using a de novo next generation sequencing approach (MiSeq <sup>R</sup> , Illumina <sup>R</sup> ) (Böhm et al., in preparation). Enterotoxin genes nheABC and hblCDA were identified via BLASTN analysis using published nhe and hbl operon sequences from the reference strains NVH 0075-95 and F837/76 as template. Alignments of the concatenated operon genes using CloneManager 7 showed that nucleotide sequences of both enterotoxin operons are highly conserved over all B. cereus strains investigated. Similarity of nheABC genes ranges from 93 to 99%, while hblCDA genes are even more conserved (95–98%). In contrast to all other hbl encoding strains of the set that harbor the major hblCDAB operon, B. cereus MHI226 solely possesses the hbl<sup>a</sup> gene variant (Beecher and Wong, 2000) with a sequence similarity of 80– 81% to the other hblCDA genes. B. cereus MHI226 does not form a functional Hbl toxin (data not shown) and was therefore reclassified in toxin profile F (only encoding nheABC). Hbl<sup>a</sup> (hblCDAa) is present in addition to the major toxin operon hblCDAB in about one third of all hbl-encoding strains (Böhm et al., in preparation), as it appears in strains 14294-3 (M6), 6/27/S and RIVM BC 126. Similarity between hbl and hbl<sup>a</sup> variants ranges from 75 to 82%.

As a consequence of the pronounced enterotoxin gene conservation, very high sequence identity can be observed in multiple amino acid (aa) sequence alignments for all enterotoxin components (Figures S2, S3). NheB is the most conserved Nhe component with 15 variable out of 402 aa (3.73%) (Figure S2B), compared to 35 out of 386 aa (9.07%) in NheA (Figure S2A) and 47 out of 359 aa (13.09%) in NheC (Figure S2C). Hbl L2 exhibits 37 variable out of 439 aa (8.43%) (Figure S3A), Hbl B 24 out of 375 (6.40%) (Figure S3C) and the highly conserved Hbl L1 6 out of 406 (1.48%) (Figure S3B).

Additionally, amino acid sequences of the putative virulence factors HlyII, InhA1, NprA, and sphingomyelinase (Cadot et al., 2010; Tran et al., 2011; Doll et al., 2013) were compared. As for Nhe, all 19 strains possess the genes for InhA1, NprA and SMase, while only 21.1% (4 out of 19, among them two high and two low toxic strains) bear hlyII (**Table 1**). Thus, hlyII was excluded from further investigations. The low toxic strain MHI226 even possesses two hlyII genes. In comparison to the enterotoxins, InhA1, NprA and SMase are less conserved harboring 11.3%, 9.0% and 16.2% variable aa positions, respectively (data not shown). Neither comparison of toxin or virulence factor genes, nor amino acid sequences displayed discriminatory differences between high and low toxic B. cereus strains or showed any clustering according to their origin (food, food poisoning, or feces). These results show that neither presence or absence nor gene and protein sequences of enterotoxin and virulence factors provide a basis to assess virulence of a B. cereus strain. Therefore, it may be suspected that differential toxicity may be due to regulatory processes at the transcriptional or translational level of toxin expression.

#### Highly Similar Growth Kinetics within the *B. cereus* Strain Set

For comparative analyses of toxin transcription, production and secretion, all 19 B. cereus strains were grown under standard laboratory conditions. In **Figure 1**, the growth kinetics of strains grouped according to their toxin profile are shown. In general, growth kinetics of all B. cereus strains were very similar except for the low toxicity strain MHI226, which had a more shallow slope and reached the lowest maximal optical density of OD<sup>600</sup> 11.9, while OD600max. of all other stains in toxin profile F ranged between OD<sup>600</sup> 15.8–23.2 (**Figure 1D**). Maximal OD<sup>600</sup> in toxin profile A (**Figure 1A**) was between OD<sup>600</sup> 15.2–27.4, between OD<sup>600</sup> 20.0–23.7 in toxin profile C (**Figure 1B**) and between OD<sup>600</sup> 14.6–20.8 in toxin profile D (**Figure 1C**). Strains isolated from human feces (6/27/S, F3175/03 (D7), RIVM BC 126, F3162/04 (D8), RIVM BC 90, 7/27/S) tended to grow to higher maximal optical densities (OD<sup>600</sup> > 20.0) than non-feces strains of the same toxin profile. However, significant differences in growth kinetics were only seen for MHI226. To account for any growth differences, further data were normalized to the optical density of the respective B. cereus strain at the time point analyzed.

#### Strain-dependent Toxin Gene Transcription

Transcription of enterotoxin genes was analyzed using qRT-PCR. As previously described, B. cereus cytotoxicity strongly correlates with NheB and Hbl L1 concentrations in the supernatant (Jeßberger et al., 2014). Therefore, transcript levels (Livak and Schmittgen, 2001) of the corresponding genes nheB and hblD were determined. Relative nheB and hblD transcription [%] of all 19 B. cereus strains is presented in **Figures 2A,B**, respectively. In general, relative toxin gene transcription increased from 2 to 6 h in CGY medium, except for B. cereus 14294-3 (M6). Relative hblD transcription was higher than nheB transcription, except for strains 14294-3 (M6), RIVM BC 934 and RIVM BC 126 which showed 2.4 and 3.9-fold increase of nheB over hblD transcription in 6 h cultures. Unexpectedly, maximum hblD transcription occurred in F528/94 (**Figure 2B**), a strain classified as low toxic in the cytotoxicity assay. Relative transcription was further normalized to the OD<sup>600</sup> and resulting transcription efficiencies (% transcription/OD600) are depicted in **Figures 2C,D**. In contrast to relative toxin transcript levels, toxin transcription efficiency did not uniformly increase from 2 to 6 h for all B. cereus strains, but showed strain-dependent kinetics. Eight strains of toxin profile A, C, D and F and of various origins showed decreased nheB transcription efficiency in 6 h cultures compared to 2 h, six strains had increased efficiency and five were similar at both growth stages (**Figure 2C**). In contrast, 80% of strains expressing Hbl showed higher hblD transcription efficiency at 2 h compared to 6 h except for SDA KA96 and

F528/94 (**Figure 2D**). Interestingly, maximal nheB transcription efficiency was observed in the low toxic strain 14294-3 (M6) after 2 h in CGY medium, followed by the high toxic strains F837/76, RIVM BC 126 and F4429/71 (**Figure 2C**). Highest hblD transcription efficiency was achieved in 2 h cultures by F837/76, followed by INRA C3 and 14294-3 (M6) (**Figure 2D**).

In principle, toxin transcript levels of both nheB and hblD reflect the classification into high and low toxin producing strains according to their cytotoxicity (**Table 1**) to a certain extent. However, this is not generally the case. Setting a random threshold of 20% relative nheB and 40% hblD transcription for strains showing high toxin transcript levels, high cytotoxic strains F3175/03 (D7), F4429/71, F3162/04 (D8), WSBC10035, and 7/27/S cannot be considered to transcribe high levels of nheB, neither RIVM BC 126 for hblD. Moreover, some strains, e.g., 14294-3 (M6) are low toxic even though toxin gene transcription efficiency is amongst the highest, indicating that posttranscriptional and/or posttranslational regulation plays an important role for the manifestation of B. cereus toxicity. Furthermore, transcription of inhA1, sph and nprA, discussed as additional virulence contributors, proved to be highly strain-dependent (data not shown). In summary, there was no correlation of a strain's origin or genetic toxin profile and its toxin or virulence factor transcript level.

#### Strain-specific Enterotoxin Production

Reciprocal titers of the enterotoxin components NheB, Hbl L2, Hbl L1, and Hbl B in the culture supernatants were determined performing specific EIAs. As a strong correlation of NheB with cytotoxicity has already been demonstrated (Moravek et al., 2006; Jeßberger et al., 2014), titers of NheA or NheC were not investigated. The productivity for all toxin components was calculated as titer/OD<sup>600</sup> to exclude growth effects (**Figure 3**). After 2 h, no Hbl and only trace amounts of NheB protein were detectable. After 6 h, NheB titers generally confirmed prior classification into high, medium and low toxin producing strains (compare **Table 1**). Only 14294-3 (M6) showed higher NheB levels than in previous experiments (**Figure 3A**). Strains producing high, medium and low amounts of NheB were found among all toxin profiles as well as among food- and food poisoning-associated strains. NheB titers of strain F3162/04 (D8) were determined using only NheB-specific mab 1E11, as this strain is negative in sandwich EIA (A. Didier, in preparation). As the indirect EIA is not as sensitive as the sandwich, NheB production of this strain most likely is higher than detected with our tools.

The toxin pattern of the three Hbl components was similar to Nhe, i.e., strains earlier classified as high showed high amounts of all Hbl components, low strains showed lower Hbl production (**Figures 3B–D**). Interestingly, the two strains 6/27/S and F3175/03 (D7) showed similar toxic activity (**Table 1** and **Figure 6**), but a different pattern. While 6/27/S produced comparably high amounts of Hbl L1 but low amounts of L2 and B, F3175/03 (D7) showed a comparably high titer and productivity for Hbl L2 but low L1 and B. Strain F528/94 showed similar Hbl titers compared to strain F837/76, though classified as a pair of low and high toxicity strains of the same genetic

toxin profile (**Table 1**), suggesting that the total toxic activity as determined by WST-1-bioassays on Vero cells depends on Nhe rather than Hbl production. As for NheB, no specific pattern of Hbl production was found among the toxin profiles or food and food poisoning strains. NheB and Hbl L1 titers were compared to the transcription levels of their corresponding genes nheB and hblD using the Spearman correlation test. Significant dependence (P values < 0.05) of secreted NheB on relative nheB transcript level could be demonstrated for all time points investigated. Spearman correlation coefficient R was 0.817 (2 vs. 2 h), 0.605 (6 vs. 6 h) and 0.600 for comparing 2 h transcription with 6 h toxin titers. Poor influence (P > 0.05) of hblD transcription on Hbl L1 titer was detected.

#### Comparable Protein Secretion Ability

To analyze whether enterotoxin production generally correlates with the amount of secreted proteins, total extra- and intracellular protein concentrations were quantified. Intracellular protein concentrations were similar among all isolates tested (data not shown). Extracellular protein contributed to up to 10% of total protein content. 2 h after inoculation, only few extracellular protein concentrations were detectable (**Figure 4A**), while extracellular proteins of about 20–60 ng/µl were isolated after 6 h of growth under the standard laboratory conditions. For MHI226 total extracellular protein concentrations of 90 ng/µl were found. Extracellular protein content was normalized to the optical density of the strains to calculate the efficiency of protein secretion. As depicted in **Figure 4B**, all strains,

except MHI226, showed comparable values. Generally, secretion efficiency was higher after 6 h of growth than at early growth phase. Only isolates 14294-3 (M6), F528/94 and NVH0075-95 revealed comparable efficiency values at 2 and 6 h. However, no significant differences in extracellular protein concentrations or efficiency of protein secretion were found between low and high toxic strains.

#### Exoprotein Profiling by DIGE Analysis

To study secretion patterns of selected B. cereus isolates, secreted proteins were isolated 6 h after inoculation and analyzed by SDS-PAGE. Extracellular protein extracts of high and low toxin producing strains of the same genetic toxin profile were labeled with fluorescent Cyanine Dyes and separated in one single lane. Analysis of exoprotein profiles revealed distinct differences between the B. cereus isolates (**Figure 5**). Various proteins were differentially regulated in high and low toxic strains. However, secretion patterns of B. cereus isolates show high variability between the strains and no marker bands specific for high or low toxic strains could be detected by DIGE analysis. Silver staining of SDS-PAGE analyzed exoprotein extracts of the remaining 9 strains

confirmed the high variability of secretion patterns (data not shown).

#### Strain-specific Cytotoxic Activity

Cytotoxicity assays were performed using the human colon carcinoma cell line CaCo-2 instead of the routinely used Vero cells. As a consequence, the classification as high, medium or low toxic (**Table 1**) was altered. 14294-3 (M6), RIVM BC 934 and MHI226, former low toxic, now showed relatively high cytotoxicity in comparison with all strains. On the other hand, for INRA C3 and F4429/71, former highly toxic strains, comparably low cytotoxicity titers were obtained (**Figure 6A**). Of special interest were the strains F837/76 and F528/94 (high and low toxic, sharing the same genetic enterotoxin profile). Although F837/76 showed 12 x higher NheB production than F528/94 (**Figure 3A**), host cell toxicity was only 1.8 x increased (**Figure 6A**). Hbl titers were similar in both strains (**Figures 3B–D**), underlining that CaCo-2 cells are more sensitive to Hbl than Nhe. Our data indicate that results of the WST-1 bioassay depend very much on the target cell line and that for studying the toxic potential of enteropathogenic B. cereus in the consumer human colon cell lines such as CaCo-2 must strictly be used to avoid underestimation of certain strains.

Additionally, propidium iodide influx tests were performed to determine the speed of pore formation by the enterotoxins. Strains were compared according to the speed of PI influx shown as the maximum linear slope of the fluorescence curves (**Figure 6B**). Compared to the WST-1 bioassay, almost no differences between high and low toxic strains appeared, whereas strains producing Nhe and Hbl caused a significantly faster PI influx than solely Nhe producing strains. This phenomenon, as observed before (Jeßberger et al., 2014), suggests that pore formation in CaCo-2 cell membranes is triggered by the presence of Hbl and additionally that even low amounts of Hbl cause rapid PI influx.

#### Discussion

So far, the reasons why B. cereus strains with identical genetic toxin profile produce such variable amounts of enterotoxins are not understood, i.e., it is yet to be investigated at which levels the varying pathogenic potential originates and is regulated. Moreover, it is unclear, whether further markers (proteins or nucleotides) can be identified for the differentiation between high and low toxic strains. To our knowledge, the presented study is the first comprehensive analysis of a representative set of enteropathogenic B. cereus strains that integrates many levels of toxin formation, from gene sequence to host cell cytotoxicity.

#### Strain Set

Currently, the most reliable method for predicting the toxic potential of enteropathogenic B. cereus is the quantification of the

FIGURE 5 | 1D DIGE exoprotein profiling of selected *B. cereus* isolates. Exoproteins were isolated 6 h after inoculation and precipitated by trichloroacetic acid overnight. Five microgram of protein extracts were labeled with CyDye Fluor minimal dyes and separated by SDS PAGE. Proteins present in low toxin producing strains are shown in red (Cy5), proteins present in high toxin producing strains are displayed in green (Cy3). Similar protein levels are indicated by yellow bands.

secreted enterotoxins (Guinebretière et al., 2002; Moravek et al., 2006). This can also be seen in ourstrain set, which can be divided into "pairs" of high and low toxic strains with identical genetic background. For example, INRA A3 and INRA C3, both clade II, toxin profile A, hlyII−, inhA1+, nprA+, share over 99% and 98% sequence similarity for Nhe and Hbl, but low vs. high amounts of the enterotoxin components are found in the respective culture supernatant. This was also observed for F528/94 (low) and F837/76 (high), both clade I, toxin profile C, hlyII+, inhA1+, nprA+, with 96% and 97% sequence similarity for Nhe and Hbl, respectively. The Nhe toxin components of MHI86 (low) and F4429/71 (high) (clade I, profile D, hlyII−, inhA1+, nprA+) even were 100% identical. So far, no connection between a certain enterotoxin gene profile, i.e., the presence of certain enterotoxin genes, and the toxic potential was found (Ceuppens et al., 2013). Even the presence of the PlcR/PapR regulatory system is no reliable indication for toxic activity (Fagerlund et al., 2007). Our data suggest that cytotoxicity of B. cereus is rather independent of toxin gene presence and sequence polymorphisms.

#### Enterotoxin Gene Transcription

In a first systematic approach we compared enterotoxin gene expression of a representative B. cereusstrain set. Previous studies compared enterotoxin expression in single B. cereus strains under varying growth temperature (Van Netten et al., 1990; Fermanian et al., 1997; Rejasse et al., 2012), oxygen availability (Duport et al., 2004; Van Der Voort and Abee, 2009), lowered oxidationreduction potential (ORP) (Duport et al., 2006; Zigha et al., 2007; Esbelin et al., 2009) and differing availability of nutrients, for instance sugars (Ouhib et al., 2006; Ouhib-Jacobs et al., 2009). In our study, quantification of nheB and hblD mRNA levels at early exponential and late exponential growth phase/transition phase by qRT-PCR revealed highly strain-dependent variations in toxin gene transcript levels and transcription efficiency. It was further shown that toxin production is only to a certain extent determined by the level of enterotoxin transcription, which suggests that further posttranscriptional and posttranslational processes are involved.

The global virulence regulator PlcR strongly induces transcription of all enterotoxin genes during entry into the stationary growth phase (Gohar et al., 2008). PlcR-driven toxin gene expression might continue until B. cereus faces nutrient limitation at late stationary growth and undergoes sporulation, while at the same time plcR transcription is repressed by the transition state regulator Spo0A (Lereclus et al., 2000). However, in the present study transcription of nheB and hblD could be detected as early as 2 h after inoculation, and increased markedly during exponential growth until 6 h. These results coincide with the findings of Frenzel et al. (2012) on high plcR transcript levels during early B. cereus growth phase and reemphasize the direct positive transcriptional regulation of the enterotoxin operons by PlcR (Frenzel et al., 2012).

Comparison of the PlcR regulator proteins of our strain set revealed minor differences, especially in the less conserved Cterminal part of the protein (data not shown), which is also truncated in B. anthracis (Agaisse et al., 1999; Gohar et al., 2005). With the exception of one strain, PlcR of all B. cereus strains investigated shared between 87 and 100% similarity. Transcription analysis of the plcR regulator gene for a subset of B. cereus strains resulted in less pronounced strain-specificity compared to toxin gene transcription, but demonstrated similar plcR transcription kinetics for toxin profile A/C and D/F. Generally, no enhanced plcR transcript levels could be measured for B. cereus isolates classified as high in contrast to low toxic strains, except for the pair MHI86 (low) and F4429/71 (high) (Figure S4). Furthermore, no correlation between plcR transcription and nheB or hblD transcript levels could be determined. However, another pattern for differentiation of isolates was found: all four strains of toxin profile A and C showed increased plcR transcription after 6 h compared to 2 h, while the remaining six strains of toxin profile D and F transcribed lower amounts of plcR and tend to decrease plcR transcription at the later growth stage (Figure S4).

The present study focused on the comparison of a B. cereus strain set under identical aerobic culture conditions. Hence, induction of environment-sensing regulator proteins Fnr, ResDE, CcpA or CodY (Duport et al., 2006; Zigha et al., 2007; Van Der Voort et al., 2008; Esbelin et al., 2009, 2012; Messaoudi et al., 2010) was assumed to be comparable between the strains, their impact on differential toxin gene expression being negligible. Furthermore, since Fnr, ResDE, CcpA, and CodY protein sequences are highly conserved (sequence similarity > 98% and identity > 96%, data nor shown), variations in functionality are unlikely to account for strain-specific differences in toxin gene activation. Also, comparison of the nheABC and hblCDAB operon promoter regions of the strain set revealed only minor DNA sequence differences between essential promoter elements and regulator binding sites, precluding mutations that could explain single strain nheB or hblD transcript levels (data not shown).

Enterotoxin gene transcription only partially reflects our isolates' actual toxicity, indicating posttranscriptional and/or translational regulation. Alternative regulation of gene expression by non-coding RNA molecules is well-known in bacteria. RNAs might act as antisense RNA or form alternative structures, so called riboswitches, that promote or inhibit target gene translation in response to small molecule binding or temperature changes (for review see Johansson and Cossart, 2003; Batey, 2006). It has been demonstrated that several virulence determinants of the Gram-positive pathogen Listeria monocytogenes are controlled posttranscriptionally by long 5′ untranslated RNAs (Wong et al., 2004; Shen and Higgins, 2005; Loh et al., 2006). Involvement of stabilized mRNA transcripts in the regulation of cryIIIA toxin gene expression has been shown in B. thuringiensis and more recently gene control by riboswitches has been characterized in B. anthracis and B. subtilis (Welz and Breaker, 2007; Irnov et al., 2010; Wilson-Mitchell et al., 2012). The observation of unusually long 5′ untranslated regions (UTR) upstream of the start codons of both enterotoxin operons that span over 600 bp for Phbl and 300 bp for Pnhe (data not shown; see also Agaisse et al., 1999) may point to a role in posttranscriptional regulation via formation of regulatory

mRNA structures. Hence, deciphering a possible function of the long 5′ UTRs in enterotoxin expression of B. cereus is subject to ongoing investigations.

#### Protein Secretion

All 19 B. cereus strains included in this study were capable of significant protein secretion. Under laboratory growth conditions all strains showed comparable extracellular protein concentrations and no differentiation between high and low toxicity strains were obvious. Thus, a defect in general protein secretion can be excluded as a potential cause for the low toxicity of certain strains.

B. cereus enterotoxins are most likely secreted via the sec translocation pathway (Tjalsma et al., 2004; Fagerlund et al., 2010; Vörös et al., 2014). Complete amino acid sequence comparisons performed in our study showed that Sec-type secretion signal peptides of all Nhe and Hbl enterotoxin components, as predicted using the SignalP method, are highly conserved, for NheB and Hbl L1 even up to 100% (Figures S2, S3). Especially, signal peptidase cleavage sites are identical with the exception of NheA, where 14294-3 (M6) showed an exchange of the basic amino acid lysine against the negatively charged, hydrophilic amino acid residue glutamate K27E (Figure S2A). It has already been shown that a modification of the signal peptide sequence of Hbl B within the hydrophobic region leads to loss of secretion and intracellular accumulation of the protein (Fagerlund et al., 2010). In our study, most amino acid exchanges could be found in the signal peptide sequence of NheC and Hbl B. While most changes are neutral for the charge of the signal peptide sequences of Nhe and Hbl, some substitutions of the neutral alanine residue to threonine render the hydrophobic region more hydrophilic: A19T (NheB), A12T (NheC), A21T (Hbl B) (Figures S2B, S2C, S3C). However, no correlation of signal peptide mutations with high and low toxicity was found. Only the signal peptide sequence of Hbl B in the highly toxic isolate RIVM BC 126 showed several amino acid residue substitutions that might explain why high levels of Hbl L1 and Hbl L2 were detected in the secretome of RIVM BC 126, but only relatively low amounts of Hbl B (**Figure 3**).

Additionally, we compared extra- and intracellular amounts of NheB and Hbl L1 to exclude potential defects in the secretory mechanisms of our strains. Though taken in an independent approach, NheB and Hbl L1 titers in the culture supernatants were comparable to those in **Figure 3**. No or only trace amounts of NheB and Hbl L1 were detectable in the cytoplasm (data not shown), suggesting that the enterotoxins of all strains are secreted immediately and effectively after translation. Furthermore, intracellular titers rose in comparison to extracellular titers after treatment of the strains with sodium azide (data not shown), which inhibits the ATPase function of SecA (according to Fagerlund et al., 2010). This indicates that the enterotoxins of all 19 strains are indeed secreted via a fully functional Sec translocation pathway.

Besides enterotoxins, B. cereus secretes a variety of other proteins such as proteases, phospholipases and other members of the PlcR regulon (Gohar et al., 2008). For B. anthracis it has been shown that own bacterial proteases contribute to virulence by directly degrading anthrax toxins or modulating concentrations and activity of other extracellular proteases (Pflughoeft et al., 2014). The proteases InhA1 and NprX can represent up to 90% of the bacterium's secretome (Chitlaru et al., 2006). The concentrations of the respective proteins in the secretome of the B. cereus strain set cannot be detected so far, as suitable detection methods (e.g., EIAs based on specific antibodies) still have to be developed. We suppose it may rather be possible that the low toxic strains generally secrete higher levels of proteases, which subsequently degrade the secreted enterotoxins, than secreting low amounts of enterotoxins and proteases. Enterotoxin gene expression of strain MHI226 was hardly detectable, but cytotoxic activity on CaCo-2 cells was comparably high. Perhaps, the lack of enterotoxin production was compensated by enhanced protein secretion. Generally, MHI226 showed quite extraordinary properties concerning growth, enterotoxin gene expression, toxicity and especially protein secretion, thus it might be a suitable candidate to further study the relationship of protein secretion and toxicity in B. cereus.

#### Enterotoxin Production and Toxicity

Enterotoxin titers generally confirmed the previous classification of high, medium and low toxin producing strains. Comparison of F3162/04 (D8) with the other strains must be done with caution, as NheB titers were obtained in indirect EIA using only mab 1E11. In preliminary experiments, this strain showed low NheB production but in comparison to that unusually high cytotoxicity (**Table 1**, Table S1). Sequence analyses revealed a point mutation in the nheB gene resulting in the exchange of amino acid residue 151 from glutamic to aspartic acid. This residue is situated in direct proximity to the predicted epitope of mab 2B11, which is used in sandwich EIAs for NheB detection (Didier et al., 2012). This mutation might contribute to a loss of 2B11 binding capacity and thus, detection of the NheB protein, but without affecting its functionality (A. Didier, in preparation). During selection of our strain set, three strains (F3162/04 (D8), MHI 124 and HWW 274- 1) out of 136 showed this phenomenon, implying that the toxic potential of 2.2% of all enteropathogenic B. cereus isolates might be underestimated when NheB production is investigated with sandwich EIAs based on mab 2B11.

Pearson correlation tests showed significant correlation of cytotoxicity, obtained with WST-1 bioassays on CaCo-2 cells, and NheB titers after 6 h of growth (correlation coefficient 0.61). The same WST test was performed on Vero cells (data not shown), with which NheB titers correlated even better (correlation coefficient 0.84). This confirmed prior observations that cytotoxicity depends significantly on NheB production and that Vero cells are even more sensitive, particularly toward Nhe and generally to B. cereus culture supernatants than CaCo-2 cells (Moravek et al., 2006; Jeßberger et al., 2014). Nevertheless, CaCo-2 cells should be used rather than Vero cells to investigate the cytotoxic activity, as B. cereus enterotoxins operate rather in the human intestine than in kidney. Furthermore, discrepancies between high and low toxic strains are bigger on Vero cells, which might lead to underestimation of putative non-pathogenic strains. Contrary to earlier studies (Jeßberger et al., 2014) no significant correlation between Hbl production and cytotoxicity was found, possibly due to the small number of strains investigated (10 hbl positive strains). We further observed that for uncovering the varying toxic potential of different B. cereus isolates the choice of cytotoxicity assay is very important. While differences between high and low toxic strains of our strain set became obvious in WST-1 bioassays, these differences were no longer detectable when PI influx tests were performed (**Figure 6**). The 10 strains expressing Nhe and Hbl showed comparably quick PI influx and thus pore formation while the 9 strains expressing only Nhe showed only poor PI influx. This supports the assumption that, although high sequence homologies between the Nhe and the Hbl components exist (Granum et al., 1999), the mode of action and the way of pore assembly and formation of these two enterotoxin complexes differs significantly.

#### Conclusion

In summary, our data suggest that the highly variable toxic potential of B. cereus is determined by complex and dynamic regulatory processes involving toxin gene transcription strictly regulated by known regulator proteins as well as strain-specific posttranscriptional or posttranslational toxin modifications affecting mRNA stability, translation initiation or protein durability and, perhaps, resistance to extracellular proteolytic degradation. Furthermore, high and low toxic strains could not be correlated with certain groups such as food or food poisoning isolates or enterotoxin gene profiles. Also, the presence of genes encoding further putative virulence factors such as hemolysin II, sphingomyelinase or the metalloproteases InhA1 and NprA provided no further information to predict high toxicity in B. cereus strains.

Besides these intrinsic bacterial factors, the condition of the host (age, immunity, etc.) as well as the composition of the contaminated food may contribute to the high variability of enterotoxin production (Ceuppens et al., 2013), which makes the toxic potential even more difficult to assess and to predict. In this study, using standard laboratory growth conditions which are routinely applied for the assessment of B. cereus toxicity, no distinct bacterial factors or mechanisms responsible for the highly variable enterotoxin production and thus cytotoxic potential were found. Therefore, the question arises whether enterotoxin production in the intestine is triggered or activated by environmental factors and whether potentially high and low toxic strains respond differentially to such factors. We showed that the pattern of high and low toxic strains was already altered when an alternative target cell line was used in cytotoxicity assays (CaCo-2 instead of Vero). It may be speculated that this pattern is further influenced by growth of bacteria under simulated intestinal instead of standard laboratory conditions or in the presence of colon epithelial cells.

#### References


#### Acknowledgments

This research project was supported by the German Ministry of Economics and Energy (via AiF) and the FEI (Forschungskreis der Ernährungsindustrie e.V., Bonn). Project AiF 17506 N. We thank Scharifa Bornschier, Christine Braig, Romy Wecko, Tatjana Svoboda and Cornelia Berger for excellent technical assistance.

#### Supplementary Material

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2015.00560/abstract


production in emetic Bacillus cereus. Appl. Environ. Microbiol. 71, 105–113. doi: 10.1128/AEM.71.1.105-113.2005


state regulator AbrB, but not by the virulence regulator PlcR. Microbiology 155, 922–931. doi: 10.1099/mic.0.024125-0


apoptosis of macrophages. Cell. Microbiol. 13, 92–108. doi: 10.1111/j.1462- 5822.2010.01522.x


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2015 Jeßberger, Krey, Rademacher, Böhm, Mohr, Ehling-Schulz, Scherer and Märtlbauer. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.