# BIOACTIVE COMPOUNDS FROM MICROBES

EDITED BY: Roberto Mazzoli, Katharina Riedel and Enrica Pessione 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 978-2-88945-185-2 DOI 10.3389/978-2-88945-185-2

### 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

# **BIOACTIVE COMPOUNDS FROM MICROBES**

Topic Editors: **Roberto Mazzoli,** University of Turin, Italy **Katharina Riedel,** University of Greifswald, Germany **Enrica Pessione,** University of Turin, Italy

Schematic representation of the interactions established between the intestinal microbiota and the host immune system. General overview of the epithelium (enterocytes; M cells; Goblet cells; antigen presentation cells; dendritic cells and macrophages) in contact with multiple species of microorganisms that constitute the intestinal microbiota; defensins, bacteriocins, and secreted IgA also play an important role in controlling the levels of the different populations of microorganisms.

Image modified from: Hevia A, Delgado S, Sánchez B and Margolles A (2015) Molecular Players Involved in the Interaction Between Beneficial Bacteria and the Immune System. Front. Microbiol. 6:1285. doi: 10.3389/fmicb.2015.01285

Microorganisms have had a long and surprising history. They were "invisible" until invention of microscope in the 17th century. Until that date, although they were extensively (but inconsciously) employed in food preservation, beer and wine fermentation, cheese, vinegar, yogurt and bread making, as well as being the causative agents of infectious diseases, they were considered as "not-existing". The work of Pasteur in the middle of the 19th century revealed several biological activities performed by microorganisms including fermentations and pathogenicity. Due to the urgent issue to treat infectious diseases (the main cause of death at those times) the "positive potential" of the microbial world has been neglected for about one century. Once the fight against the "evil" strains was fulfilled also thanks to the antibiotics, industry began to appreciate bacteria's beneficial characteristics and exploit selected strains as starters for both food fermentations and aroma, enzyme and texturing agent production. However, it was only at the end of the 20th century that the probiotic potential of some bacteria such as lactic acid bacteria and bifidobacteria was fully recognized. Very recently, apart from the probiotic activity of *in toto* bacteria, attention has begun to be directed to the chemical mediators of the probiotic effect. Thanks also to the improvement of techniques such as transcriptomics, proteomics and metabolomics, several bioactive compounds are continuously being discovered. Bioactive molecules produced by bacteria, yeasts and virus-infected cells proved to be important for improving or impairing human health. The most important result of last years' research concerns the discovery that a very complex network of signals allows communication between organisms (from intra-species interactions to inter-kingdom signaling). Based on these findings a completely new approach has arisen: the system biology standpoind. Actually, the different organisms colonizing a certain environmental niche are not merely interacting with each other as individuals but should be considered as a whole complex ecosystem continuously exchanging information at the molecular level. In this context, this topic issue explores both antagonistic compounds (i.e. antibiotics) and "multiple function" cooperative molecules improving the physiological status of both stimulators and targets of this network. From the applicative viewpoint, these molecules could be hopefully exploited to develop new pharmaceuticals and/or nutraceuticals for improving human health.

**Citation:** Mazzoli, R., Riedel, K., Pessione, E., eds. (2017). Bioactive Compounds from Microbes. Lausanne: Frontiers Media. doi: 10.3389/978-2-88945-185-2

# Table of Contents


### **Chapter 4: Antibiotic compounds**

*112 The broad-spectrum antibiotic, zeamine, kills the nematode worm* **Caenorhabditis elegans**

Josephine E. E. U. Hellberg, Miguel A. Matilla and George P. C. Salmond


# Editorial: Bioactive Compounds from Microbes

#### Roberto Mazzoli <sup>1</sup> , Katharina Riedel <sup>2</sup> and Enrica Pessione<sup>1</sup> \*

*<sup>1</sup> Laboratory of Microbial and Applied Biochemistry and Proteomics, Department of Life Sciences and Systems Biology, University of Turin, Turin, Italy, <sup>2</sup> Department for Microbial Physiology and Molecular Biology, Institute of Microbiology, University of Greifswald, Greifswald, Germany*

Keywords: gut-brain axis, antitumor activity, viral immune-escape, Bacteriocins, human-microbes cross talk

#### **Editorial on the Research Topic**

#### **Bioactive Compounds from Microbes**

Microorganisms are ubiquitous and essentially interact with all the other organisms present in the biosphere, sometimes creating a network of signals that constitutes the basis for life on the Earth. Clarifying the nature of these molecular signals, their targets and the pathways underlying their production, constitutes the essential pre-requisite for deciphering inter-kingdom communications, adaptive responses and systems biology (Nicholson et al., 2012; O'Mahony et al., 2015; Cani and Knauf, 2016). The insights acquired in the last two decades about human microbiota, and its fundamental role in maintaining a healthy physiological status, have opened the way to understanding the complex reciprocal talk between bacteria and humans (Hughes and Sperandio, 2008; Lyte and Freestone, 2010; Mayer, 2011; Cryan and Dinan, 2012; Mazzoli, 2014; Halang et al., 2015; Kelly et al., 2015). In parallel with these aspects, new bioactive molecules from the microbial world that interact with different cellular models continue to be discovered.

The times and the technological advances start now to be suitable for intercellular/interorganism communication to be studied at the molecular level. The aim of the present topic issue is to try to describing the mediator molecules of a network of signals which is still largely underexplored and underexploited.

As an example, some soil bacteria (such as Serratia spp.) can have antagonistic actions toward worms and the molecule involved, zeamine, is effective against yeasts and other biological systems, as well. Prodigiosin, the well-known pigment produced by the marine bacterium Vibrio ruber, has a broad antimicrobial spectrum and induces autolytic activity in the target cells (i.e., Bacillus subtilis). Lantibiotics are class I bacteriocins produced by Gram-positive bacteria that can be bioengineered to both enhance their effectiveness against a larger number of bacterial strains and to improve their stability during the gastric transit that is by rendering them protease-resistant. These overall data open new possibilities for antibiotic therapy in a period in which the phenomenon of antibiotic resistance is considered as a major threat to public health (World Health Organization, 2014) since it is widespread in pathogenic, commensal, and food bacteria (Laxminarayan et al., 2013). Furthermore, the appearance of multiresistant bacterial strains (the so-called superbugs), often causing death, clearly constitutes a serious problem to be solved. Exploiting the microbial world and its huge potential in finding new antimicrobial drugs is an urgent concern and some chapters of this topic issue deal with these aspects.

Other interesting molecules are produced by cytomegalovirus-infected cells: these compounds of viral origin (essentially proteins) can promote virus dissemination, persistence, and pathogenesis by counteracting host innate and adaptive immune responses. Conversely, some beneficial microbes, like Lactic Acid Bacteria (LAB) and Bifidobacteria can modulate the immune system controlling inflammation by means of proteins and non-proteinaceous compounds (Pessione, 2012).

#### Edited by:

*George Tsiamis, University of Patras, Greece*

#### Reviewed by:

*George Tsiamis, University of Patras, Greece Spyridon Ntougias, Democritus University of Thrace, Greece*

#### \*Correspondence:

*Enrica Pessione enrica.pessione@unito.it*

#### Specialty section:

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

Received: *20 December 2016* Accepted: *24 February 2017* Published: *08 March 2017*

#### Citation:

*Mazzoli R, Riedel K and Pessione E (2017) Editorial: Bioactive Compounds from Microbes. Front. Microbiol. 8:392. doi: 10.3389/fmicb.2017.00392*

Several microbiota-derived compounds can contribute to control host physiological and pathological states: Metabolomic profiling of gut bacteria can allow to decipher several molecules (among which short chain fatty acids, for instance those produced by bacteria belonging to the clostridial cluster IV and XIV, vitamins and aromatic compounds) controlling cholesterol synthesis, obesity, cardiovascular diseases, metabolic syndrome. Curiously, some bioactive peptides or amino acids can be delivered by food in which LAB are present as fermentation starters. These bacteria can decrypt peptides from food proteins, which have anti-oxidant or metal chelating function, immunomodulatory properties as well as peptides controlling hyperglycemia, hypercholesterolemia, hypertension, cell cycle, apoptosis, and even having a refolding action on damaged proteins. An interesting tannin-degrading activity of bacteria (e.g., Streptococcus spp. and Fusobacterium spp.) can generate gallic acid and pyrogallol, both having an anti-carcinogenic role. Finally, microbial-derived amines can modulate a series of patho/physiological functions such as allergy, smooth muscle relaxation, anxiety, appetite, depression (Pessione et al., 2005). LAB are good producers of gamma-amino butyric acid (GABA)

### REFERENCES


(Laroute et al., 2016) that contributes to gut-to-brain signaling through different pathways involving enteric neurons, enteroendocrine cells and immune cells. Yeasts, mainly Saccharomyces cerevisiae, can convert tryptophan into melatonin and serotonin that are informational molecules related to circadian rhythm but also promising agents for the prevention of neurodegenerative diseases.

This intense network of molecules, allowing communication among bacteria, viruses, and eukaryotic cells has evolved to guarantee optimal life in different ecological niches to each component of the ecosystem and is based upon effector-receptor model. In-depth knowledge of all these biochemical signals, as well as the underlying finalism are still far to be fully elucidated. Nevertheless, some mechanistic aspects highlighted in the present topic issue can open new perspectives in medicine but can also shed light on evolution strategies.

### AUTHOR CONTRIBUTIONS

EP wrote the manuscript which was then reviewed by RM and KR.


**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 Mazzoli, Riedel and Pessione. 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.

## Bioactive Molecules Released in Food by Lactic Acid Bacteria: Encrypted Peptides and Biogenic Amines

#### Enrica Pessione\* and Simona Cirrincione

Laboratory of Biochemistry, Proteomics and Metabolic Engineering of Prokaryotes, Department of Life Sciences and Systems Biology, University of Torino, Torino, Italy

#### Edited by:

Steve Lindemann, Pacific Northwest National Laboratory, USA

#### Reviewed by:

Giulia Tabanelli, CIRI Agroalimentare, Italy Giovanna Suzzi, Università degli Studi di Teramo, Italy

> \*Correspondence: Enrica Pessione enrica.pessione@unito.it

#### Specialty section:

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

Received: 09 February 2016 Accepted: 24 May 2016 Published: 09 June 2016

#### Citation:

Pessione E and Cirrincione S (2016) Bioactive Molecules Released in Food by Lactic Acid Bacteria: Encrypted Peptides and Biogenic Amines. Front. Microbiol. 7:876. doi: 10.3389/fmicb.2016.00876 Lactic acid bacteria (LAB) can produce a huge amount of bioactive compounds. Since their elective habitat is food, especially dairy but also vegetal food, it is frequent to find bioactive molecules in fermented products. Sometimes these compounds can have adverse effects on human health such as biogenic amines (tyramine and histamine), causing allergies, hypertensive crises, and headache. However, some LAB products also display benefits for the consumers. In the present review article, the main nitrogen compounds produced by LAB are considered. Besides biogenic amines derived from the amino acids tyrosine, histidine, phenylalanine, lysine, ornithine, and glutamate by decarboxylation, interesting peptides can be decrypted by the proteolytic activity of LAB. LAB proteolytic system is very efficient in releasing encrypted molecules from several proteins present in different food matrices. Alpha and beta-caseins, albumin and globulin from milk and dairy products, rubisco from spinach, beta-conglycinin from soy and gluten from cereals constitute a good source of important bioactive compounds. These encrypted peptides are able to control nutrition (mineral absorption and oxidative stress protection), metabolism (blood glucose and cholesterol lowering) cardiovascular function (antithrombotic and hypotensive action), infection (microbial inhibition and immunomodulation) and gut-brain axis (opioids and anti-opioids controlling mood and food intake). Very recent results underline the role of food-encrypted peptides in protein folding (chaperone-like molecules) as well as in cell cycle and apoptosis control, suggesting new and positive aspects of fermented food, still unexplored. In this context, the detailed (transcriptomic, proteomic, and metabolomic) characterization of LAB of food interest (as starters, biocontrol agents, nutraceuticals, and probiotics) can supply a solid evidence-based science to support beneficial effects and it is a promising approach as well to obtain functional food. The detailed knowledge of the modulation of human physiology, exploiting the health-promoting properties of fermented food, is an open field of investigation that will constitute the next challenge.

Keywords: opioid, antioxidant, chaperon-like, cell-cycle control, antimicrobial peptides, beta-phenylethylamine, tyramine

### INTRODUCTION

fmicb-07-00876 June 7, 2016 Time: 13:41 # 2

Human health is the result of a correct physiological status often resulting from the reciprocal interaction of gene-derived signals (genetics) and environment–generated information (epigenetics). Recognizing gene signals is relatively easy whereas environmental stimuli are often multiple, complex, and reciprocally interacting. Temperature, pH, redox balance, sleep, diet, drugs, and psychological status can deeply affect gene expression, metabolic pathways, and homeostasis (Carey, 2012). However, human genes and environment are not the only players: microorganisms can take part to the complex molecular cross-talk existing between external world and "self."

Firstly, endogenous symbiont microorganisms, the so-called microbiota, can modulate gene expression, induce preferential food intake (Aydin, 2007), influence pH, redox balance and the ratio between pro-inflammatory and anti-inflammatory cytokines (Belkaid and Hand, 2014). Briefly, control brain, metabolism, immune system, and several homeostatic routes.

Secondarily, food-derived bacteria and yeasts can exogenously affect nutritional parameters, metabolism, oxidative status, immunity, blood pressure, appetite, behavior, also controlling the endogenous microbiota (Arora et al., 2013) by altering the ratio among saccharolytic and proteolytic species, by modulating symbionts gene expression and several other functions (O'Flaerty, 2014). Hence, all fermented food, containing living organisms, can contribute to modulation of the host physiological balance and it constitutes an opportunity to enrich the diet with new bioactive molecules finally resulting in phenotypic effects (appetite, cholesterol and blood pressure lowering, improvement of mood, antioxidant, and immune defenses) on humans (Pessione, 2010). Actually, all types of modulations occur via a complex network of signals, among which proteinaceous compounds play a crucial role.

Microorganisms are able to synthesize a large number of metabolites with assessed beneficial or detrimental properties for human health. Among these, nitrogen-bearing molecules such as amino acids, amino acid derivatives and oligopeptides have received great attention since they can affect human physiology in multiple ways.

As an example, amino acid derivatives such as selenocysteines and selenomethionines, have recently been reported to be biosynthesized in both Lactobacilli (Lamberti et al., 2011) and yeasts (Porto et al., 2015). Although selenoaminoacids are not true "bioactive compounds," directly stimulating receptors on human cells, they can trigger effects deeply affecting human health. The bioactive role of seleno-fixing microorganisms lies in the fact that diet-derived inorganic selenium is toxic (selenate and selenite) or poorly active (elemental selenium) whereas fixed selenium forms (selenomethionines and selenocysteines) are the only bioavailable for humans. On the other hand, only bacteria and yeasts can produce seleno-amino acids from inorganic selenium. Once properly inserted into selenoproteins (i.e., glutathione peroxidase), they can counteract oxidative stress. Besides this well-known antioxidant function, there are data indicating that selenoproteins can modulate immune system (Huang et al., 2012) and activating anabolic circuits such as thyroid hormone biosynthesis (Mullur et al., 2014). Furthermore, epidemiological studies show an inverse correlation between selenium level in blood and cancer mortality, and laboratory experiments have shown a selenium protective effect against cancer initiation and development (Gromadzinska et al., 2008 ´ ). In Lactobacillus reuteri exoproteome studies have demonstrated that two secreted proteins (GAPDH and Phosphoketolase) contain selenocysteines opening the way to employ this strain to supply organic bioavailable forms of selenium (Galano et al., 2013; Mangiapane et al., 2014a,b).

Among amino acid derivatives found in food, biogenic amines are worth of a special mention. Such compounds, although sometimes naturally present (especially in vegetal food) are often the result of the bacterial decarboxylative activity on free amino acids in food. Biogenic amines can be present in non-fermented food, like fish, due to spoilage bacteria that during protein putrefaction release free amino acids undergoing decarboxylation. E. coli can produce cadaverine from lysine and putrescine from ornithine (Applebaum et al., 1975). Proteus can produce putrescine from ornithine as a communication signal (Visick and Fuqua, 2005). However, also not-spoiled food, such as fermented food, can present the risk of biogenic amine accumulation. Although starters, exogenously added to perform controlled fermentations, are accurately typed to avoid any risk, autochthonous or contaminant lactic acid bacteria (LAB) can contribute to amine release. LAB are strong amine producers since they use this metabolic pathway (at the place of respiration) to both create a proton gradient and hence energy (for exhaustive review, see Pessione et al., 2010) and to alkalinize the environment, very acidic since their main fermentation products are acids (lactic acid for homofermenter LAB and lactic + formic + acetic acid in heterofermenters).

Many experimental evidences demonstrate that some LAB strains also produce anti-hypertensive, anti-thrombotic, cholesterol-lowering, metal-chelating, antimicrobial, antioxidant, immune-modulating, chaperone-like and opioid/opioid antagonist peptides from food proteins (Pessione, 2012), and they can modulate the concentration of opioid and cannabinoid receptors in the gut epithelium (Hayes et al., 2007b). The next sections will focus on some of the referred compounds and will illustrate the main effects exerted on human health.

### ENCRYPTED BIOACTIVE PEPTIDES

Several bioactive peptides, lacking activity when proteinencrypted but acquiring their biological effects when proteolytically released, have been shown to have healthpromoting properties as anti-microbials, hypocholesterolemics, opioid and opioid antagonists, angiotensin-converting enzyme inhibitors, anti-thrombotics, immunomodulators, cytomodulators, and anti-oxidants (Hayes et al., 2007b).

The human pool of digestive proteases and peptidases can liberate food-encrypted bioactive peptides that can be absorbed by the gut and then reach peripheral organs. However, the enzymatic activity of LAB largely contribute to their release, either into the food matrix (starter or autochthonous LAB) or in

the gut (endogenous microbiota or probiotics). LAB are ancient organisms adapted to an anoxic environment that never evolved the capability to biosynthesize heme and hence to have functional cytochromes, catalases, and peroxidases. They are very sensitive to oxygen and devote most of their genes to oxygen stress counteracting. Due to the limited length of the overall genome, the biosynthetic abilities of LAB are very limited especially in amino acid synthesis (Pessione, 2012). Therefore, LAB evolved a complex and sophisticated proteolytic system allowing them to get amino acids from the proteins present in the external environment. A schematic representation of this system, which includes proteases, peptidases and membrane transporters, is referred in **Figure 1**.

### LAB Proteolytic System

Extracellular protein hydrolysis into various long oligopeptides is initiated by a cell-envelope proteinase (CEP). Oligopeptides generated by this first cleavage are subsequently taken up by cells via specific transport systems and they undergo further degradation into shorter peptides (bioactive or possible precursors of bioactive compounds) and amino acids in the cytoplasm (Savijoki et al., 2006). In LAB the oligopeptide transporter system (Opp) is the main transporter, belonging to a superfamily of highly conserved ATP-binding cassette transporters (Doeven et al., 2005). The Opp system of L. lactis transports peptides up to at least 18 residues (Savijoki et al., 2006). It should be underlined that most bioactive peptides are released into the external environment only when cells undergo autolysis, because only occasionally the longer peptides (originated by the first hydrolytic step) possess biological activity (Meisel and Bockelmann, 1999). Nevertheless, some authors reported that CEP from Lactobacillus delbrueckii subsp. lactis are suitable to liberate both metal-chelating and anti-hypertensive peptides directly from caseins and without necessity of prior cell autolysis. Maximum CEP activity was a little bit decreased by the addition of NaCl (1%) and glycerol (5%; Hébert et al., 2008). However, proteolytic enzymes released by LAB proved to be very different in the different LAB species and strains, giving rise to a different pool of bioactive peptides (Hébert et al., 1999).

### Food Matrices Containing Encrypted Peptides

Lactic acid bacteria proteolytic system is suitable to produce bioactive peptides from several food proteins, especially caseins that constitute the main nitrogen substrate present in their habitat (milk and milk derivatives; Clare and Swaisgood, 2000). Casein consists of four main proteins (whose ratio is about 38:11:38:13): alpha s1 casein, alpha s2 casein, beta-casein and k-casein, differing in amino acid sequence, hydropathicity index, glycosylation and phosphorylation degree (Pessione, 2012). Usually, CEPs have a strong preference for hydrophobic caseins and Lactococcus CEPs have been classified into several types and subtypes depending on their substrate specificity for αS1-, β-, and κ-caseins (Kunji et al., 1996). Casein hydrolysis can give rise mainly to opioid/anti-opioid peptides, antithrombotic and antihypertensive, immunomodulatory, mineral-binding and antimicrobial peptides (Thakur et al., 2012) easily detectable by means of chromatographic techniques after microbial food digestion (Saraswat et al., 2012).

Hydrolytic cleavage of milk whey proteins (alphalacto-albumine, beta-lacto-globulin, lactoferrin, and immunoglobulins) can also generate bioactive molecules, such as hypocholesterolemic peptides, which can decrease absorption and enhance fecal extrusion of cholesterol. Actually, milk whey proteins display a greater cholesterol-lowering activity than casein, in particular beta-lactoglobulin (Nagaoka et al., 2001). Beta-lactoglobulin-released peptides also exhibit antioxidant (Hernández-Ledesma et al., 2005) and immune-modulating (Prioult et al., 2004) activity.

Bio-active molecules, especially ACE-inhibiting peptides, can also be found encrypted in bovine meat proteins, such as hemoglobin (i.e., hemorphins) and serum albumin (i.e., serorphin), collagen, elastine, fibrinogen (Minkiewicz et al., 2011; Lafarga et al., 2015).

However, also food of vegetal origin can be a source of bioactive peptides. In soy β-conglycinin β-subunit (Kaneko et al., 2010) and in the large subunit of spinach rubisco, both antihypertensive (Yang Y. et al., 2003) and opioid peptides (i.e., soymorphins-5,-6, and -7, and rubiscolin-5 and -6, respectively), displaying anxiolytic effects, food intake controlling action and enhancement of memory, have been described (Yang et al., 2001; Yang S. et al., 2003; Hirata et al., 2007).

A special mention needs gluten: opioid peptides (i.e., gluten exorphins A4, A5, B4, B5, and C), have been detected and characterized in wheat gluten since 1992 (Fukudome and Yoshikawa, 1992). Some of these peptides have analgesic action on the CNS (Takahashi et al., 2000) but some can also act on opioid receptors located outside the blood–brain barrier, inducing prolactin secretion (Fanciulli et al., 2005). Both antihypertensive (Rizzello et al., 2008) and antioxidant (Coda et al., 2012) peptides are the result of LAB activity on wheat and other cereals. Generally, the high proline content of gluten alpha gliadins prevent hydrolysis by enzymes of the gastrointestinal tract, whereas LAB possess proline-specific peptidase systems. Alternatively, peptide degradation is achieved by the combined action of peptidase from different LAB strains simultaneously present in fermented food (Gerez et al., 2008).

To date, LAB strains able to release bioactive peptides from food proteins, with particular reference to milk caseins, include L. helveticus CP790, L. rhamnosus GG, L. bulgaricus SS1, and L. lactis subsp. cremoris FT4 (Gobbetti et al., 2002). The hydrolytic ability is related to both the protein substrate, i.e., its amino acid sequence and the proteolytic enzyme panel of each microbial strain (Griffiths and Tellez, 2013). It is possible to enhance (100-fold) the proteolytic potential of LAB by growing them in milk (Wakai and Yamamoto, 2012). Actually it has been demonstrated that peptide-rich media like MRS supplying a ready-to-eat nitrogen source are unfavorable to the induction of most proteolytic enzymes for instance those encoded by Pep N and PepX (Smeianov et al., 2007). On the other hand, CEP are able to hydrolyze caseins only after growth on skim milk but not after MRS pre-culture (Jensen et al., 2009). Generally, lactobacilli genome encodes a larger number of proteases, peptidases, amino

acid permeases, and Opp transport systems than lactococci (Klaenhammer et al., 2005). However, more studies on a larger number of species and strains are necessary to establish the true potential of each strain.

Finally, it is worth considering that most information on LAB proteolytic system has been referred to food matrix isolated microflora and to its action on bovine milk casein (Korhonen and Pihlanto, 2003). However, endogenous or probiotic or food – derived LAB can also produce bioactive peptides from the different protein pool present in the host gut, opening interesting perspectives for improving human health (Sharon et al., 2014). In the following paragraphs, the best known bioactive molecules generated from LAB proteases and peptidases are described.

### Antimicrobial Peptides

Lactic acid bacteria proteolytic activity on caseins gives rise to small peptides displaying antimicrobial (both bactericidal and bacteriostatic) activity. Alpha s1 casein and K-casein hydrolysis gives rise to isracidin and k-casecidin, respectively (McCann et al., 2006). Both peptides display inhibitory action on S. aureus growth (Meisel and Bockelmann, 1999; Atanasova et al., 2014). K-casein can also originate kappacin, a long bioactive peptide active against Streptococcus mutans. This peptide, although deriving from K-casein, is phosphorylated but not glycosylated, and displays the ability to prevent bacterial adhesion to gingival mucosa and to bind enterotoxins (Malkoski et al., 2001). The alpha-s-2 casein fragment 165-203, called casocidin-I can inhibit both S. carnosus and E. coli (Zucht et al., 1995). Lactobacillus helveticus can hydrolyze beta casein, by means of a PR4 proteinase, decrypting an antimicrobial peptide active toward several strains of Gram-positive bacteria (including S. aureus and Listeria innocua) but also against Gram-negative pathogens such as E. coli, Salmonella, Yersinia, (Minervini et al., 2003). Homology studies revealed that this peptide has a similar length to isracidin, but a stronger hydrophobic nature (lacking positive charges) allowing a better activity on Gram-negative bacteria. Furthermore, it bears some proline residues near the C-terminal, which can prolong its half-life rendering this molecule more resistant to the peptidolytic action (Gobbetti et al., 2004). The critical question concerning this peptide is that it has been decrypted from human milk but not from other caseins present in food (cow, buffalo, goat, ship).

Different peptides produced from caseins by different Lactobacillus species (L. acidophilus, L. helveticus, L. plantarum, and L. rhamnosus) and by Lactococcus lactis are active against Gram-negative rods such as Enterobacter sakazakii (Hayes et al., 2006). An interesting antimicrobial peptide, named Lactoferricin B, having broad spectrum of action including Gram-negative and Gram-positive bacteria, yeasts and filamentous fungi, has been described as the result of the hydrolytic action on bovine lactoferrin (Dionysius and Milne, 1997; Kitts and Weiler, 2003). This cationic peptide alters membrane permeability with consequent dissipation of the proton gradient and has therefore bactericidal action (FitzGerald and Murray, 2006). It was cloned and expressed in E. coli, showing activity against Klebsiella pneumoniae, Streptococcus Mutans, and S. aureus (Luo et al., 2007). Lactoferricin B is especially active toward the enterohemorrhagic strain of E. coli 0157h:7 (Muro-Urista

et al., 2011) and proved to be active also against winespoilage LAB such as Oenococcus oeni, Pediococcus damnosus, and Lactobacillus brevis (Enrique et al., 2009). It is difficult to establish the real LAB potential in hydrolyzing lactoferrin. Some experiments were performed by using both microbialderived enzymes and fermentation with proteolytic starter cultures (Sharma et al., 2011), however, the true potential of each strain has to be experimentally proved since a great variability was observed among strains (Korhonen, 2009). More recently, the proteolytic activity of different LAB strains on goat milk beta casein and beta-lactoglobulin were tested. Lactococcus lactis l598, Lactobacillus lactis 1043, Streptococcus thermophilus t3D1, Dt1and Lb. delbrueckii subsp. bulgaricus b38, b122 and b24 revealed a significant potential in producing antimicrobial peptides inhibiting S. aureus, L. monocytogenes, Listeria innocua, Enterobacter aerogenes, and Salmonella enteritidis (Atanasova et al., 2014).

Concerning the mechanism of action of such antibacterial compounds, several models have been proposed. Being amphipathic molecules, bearing a hydrophobic moiety and a strongly positively charged domain, this peptides can firstly bind to teichoic acids in Gram-positive-, or to LPS in Gramnegative bacteria (Vorland, 1999) and then interact with the negatively charged bacterial membranes. Most of them act as proton gradient perturbing molecules, (causing membrane depolarization like polymyxin B and colistin; Lohner and Blondelle, 2005), however, some authors suggest that also removal/chelating of membrane-bound iron can affect bacterial viability (Sharma et al., 2011). For what concern lactoferrinderived peptides, however, it has long been established (Bellamy et al., 1992) that the domains of lactoferrin involved in the antibacterial activity are different from those involved in iron binding. Complex mechanism, including inhibition of the synthesis of macromolecules (Ulvatne et al., 2004) as well as synergic action with host innate immunity compounds were also reported (Brogden, 2005).

These antimicrobial compounds are appreciated in the food industry (especially dairy industry) as natural preservatives counteracting undesired contamination. This allows to reduce the amount of sugar and salt with excellent benefits for diabetics, obese, and hypertensive subjects. Furthermore, their stability in blood and serum render them promising infection control agents. Some of them are also good candidates to be tested for their antiviral potential. Bojsen et al. (2007) reported the antiviral effect of bovine whey proteins, whereas Civra et al. (2015) very recently identified from donkey milk lactadherin a peptide displaying anti-Rotavirus activity. The actual role of LAB proteolysis in the decrypting of the antiviral potential is far to be fully elucidated. An overview of antibacterial peptides is proposed in **Table 1**.

### Metal-Binding Peptides

Some LAB strains digest casein, releasing casein phosphopeptides (CPP) phosphorylated on Ser residues. CPP have a strong anionic character and hence are very resistant to further proteolytic degradation (Sharma et al., 2011). They have been identified during cheese ripening due to microbial protease activity (Singh et al., 1997). The typical cluster is Ser(P)-Ser(P)Ser(P)-Glu-Glu, but they may contain also phosphorylated cysteines. Even if this motif is crucial for metal-binding, various phosphopeptide fractions revealed significant differences in this capability, which may be due to variant amino acid composition around the phosphorylated region. Hence, other factors can affect CPP-metal interaction, namely the total number of amino acids and the total negative charge (Cross et al., 2005).

It has long been established (Gagnaire et al., 1996) that beta and k-caseins (being poor in hydroxylated amino acids and therefore less suitable for phosphorylation) generate peptides less active in metal binding, whereas the Ser/Thr rich alpha s1 and alpha s2 caseins are the most suitable for chelating cations by means of CPP. However, alpha s1 casein-derived peptide (59–79), although bearing five phosphate groups, is less active in allowing mineral uptake by the human tumor cells HT29 than beta casein-derived peptides (1–25). Actually, besides the typical cluster sequence, specific secondary structure motifs in the bioactive peptide (such as beta-turn and loop), as well as the degree of aggregation of CPP in presence of divalent cations, are required for a correct mineral absorption (Ferraretto et al., 2003).

Casein phosphopeptides can chelate up to 250 mgCa/g with a dissociation constant in the mM range, with an affinity even higher for trace elements like zinc, iron, and copper (FitzGerald, 1998). Their mechanism of action rely on the fact that they can form soluble complexes with calcium even at alkaline pH (Berrocal et al., 1989). This results in increased absorption of calcium in the intestine, useful in the treatment of osteoporosis. Re-calcification of dental enamel and prevention of dental caries were also reported (Reynolds et al., 2008). Enhanced gut bioavailability and potential higher absorption of enzyme co-factors such as iron, zinc, copper, selenium, magnesium, and manganese, due to CPP was also described (Meisel and Olieman, 1998). On the other hand, controversial clinical results concerning both calcium (Korhonen and Pihlanto, 2006) zinc (Miquel and Farré, 2007) and calcium and iron absorption obtained on patients receiving CPP supplementation and controls, underline the difficulty to standardize the different supplements (Teucher et al., 2006). Finally, calcium availability is strongly dependent on meal composition (i.e., phytates) and complex interactions between different foods ingested.

In spite of all these considerations, all described peptides may constitute interesting tools for microelements nutritional improvement. In our aging western society bone disease is the third cause of nursery care and osteoporosis is not fully controlled by a healthy lifestyle (active movement and diet; Wei et al., 2003). Therefore, calcium-enriching supplements such as CPP open new nutraceutical perspectives. However, the need to use purified and well-characterized CPP (for having a reproducible nutritional impact) highly affects the costs. Hence, the use of LAB as probiotics or food supplements to obtain functional food, especially milk based drinks and yogurts will be a promising strategy. Lactobacillus helveticus LA, is a LAB adapted to the cheese environment displaying good proteolytic activity. It can decrypt from alpha casein a peptide showing calcium-binding activity (Dimitrov, 2009). Lactobacillus helveticus LBK16H can decrypt the tripeptides Ile-Pro-Pro and

#### TABLE 1 | Main antibacterial peptides from LAB.


ND, not determined.

Val-Pro-Pro, both exerting anabolic effect on bones (Griffiths and Tellez, 2013).

### Antioxidant Peptides

The existence of antioxidant peptides in vegetal food (especially cereals and legumes) is a well-established concept, recently reviewed by Malaguti et al. (2014). Marine food, such as algae (Sheih et al., 2009) and squid (Rajapakse et al., 2005b), as well as unusual proteins like royal jelly-derived (Guo et al., 2009), yak milk casein (Mao et al., 2011), goat milk casein (Li et al., 2013), donkey milk (Piovesana et al., 2015), but also egg white lysozyme (Rao et al., 2012), can be a source of antioxidant peptides. Bacteria have the capability to decrypt antioxidant peptides from several food proteins. Antioxidant activities were described in fermented mussel sauce (Rajapakse et al., 2005a), in LAB-fermented milk whey (Virtanen et al., 2007), in fermented shrimps (Faithong et al., 2010), during sourdough fermentation of wheat, rye, and kamut flours (Coda et al., 2012) and in peptides derived from bovine casein after 24 h hydrolysis with Bifidobacterium longum (Chang et al., 2013). Antioxidant efficacy was also found in cell-free extracts of Lactobacillus plantarum isolated from traditional Chinese food (tofu and kefir) but it is rather related to pseudo catalase enzymes, exopolysaccharides and lipoteichoic acids than to bioactive peptides (Li et al., 2012). Recently, mixed cultures of LAB and yeast proved to be able to produce antioxidant peptides from cow milk, however, the peptide sequences having activity were not described (Li et al., 2015).

Antioxidant peptides can control both enzymatic and nonenzymatic peroxidation of fatty acids thus preventing membrane lipid peroxidation (Hartmann and Meisel, 2007). They can act as direct radical scavengers, non-radical oxygen quenchers or metal chelators, but they also can eliminate radical precursors, hence controlling ROS damage to cells and tissues, resulting in a beneficial action on the whole organism. The redox balance between ROS and antioxidant endogenous defenses is highly impaired during chronic inflammation (Zhang et al., 2015) thus, antioxidants are essential in controlling age-related chronic degenerative diseases concerning both the cardiovascular and the central nervous system.

Natural antioxidant peptides such as GSH (Cys-Glu-Gly) or carnosine (beta-Ala-His) play important roles in radical scavenging (Power et al., 2013). Antioxidant properties have been demonstrated for Tyr, Trp, His, and Lys, in particular when Tyr and Trp are in the C-terminus and the hydrophobic residues Val and Leu at the N-terminus of a peptide. Trp and Tyr behave as antioxidant molecules since their indolic and phenolic groups can act as hydrogen donors (Hernández-Ledesma et al., 2008). Basic amino acids can chelate metallic ions and cysteine acts as a proton donor, thanks to its thiolic group but the highest antioxidant activity was demonstrated for the tripeptide Pro, His, His (Malaguti et al., 2014). Furthermore it was demonstrated that Leu-His-His and Pro-His-His can both act synergically with non-peptide antioxidants such as butylated hydroxyanisole (Kitts and Weiler, 2003).

### Immunomodulatory Peptides

It is well established that some milk proteins (caseins, whey proteins, lactoferrin, lactoperoxidase) are able to control lymphocyte proliferation (Möller et al., 2008). Milk-derived peptides such as the C-terminus sequence of beta casein (193– 209) can increase proliferation of rat lymphocytes (Meisel and Bockelmann, 1999), whereas alpha casein C-terminal hexapeptide (194–199) and beta casein fragments (63–68 and 91–93) induce macrophage maturation and phagocytosis enhancement in vitro (Fiat and Jollès, 1989). In agreement with these data, studies in vivo show that alpha casein-derived peptides can stimulate macrophages phagocytosis exerting a protective effect against Klebsiella pneumoniae infection in mice (Migliore-Samour et al., 1989). Short-chain peptides from milk whey proteins also stimulate the proliferation of murine spleen lymphocytes in vitro (Mercier et al., 2004). More detailed studies report that proliferation of human peripheral blood lymphocytes is induced by the k-casein and alpha lactalbumin fragment Tyr-Gly-Gly (Kayser and Meisel, 1996; Sharma et al., 2011). Natural killer (NK) cell activity and antibody synthesis can also be enhanced by food–derived bioactive peptides (Hartmann and Meisel, 2007). However, milkderived peptides do not act only as simply stimulators of the immune system: immune-modulating activities including cytokines regulation and attenuation of allergic reactions has also been described (Korhonen and Pihlanto, 2003; FitzGerald and Murray, 2006).

Transferring clinical data to food is not so simple. Actually, in this scenario, LAB (especially probiotic LAB) play a very central role: beta-casein medium fermented with LAB gives rise to bioactive peptides acting on monocytes, macrophages and T helper cells, particularly with Th1 like cells (Laffineur et al., 1996). Lactobacillus paracasei hydrolyzes beta-lactoglobulin originating peptides stimulating Interleukin 10 (IL-10) production and downregulation of IL-4 and gamma interferon secretion (Prioult et al., 2004). Casein hydrolysated by Lactobacillus rhamnosus GG displays modulating effects (both stimulation and suppression) on lymphocyte proliferation (Möller et al., 2008). Furthermore, casein hydrolysates from Lactobacillus rhamnosus GG can induce enhancement of anti-inflammatory cytokines from Th1 lymphocytes and a parallel decrease of pro-inflammatory cytokines and immunoglobulins produced by Th2 lymphocytes, thus controlling allergic reactions (Delcenserie et al., 2008). Lactobacillus helveticus proteolysis products down-regulate cytokines production and stimulate macrophage phagocytosis (Matar et al., 2001) and also show protective effects against E. coli O157 (LeBlanc et al., 2004) and Salmonella typhimurium (Tellez et al., 2011) infection. All the non-proteolytic strains of Lactobacillus helveticus failed to trigger immune system modulation. Three immune-active peptides derived both from beta casein and from alpha lactalbumin after milk fermentation with proteolytic strains of Lactobacillus helveticus were analyzed: they contain high percentage of proline histidine and lysine (Tellez et al., 2010). Establishing how, when and in what extent proteolysis occurs in food (and in the gut) is still an open question.

## Cell Cycle and Apoptosis Modulating Peptides

Milk-derived peptides, sometimes resulting from microbial proteolytic activity (Roy et al., 1999), can regulate cell growth, differentiation, and apoptosis (Roy et al., 2002). De Simone et al. (2008) demonstrated that waste milk whey is able to induce apoptosis in CaCo2 cancer cells, but also interfere with the cell cycle (reducing the number of cells in S-phase and enhancing the number of cells in G1 phase), thus inhibiting proliferation. Cooked milk whey loose this property probably due to irreversible denaturation of proteases decrypting the peptides of interest. The endogenous microflora of milk (both starters and contaminating bacteria) seems to be the main responsible of the phenomenon observed, since fresh milk does not exhibit this behavior (De Simone et al., 2008). Although the involved peptides were not characterized, it is very well known that some sequences such as Arg-Gly-Asp-Asp-Asp-Asp-Asp-Asp-Asp-Asp-Asp can have anti-proliferative effects on cancer cells models: the triplet Arg-Gly-Asp can account for adhesion to the extracellular matrix whereas the eight Asp residues can act intracellularly by binding chromatin (Galvez and de Lumen, 1999). Meisel and FitzGerald (2003) also demonstrated that normal cells are less sensitive to the apoptotic induction than malignant cells. Tumor cells apoptosis was also detected in an in vivo model in which mice were treated with milk that underwent fermentation by a highly proteolytic strain of Lactobacillus helveticus. On the contrary, the use of lowproteolytic strains did not gave satisfactory results in terms of breast cancer control in the same mice (De LeBlanc et al., 2005). On the other hand, not all anticancer effects of milk are related to proteolysis. It has been demonstrated that alpha lactalbumin can form a complex with oleic acid in acid conditions. This modified protein can induce apoptosis in cancer cells in vitro (Svensson et al., 2000). Therefore, is it tempting to hypothesize that acidification caused by LAB metabolism can also contribute to these effects, and further experiments proving this will be crucial to confirm such hypothesis and opening new perspectives for cancer control. Taken together all these results suggest that it is possible to increase host defenses against tumor degeneration by using food fermented with LAB.

### Opioid and Anti-opioid Peptides

These molecules constitute a promising frontier for treating both stress-related behaviors such as anxiety and depression and season-related mood disorders. They act on the appetite/satiety as well, as demonstrated by Pfluger et al. (2012) who named these molecules "nutropioids." In general, these peptides are able to control the gut–brain axis at several levels, including gut– brain communication, brain cognitive function, and behavior (Cryan and Dinan, 2012). Their mechanism of action includes stimulation of receptors (K, delta, and mu) present in both the central and the peripheral nervous system and subsequent inhibition on adenylate cyclase activity. Some of these compounds are summarized in **Table 2**.

The best studied myorelaxant peptides are beta-casomorphins (which act on the mu receptors) and alpha s1 casein-derived peptide (which acts on the delta receptor). Their primary

TABLE 2 | Overview of the opioid and anti-opioid peptides.


sequences consist of 4–10 amino acids whose N-terminal residue, (essential to trigger biological activity), is Arg in alpha s1 casein-derived peptide and Tyr in beta-casomorphins (Loukas et al., 1983). This difference, together with the presence of six proline residues in beta-casomorphins, can account for the higher resistance to enzymatic digestion of the latter, resulting in enhanced half-life and absorption in human digestive tract. Once in blood, beta-casomorphins can reach receptors in the brain and in the peripheral tissues thus exerting a relaxing action inducing calmness and sleeping (Chabance et al., 1998). Another interesting peptide derived from alpha s1casein is alpha-casozepine, displaying anxiolytic action but not directly interacting with the mu and delta receptors. It has been demonstrated that the biological effect is mediated by activation of serotonin and GABA<sup>A</sup> receptors, causing release of endogenous serotonin, dopamine, and GABA (Mizushige et al., 2013). Conversely, casoxins peptides (originated from k-casein hydrolysis) behave as opioid-antagonists over both the mu and k-type receptors. Due to this different physiological role, casoxins can be employed to counteract depression (Chiba and Yoshikawa, 1986).

Peptides produced with the contribution of LAB protease system on dairy proteins are named exorphins and casoxins, the former having opioid–like and the latter opioid-antagonist function (Chabance et al., 1998). However, other foods may contain opioid/anti-opioid peptides. It has long been established that wheat gluten is a good source of opioid peptides (Fukudome and Yoshikawa, 1992), and more recently, it has been shown that gluten exorphin B5 can enhance prolactine secretion by acting on receptors present outside the blood brain barrier. In this case, the biological effect is linked to a reduced dopaminergic tone (Fanciulli et al., 2005). Soymorphins are very interesting opioid peptides acting on mu-receptors present in the gut and connected with the serotonin, dopamine, and GABA receptors systems. It was shown that soy-derived opioid peptides act as anorexigenics and also reduce gastrointestinal motility (Kaneko et al., 2010). A delta-opioid peptide called rubiscolin was decrypted from the large subunit of spinach rubisco (Yang et al., 2001). The rubiscolin amino acid sequence is Tyr-Pro-Leu-Asp-Leu-Phe and it displays an analgesic effect besides being involved in memory consolidation (Yang S. et al., 2003). Hirata et al. (2007) also report anxiolytic action of this natural peptide by activation of the sigma1 and D1 dopamine receptors, when delivered orally (100 mg/Kg). The replacement of Leu<sup>3</sup> with Ile or Met enhances by a factor four the opioid activity, whereas the replacement of Phe with Val potentiates the opioid activity more than 10-fold. The overall activity of the sequence Tyr-Pro-Met-Asp-Leu-Val is 20 times higher than the natural rubisco derivative (Yang S. et al., 2003).

### Antithrombotic Peptides

Blood coagulation occurs through a complex system of proteins undergoing a sequential proteolytic cascade. These proteins include pro-enzymes such as fibrinogen that is proteolytically activated to fibrin by means of thrombin. Bovine k-casein derived oligopeptides (fragment 106–116, 11 amino acids long) named casoplatelins (or thrombin inhibitory peptides), can inhibit the binding of human fibrinogen (gamma chain) on the platelet surface receptor, due to sequence homology between fibrinogen gamma chain and K-casein. Actually, the mechanism involved in milk clotting, defined by interaction of k-casein with chymosin bear a remarkable similarity to the process involved in blood clotting, defined by interaction of fibrinogen with thrombin (Qian et al., 1995). As a result, they prevent aggregation of ADP-activated platelets. Also the glycopeptide present on the N-terminus of k-casein, a smaller molecule called casopiastrin, displays inhibiting activity on fibrinogen binding on the platelet membrane (Ricci-Cabello et al., 2012). Although these molecules are mainly the result of human proteolysis on milk, LAB proteolytic activity on casein can contribute to their release both in food and in vivo. Casoplatelins are present in high amounts in the gut but, after absorption, significant concentrations are still bio-available in blood thus contributing to thrombosis control in humans.

### Antihypertensive Peptides

Lactobacillus helveticus, Lactobacillus delbrueckii subsp. Bulgaricus SS1, Lactobacillus delbrueckii subsp. lactis, and Lactococcus lactis subsp. cremoris FT4 can modulate blood pressure by producing angiotensin 1-converting enzyme inhibitory peptides (ACE inhibitors) from milk proteins. ACE is a carboxypeptidase converting angiotensin I (a decapeptide) into angiotensin II (an octapeptide) having a strong vasoconstrictor action (Hartmann and Meisel, 2007). Lactobacillus delbrueckii subsp. lactis hydrolyzes both alpha s1 and beta casein (but not k-casein) by means of a cell envelope proteinase releasing antihypertensive peptides: the proteolytic activity was maximal during the logarithmic growth and addition of NaCl and glycerol prevents correct proteolysis (Hébert et al., 2008). Casein-derived anti-hypertensive peptides are 2–6 amino acids long oligopeptides. Generally, LAB peptidases, by shortening the poly/oligopeptide chain, contribute to enhance the anti ACE potential. Actually the 6 (alpha s1 casokinin-6 = Thre-Thre-Met-Pro-Leu-Trp) and the 5 (alpha s1 casokinin-5 = Phe-Phe-Val-Ala-Pro and betacasokinin = Lys-Val-Leu-Pro-Val) amino acids long oligopeptides are less active than the tripeptides made up of Val-Pro-Pro and Ile-Pro-Pro. A very small dipeptide Tyr-Pro, proved to be effective in blood pressure control as well

(Yamamoto et al., 1999). The final active short peptides are resistant to both pH variations and human digestive enzymes thus opening interesting applicative perspectives (Gobbetti et al., 2004).

### Cholesterol-Lowering Peptides

Hypocholesterolemic peptides can be originated by betalactoglobulin, casein and soy proteins proteolysis (Hori et al., 2001). They can lower the total cholesterol of rats in vivo. Their mechanism of action probably consists in the reduction of the micellar cholesterol solubility or also to an enhanced ability to bind taurocholate (Nagaoka et al., 2002). They overall action prevents cholesterol absorption by CaCo2 cells in vitro and enhances fecal steroid excretion in vivo. The sequences displaying such activities are Ile-Ile-Ala-Glu-Lys, Ala-Leu-Pro-Met-His and Gly-Leu-Asp-Ile-Gln-Lys (Nagaoka et al., 2001).

### Glucose-Uptake Stimulating Peptides

It has been demonstrated that dipeptides containing branched chain amino acids (Ile-Ile, Leu Leu, Ile-Leu, Leu-Ile, Ile-Val, Leu-Val, Val-Leu) can enhance glucose uptake in skeletal muscle cells, favoring glycogen synthesis and thus controlling hyperglycemia (Morifuji et al., 2010). Generally, glucose uptake by skeletal muscle cells is induced by exercize-training by means of increased expression or activity of key-signaling proteins (Goodyear and Kahn, 1998). Branched-chain amino acids are the main nitrogen source for skeletal muscles and milk whey proteins are particularly rich in branched chain amino acids (22.3%) as compared to caseins (20.3%), soy proteins (17.5%), or wheat gluten (14.1%; Morifuji et al., 2010). The dipeptide Ile-Leu is the most abundant after milk whey hydrolysis. Therefore, assessing the proteolytic activity of LAB toward milk whey proteins (that constitute natural nitrogen substrates for LAB), is a promising strategy to get nutritional supplements improving health. Proteomic approaches using LC-MS-MS can be a valuable help for identifying small nutritionally relevant peptides and a new discipline, called food peptidomics, is now expanding (Lahrichi et al., 2013).

### Chaperone-Like Peptides

A very interesting class of protein-encrypted peptides are those displaying chaperone-mimetic action (Bumagina et al., 2009; Artemova et al., 2010). Synthetic peptides were constructed on the model of bovine hemorphin-6, wheat gluten exorphin C, and spinach rubiscolin-5 (all opioid peptides), and their ability in refolding model proteins was investigated. The target proteins to be refolded were heat-treated or DTTaggregated carbonic anhydrase, ADH and bovine milk alpha lactoglobulin. The results clearly indicate that the peptides are able to refold damaged proteins thus opening the way to the interesting hypothesis that food proteins, besides their nutritional role, can act as systems involved in quality control of proteins especially during stress (Artemova et al., 2010).

## Mixed Function Peptides

Some food encrypted peptides, in case decrypted by LAB proteolytic action, possess mixed function. In **Table 3** are reported the peptides displaying two or more biological activities.

As an example the tripeptides Val-Pro-Pro and Ile-Pro-Pro, released from beta and K-casein by Lactobacillus helveticus are immunomodulatory and hypotensive (Hartmann and Meisel, 2007). This effect can be explained by the fact that ACE inhibition favors bradykinin production, and bradykinin plays a crucial role in the inflammation process by stimulating macrophages and increasing lymphokines secretion by lymphocytes (Sharma et al., 2011).

Some immunostimulatory peptides also exhibit antimicrobial features: beta-lactoglobulin-derived peptides can improve phagocytosis but also stimulate the microbial autolytic system. This is not only restricted to bacteria but also to fungi and both naturally autolyzing and non-autolyzing strains are sensitive to this effect (Hernández-Ledesma et al., 2008).

It has long been established that opioid peptides such as casomorphin agonists also decrease cell proliferation by acting on somatostatin receptors (Hatzoglou et al., 1996a,b). Since opioid and somatostatin receptors are present on different cells including central nervous, endocrine and immune system (De Simone et al., 2008) the action of such peptides is more complex than expected.

Several peptides are opioid and hypotensive, especially those derived from bovine milk proteins (casomorphins, lactorphins) and beta-casomorphin 7 is opioid, hypotensive, and immunomodulatory: this last function is targeted on lymphocyte proliferation that is sometimes stimulated and sometimes inhibited depending on the peptide concentration (Meisel and Bockelmann, 1999).

Lactoferricin, the lactoferrin-derived antimicrobial peptide, besides being an antibacterial, antifungal and antiviral molecule, can also control carcinogenesis by means of its anti-inflammatory and immune-modulating properties. This effect seems to be due to a positively charged region of the peptide that is very rich in tryptophan and arginine residues (Sharma et al., 2011)

The reason of these mixed functions lies in the fact that some regions in the primary structure of caseins contain overlapping peptide sequences, exerting different physiological functions (Meisel and Bockelmann, 1999). These domains, highly hydrophobic and rich in proline residues, have been defined as "strategic zones" and are resistant to proteolytic attack (Fiat and Jollès, 1989).

In conclusion, of this paragraph it is important to underline that not all bioactive peptides released by the combined action of LAB proteolytic system and digestive proteases in food and in the gut can be available for humans. Actually, bioactive peptides may be degraded during digestion, may be poorly absorbed and hence reach the target tissues at a concentration lower than the one necessary to exert their biological function. As far as degradation is concerned, supplying bioactive peptides by means of GRAS bacteria (like LAB are) plus protein complements is simpler than encapsulate them and this constitutes therefore a promising strategy to transport the molecules to the intestine, preventing degradation


#### TABLE 3 | Overview of peptides with mixed functions.

in the upper digestive tract. Efforts have been made also to assess the real in vivo absorption of bioactive peptides: blood levels of antihypertensive peptides have been quantified (by LC ESI triple) after ingestion by human volunteers (van Platerink et al., 2006) but also in vitro studies measuring epithelial translocation across CaCo2 monolayers by means of ESI-LC-MS-MS can supply reasonable information about absorption. One challenging aspect of these approaches is the need of improved "omics" methods aimed to identify lowmolecular weight bioactive peptides. MALDI (matrix-assisted laser desorption ionization) TOF-MS has proved to be inadequate because of matrix interference (Saavedra et al., 2013) and matrix-free methods such as NALDI (nanostructure-assistedlaser desorption ionization; Kütt et al., 2011) have been reproposed.

### BIOACTIVE AMINES

A number of microorganisms synthesize, as decarboxylation products of precursor amino acids, biogenic amines. In LAB, expression (by transcriptional induction) and/or activation (by catalytic modulation) of amino acid decarboxylation systems are non-essential adaptive responses to energy depletion but also strategies to counteract acid stress (Pessione, 2012).

However, not all biogenic amines are bioactive molecules since most of them only act as spoilage compounds (putrescine, cadaverine) sometimes enhancing the toxicity of true bioactive molecules (Medina et al., 2003). On the contrary, the decarboxylation of certain amino acids, as tyrosine and histidine, can give rise to bioactive molecules (tyramine and histamine) involved in several pathogenic syndromes, more extensively described in the following paragraphs. On the other hand, not all bioactive compounds negatively affect human health: as an example, beta-phenylethylamine derived from phenylalanine or tryptamine originated from tryptophan can exert beneficial actions (such as mood control and appetite/satiety balance regulation) when administered to humans (Shimazu and Miklya, 2004). Similarly, it has been reported that several LAB (L. bulgaricus, L. acidophilus, L. casei, L. plantarum) can synthesize melatonin, a bioactive molecule, deriving from tryptophan decarboxylation and serotonin metabolism, acting on sleep and reproductive behavior but also controlling immunity, inflammation and carcinogenesis (Tan et al., 2014). A particular case is the glutamate product gamma amino butyrate (GABA): this molecule can be naturally present in food but also artificially added to enhance the nutraceutical value of a certain food (GABA-tea, GABA-rice; Abe et al., 1995; Oh, 2003) due to its relaxing action on muscles and to its overall beneficial effects on the nervous system.

### Bioamines in Food

Cheese is the food most frequently associated with a too high biogenic amine content. It is very rich in free amino acids due to the desired proteolytic action performed by both bacteria and fungi during "maturation," when colonized by decarboxylasepositive LAB (either autochthonous or contaminant bacteria) it can be a source of tyramine, histamine, putrescine, and beta-phenylethylamine (Pessione et al., 2009). Tyramine and beta phenylethylamine are particularly abundant in food from animal origin (including fermented sausages) since these matrices are very rich in the precursor amino acids, tyrosine, and phenylalanine, used by animals to synthesize catecholamines. Both tyramine and beta-phenylethylamine are bioactive compounds whose action will be better described in the following paragraphs.

As far as fermented meat products are concerned, tyramine, cadaverine, putrescine, and histamine can be found (Singh et al., 2012). Poor quality processing favoring contamination is the main cause of a too high bio-amine content of meat, however starter strains possessing the capability to form biogenic amines, like Lactobacillus curvatus, have been described as well (Singh et al., 2012). To prevent such a risk, selection of particular starter cultures possessing either amino oxidase activity (Alvarez and Moreno-Arribas, 2014) or bacteriocin synthesis capability (Somda et al., 2011) should be taken into concern to contain the undesired consequences of spontaneous fermentations.

Other fermented foods are alcoholic beverages such as beer, ciders, and wine, that being of vegetal origin, are generally poor of aromatic amino acid precursors (Lehtonen, 1996). At the end of alcoholic fermentation the only amino acid present in wine in significant amount is arginine (2,4 g/L): this can be converted by the arginine decarboxylase into agmatine and then putrescine, or rather be directed to the arginine deiminase pathway (ADI) generating ammonia, ornithine (generating putrescine) and carbamoyl phosphate (**Figure 2**) (Costantini et al., 2013). Both putrescine and ornithine are not true bioactive compounds even if they can

alter the organoleptic properties of wine and enhance the toxicity of other biogenic amines (Costantini et al., 2013). Carbamoyl phosphate can combine with ethanol in wine generating ethyl carbamate a carcinogenic molecule (Jiao et al., 2014).

However, if the oenological procedures are not correctly performed, other bioactive amines can be generated. Actually, at the end of the alcoholic fermentation when all the grape sugars are exhausted, yeast cells die and undergo autolysis, releasing their own proteins and proteases (from lysosomes). Hence, if dead yeasts cells are not immediately separated from wine, free amino acids (from yeast proteins) become available for autochthonous LAB able to perform malolactic fermentation (MLF; Coton et al., 1999). This secondary fermentation is widely appreciated for its improvement of wine flavor, and it is usually performed by the predominant species Oenococcus oeni or, more occasionally, by other LAB species belonging to the genera Lactobacillus and Pediococcus, which may develop during the winemaking process (Lonvaud-Funel, 1999). Unfortunately, some LAB, besides TPP-dependentacid decarboxylases, also bear PLP-or pyruvoil-dependent amino acid decarboxylases that can generate undesired amines, depending on the precursor amino acid present at this fermentation stage (Recsei and Snell, 1985). Therefore, histamine, cadaverine, tyramine, and beta-phenylethylamine (whose action will be better described in the following paragraphs) can be produced.

The total content of biogenic amines in food varies from 1 to 50 mg/Kg (alcoholic and no-alcoholic beverages, soy products) to 20–1,500 mg/Kg [sausages, cheeses; EFSA Panel on Biological Hazards (BIOHAZ), 2011] At high concentrations, BA are risk factors for food intoxication, while moderate levels may lead to food intolerance (Ladero et al., 2010) It is very difficult to establish a uniform maximum limit for ingested BA since their toxic effect depends on the type of BA in question. The unique BA for which the European Food Safety Authority (EFSA) has legally set maximum limits is histamine and only related to fish and fish products (Linares et al., 2016). Furthermore, the toxic effects exerted by amines possessing biological activity in humans, strongly depend on individual sensitivity to them (Santos, 1996). As an example, tyramine tolerance is lower in patients treated with antidepressant monoamino-oxidase inhibitory drugs (IMAO), which interact with amino-oxidase enzymatic systems (MAO, DAO), since these patients have an impaired ability to oxidize amines. For the same reason, the toxic effects of bioactive amines in wine consumers is enhanced by the presence of ethanol, which is one of the most effective inhibitors of amine oxidases (Sessa et al., 1984).

### Factors Affecting Amines Production in LAB

When considering the risk of bioactive amine release into food, it is important to underline that maximum energy depletion and acidification occurs during the stationary growth phase. If (theoretically) LAB can be removed from food before resting, no amine accumulation takes place. This statement is supported by proteomic investigations that revealed that, in L. hilgardii, histamine production is induced only during stationary phase (Pessione et al., 2005). On the contrary, northern blotting experiments showed activation of the HDC (histidine decarboxylase) gene already during exponential growth phase in a different L. hilgardii strain (Landete et al., 2008). However, the same authors clearly demonstrated an inhibitory effect of both glucose and fructose (still high during exponential growth phase) over the genes coding for HDC in L. hilgardii, O. oeni, and P. parvulus (Landete et al., 2008). Furthermore, in both L. hilgardii and L. brevis, it has been demonstrated that tyramine accumulation is maximal in presence of either fructose or fructose plus L-malic acid (Moreno-Arribas et al., 2003) thus supporting the idea that it is not necessary to reach the stationary phase to trigger amine accumulation. Actually, also in a different bacterial model (E. faecalis) studies involving sub-proteomes profiling demonstrated a high tyramine content already in the exponential phase when sugars are not yet consumed (experiments performed in chemically defined medium having glucose as sugar carbon source; Pessione et al., 2009). Therefore, the evidences are still controversial and in part related to specific amino acid/amine reaction, in part linked to a strain difference.

A further strong evidence is the fact that amine production is related to the availability of the precursor amino acid. Supplements of tyrosine in the medium can increase about 10-fold tyramine accumulation and about twofold tyrosine decarboxylase (TDC) activity in L. brevis (Moreno-Arribas et al., 2000). Similarly, in Enterococcus faecalis, TDC was found to be more abundant in the membrane extracts when the strain was exposed to free tyrosine (Pessione et al., 2009). Histamine is biosynthesized only in presence of histidine: the cytosolic enzyme HDC was found more abundant in cells grown in histidine medium (stimulated condition) in Lactobacillus hilgardii, and Lactobacillus 30a (Pessione et al., 2005; Mazzoli et al., 2009) as compared to control conditions.

The only exception to this rule seems to be the GAD enzyme that can allow GABA accumulation even in the absence of glutamate. However, the chemically defined-medium medium (CDM) used for growing L. lactis contains glutamine, so it is

possible that part of it is converted to glutamate that can undergo decarboxylation to GABA. On the other hand, a transcriptional control exists since addition of glutamate can trigger a light enhance of GABA biosynthesis (Mazzoli et al., 2010).

A critical concern connected with the necessity of free amino acid availability is how bacteria can get them. As far as LAB are concerned, often in the same ecological niche where they live proteins are available (cheese) or microbial autolysis (wine) generates free proteins that can undergo proteolytic breakdown by LAB (Lamberti et al., 2011).

Interestingly, other factors can affect bioactive amine production and, among these, acids (that can undergo or not a similar decarboxylative process). It has been demonstrated that L-lactic acid does not influence both histamine accumulation and HDC biosynthesis (Landete et al., 2008). Similarly, malate has no effect on histamine accumulation (Mangani et al., 2005) or histidine consumption (Mazzoli et al., 2009). In agreement with these data, Moreno-Arribas et al. (2000) showed that tyramine production was not affected by either malate or citrate, whereas Lonvaud-Funel (2001) observed that both tyramine and histamine accumulation was lower during MLF and higher after it. The utilization of the amino acids tyrosine and histidine only after malate does not necessarily indicate malate inhibition, but simply that bacteria prefer utilize carbon sources before consuming nitrogen compounds.

Other factors affecting amine accumulation in food include temperature, spices and salt content, generally controlling microbial growth. The optimum growth for most amineproducing bacteria is 20–37◦C. Below this temperature, the risk to find biogenic amines in food decreases (Karovicová ˇ and Kohajdová, 2005). The sausage diameter as well as the addition of spices can control a too high amine content in meat: the larger the size, the higher the risk of amines due to higher water activity and lower salt content. However, NaCl not only negatively affects bacterial growth, proteolytic activity and histidine decarboxylase activity but also stimulates both tyrosine and glutamate decarboxylase (GAD) synthesis acting as operon inducer (Singh et al., 2012).

### Histamine

The ingestion of histamine-rich food may cause heterogeneous symptoms such as diarrhea, arrhythmia, headache, and allergic syndromes such as respiratory distress (asthma), rhinoconjunctivitis, urticaria, pruritus, and flushing (Manzotti et al., 2016). Histamine is a strong hypotensive compound also acting on many receptors present in the gastro intestinal tract (Santos, 1996). Hence, high content of histamine is responsible of the toxicity of spoiled or not properly stored fish. In black carp fillets, for example, the contents of BAs remained invariable during the early stages of storage and began to increase at the later stages of storage occurs during food storage. In particular, a significantly higher concentration of histamine (132.05 mg/kg on the third day) was detected in the black carp filets stored at 20◦C, indicating that this temperature favors the formation of histamine (Hongbing et al., 2016). In wines, especially red wines undergoing spontaneous MLF by LAB, are frequently responsible of these syndromes. Amino acids generating bioactive amines (histidine, tryptophan, glutamate) can be released during the, post-alcoholic fermentation, by yeasts autolysis as referred above (Alexandre et al., 2001). The concentration of free amino acids in wine is also dependent on the proteolytic activity of the strain concerned with MLF (Alexandre et al., 2001).

### Tyramine

Tyramine is a bioactive molecule acting both at vascular level (hypertension, vasal-constriction) and on the central nervous system (headache). Despite it is generally assumed that histamine is the most toxic BA, a recent work reveals that unexpectedly, tyramine is more cytotoxic than histamine on an in vitro model of the human intestinal epithelium. Furthermore, the concentrations found to be toxic are commonly reached in BArich foods (Linares et al., 2016). Severe syndromes have been described in patients lacking a correct amino-oxidase activity because in treatment with mono-amino-oxidase inhibitory drugs (MAOI). MAOI are used as antidepressant because they prevent the oxidative catabolism of brain-active amines, such as dopamine, serotonin, tryptamine, and β-phenylethylamine. Unfortunately, MAO-inhibitors cannot discriminate among the different MAOs, and their overall effect results in high concentrations of tyramine that can cause hypertensive crises, brain hemorrhage and sometimes death (Pessione et al., 2009). Since tyramine is the biogenic amine most frequently found in cheese (due to the referred high tyrosine content of dairy products) this syndrome is also called "cheese reaction." For the reasons explained above, MAOI development has led to compounds with improved tolerability profiles compared to the initial irreversible MAOIs, particularly in regard to tyramineinduced hypertension and dietary restrictions (Yanez et al., 2012). The huge number of non-controlled autochthonous bacteria involved in cheese manufacturing and the accidental contamination by unwanted strains during storage render cheese a risky food for tyramine ingestion, although starters are accurately typed. TDC activity has been described in several LAB strains of the genera Lactobacillus, Carnobacterium, and Leuconostoc. The enzyme from Lactobacillus brevis and Enterococcus faecalis has been fully characterized. Proteomic studies reveal the membrane location of such a protein in Enterococcus faecalis (Pessione et al., 2009), thus confirming the high hydrophobic nature of this enzyme (previously predicted on the basis of gene analyses suggesting large hydrophobic domains but also because of the harsh procedures needed to purify it). Interestingly, Enterococcus faecalis TDC proved to be able to produce not only tyramine from tyrosine but also betaphenylethylamine from phenylalanine (Pessione et al., 2009).

### Beta-Phenylethylamine

β-Phenylethylamine, the decarboxylation product of phenylalanine, behaves like an endogenous amphetamine. It can control appetite/satiety thus being useful to treat obesity and related syndromes as well as in weight-control diets. Like

amphetamine it can also positively affect mood. However, controversial effects including insomnia, anxiety, and a MAOblocking activity that may extend the half-life of tyramine thus potentiating its toxicity (Millichap and Yee, 2003) have been described. Evidence of a specific phenylalanine decarboxylase enzyme in bacteria, has been reported only for Bacillus cohnii, a species taxonomically distant from LAB. Nevertheless, both Lactobacilli and Enterococci, frequently found in fermented food, display a clear β-phenylethylamine accumulation. Originally, some authors (Millichap and Yee, 2003) suggested that the β-phenylethylamine found in fermented food is the result of the activity of tyrosine decarboxylating bacteria. It is interesting to underline that in mammals and in several eukaryotic cells models a low-selective aromatic L-amino acid decarboxylase (EC 4.1.1.28) exists. In bacteria, enzymatic studies proved that TDC from E. faecalis is active also toward DOPA, but not toward other amino acids like phenylalanine while Lactobacillus bears a highly tyrosine selective TDC. More recently, the presence in E. faecalis of phenylalanine-decarboxylating TDC (EC 4.1.1.25) has been clearly demonstrated by a bi-phasic kinetic of tyramine and β-phenylethylamine production (Pessione et al., 2009). Tyrosine is consumed first and it is fully converted into tyramine (during logarithmic growth). Then phenylalanine is decarboxylated to β-phenylethylamine, although with lower conversion rate and yield. These results are consistent with a double catalytic activity of the same protein (TDC) having higher specificity for tyrosine and only low affinity for phenylalanine. On the other hand, Marcobal et al. (2006) observed β-phenylethylamine accumulation in a recombinant E. coli expressing the TDC gene of E. faecium.

## GABA

γ-Aminobutyric acid is one of the bioactive amines having true positive effects on human health. It acts as a neurotransmitter in the central nervous system of most vertebrates and plays additional roles on the overall human physiology such as lowering the blood pressure in mild hypertensive patients and acting as smooth muscles relaxation system (Mazzoli et al., 2010). Bacteria can produce GABA by the decarboxylation of glutamate through GAD. Glutamate and GABA act as opposite molecules having excitatory and inhibitory roles, respectively (Hyland and Cryan, 2010). Lactococcus lactis NCDO 2118 can obtain GABA from glutamine opportunely converted into glutamate (Mazzoli et al., 2010). It has long been established that L. lactis possess only one gene encoding GAD and that this enzymes is highly specific for glutamate (apparent Km 0.51 mM). Its molecular weight is about 54 kD and it has maximal activity at pH 4.7, whereas no activity is present at neutral pH (Nomura et al., 1999). Differences have been reported between Lactococcus lactis subsp. lactis and L. lactis subsp. cremoris: both possess gadCB genes encoding for GAD, however, onebase deletion of adenine and one-base insertion of thymine were detected in the coding region of the latter rendering it unable to synthesize a functional GAD enzyme (Nomura et al., 2000). Lactobacilli are the best GABA-producers, however, Lactococci, Streptococci, and Bifidobacteria can synthesize GABA as well (Lyte, 2011). Several strategies have been proposed to obtain high GABA amounts from cheap substrates through LAB sourdough fermentation (Coda et al., 2012), batch fermentation using immobilized (Huang et al., 2007) or entrapped (Choi et al., 2006) Lactobacillus brevis cells for the construction of functional food. To achieve maximum GABA production, temperature and pH are crucial (Dhakal et al., 2012). Nowadays, several nutraceutical preparations, such as GABA tea (Abe et al., 1995), GABA rice (Oh, 2003), GABA soymilk (Tsai et al., 2006), and GABA chocolate (Nakamura et al., 2009) are available, especially in Asia. GABA natural content of cacao is about 0.009%, however, this value is lower when considering the market-available chocolate also containing milk, sugar, and other ingredients. Functionalized chocolate, containing 0.28% GABA, proved to be able to lower the level of salivary chromogranin A and cortisol, thus reducing psychological stress (Nakamura et al., 2009).

## CONCLUSION

The knowledge on how, when and where bioactive amines are produced by LAB present in fermented food will be helpful by one side to prevent food borne diseases but on the other side also to use food as a nutraceutical delivery system to control mood, appetite and smooth muscle relaxation.

The knowledge about food-encrypted peptides and their potential can offer, in the coming decades, new tools for valuable therapies for the treatment of infectious diseases as well as metabolic, nutritional and psychiatric disorders, cardiovascular diseases, allergies and cancer, as suggested by Lien and Lowman (2003). The increasing availability of mass spectrometry facilities together with homology-based identification of potential bioactive peptide domains on protein sequences allows in silico prediction of encrypted peptides and will in the future be essential for targeting the right proteins. Several data banks are available for the bioinformatics analysis of potential bioactive peptides (Iwaniak et al., 2005; Lahrichi et al., 2013), especially antibacterial peptides, due to the urgent issue to face antibiotic resistance (Khaldi, 2012; Mooney et al., 2013). APD2 and CAMP are databases set up to classify not only antibacterial but also antifungal and antiviral peptides (Wang et al., 2009; Thomas et al., 2010). An even larger target is supplied by BIOPEP, Pep Bank, and Peptide DB, also classifying peptide hormones, growth factors and cytokines (Shtatland et al., 2007).

A new term, "nutritional peptidomics" has been proposed by Panchaud et al. (2012) to underline the importance of characterizing (mainly by mass spectrometry) food-encrypted peptides displaying health effects. These approaches allow to set up procedures to liberate the hidden bioactive molecules by means of different proteolytic enzymes (pepsin, trypsin, chymotrypsin) but also to synthesize artificial peptides displaying bioactivity (Zambrowicz et al., 2013). Microbial fermentation is one of the most promising strategies to generate bioactive peptides, hence genomic and proteomic characterization of new strains to predict

their proteolytic profile is a challenging approach in view of obtaining functional food. As far as LAB are concerned, combining the substrate specificity and the cleavage pattern of LAB proteases can highlight the strain potential to release healthpromoting molecules (or possible dangerous compounds) in food (Hayes et al., 2007a,b). This is a valuable strategy for selecting the right starters and to enhance the number of LAB strains to be used as food additives or probiotics. Moreover, the use of industrial by-products, such as milk whey, as growth substrates can lower the cost of functional food production and furtherly improve nutraceutical food availability.

### REFERENCES


### AUTHOR CONTRIBUTIONS

EP: Overall organization of the paper and content. SC: Figure and table management and critical reading.

### ACKNOWLEDGMENTS

SC benefits from a grant from CRT Foundation; Turin, Italy. The research on bioactive peptides has been supported by Ricerca Locale (ex 60%).



of peptides with antihypertensive activity. Acta Aliment. 34, 417–425. doi: 10.1556/AAlim.34.2005.4.9


Clin. Vaccine Immunol. 11, 1171–1181. doi: 10.1128/CDLI.11.6.1171-1181. 2004


helveticus and its non-proteolytic variant. J. Dairy Res. 68, 601–609. doi: 10.1017/S0022029901005143




**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 Pessione and Cirrincione. 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.

# Identification of novel esterase-active enzymes from hot environments by use of the host bacterium *Thermus thermophilus*

Benedikt Leis <sup>1</sup> , Angel Angelov <sup>1</sup> , Markus Mientus <sup>1</sup> , Haijuan Li <sup>1</sup> , Vu T. T. Pham<sup>1</sup> , Benjamin Lauinger <sup>2</sup> , Patrick Bongen<sup>2</sup> , Jörg Pietruszka<sup>2</sup> , Luís G. Gonçalves <sup>3</sup> , Helena Santos <sup>3</sup> and Wolfgang Liebl <sup>1</sup> \*

<sup>1</sup> Department of Microbiology, Technische Universität München, Freising, Germany, <sup>2</sup> Research Center Juelich, Institute of Bioorganic Chemistry, Heinrich-Heine-Universität Düsseldorf, Juelich, Germany, <sup>3</sup> Instituto de Tecnologia Química e Biológica, Universidade Nova de Lisboa, Oeiras, Portugal

#### *Edited by:*

Katharina Riedel, University of Greifswald, Germany

#### *Reviewed by:*

Rene Sutherland, University of Pretoria, South Africa Justin Joseph Donato, University of St. Thomas, USA

#### *\*Correspondence:*

Wolfgang Liebl, Department of Microbiology, School of Life Sciences Weihenstephan, Technische Universität München, Emil-Ramann-Str. 4, 85354 Freising, Germany wliebl@wzw.tum.de

#### *Specialty section:*

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

> *Received:* 18 December 2014 *Accepted:* 19 March 2015 *Published:* 08 April 2015

#### *Citation:*

Leis B, Angelov A, Mientus M, Li H, Pham VTT, Lauinger B, Bongen P, Pietruszka J, Gonçalves LG, Santos H and Liebl W (2015) Identification of novel esterase-active enzymes from hot environments by use of the host bacterium Thermus thermophilus. Front. Microbiol. 6:275. doi: 10.3389/fmicb.2015.00275 Functional metagenomic screening strategies, which are independent of known sequence information, can lead to the identification of truly novel genes and enzymes. Since E. coli has been used exhaustively for this purpose as a host, it is important to establish alternative expression hosts and to use them for functional metagenomic screening for new enzymes. In this study we show that Thermus thermophilus HB27 is an excellent screening host and can be used as an alternative provider of truly novel biocatalysts. In a previous study we constructed mutant strain BL03 with multiple markerless deletions in genes for major extra- and intracellular lipolytic activities. This esterase-diminished strain was no longer able to grow on defined minimal medium supplemented with tributyrin as the sole carbon source and could be used as a host to screen for metagenomic DNA fragments that could complement growth on tributyrin. Several thousand single fosmid clones from thermophilic metagenomic libraries from heated compost and hot spring water samples were subjected to a comparative screening for esterase activity in both T. thermophilus strain BL03 and E. coli EPI300. We scored a greater number of active esterase clones in the thermophilic bacterium than in the mesophilic E. coli. From several thousand functionally screened clones only two thermostable α/β-fold hydrolase enzymes with high amino acid sequence similarity to already characterized enzymes were identifiable in E. coli. In contrast, five further fosmids were found that conferred lipolytic activities in T. thermophilus only. Four open reading frames (ORFs) were found which did not share significant similarity to known esterase enzymes but contained the conserved GXSXG motif regularly found in lipolytic enzymes. Two of the genes were expressed in both hosts and the novel thermophilic esterases, which based on their primary structures could not be assigned to known esterase or lipase families, were purified and preliminarily characterized. Our work underscores the benefit of using additional screening hosts other than E. coli for the identification of novel biocatalysts with industrial relevance.

Keywords: *Thermus thermophiles*, novel screening host, functional metagenomics, comparative screening, novel metagenomic esterases

## Introduction

As only a small fraction of all microorganisms from environmental samples can be isolated and studied by classic microbiological methods, culture-independent strategies, such as metagenomics, have been employed to overcome these restrictions (Torsvik et al., 1990; Amann et al., 1995; Handelsman et al., 1998). The cloning and analysis of environmental DNA has led to the identification of various new genes for valuable enzymes (for a recent review see Leis et al., 2013). Sequence-based metagenomics strategies have been successful in many cases, but sequences without similarity to known genes cannot be identified using this approach (Steele et al., 2009; Liebl, 2011; Liebl et al., 2014). Function-based strategies on the other hand rely on the direct detection of functional gene products via diverse screening setups. Until today, E. coli is the most commonly used bacterial host in functional metagenomics. Since only a fraction of the genes from random (meta)genomic gene libraries are expressed (Gabor et al., 2004; Sorek et al., 2007), the identification of truly novel genes from metagenomes is hampered when a single host is used (Jenney and Adams, 2008). Other hosts than E. coli (e.g., diverse Proteobacteria, Actinobacteria, and eukaryotic fungi) have been shown to be good alternatives and new broad-range expression vectors have been developed for use in functional metagenomics (Courtois et al., 2003; Martinez et al., 2004; Aakvik et al., 2009; Craig et al., 2010; Damon et al., 2011; Kakirde et al., 2011; Kellner et al., 2011; Parachin and Gorwa-Grauslund, 2011; Liebl et al., 2014).

In previous work we have shown that Thermus thermophilus HB27 can be used for screening of genomic libraries using the shuttle fosmid pCT3FK. These studies resulted in the identification of several genes that would not have been found by using the mesophilic E. coli host alone (Angelov et al., 2009; Leis et al., 2013).

In this study we present the generation of large-insert metagenomic libraries and their subsequent screening for lipolytic activities in E. coli and T. thermophilus HB27. For the screenings in T. thermophilus, a multiple clean deletion mutant was used (termed BL03) which lacks several characterized extracellular and putative esterase encoding genes (Leis et al., 2014). The knockout strain had a substantially lower esterase activity than the wild type and the deletions abolished its ability to grow on SH minimal medium supplemented with tributyrin. Therefore, we decided to use this esterase-deficient strain in a high-throughput screening and selection setup where complementation of growth on tributyrin minimal medium by cloned heterologous metagenomic genes served to select for new environmental genes encoding tributyrin-cleaving esterases.

### Materials and Methods

### Bacterial Strains and Growth Conditions

Transformation and propagation of recombinant plasmids were performed using Escherichia coli XL1-Blue, XL10 GOLD <sup>R</sup> ultracompetent cells (Stratagene, La Jolla, USA) and DH10B (Invitrogen, Carlsbad, USA). E. coli EPI300-T1 <sup>R</sup> (Epicentre, Madison, USA) was used for the generation and screening of metagenomic libraries cloned in pCT3FK large-insert fosmids (Angelov et al., 2009) based on pCC1FOS) upon addition of fosmid autoinduction solution (Epicentre). E. coli strain BL21(DE3) was used for the expression of target proteins in pET21a(+) (Novagen, Merck KGaA, Darmstadt, Germany). Bacterial cultures were grown in lysogeny broth (LB) or on LB agar plates supplemented with ampicillin (100µg/ml), kanamycin (20µg/ml) and chloramphenicol (12.5µg/ml) as necessary. For T. thermophilus HB27 (DSMZ 7039) and T. thermophilus BL03 (genotype 1TT\_P0042, 1TT\_C0340-1, 1TT\_C0904, 1TT\_C1787), culturing and transformations were done as described in Angelov et al. (2009), using TB complex medium (8 g/l trypticase peptone, 4 g/l yeast extract, and 3 g/l NaCl) (Ramírez-Arcos et al., 1998) and SH minimal medium (Leis et al., 2014) with high-carbonate mineral water Aqua Purania (TSI, Zeven, Germany) at pH 7.5. Cells were cultured at 70◦C or at 60◦C when using kanamycin (final concentration of 20µg/ml) for the propagation of transformants. For the growth selection and complementation assay, bacterial suspensions from overnight-grown liquid cultures were washed three times in 50 mM phosphate buffer (pH 7.5) and adjusted to equal optical densities before spotting 5 or 10µl on complex and minimal medium. For screening and selection purposes on solid medium, tributyrin with a final concentration of 1.0% (v/v) was emulsified with an Ultra-Turrax homogenizer (IKA, Staufen, Germany) immediately before autoclaving of the medium.

### Generation of a Large-Insert Fosmid Library and Functional Screenings

Water, sediment and biofilms were sampled from hot springs in the town of Furnas Azores, Portugal (37◦ 46′ 21.78′′N 25◦ 18′ 14.673′′W). Compost heap samples were collected from a composting plant (Bioenergiezentrum GmbH, Göttingen, Germany). After separation of microorganisms from the samples' matrices (filtration and scraping off cells from organic matter), the metagenomic DNA was prepared using the PowerSoil, PowerMax Soil and UltraClean Fecal DNA Isolation Kit (MO BIO, Carlsbad, USA). The generation of a large-insert DNA library was done using the pCC1FOS-derived pCT3FK T. thermophilus/E. coli shuttle fosmid (Angelov et al., 2009) according to the instructions of the pCC1FOS Fosmid Library Production Kit (Epicentre, Madison, USA). Arrayed clones from E. coli were cultured in 96-well microtiter plate format. To overcome labor-intensive transfer of single fosmid clones from E. coli to T. thermophilus, a high-throughput screening strategy was chosen based on growth complementation (Leis et al., 2014). Therefore, each pool of 96 single fosmids from E. coli was transferred in the screening strain T. thermophilus BL03 by a single transformation reaction. The transformants of each reaction were grown on TB plates containing kanamycin as antibiotic. After 2 days of incubation at 60◦C, colonies were suspended from the plates and washed three times using 50 mM phosphate buffer (pH 7.5). The washed cell suspensions (each comprising a pool of 96 T. thermophilus transformants obtained by transformation of strain BL03 with pooled fosmids from 96 E. coli clones) were spotted onto SH minimal medium containing tributyrin (1% v/v) and kanamycin. Candidate clone pools were able to grow on tributyrin due to heterologous growth complementation after 2–3 days of incubation at 60◦C. In order to determine single clone(s) whose fosmid-derived

inserts conferred tributyrase activity within each positive pool, T. thermophilus BL03 was transformed individually with each of the 96 E. coli fosmids in deep well plate format as described before (Angelov et al., 2009). Suspensions of the transformants were washed several times and spotted on minimal medium containing tributyrin. T. thermophilus clones with candidate fosmid inserts grew after several days of incubation at 60◦C. The stability of each phenotype was assessed by repeated streaking and growth of the clones on SH with tributyrin.

Each candidate clone was analyzed for its ability to hydrolyze the substrate in comparison to the control containing only the empty fosmid DNA. The corresponding halo formation was calculated as the area of the clearing zone without the area of the corresponding colony size according to the following equation:

Halo area formed =

[diameter(halo)<sup>2</sup> − diameter(colony)<sup>2</sup> ] × 0.25 × π

### Sequencing and *in silico* Analysis of Sequence Data

Fosmids conferring significant halo formation were sequenced using Roche's GS-FLX series 454 pyrosequencer at the Göttingen Genomics laboratory (G2L) and LGC Genomics (Berlin, Germany). For phylogenetic analysis, bacterial and archaeal 16S rDNA sequences were amplified by PCR using degenerate primer pairs 616Valt, 100K, AC165, and AC1601, respectively (Supplementary Table S1). Phylogenetic analysis was done using ARB (Ludwig et al., 2004) and the Ribosomal Database Project II (RDP-II) server (Cole et al., 2009). Assembled contig sequences were analyzed in silico using SEED (Overbeek et al., 2005). BlastP (Altschul et al., 1990), conserved domain database (CDD, Marchler-Bauer et al., 2011), and the Pfam database (Punta et al., 2012) were used for similarity searches and annotation. The presence of predicted signal peptides was analyzed with the SignalP server (Petersen et al., 2011). Regular expression algorithm was used to unravel potential esterase candidate genes by GXSXG conserved motif searches. For the generation of multiple sequence alignments and similarity tree reconstruction, representative protein sequences from classified esterases/lipases (Chow et al., 2012) were used. Tree calculations based on multiple sequence alignments were performed with the T-REX webserver (Boc et al., 2012) and the resulting Newick-formatted data was visualized with MEGA Version 5.2 (Tamura et al., 2011).

### Cloning, Expression, and Purification of Metagenomic Esterases

Candidate genes were PCR amplified with Phusion DNA polymerase (ThermoFisher Scientific, Waltham, USA) using primers listed in Supplementary Table S1 (oligonucleotides EstA2\_for and EstA2\_rev for cloning EstA2, EstB1\_for and EstB1\_rev for EstB1, respectively). The PCR products were cloned using a Clone-JET vector (Fermentas) backbone. Insertion of the polyhistidine (6xHis) sequence was performed by the Change-IT Multiple Mutation Site Directed Mutagenesis Kit (Affymetrix, Santa Clara, USA) with 5′ -phosphorylated oligonucleotides (EstA2\_PHO\_F and EstB1\_PHO\_F) according to the manufacturer's instructions. Metagenomic esterase-encoding genes (EstA2 and EstB1) were cloned in pET21a(+) (Novagen) using NdeI and NcoI, and the same strategy was used for cloning in the T. thermophilus vector pMK18 (de Grado et al., 1999) by isothermal DNA assembly using the Gibson Assembly Master Mix (New England Biolabs, Ipswich, USA) with vector primers pMK18\_RBS\_for and pMK18\_RBS\_rev. The insert was amplified with Est\_rev and corresponding forward primers of the esterases. For expression, E. coli BL21(DE3) Star cultures (250 ml) were grown in Erlenmeyer flasks to mid-log phase (absorbance at 600 nm ranging from 0.6 to 0.7) and induced with isopropyl-β-D-thiogalactopyranoside (IPTG) at a final concentration of 1.0 mM. After 4 h at 37◦C, the cultures were harvested by centrifugation and disrupted by sonication (Hielscher Ultrasonics GmbH, Teltow, Germany). After removal of the cell debris, the lysate was subjected to heat treatment (at least 60◦C for 20 min). The supernatant containing the thermostable and soluble protein of interest was purified using Protino Ni-IDA 2000 protein purification system (Macherey-Nagel) according to the manufacturer's instructions. Protein separation and purity was determined with sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) based on Laemmli (1970).

### Enzyme Assays

E. coli and T. thermophilus cells from 5 ml overnight grown cultures were harvested by centrifugation and washed twice in 50 mM Tris-HCl (pH 7.5). All steps were performed on ice. Cells were resuspended in half volume of Tris buffer and disrupted by sonication until the turbid suspensions became clear. Cell debris was removed by centrifugation (21,000 × g, 10 min at 4◦C) and supernatants were used for protein quantification and measurement of intracellular lipolytic activities. The paranitrophenyl (pNP) assays were performed according to Leis et al. (2014) by using 0.1 ml of crude lysates incubated in 50 mM Tris-HCl (pH 7.5) buffer with 1.25 mM pNP-caprylate, -laurate, and palmitate at 60◦C. After stopping the reaction with 0.25 ml of 2 M sodium carbonate and incubation on ice for 15 min, the mixture was centrifuged at 21,000 × g for 5 min at 4◦C. Esterase activity measurements for purified enzyme preparations were performed using 1.25 mM final concentrations of different para-nitrophenyl-substrates (pNP-propionate (C3), -butyrate (C4), -valerate (C5), -caproate (C6), -caprylate (C8), caprate (C10), -laurate (C12), -myristate (C14), and -palmitate (C16), all obtained from Sigma-Aldrich). The influence of the additives NaCl, KCl, CaCl2, MgCl<sup>2</sup> (salts), and inhibitor EDTA (chelating agent) and phenylmethylsulfonylfluorid (PMSF), an agent known to inhibit serine/cysteine proteases as well as acetylcholinesterases, was tested at final concentrations of 1 mM. Other pNP-substrates were derived by esterification with 2 methyldecanoate (Reetz et al., 1997; Franken et al., 2010; Wu et al., 2013), Ibuprofen, Naproxen (Reetz et al., 2010; Sandström et al., 2012), 6-methyl-2-(ortho-tolyl)hept-5-enoate (Gaich and Mulzer, 2009), and indancarboxylic acid (Pietruszka et al., 2009) and the activity assays were performed at 60, 70, and 80◦C under the conditions reported by López et al. (2014) and Chow et al. (2012).

One unit of specific enzymatic activity was defined as 1.0 µmol of para-nitrophenol per minute and mg protein released from the substrate. All measurements were performed at least in duplicates (n = 2).

### Synthesis of Assay Substrates—Representative Procedure

Synthesis of 4-nitrophenyl 6-methyl-2-(ortho-tolyl)hept-5-enoate: 4-Nitrophenol (538 mg, 3.87 mmol), 4-dimethylaminopyridine (47.3 mg, 387µmol), and dicyclohexylcarbodiimide (798 mg, 3.87 mmol) were added to a solution of 6-methyl-2-(ortho-tolyl)hept-5-enoic acid (900 mg, 3.87 mmol) in 25 mL of dichloromethane under nitrogen. The mixture was stirred at room temperature. After 24 h the precipitate was filtered off and the solvent was removed under reduced pressure. The residue was purified by flash chromatography (petroleum ether/ethyl acetate, 99/1) to give the product as a yellow oil (1.07 g, 3.03 mmol, 78%). <sup>1</sup>H-NMR (600 MHz, CDCl3,): δ<sup>H</sup> [ppm] = 1.55 (s, 3 H, 1-H), 1.71 (s, 3 H, 2-H), 1.88 [dddd, <sup>2</sup> J(6a,6b)=13.7 Hz, 3 J(6a,5b) =7.2 Hz, <sup>3</sup> J(6a,5a) = 7.2 Hz, <sup>3</sup> J(6a,7) = 6.9 Hz, 1 H, 6-Ha], 2.07 [dd,<sup>3</sup> J(5,6) = 7.2 Hz, <sup>3</sup> J(5,4) = 7.2 Hz, 2 H, 5-H], 2.26 [dddd, <sup>3</sup> J(6b,7) = 6.7 Hz, <sup>3</sup> J(6b,5a) = 7.2 Hz, <sup>3</sup> J(6b,5b) = 7.2 Hz, 2 J(6b,6a) = 13.6 Hz, 1 H, 6-Hb], 2.44 (s, 3 H, arom.-CH3), 4.12 [dd, <sup>3</sup> J(7,6a) = 6.8 Hz, <sup>3</sup> J(7,6b) = 6.7 Hz, 1 H, 7-H], 5.14 [t, 3 J(4,5) = 7.2 Hz, 1 H, 4-H], 7.15- (m<sup>c</sup> , 2 H, arom.-CH), 7.18–7.25 (m, 3 H, arom.-CH), 7.35 (mc, 1 H, arom.-CH), 8.22 (mc, 2 H, arom.-CH). <sup>13</sup>C-NMR (151 MHz, CDCl3): δ<sup>C</sup> [ppm] = 17.7 (C-1), 19.8 (arom.-CCH3), 25.7 (C-5), 25.8 (C-2), 32.8 (C-6), 46.3 (C-7), 122.3 (arom.-CH), 122.3 (arom.-CH), 123.1 (C-4), 125.1 (arom.-CH), 125.1 (arom.-CH), 126.6 (arom.- CH), 127.5 (arom.-CH), 130.8 (arom.-CH), 130.8 (arom.-CH), 133.2 (C-3), 136.6 (arom.-Cipso), 136.6 (arom.-Cipso), 145.3 (COOpNP), 155.6 (arom.-Cipso), 171.9 (arom.-CNO2). HR-MS (ESI, cation): calculated [C21H23NO4+Na+]: m/z = 376.15193; found [C21H23NO4+Na+]: m/z = 376.15211.

### Results

### Generation of Fosmid Libraries and Phylogenetic Analysis of Hot Spring and Compost Samples

Environmental samples were obtained from strong acidic and neutral hot spring water sediments and biofilms (pH ranging between 2.0 and 7.0, temperature from 60 to 62◦C) from the São Miguel island, Azores, Portugal. In addition, a naturally heated lumber waste compost heap (63.3◦C after 4 days) sample was obtained from a compost facility in Göttingen (Germany) (**Table 1**). High molecular weight DNA from both samples could be obtained and enabled the generation of a large-insert metagenomic library in the pCT3FK shuttle fosmid. The library comprised around 6048 single fosmid clones for the Azores samples and 1920 clones for the compost sample with each clone carrying approximately 35–40 kbp metagenomic inserts (**Table 1**). Phylogenetic analysis of the metagenomic DNA inserts

TABLE 1 | Overview of the constructed metagenomic libraries and the esterase screening results in *E. coli* EPI300 and *T. thermophilus* BL03.


TABLE 2 | Frequency of operational taxonomic units (OTUs) identified by 16S rDNA-analysis of the metagenomic samples.


was done by sequencing of PCR products obtained with degenerate primers which amplify bacterial and archaeal 16S rDNA sequences (Supplementary Table S1). This analysis showed a very low archaeal diversity in the hot spring water, with only representatives of the genera Stygiolobus azoricus DSM 6296 (99% sequence identity) and Sulfolobus species (e.g., S. acidocaldarius DSM 639 with 95% sequence identity) found. In the compost sample a more diverse representation of bacteria was observed with Firmicutes being the most highly represented phylum in which the majority of sequences were assigned to the orders Bacillales and Clostridiales (**Table 2**).

### Functional Screening of Fosmid Libraries in *E. coli* and *T. thermophilus*

After screening of the generated libraries in E. coli using a plate assay at 60◦C, only two halo-forming colonies were identified on tributyrin agar plates, one from the Azores sample (termed AZ2-4-B6) and one from the compost sample (M12-4-D9). In order to perform the screening in T. thermophilus, the BL03 strain was transformed with 83 fosmid pools, each prepared from 96 fosmid clones. The transformant pools obtained in this way were examined for their ability to grow on minimal medium plates supplemented with tributyrin. A total of 13 pools showed

FIGURE 1 | (A) Growth complementation screening results of T. thermophilus BL03 transformed with metagenomic fosmid DNA (upper scan picture) and corresponding tributyrase halo formation capabilities on substrate plates (lower scan picture). (B) Tributyrase activity of 8 single fosmids in E. coli EPI300 (upper part of the table) and T. thermophilus BL03 (below). In this representation, the increase of halo formation (black bars, halo area was calculated as described in the Materials and Methods section) on 1% (v/v) tributyrin substrate plates after 3 days of incubation at 60◦C is shown (from triplicate measurements, n = 3; error bars indicate the standard deviation). In T. thermophilus, heterologous growth complementation results after 3 days of growth at 60◦C on minimal medium are depicted. Comparative pNP-activity data from crude extracts (specific activity mU × mg−1) is shown for each single fosmid clone as average values from duplicate measurements (± standard deviation).

substantial growth on these plates (namely fosmid pools AZ2- 4; AZ3-14; AZ3-25; AZ3-30; AZ3-32; AZ3-33; AZ3-37; AZ3-38; AZ3-47; AZ4-2; M12-3; M12-4; and M12-6). From these candidate pools, single fosmid transformations were performed in 96 well plate format. Again, growth on minimal medium was monitored, and positive clones were selected. After re-streaking them two more times to confirm the stability of their phenotype, six fosmid-containing T. thermophilus clones could be isolated from five different fosmid pools. They stably grew on minimal medium agar supplemented with tributyrin, and hydrolysis halos on tributyrin substrate plates were monitored in comparison with the BL03 strain carrying an empty fosmid (**Figure 1A**). In addition, the esterase activity of crude cell extracts obtained from the respective strains was measured with various pNP-substrates (**Figure 1B**). Almost all fosmid clones showed preferences for medium to short acyl chain fatty acid pNP-esters. Crude extract from the strain AZ3-14-D2 exhibited the highest activities of all crude extracts tested. This extract was most active over a broad range of pNP-substrates tested and its long-chain palmitate fatty acid hydrolysis was approximately three-fold higher than in the BL03 empty fosmid control. The two fosmids that conferred tributyrase activity in E. coli, AZ2-4-B6 and M12-4-D9, were also found to confer activity in T. thermophilus when tested in pNPassays. Fosmid M12-4-D9 could not be identified by heterologous growth complementation screenings in T. thermophilus BL03, but showed higher activity on pNP-C<sup>8</sup> compared to E. coli. In general, the pNP-substrate preference of the fosmid-encoded esterases for short-chain acyl esters was in agreement with the observed capabilities of halo formation on tributyrin agar plates.


†Sequenced separately at Göttingen Genomics Laboratory (G2L), no assembly statistics available, fosmid backbone pCT3FK sequence not included. ‡ Identified from shotgun library in E. coli only.

n.d., no detectable candidate ORF.

### Sequencing and *in silico* Analysis of the Esterase Activity-Conferring Fosmid Inserts

The candidate fosmids were subjected to 454-sequencing and plasmids containing shotgun libraries of positive fosmids were sequenced with Sanger technology. The bioinformatic analysis of the fosmids is summarized in **Table 3**. Annotated α/β hydrolase 6 family proteins sharing high similarity to already characterized proteins were predicted from the sequence of two fosmids, namely M12-4-D9 and AZ2-4-B6. Not all positive fosmids from high-throughput T. thermophilus screenings could be completely assembled and sequenced (in average, 32.86 kb sequence information was obtained from each fosmid). The assembly data is summarized in Supplementary Table S2. Only in the case of two fosmids, AZ3-33-C12 (GenBank accession number KP892656) and M12-4-G10 (GenBank accession number KP892657) complete sequence assemblies where available for bioinformatic analysis. By motif search, the conserved GXSXG pentapeptide signature was identified in two of the predicted proteinencoding ORFs (EstA2 from fosmid AZ3-33-C12 and EstB1 from M12-4-G10). Both candidates showed no or only weak sequence similarity to lipolytic enzymes. Pfam prediction did not reveal any esterase/lipase-encoding function. For fosmid clones AZ3-32-E5, AZ3-14-D2, and AZ3-14-D11, sequence information was incomplete, therefore hampering further in silico analysis.

### Heterologous Expression and Preliminary Characterization of Two Novel Metagenomic Esterases

From all candidate esterase-encoding ORFs, we selected EstA2 and EstB1 for expression, purification and characterization using T. thermophilus as an expression host. Expression of these ORFs in pMK18 (de Grado et al., 1999) led to significant pNPbutyrate-cleaving activity (measured at 60◦C). Further investigation revealed that short-chain acyl esters were the preferred substrates, such as pNP-butyrate, that was hydrolyzed most

optimally around 80–85◦C and pH 8.0. EstA2 (encoded on fosmid AZ3-33-C12) resulted in a significant activity increase in the extracellular fraction, resulting in a 1.54-fold induction of activity, while EstB1 showed a 1.53-fold increased activity level compared to 338.4 ± 4.8 mU/mg protein in BL03 with the empty pMK18 vector. However, no recombinant protein band could be detected by SDS-PAGE analysis. We then attempted to produce C-terminally tagged versions of both proteins in E. coli by using the pET21a(+) expression vector. For both proteins, purification

FIGURE 3 | Characterization of the purified esterase-active proteins EstA2 and EstB1. (A) Substrate specificity of various acyl chain length pNP-esters. (B) Temperature and (C) pH optimum. The assays were performed in 50 mM phosphate buffer (pH 6–7) or 50 mM Tris-HCl buffer (pH 7.2–8.8) under optimal activity parameters (80◦C for EstA2 and 75◦C for EstB1, respectively) against 1.25 mM pNP-butyrate. 0.065–1.3 µg of purified enzyme was used for the assays. Data represents average values and standard deviations (error bars).

using heat treatment followed by IMAC gravity flow columns resulted in functional protein preparations (**Figure 2**).

The characterization of the recombinant enzymes revealed similar substrate preferences for short acyl chain substrates (pNP-C<sup>3</sup> > C<sup>4</sup> > C<sup>5</sup> > C<sup>6</sup> > C8) as the ones observed in T. thermophilus, while activities on long-chain substrates (over C10) were comparably low (**Figure 3**). Under the assay conditions used here, optimum activity of EstA2 was observed around 80◦C and pH 8.0, while EstB1 was most active at 75◦C and pH 8.0. The specific activities of both enzymes on para-nitrophenyl butyrate were 44.05 ± 2.06 U/mg protein for EstA2 and 4.13 ± 0.71 U/mg for EstB1. The influence of several additives, NaCl, KCl, CaCl2, MgCl2, EDTA, and PMSF at 1–10 mM final concentration was negligible. Our data suggests both proteins to be esterases rather than lipases due to their characteristic substrate specificities.

EstA2 and EstB1 were tested toward various racemic mixtures of a selection of pNP-esters (**Table 4**). Activity was measured over a wide range of temperatures (60–80◦C) to overcome constraints by autohydrolysis of the substrates at elevated temperatures (Chow et al., 2012; López et al., 2014). In general, the activities measured were comparably low (mU range). The highest activities under the conditions tested were found with the 4-nitrophenolates of-Naproxen and Ibuprofen. No activity could be detected with the pNP-esters of indancarboxylic acid while 2-methyldecanoate and 6-methyl-2-(ortho-tolyl)hept-5 enoate were only hydrolyzed weakly in the presence of EstB1 at the highest temperature tested so far.

### Classification of Lipolytic Enzymes

In order to classify the gene products identified in our metagenomic screenings, we performed multiple sequence alignments with known esterases and lipases from families I to VIII (Arpigny and Jaeger, 1999). Recently, the number of known lipolytic enzyme families has been extended by two new families, termed LipS and LipT (Chow et al., 2012). From the positive fosmids identified in E. coli (AZ2-4-B6 and M12-4-D9), both encoding α/β-hydrolase ORFs were identified as esterase class V and LipS family protein, respectively. The lipolytic enzymes discovered by using T. thermophilus for functional library screening, EstA2 and EstB1, could not be assigned to any known esterase/lipase family (**Figure 4**).

### Discussion

In the past, E. coli has been shown to be a good provider of interesting biocatalysts and enzymes from metagenomic screenings. Although it is clear that E. coli has restricted expression capabilities, it is the best developed and most commonly used host until today. Here, we show that using an additional host in functional screenings of metagenomic libraries can lead to the discovery of enzymes that would have remained hidden if only E. coli had been used.

How heterologous sequences are recognized and expressed into functional proteins by the transcriptional and translational machinery of an expression host is poorly understood. Factors influencing these processes, amongst other unknown factors, are: (i) toxicity of the foreign DNA and encoded proteins (Sorek et al., 2007), (ii) recognition of transcriptional signals (promoters) (Warren et al., 2008), (iii) ribosomal binding and translation initiation (Sørensen and Mortensen, 2005; Villegas and Kropinski, 2008), (iv) mRNA stability (Kudla et al., 2009), (v) codon usage (Sørensen and Mortensen, 2005), (vi) (post-translational) protein modification and secretion (Mergulhão et al., 2005). To overcome these restrictions, other screening hosts have been implemented in function-based metagenomic screening approaches in order to increase the detection frequency of active clones. First attempts using additional screening hosts involved Streptomyces and Pseudomonas species (Courtois et al., 2003; Martinez et al., 2004; Lussier et al., 2011). By the use of broad-host range vectors, the expression could recently be extended to other members of the Proteobacteria, i.e., the same taxon as E. coli (Aakvik et al., 2009; Craig et al., 2010; Kakirde et al., 2011), and to eukaryotic expression hosts (Damon et al., 2011; Kellner et al., 2011; Parachin and Gorwa-Grauslund, 2011). Our approach is distinguished from these "broad-host vectors" strategies by performing the library construction steps in the well-established E. coli system, followed by transfer of the metagenomic fosmids in the alternative host (Angelov et al., 2009). In order to facilitate the detection of esterase-encoding fosmid inserts, we used a selection-based strategy which relies on the complementation of growth of a custom, esterase-deficient strain of T. thermophilus (strain BL03, Leis et al., 2014). The BL03 strain is a multiple esterase deletion mutant which exhibits significantly reduced extracellular lipolytic activities compared with the parent strain HB27 and was shown


Activity tests were performed in 850µL of 50 mM Sørensen buffer containing 0.1% (w/v) gum arabic and 5 mM sodium deoxycholate (pH 8.0 at 60, 70, and 80◦C, respectively), mixed with 120µL of DMSO and 20µL of 10 mM of the pNP-ester substrates. Within 10µL residual volume, 0.4µg of the enzymes were added to the assay, which was performed for 30 min. Absorbance at 400 nm was measured in cuvettes, enzyme-free samples were measured as reference (the molar extinction coefficient was determined as ε = 14,000 M-1 × cm-1).

to be severely impaired in growth on minimal medium supplemented with tributyrin (Leis et al., 2014). The growth could be regained via heterologous complementation when metagenomic DNA was expressed by the host. Low detection limits of classical phenotypic plate screenings (clearing zones around the colonies) could thus be overcome, as even weak gene expression levels yield enough metabolites to reconstitute growth. By implementing a high-throughput pooling strategy, approximately 8000 single fosmid clones yielded six single candidate fosmid clones, which represents a detection frequency that is comparable with screening results obtained from other metagenomics studies aimed at lipolytic activities from hot environments (Rhee et al., 2005; Chow et al., 2012). To our knowledge, this is the first report of identifying novel metagenome-borne esterases by the use of a screening organism other than traditional E. coli.

Metagenomic esterases having similarity to β-lactamase family proteins have been reported from various environmental or anthropogenic sources, e.g., leachate samples (Rashamuse et al., 2009), arctic soil (Yu et al., 2011), alkaline compost (Kim et al., 2010), drinking water and soil metagenomes (Elend et al., 2006). They share the conserved sequence motif common for class C β-lactamases (Ser-Xaa-Xaa-Lys, whereas Xaa is any amino acid) that is important for catalytic activity of esterases with similarity to penicillin-binding proteins (Petersen et al., 2001), and have been assigned to the less characterized family VIII carboxylesterases (Arpigny and Jaeger, 1999). EstB1 lacks this conserved sequence motif and does not belong to the class C β-lactamase protein family. Instead, it could be assigned to class B metallo-β-lactamase proteins (PF00753). This group comprises hydrolases acting on thioester (thioesterase) and sulfate ester (sulfatase) bonds. Furthermore, thioesterases have also been shown to convert para-nitrophenol esters (Shahi et al., 2006; Kotowska et al., 2009). EstB1 shared 30% sequence identity with SdsA1 from Pseudomonas aeruginosa, which is a characterized alkyl-sulfatase (Hagelueken et al., 2006). The other esterase-acting enzyme identified in our metagenomic studies was EstA2. It is most similar to hypothetical proteins from diverse archaeal origins. One characterized enzyme, to which EstA2 shows very weak similarity, is a phospholipase from the hyperthermophilic archaeon Aeropyrum pernix K1 encoded by ORF APE2325 (Wang et al., 2004). Also being a small protein with 18 kDa in size, its reported substrate spectrum and activity pattern is comparable to EstA2, as it also preferred short acyl chain pNP-substrates and was most active at elevated temperatures (above 80◦C).

In this study we have demonstrated that T. thermophilus is an excellent host bacterium for the detection of novel metagenomeborne enzymes that could not readily have been detected by the use of E. coli or by in silico analysis. To our knowledge, this is the first report that metagenome-derived esterases could be identified in an expression host other than E. coli. The functional screening implementing T. thermophilus BL03 could uncover several lipolytic enzymes from underrepresented species and archaeal origin, and most strikingly, some of them could not be assigned to classical esterase and lipase enzyme families and therefore represent truly novel esterase-active proteins. The confirmation of predicted ORFs responsible for the lipolytic activity of the remaining fosmids will be addressed in the future. The use of

### References


complementation screens in T. thermophilus with shotgun subclones will help to identify more active genes that do not share any known sequence signatures at all.

### Acknowledgments

This work was supported by the German Research Foundation (DFG) and the Technische Universität München within the funding programme Open Access Publishing. The authors gratefully acknowledge the support by the Faculty Graduate Center Weihenstephan of TUM Graduate School at Technische Universität München, Germany. This work was funded by the German Federal Ministry of Education and Research (BMBF) in the collaborative project ExpresSys within the funding measure GenoMik-Transfer. JP thanks T. Pham, C. Lorenz, M. Schölzel, R. C. Simon, D. Schumacher, and K. Morka for supporting experiments.

### Supplementary Material

The Supplementary Material for this article can be found online at: http://www.frontiersin.org/journal/10.3389/fmicb. 2015.00275/abstract


of Saccharomyces cerevisiae mutants. Microbiol. Res. 166, 360–368. doi: 10.1016/j.micres.2010.07.006


**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 Leis, Angelov, Mientus, Li, Pham, Lauinger, Bongen, Pietruszka, Gonçalves, Santos and Liebl. 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.

**PERSPECTIVE ARTICLE** published: 05 December 2014 doi: 10.3389/fmicb.2014.00684

## Bioactive compounds produced by gut microbial tannase: implications for colorectal cancer development

### *Félix López de Felipe\*, Blanca de las Rivas and Rosario Muñoz*

Laboratory of Bacterial Biotechnology, Institute of Food Science, Technology and Nutrition – Spanish National Research Council (ICTAN-CSIC), Madrid, Spain

#### *Edited by:*

Enrica Pessione, University of Torino, Italy

#### *Reviewed by:*

Rene Sutherland, University of Pretoria, South Africa Robert Kourist, Ruhr-University Bochum, Germany

#### *\*Correspondence:*

Félix López de Felipe, Laboratory of Bacterial Biotechnology, Institute of Food Science, Technology and Nutrition – Spanish National Research Council (ICTAN-CSIC), Ciudad Universitaria s/n, Madrid 28040, Spain e-mail: fxlopez@ictan.csic.es

The microorganisms in the human gastrointestinal tract have a profound influence on the transformation of food into metabolites which can impact human health. Gallic acid (GA) and pyrogallol (PG) are bioactive compounds displaying diverse biological properties, including carcinogenic inhibiting activities. However, its concentration in fruits and vegetables is generally low. These metabolites can be also generated as final products of tannin metabolism by microbes endowed with tannase, which opens up the possibility of their anti-cancer potential being increased. Patients with colorectal cancer (CRC) display an imbalanced gut microbiota respect to healthy population.The recent use of next generation sequencing technologies has greatly improved knowledge of the identity of bacterial species that colonize non-tumorous and tumorous tissues of CRC patients.This information provides a unique opportunity to shed light on the role played by gut microorganisms in the different stages of this disease.We here review the recently published gut microbiome associated to CRC patients and highlight tannase as an underlying gene function of bacterial species that selectively colonize tumorous tissues, but not adjacent non-malignant tissues. Given the anti-carcinogenic roles of GA and PG produced by gut tannin-degrading bacteria, we provide an overview of the possible consequences of this intriguing coincidence for CRC development.

**Keywords: microbial tannase, bioactive compounds, colorectal cancer, gallic acid, pyrogallol**

### **INTRODUCTION**

An imbalance in intestinal bacteria potentially leading to diseases termed colonic dysbiosis is associated to CRC, however, the possible role of intestinal microorganisms in the different CRC-stages is currently under debate.

Host diet is a crucial variable to shape the gut microbiota and a risk factor for CRC. Increased fruit and vegetable consumption is a dietary habit modification aimed to prevent gastrointestinal cancer (WHO Technical Report Series, No. 916 [TRS 916], 2003). Plant foods are main sources of tannins, dietary phytonutrients that display pro-apoptotic and antimetastasic properties in animal and *in vitro* studies (Serrano et al., 2009). Despite its promising chemopreventive potential, results from human studies to evaluate the association between the intake of foods rich in tannins and cancer risk are more dissuading. For example, inverse association (Theodoratou et al., 2007) but also no association (Kyle et al., 2010) between dietary proanthocyanidin intake and CRC risk have been reported. These apparently conflicting results may reflect interindividual differences in gut microbiota and reveal the difficulty to infer biological effects from dietary intake records without considering tannin-microbiota interactions. The microorganisms in the human gastrointestinal tract have a profound influence on the transformation of food into metabolites which can impact human health (Nicholson et al., 2012). Thus, the highly individual gut microbial activity, in particular the tannin-metabolizing activity of CRC-associated microbiota, will probably determine the bioavailability and physiological effects of tannins and their degradation products on tumorous tissues.

### **TANNASE: A COMMON FUNCTION IN THE MICROBIOTA COLONIZING COLORECTAL TUMOROUS TISSUES**

Gut microbial tannase activity is likely crucial to determine the tannin-metabolizing phenotype of CRC-associated microbiota. Tannase (tannin acylhydrolase) transforms the gallate esters of tannins and other phenolic compounds, such as epigallocatechin gallate, into GA. GA can, subsequently, be decarboxylated by gallate decarboxylase to yield PG as final product of tannin metabolism. The capacity for tannin degradation is a survival advantage over other gut microorganisms because tannins display antimicrobial activity (Chung et al., 1998), reduce the species richness in the gut (Tan et al., 2013) and probably improve the adaptation of tannin-degrading bacteria to the intestinal environment (Reverón et al., 2013).

Recently the CRC microbiome encompassing all DNA within the microbiota of late stage CRC patients has been disclosed (Tjalsma et al., 2012). This information enabled these authors to flesh out the microbiota that colonizes tumorous or adjacent non-malignant sites. If one focuses on these intestinal microorganisms and perform a search for homologous genes to the known *Lactobacillus plantarum* tannase gene (KEGG database http://www.genome.jp/kegg/), only bacterial spp. specifically overrepresented in tumorous tissues, but not in adjacent nonmalignant tissues, display tannase-homologous genes (**Table 1**).

**Abbreviations:** CRC, colorectal cancer; GA, gallic acid; PG, pyrogallol; ROS, reactive oxygen species.

#### **Table 1 | Gut microbial tannase in late stage CRC-patients.**


<sup>a</sup>On tumor: bacteria inhabiting tumorous tissues; Off tumor: bacteria inhabiting adjacent healthy tissues; On/Off tumor: bacteria inhabiting indistinctly tumorous and adjacent healthy tissues.

<sup>b</sup>Homologous genes to L. plantarum tannase lp\_2956.

<sup>c</sup>Two Clostridium spp. (C. botulinum E3 and C. strain SY8519) displayed genes with homology to L. plantarum tannase. However, this spp. are not reportedly associated to colorectal malignancies.

NA. Sequence not available.

This intriguing coincidence could be consistent with the selectivity of plant phenolic compounds to target cancer cells (Azmi et al., 2013) and suggest that the presence of gut microbial tannase activity may have potential implications on the colonic epithelial health.

### **IMPORTANCE OF TANNASE-GENERATED METABOLITES IN CRC DEVELOPMENT**

Recent findings have highlighted the potent DNA-damaging activities of tannins and tannase-derived metabolites, including PG, GA, and epigallocatechin gallate (Hossain et al., 2013). The genotoxic behavior of these metabolites is determined by its prooxidant activity which is in line with the pro-oxidant activity displayed by a number of further anti-cancer phenolic compounds in cellular/biological systems (Azmi et al., 2013).

Although the antioxidant capacity of tannins, PG and GA, is generally accepted, its pro-oxidant behavior should not be surprising. Thus, plants elaborate these natural phenolics to have, among other important functions, antimicrobial activity and antimicrobials share a common mechanism to kill bacteria: through the overproduction of ROS (Kohanski et al., 2007).

It is to be noted that the pro-oxidant behavior of tannasederived metabolites in CRC-tissues can be boosted in the colon compared to the small intestine due to a more pronounced pro-oxidant environment in the former (Sanders et al., 2004). The presence of cooper, a transition metal participating in the Fenton-like chemistry, is another factor for inducing the chemopreventive, pro-oxidant behavior of certain plant phenolic compounds (Azmi et al., 2013), including GA (Hossain et al., 2013), especially when it has been found in particularly augmented concentrations in a number of malignancies (Linder, 2012). Nevertheless some phenolics, such as PG, do not necessarily require a metal ion catalyst to produce OH radicals (Hossain et al., 2013).

Given the pro-oxidant activity of plant phenolics has been proposed as one of its anti-cancer properties (Azmi et al., 2013), the question arises as to whether the production of GA and PG by tannase-producing CRC-associated bacteria affects CRC development. There are several hints that the answer is yes. The potent ROS generation by GA and particularly PG strongly activates the tumor suppressor p53 in RKO CRC cells (Hossain et al., 2013). In addition GA and PG are known glutathione depletors (Park et al., 2007; You and Park, 2010) and could directly exert a cancer cell killing action via potent ROS generation. Hence, the selective production of GA/PG by the CRC-associated microbiota could promote pro-oxidative stress in the tumorous tissues to increase the direct or apoptotic killing (via p53 activation) of cancerous cells. Consistent with this hypothesis recent findings have suggested that tannase-mediated biotransformation of green tea extracts or epigallocatechin gallate *in vitro* modulates the expression of several pro and anti-apoptotic genes to enhance the pro-apoptotic behavior of these natural compounds (Macedo et al., 2012). Although details on these regulatory effects were not reported, an increased bioavailability of GA produced by tannase could account for the observed differences. Besides influencing apoptotic processes, GA modulates the expression of genes involved in cell cycle, angiogenesis, and metastasis inducing death in various cancer cell lines (Verma et al., 2013). Among other processes GA significantly modulates NF-κB, Akt, and ATM kinase signaling pathways to prevent the processes of carcinogenesis and it is a competitive inhibitor of the pro-inflammatory mediator COX-2.

### **ENHANCING THE ANTI-CARCINOGENIC POTENTIAL OF GALLIC ACID AND PYROGALLOL BY THE GUT TANNIN-DEGRADING BACTERIA COLONIZING COLORECTAL TUMOROUS TISSUES**

Gallic acid and PG display anticancer properties, however, it must be noted that its concentration in fruits and vegetables is generally low (Tomas-Barberán and Clifford, 2000). Hence the tannic acid transformation performance by gut tannin-degrading bacteria will likely play a key role to increase the concentration and bioactivity of GA and PG in the colon. The bioaccumulation of GA and PG in microbial biofilms may be another factor contributing to increase the anti-carcinogenic effects of these bioactive compounds. In this respect, biofilm formation on the collagen-rich surfaces accessible in the displaced mucosal epithelium of carcinomas is the most outstanding characteristic of *Streptococcus gallolyticus*, a tannin-degrading bacterium unquestionably associated to CRC (Tjalsma et al., 2012). This bacterium has been proposed to increase, via adhesion, colonization of tumorous tissues by members of *Fusobacterium* spp. [another bacterium with capacity to produce tannase (**Table 1**)] which in turn own capacity to adhere to other bacterial ssp. and attract them to the tumor (Tjalsma et al., 2012). According to this colonization pattern and given bacterial taxa of the microbiome colonizing colon tumor tissues generally own the capacity to produce tannase (**Table 1**), it is conceivable to speculate on the formation of multispecies biofilms on cancerous tissues settled by bacteria able to process dietary tannins. Microbe-generated GA and PG accumulated in these biofilms may thus selectively increase its bioavailability and bioactivity in the tumorous tissues. While this scenario is probably correct for tumorous tissues of late stage CRC patients, it is doubtful to anticipate the same in early and adenomatous stages as the occurrence of tannase-producing opportunistic pathogens in the healthy human gut, such as *S. gallolyticus* or *Fusobacterium* spp., is very rare (Tjalsma et al., 2012). In this regard, unraveling the microbiomes associated to the early and adenomatous stages of CRC will be crucial to define the specific role of tannase-producing bacteria in CRC development. In addition, the identification of the gut microbiota in CRC patients in epidemiological studies addressing the association between dietary tannin intake records and tumor recurrence or regression, may be critical in understanding the role of gut bacteria on the anti-cancer effects of dietary polyphenols.

### **CONCLUSION**

Tannase is an underlying gene function of gut microorganisms that selectively colonize tumorous tissues of late stage CRC patients. Based on the anti-carcinogenic inducing roles reported for GA and PG, the observed colonization pattern suggests that these gut tannin-degrading bacteria are involved in the prevention rather than promotion of tumor progression. Tannase then could be submitted as a link in the yet elusive relationship between dietary factors, CRC-associated gut microbiota and CRC progression. A drawback to this hypothesis is that the ROS-mediated anti-cancer effects of the GA and PG produced by the tannase-producing bacteria naturally colonizing tumorous tissues could be blocked by the high levels of ROS-destroying antioxidants usually possessed by the cancerous cells. Nevertheless, therapeutic microbiology strategies considering the affinity of tannase-producing bacteria for colonic tumorous tissues could be a powerful weapon to efficiently produce ROS-generating drugs that selectively kill CRC cells.

### **ACKNOWLEDGMENT**

This work was supported by grant AGL2011-22745.

### **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.

*Received: 06 October 2014; paper pending published: 03 November 2014; accepted: 20 November 2014; published online: 05 December 2014.*

*Citation: López de Felipe F, de las Rivas B and Muñoz R (2014) Bioactive compounds produced by gut microbial tannase: implications for colorectal cancer development. Front. Microbiol. 5:684. doi: 10.3389/fmicb.2014.00684*

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

*Copyright © 2014 López de Felipe, de las Rivas and Muñoz. 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.*

# Gut Microbiota Profiling: Metabolomics Based Approach to Unravel Compounds Affecting Human Health

#### Pamela Vernocchi <sup>1</sup> \*, Federica Del Chierico<sup>1</sup> and Lorenza Putignani 1, 2

<sup>1</sup> Unit of Human Microbiome, Genetic and Rare Diseases Area, Bambino Gesù Children's Hospital, IRCCS, Rome, Italy, <sup>2</sup> Unit of Parasitology, Bambino Gesù Children's Hospital, IRCCS, Rome, Italy

The gut microbiota is composed of a huge number of different bacteria, that produce a large amount of compounds playing a key role in microbe selection and in the construction of a metabolic signaling network. The microbial activities are affected by environmental stimuli leading to the generation of a wide number of compounds, that influence the host metabolome and human health. Indeed, metabolite profiles related to the gut microbiota can offer deep insights on the impact of lifestyle and dietary factors on chronic and acute diseases. Metagenomics, metaproteomics and metabolomics are some of the meta-omics approaches to study the modulation of the gut microbiota. Metabolomic research applied to biofluids allows to: define the metabolic profile; identify and quantify classes and compounds of interest; characterize small molecules produced by intestinal microbes; and define the biochemical pathways of metabolites. Mass spectrometry and nuclear magnetic resonance spectroscopy are the principal technologies applied to metabolomics in terms of coverage, sensitivity and quantification. Moreover, the use of biostatistics and mathematical approaches coupled with metabolomics play a key role in the extraction of biologically meaningful information from wide datasets. Metabolomic studies in gut microbiota-related research have increased, focusing on the generation of novel biomarkers, which could lead to the development of mechanistic hypotheses potentially applicable to the development of nutritional and personalized therapies.

Keywords: gut microbiota, metabolome, state of health, diseases, dietary habits, omic approach

### INTRODUCTION

The gut microbiota is exclusively responsible for several metabolic important functions, including vitamin and short chain fatty acid (SCFAs) production, amino acid (AAs) synthesis, bile acid biotransformation, hydrolysis and fermentation of non-digestible substrates (Putignani et al., 2015). The beneficial effects of gut microbiota include: (i) immune-cell homeostasis and development (Th1 vs. Th2 and Th17), (ii) epithelial homeostasis, (iii) enteric nerve regulation, (iv) support of angiogenesis, food digestion, and fat metabolism (Holmes et al., 2011).

The gut microbiota, through metabolite production/fermentation, modulates signaling pathways involved in the homeostasis of intestinal mucosa. When a balanced interaction between the gastrointestinal (GI) tract and the resident microbiota is disrupted, intestinal and extraintestinal

#### Edited by:

Enrica Pessione, University of Turin, Italy

#### Reviewed by:

Carl James Yeoman, Montana State University, USA Miguel Gueimonde, Spanish National Research Council, Spain Justin Joseph Donato, University of St. Thomas, USA

\*Correspondence:

Pamela Vernocchi pamela.vernocchi@opbg.net

#### Specialty section:

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

Received: 03 February 2016 Accepted: 08 July 2016 Published: 26 July 2016

#### Citation:

Vernocchi P, Del Chierico F and Putignani L (2016) Gut Microbiota Profiling: Metabolomics Based Approach to Unravel Compounds Affecting Human Health. Front. Microbiol. 7:1144. doi: 10.3389/fmicb.2016.01144 diseases may develop (Putignani et al., 2015), such as allergy, inflammatory bowel disease (IBD), obesity, cancer and diabetes, metabolic disorders, cardiovascular dyslipidemia, and neuropathology (Holmes et al., 2011).

The advent of the omics-based systems biology era has opened a new scenario in the comprehension of the gut ecosystem by shedding light on its shape, modulation and interplay with microorganisms, food functionality, and the role of nutrients in health (Moco et al., 2013; Putignani et al., 2015).

The "omics" technologies are presently applied to: (i) determine specific disease markers and novel diagnostic targets; (ii) discover functional alterations in the physiopathology of several diseases; (iii) discover the relationship between the gut microbiota and the host metabolisms (De Preter and Verbeke, 2013). In particular, the use of metabolomics, being a wellestablished and powerful top–down systems biology approach, is crucial to unravel the genetics–environment–health relation, as well as the typical clinical biomarkers of the different diseases (Moco et al., 2013). In fact, metabolomics changed the concept that the cellular metabolism profile is complete (Dettmer et al., 2007). Therefore, metabolomics is useful to elucidate the complex interactions of components, to understand the whole system, and to discover new metabolites in order to provide both a different perspective on cellular homeostasis (Liu et al., 2010), and new, unexpected pathways which may have a key physiological role (Zamboni et al., 2015). The metabolomics experiment (sampling, sample preparation, instrumental analysis, data processing, and data interpretation) fulfill the goal of improving the current status of biological information associated to the metabolome and, more generally, to functional genomics (Harrigan and Goodacre, 2003). Nowadays, metabolomics is used to: (i) identify biomarkers that could indicate the presence of a diseases, or a response to drug intervention (Dunn and Ellis, 2005); (ii) determine biochemical or environmental stresses (Le Gall et al., 2003); (iii) characterize microbial metabolism (Vaidyanathan et al., 2002); and (iv) characterize human health or disease (Holmes et al., 2011).

Indeed, the metabolomics approach has been applied to several studies on the gut microbiota, mostly focused on the exploration of disease-related metabolites in order to obtain detailed information on the gut metabolic pathways. In fact, the gut microbiota is involved in several biochemical functions directly associated to the perturbation of specific gut microbial populations, which may lead to the development of diseases (De Preter, 2015; De Preter et al., 2015). In other words, as the gut microbiota interacts with the host metabolism and affects physiological or pathological conditions, (**Figure 1**; **Table 1**; Del Chierico et al., 2012) the study of its composition helps discriminate between unhealthy and healthy subjects.

Moreover, the identification of metabolites may highlight how lifestyle and dietary habits affect specific disease conditions (Vernocchi et al., 2012).

Finally, metabolomics represents an unprecedented approach to collect the complex metabolic interactions between the host and its commensal microbial partners, offering the opportunity to define individual and population phenotypes (Moco et al., 2013). In fact, several cellular metabolites are associated with the phenotypes of living organisms (i.e., human, mice, bacteria), and they represent the substrates and products of different biochemical pathways reflecting genetic and environmental factors (Kim et al., 2016).

Furthermore, these data will serve as a basis to comprehend, at the cellular and molecular levels, the relationships between nutritional status and disease risk predisposition, thus allowing to formulate nutritional recommendations.

This review is focused on the application of MS- and NMRbased metabolomic techniques to describe the gut microbiota metabolome and human physiology in relation to nutritional programs and therapies.

### VOLATILE AND NON-VOLATILE COMPOUNDS: DETECTION METHODS AND DATA ANALYSIS

Metabolomics uses high throughput techniques to characterize and quantify small molecules in several biofluids (urine, serum, plasma, feces, saliva), revealing a unique metabolic signature (Nicholson and Lindon, 2008).

However, due to the chemical diversity, the different properties of metabolites, and the large dynamic range of metabolite concentrations in samples, it is almost impossible to measure the complete metabolome with only one technique (De Preter and Verbeke, 2013; Smirnov et al., 2016). Considering that the amount of predictable metabolites and derivatives in mammals, plants and bacteria is unknown (Weckwerth and Morgenthal, 2005), there is the need of different analytical platforms and complex integrated computational pipelines, adjusted by analytical and chemical parameters, to cover complete metabolome pathways in the different biofluids (Savorani et al., 2013). Moreover, the collection and preparation of samples, and the selection of the appropriate analytical platforms are fundamental requisites for reproducibility of sample manipulation (Dunn and Ellis, 2005). Besides, the storage and the continuous sample freeze/thawing may alter the composition and stability of the samples and consequently the precision and accuracy of results (Roessner et al., 2000).

Finally, the fundamental requirements of metabolomics studies are: accurate study design; sample treatment and platform set up, corroborated by data analysis; integration of results; and biological interpretation (Smirnov et al., 2016).

Metabolomics can be divided into two different groups: targeted analysis and non-targeted discovery analysis (Dettmer et al., 2007). In particular, the targeted approach is related to the analysis of the different classes of molecules (i.e., carbohydrates, lipids, aminoacids), while non-targeted analysis gives a rapid snapshot of the metabolic profile of samples by using technologies able to detect a wide number of metabolites (Smirnov et al., 2016).

### Detection Methods

At present, we are able to separate, detect, characterize, and quantify metabolites and their relevant metabolic pathways thanks to the rapid development of a range of

#### TABLE 1 | Role of gut microbiota metabolites on health and disease.


analytical platforms, including gas chromatography (GC), liquid chromatography (LC), high pressure LC (HPLC), ultra pressure LC (UPLC), Fourier transform infrared spectroscopy (FTIR), ion cyclotrone resonance-FT (ICR-FT), capillary electrophoresis (CE) coupled to mass spectrometry (MS), and nuclear and proton nuclear magnetic resonance spectroscopy (NMR-1H-NMR) (Zheng et al., 2011; Vernocchi et al., 2012).

MS-based metabolomics allows targeted and untargeted metabolome analysis and it has become an indispensable tool in metabolome analysis (Milne et al., 2013); moreover, MS has a broader dynamic range, exhibits a high sensitivity and selectivity (Zhao and Hartung, 2015), and allows determining metabolite fingerprints for establishing metabolome libraries, which facilitate the identification of metabolites (Martins-de-Souza, 2014).

### Gas Chromatography Mass Spectrometry

GC-MS is a combined system, with which thermally stable and volatile compounds are separated by GC and then eluting metabolites are detected by electron-impact (EI) mass spectrometers. GC-MS is considered as the gold standard in metabolomics (Harrigan and Goodacre, 2003). Even if GC has several advantages, such as high efficiency, reproducibility and sensitivity, it shows also some drawbacks. In fact, it can only be performed for volatile compounds, or those that can be made volatile, or made stable by derivatization (Roessner et al., 2000; Vernocchi et al., 2012).

Volatile organic compounds (VOCs) are important components of the metabolome (i.e., alcohols, esters, aldehydes, ketones, SCFAs) and are found in biological samples (Mills and Walker, 2001). The sample preparation methods consist of liquid or solid phase extraction (SPE) (Dettmer et al., 2007). Another rapid and solvent-free sample preparation technique is headspace-solid phase microextraction (HS-SPME) (Pawliszyn, 1997), for which the different types of stationary phases (polar and non-polar) used as fiber coatings are commercially available. On the other side, to stabilize the metabolites, two stages of derivatization with different kinds of reagents need to be performed (Roessner et al., 2000). During these processes, small aliquots of samples are analyzed by split or splitless mode on GC columns of various polarities, thus obtaining high chromatographic compound resolution and sensitivity, even if the resulted chromatograms are complex (i.e., multiple derivatization products), contain many metabolite peaks, and need a long run time (longer than 60 min) (Roessner et al., 2000). Therefore, coupling GC with time-of-flight (TOF)-MS, which has high scan rates and produces accurate peak deconvolution of complex samples in faster times, allows improving conventional GC–MS techniques in the analysis of ultra-complex samples (Dunn and Ellis, 2005; Dettmer et al., 2007).

Finally, metabolite quantification is obtained by external calibration or response ratio (peak area of metabolite/peak area of internal standard), while metabolite identification is obtained by matching retention time and mass spectrum of the sample peak with a pure compound previously analyzed, under identical instrumental conditions, with the same or different instruments (Fiehn et al., 2000), or by matching the metabolite against commercial databases (i.e., NIST, WILEY, EPA, NIH).

GC-MS can be used in several fields, such as plant metabolomics, as reported by Stashenko et al. (2004), who used SPME-GC-MS for sampling the volatile plant metabolites, or for example, by Akhatou et al. (2016), who combined GC-MS with multivariate statistical techniques to characterize the primary metabolome of different strawberry cultivars, and to study the influence of multiple agronomic conditions. Moreover, as reported by Currie et al. (2016), GC-MS has proved useful in microbial metabolomics related to pharmaceutical studies to analyze the endogenous metabolite levels produced by Pseudomonas putida in response to six pharmaceuticals; or in food studies, when used to characterize the microbial metabolite production in: cheese (Vannini et al., 2008; Pisano et al., 2016), probiotic food (Patrignani et al., 2009; Tabanelli et al., 2015b), sourdough (Guerzoni et al., 2007), wine (Vernocchi et al., 2011; Patrignani et al., 2016), sausages (Tabanelli et al., 2015a), and ready to eat products (Siroli et al., 2015). Moreover, GC-MS is used in clinical applications, for example to analyze volatile compounds (SPME-GC-MS) in urine, blood, feces, hair, breath and saliva (Mills and Walker, 2001), or to evaluate biomarkers in several diseases, such as asthma (Gahleitner et al., 2013; Chang et al., 2015), schizophrenia (Liu et al., 2010), depressive disorders (Ding et al., 2014), ulcerative colitis (Kohashi et al., 2014), and neonatal sepsis (Fanos et al., 2014). In the last years, GC-MS has become one of the most used techniques to study the modulation of gut microbiota as a result of nutrition (i.e., diet, nutraceutical food consumption), diseases, drug, and probiotic administration. Garner et al. (2007) qualitatively and quantitatively analyzed the fecal metabolome to identify potential biomarkers in GI diseases; Di Cagno et al. (2011) characterized the fecal metabolome of celiac children subjected to glutenfree diet, compared to healthy children; Francavilla et al. (2012) evaluated the gut metabolome of allergic children; Vitali et al. analyzed the effects of symbiotic or prebiotic foods and probiotic foods on the human gut metabolome profile (Vitali et al., 2010, 2012); De Filippis et al. evaluated the effects of the Mediterranean diet on the gut microbiota metabolome (De Filippis et al., 2015).

Moreover, De Preter (2015) and De Preter et al. (2015) applied this technique to make a clinical diagnosis of IBD, and to determine the impact of prebiotics on Chron's disease. Del Chierico et al. characterized the gut microbiota of non-alcoholic fatty liver disease (NAFLD), and obese pediatric patients to unravel disease signatures (Del Chierico et al., 2016). Finally, De Angelis et al. analyzed the fecal metabolome of children with autism and pervasive developmental disorders (De Angelis et al., 2013).

### Liquid Chromatography Mass Spectrometry

HPLC separation may cover a wide range of compounds' determination, even though its resolution is low. LC is probably the most flexible separation method, as it allows to separate compounds with little effort in few pre-analytical steps (compared to GC-MS) (Moco et al., 2007). Usually, the metabolite separation obtained with LC is followed by electrospray ionization (ESI) or, to a lesser extent, by atmospheric pressure chemical ionization (APCI) (Bakhtiar et al., 2004). The combination of LC with MS allows to analyze polar, non-polar and neutral compounds separately in a complex matrix (Smirnov et al., 2016). This technique diverges from GC-MS for the lower temperatures of analysis, and it does not require sample volatility, thus entailing an easier sample preparation (Dunn and Ellis, 2005).

LC/MS is an excellent technique showing sensitivity, specificity, resolving power, and capability to extract additional information about metabolites from their retention time (RT) domain (Forcisi et al., 2015).

Sample derivatization is commonly not necessary, although it can be helpful to improve the chromatographic sensitivity and resolution (Leavens et al., 2002), or to produce ionisable groups of metabolites otherwise not detectable by electrospray ionization (ESI) MS. Metabolite quantification is obtained by external calibration or response ratio, and metabolite identification is more time intensive. Moreover, ESI does not produce molecular ion fragmentation as it occurs by electron impact MS, so it does not provide direct metabolite identification by ESI mass spectra comparison, as ESI mass spectral libraries are not generally available. Nevertheless, accurate mass measurements can be obtained by coupling MS/MS using metabolite identification (Lenz et al., 2004). The advent of HPLC and UPLC allowed to shorten the analyzing time, provided higher resolution, sensitivity and efficiency, and permitted to reduce the quantity of samples and solvent necessary for the analysis (Smirnov et al., 2016).

The application of LC/MS allows the identification of target metabolites within a complex sample, not only with the information about monoisotopic mass, but also providing advice on the metabolite structure (Villas-Bôas et al., 2005).

LC-MS applications mainly concern the clinical and pharmaceutical fields (Bakhtiar et al., 2004). Nardotto et al. (2016) used LC/MS/MS systems to investigate patients with type 2 diabetes mellitus treated with an oral dose of racemic carvedilol, who showed accumulation in plasma. Mueller et al. applied this technique to measure plasma concentrations of trimethylamine-N-oxide, betaine and choline in the evaluation of patients with suspected coronary artery disease (Mueller et al., 2015).

An example of LC and GC technologies' combination, is also given by Chow et al. who studied the fecal metabolome, including the application of non-targeted metabolomics to separate breastfed from formula-fed infants by using GC/MS and LC/MS/MS analysis to identify the various metabolites undergoing change (Chow et al., 2014).

### Capillary Electrophoresis Mass Spectrometry

CE may offer high-analyte resolution and detect a wider spectrum of (polar) compounds compared to HPLC, but it is properly applicable only to charged analytes (Ramautar et al., 2013). However, only a few studies on this subject have been published to date, such as Soga et al. (2003), who separated cationic, anionic nucleotidis, and CoA metabolites to describe the coverage of the metabolome. These authors analyzed 1692 metabolites in bacterial extract (Chow et al., 2014).

### Fourier Transform Infrared Spectroscopy

FT-IR spectroscopy allows rapid, non-destructive and highthroughput determination of different sample types. In particular, it can simultaneously detect different molecules, such as lipids and fatty acids (FAs), proteins, peptides, carbohydrates, polysaccharides, nucleic acids, (Harrigan and Goodacre, 2003; Dole et al., 2011), but sensitivity and selectivity of this technique are not high. On the contrary, ICR-FT/MS offers an ultrahigh mass resolution able to distinguish slight variations in a wide number of mass signals (Rosselló-Mora et al., 2008), and allowing to obtain the structural identification of new biomarkers (Jansson et al., 2009). In fact, Jansson et al. (2009) used ICR-FT/MS to distinguish between the masses of fecal metabolites in Chron's disease patients and healthy subjects.

Furthermore, FT-IR is principally useful for the identification of functional groups (Vernocchi et al., 2012). In fact, FT-IR has been used to assist infrared imaging in the diagnosis of many diseases, such as Parkinson, cancer, Alzheimer, kidney stone, arthritis (Dole et al., 2011), diabetes, and early stage insulin resistance (Chen et al., 2008).

Moreover, it is possible to combine the LC and FT-IR techniques, as performed by Walker et al. (2014), who identified taurine and sulfate conjugated fatty acids in feces of diabetic mice by coupling ICR-FT/MS and UPLC-MS.

Finally, MS-based metabolomic techniques offer high selectivity and sensitivity for metabolites' identification and quantification. In fact, they are considered as the most appropriate techniques for the detection of large numbers of metabolites, and, in combination with advanced and highthroughput platforms, they may help decrease the complexity of metabolite separation (Zheng et al., 2011; Zhao and Hartung, 2015). In particular, for the total screening of the small molecules in a biological system, MS non-targeted metabolomics is a powerful tool for the identification of metabolite signals present in spectra (Naz et al., 2014). The identification can be partly completed by matching against metabolite and spectral databases, such as METLIN (Smith et al., 2005), HMDB (Wishart et al., 2013), or ChemSpider (Pence and Williams, 2010). To describe the metabolic pathway of biological systems, it is also possible to refer to databases such as KEGG (Kanehisa et al., 2007), or Meta-Cyc (Caspi et al., 2007).

### Nuclear Magnetic Resonance Spectroscopy

NMR spectroscopy, instead, uses the intramolecular magnetic field around atoms in molecules to change the resonance frequency, thus allowing access to details of molecules' electronic structure and obtaining information about their dynamics, reaction state, and chemical environment. Moreover, a minimal sample preparation is necessary for biofluids, except for feces or gut luminal content, which require the removal of undigested material, dead microbes and other particles (Smirnov et al., 2016). For feces in particular, samples are prepared using methanol (for lipophilic compounds as lipids, cholate, small phenolic acids) or water (for aminoacids, glucose, glycerol) (Jacobs et al., 2008; Lamichhane et al., 2015).

NMR spectroscopy is mainly useful to determine metabolic fingerprints leading to the identification and quantification of compounds in a non-targeted large-scale, in a non-destructive way, and with a high reproducibility (Lenz et al., 2004; Smolinska et al., 2012).

However, it is a relatively insensitive technique, and can only detect metabolites in high concentrations. The annotation is restricted to a limited number of low molecular weight molecules, and this is the major pitfall of this application (Jansson et al., 2009; Zhao and Hartung, 2015). Indeed, sensitivity depends on the natural concentration of the atoms in the matrix and, to improve sensitivity, long times of analysis, higher magnetic fields, and cryogenic probes are needed (Keun et al., 2002).

Another type of NMR spectroscopy is <sup>1</sup>H (proton) NMR, which is unbiased to particular metabolites (Dunn and Ellis, 2005), unlike the other techniques discussed above. The NMR spectrum (chemical shift) depends on the shielding from electrons orbiting the nucleus, whereas for <sup>1</sup>H-NMR the chemical shift is arranged as the difference between the resonance frequency of the observed proton and that of a reference proton in a reference metabolite (tetramethylsilane in solution, set at 0 ppm) (Dunn and Ellis, 2005).

The obtained spectra are complex and contain a wide number of signals, and frequently pure metabolites can be added to give a more in-depth clarification. This technique is frequently employed in clinical and pharmaceutical research and applications, in particular in the analysis of biofluids and tissues, where <sup>1</sup>H-NMR is used to detect the modulation of metabolites in response to cellular stresses (Lindon et al., 2003). This has also been reported by Bro et al. (2015), who used plasma to determine breast cancer biomarkers, or by Villaseñor et al. (2014), who described the global metabolic phenotyping of acute pancreatitis, and by Dumas et al. (2016), who used this technique to study metabolic syndrome and fatty liver disease.

Moreover, when there is the need to understand how the diet or other external stimuli or diseases affect the microbiome composition (i.e., gut, urine, salive), metabolite detection is performed using this functional technique. In fact, several research studies have been conducted using NMR and <sup>1</sup>H-NMR, for instance: Ndagijimana et al. (2009) described the effects of symbiotic food on human gut metabolic profile; Martin et al. (2012) studied the influence of gut metabolome on health and diseases; Bjerrum et al. (2015) investigated the gut metabolic biomarkers characterizing Chron's disease, ulcerative colitis and healthy subjects; Zhang et al. (2015) studied how the gut microbiota metabolome could alleviate obesity in children; Laghi et al. (2014) studied the antibiotic effect on vaginal microbiome; and Holmes et al. (2008) analyzed urine samples to discriminate metabolites across populations in order to identify major risk factors for coronary heart disease and stroke.

Hence, MS and <sup>1</sup>H-NMR are by far the most frequently applied and the most powerful techniques in metabolomics (Collino et al., 2012). Recent advances in NMR and MS have allowed to evaluate at the same time thousands of metabolites related to the "metabolome," and to define the end-points of metabolic processes in living systems (Nicholson et al., 2005).

In particular, <sup>1</sup>H-NMR is currently the most used analytical technique for metabolite profiling compared to MS, while the combination of <sup>1</sup>H-NMR and MS technologies would result in a better coverage of the complete metabolome (Bjerrum et al., 2015; De Preter, 2015; Wissenbach et al., 2016). In fact, by coupling metabolite separation technologies with spectrometry and spectroscopy, it is possible to reach a multidimensional approach leading to the structural identification of new metabolites (Chen et al., 2008).

However, the challenge for metabolomics is not only to discover unknown chemical structures, but also to generate meta-information, (i.e., sample origin, tissue, experimental conditions) in an accessible format (Weckwerth and Morgenthal, 2005). Thus, the structural identification of metabolites as potential biomarkers associated with diseases will be a major task of biological interpretation (Nassar and Talaat, 2004). In fact, small molecule metabolites are able to provide new mechanistic information on novel disease biomarkers, which is extremely important, given the paucity of existing markers. Moreover, metabolomics can induce significant progress in the identification of metabolomic fingerprints (which could be used as crucial diagnostic biomarkers) by producing a comprehensive map of metabolic pathway regulations, which represent the downstream expression of genome, transcriptome, and proteome. This comprehensive map may help define the phenotype of an organism at a specific time (Zhang et al., 2012). Therefore, the analysis of metabolic differences between unperturbed and perturbed pathways could provide insights on the underlying disease prognosis and diagnosis (Zhang et al., 2012; **Figure 2**).

### Data Analysis

Hence, statistic and bioinformatic techniques are used for data mining complex metabolic profiles containing information related to genetics, environmental factors, gut microbiota activity, lifestyle, and eating habits. These strategies support the complicated process of identifying new biomarkers, which could indicate the individual response to specific physiological factors and/or nutritional interventions, and manage the relevant biological outcomes (Moco et al., 2013).

The application of biostatistics and mathematical approach has a key role in the extraction of biologically meaningful information from wide datasets. In computational analysis, the problems derive from a small batch of samples in contrast with the high number of detected metabolites, and in the consequent high dimensionality of the data matrix (Weckwerth and Morgenthal, 2005). Therefore, different statistical tools can be employed to discriminate among the samples and within the sample set (Worley and Powers, 2012).

In particular, diverse univariate and multivariate methods can be used as parametric (i.e., Student t-test, multivariate linear regression) or non-parametric (i.e., Mann-Whitney, Kruskal-Wallis) tests. Moreover, these methods can be divided into unsupervised techniques (i.e., principal component analysis (PCA), hierarchical cluster analysis), and supervised techniques

(i.e., linear discriminant analysis, k-nearest neighbor). Among the supervised multivariate techniques, the partial least squares discriminant analysis (PLS-DA) has proved to be a particularly useful tool, in the presence of an irresolute sample separation obtained from PCA as it offers the possibility to use a priori information based on replicates (Raamsdonk et al., 2001; Gromski et al., 2015). PLS-DA is also used for its ability to identify potential biomarkers.

Moreover, statistical analyses of multivariate datasets allow to visualize the biological and molecular consistency. This consistency is based on the correlated functioning of metabolites in response to external conditions. Finally, correlation networks represent fingerprints of biochemical interactions, like the regulation of enzyme activity, and the interplay of anabolism and catabolism between the host and gut microbes (Weckwerth and Morgenthal, 2005).

### METABOLITES ASSOCIATED WITH MICROBIAL METABOLISM OR MICROBIAL–HOST COMETABOLISM

The gut microbiota operates in a combined way with the host through the metabolic communication enacted by the different bacterial genera and species responsible for metabolite production (**Table 2**).

### Bacterial Metabolism SCFAs

Several substances present in the large intestine, including indigestible oligosaccharides, dietary plant polysaccharides or fibers, non-digested proteins and intestinal mucin are fermented by the microbiota populations to produce SCFAs (Arora and Sharma, 2011). In particular, clostridial clusters IV, XIVa (e.g., Eubacterium, Roseburia, Faecalibacterium, and Coprococcus spp.) and Lactobacillus belonging to the phylum Firmicutes and the groups of Actinobacteria (Bifidobacterium spp.) are the main bacteria playing a central role in SCFAs metabolism (Nicholson et al., 2012; van Zanten et al., 2014).

Acetate is an important SCFA present in the colon, which could have a trophic effect on the colonic epithelium not only by local action, but also by raising the mucosal blood flux. Moreover, after transport to the portal circulation across the colonic mucosa, acetate passes through the liver and is regained in peripheral blood. Acetate's effect at the ileal level exceeds that of mixed SCFA (Scheppach, 1994). However, acetate is adsorbed by tissues involved in the rise of cholesterol synthesis (Scheppach et al., 1991). On the other side, propionate inhibits cholesterol TABLE 2 | Metabolites associated with microbial metabolism or microbial–host cometabolism.


synthesis (Scheppach, 1994; Wong et al., 2006). In fact, substrates that can decrease the acetate/propionate ratio may diminish serum lipids and consequently decrease the risk of cardiovascular disease (Wong et al., 2006). Butyrate represents the major energy source for colonocytes and has been studied for its role in nourishing the colonic mucosa and preventing colon cancer by promoting cell differentiation, cell-cycle arrest and apoptosis of transformed colonocytes (Scheppach, 1994; Velázquez et al., 1997; Walton et al., 2013).

Furthermore, butyrate improves insulin sensitivity and raises energy consumption in obese mice submitted to dietary regimen (Gao et al., 2009), and butyrate irrigation (enema) improves inflammation in diversion colitis (Scheppach et al., 1992; Wong et al., 2006). Butyrate and propionate, but not acetate, induce the production of gut hormones and reduce food intake (Lin et al., 2012). The treatment with acetate induces a marked reduction in lipid accumulation in the adipose tissue, protects against accumulation of fat in the liver, and improves glucose tolerance (Yamashita et al., 2007). In obese subjects, propionate significantly increases the release of post-prandial plasma peptide YY and glucagon-like peptide-1 from colonic cells, and reduces energy intake (Chambers et al., 2015). Chambers et al. (2015) found that inulin-propionate ester administrated to overweight adults significantly reduced weight gain, intra-abdominal adipose tissue distribution, and intrahepatocellular lipid content, and improved insulin resistance in the inulin control group.

Furthermore, other clinical studies demonstrated that the administration of SCFAs has a positive effect on the treatment of ulcerative colitis, Crohn's disease, and antibiotic-associated diarrhea and obesity, metabolic syndrome, bowel disorders, and cancer (Harig et al., 1989; Sabatino et al., 2005; Binder, 2010; Donohoe et al., 2011; Fukuda et al., 2011; Chambers et al., 2015). The degradation of proteins and amino acids by gut microbes also forms small amounts of branched chain FAs (isobutyrate, valerate and iso-valerate) (Macfarlane and Gibson, 2004). SCFAs can be detected by using both the GC-MS and <sup>1</sup>H-NMR spectroscopy techniques.

### Organic Acids

Several organic acids result from bacterial metabolism of dietary polyphenols or unassimilated AAs or carbohydrates (Lord and Bralley, 2008). High levels of organic acids in urines are associated with microbial overgrowth (Lord and Bralley, 2008). In particular, the hyperproduction of organic acids is associated with the overgrowth of Clostridium difficile, Faecalibacterium prausnitzii, Bifidobacterium spp., Subdoligranulum spp., Lactobacillus (Lord and Bralley, 2008; Nicholson et al., 2012).

Amongst organic acids, urinary hippuric acid may be a biomarker of hypertension and obesity in humans, while urinary 4-hydroxyphenylacetate and phenylacetate are potential biomarkers of colorectal cancer (Nicholson et al., 2012).

Lactic acid is the main product in the lactic acid bacteria (LAB) fermentation process. LAB are Gram+ and constitute a heterogeneous group of microorganisms that can also produce proteinaceous antimicrobial molecules, known as bacteriocins, that can help the producer microorganism to outcompete other bacterial species (Alvarez-Sieiro et al., 2016). Moreover, lactic acid represents a secondary metabolite that can be converted by clostridial cluster XIVa species into butyrate, or by clostridial cluster IX into propionate (Louis et al., 2007), thus inducing benefits by inhibiting both the propagation of harmful bacteria, and the production of putrefactive intestinal products. Lactic acid also participates in the intestinal peristalsis regulation (Sugawara et al., 2016). Furthermore, lactic acid is correlated to healthy vaginal microbiota, in fact it decreases in bacterial vaginosis, and it's produced by microbial species such as Lactobacillus crispatus and Lactobacillus jensenii (Vitali et al., 2007, 2015; Cruciani et al., 2015; Srinivasan et al., 2015). The detection of organic acids is most commonly obtained using LC-MS and <sup>1</sup>H-NMR spectroscopy platforms.

### Vitamins

Vitamins are indispensable micronutrients, essential for biochemical reactions in all organisms. Humans are unable to synthesize most vitamins, hence, most of them need to be obtained exogenously, and some are produced by the gut microbiota (Stanton et al., 2005; Rossi and Amaretti, 2010).

Recently, Magnúsdóttir et al. (2015), using the PubSEED platform, assessed the genomes of 256 human gut bacteria involved in the biosynthesis of eight B-vitamins: biotin, folate, cobalamin, niacin, pantothenate, riboflavin, pyridoxine and thiamin. In particular, the authors demonstrated that each of the reported vitamins was produced by 40–65% of the 256 human gut microbes (Magnúsdóttir et al., 2015). Moreover, the absorption of some vitamins occurs in the small intestine after conjugation of vitamins with molecules (intrinsic factors) which are produced in the stomach. Since some vitamins are synthetized by the colonic microbiota, they are not adsorbed by the colon but are excreted in feces (Wilson, 2005).

Bifidobacteria strains have been recognized to be the strongest vitamin producers (Deguchi et al., 1985; Noda et al., 1994; Pompei et al., 2007), and in particular Bifidobacteria and Lactobacilli have been proposed as possible folate producers (Pompei et al., 2007; Kleerebezem and Vaughan, 2009). Folate (vitamin B9) is involved in various essential metabolic functions, such as DNA replication, repair and methylation, and synthesis of nucleotides, vitamins and certain AAs (LeBlanc et al., 2013). Folate is contained in leaf vegetables, cereals and liver.

The biosynthesis of thiamin (vitamin B1) consists of two pathways that unite in the final step of thiamin monophosphate production. Although thiamin diphosphate is the functional version of thiamin, all phyla (in particular Bacteroidetes and Fusobacteria), except Firmicutes are producers of thiamin monophosphate (Magnúsdóttir et al., 2015). Vitamin B1 is contained in pork meat, oatmeal, brown rice, vegetables, potatoes, liver, and eggs.

Biotin (vitamin B8) can be synthesized de novo from two pimeloyl precursors, namely malonyl-ACP and pimelate. Fusobacteria, Bacteroidetes and Proteobacteria synthetize biotin by different biochemical pathways, while Actinobacteria genomes lack the essential role of biotin biosynthesis (Magnúsdóttir et al., 2015). Vitamin B8 is contained in raw egg yolk, liver, peanuts and green leafy vegetables.

Riboflavin (vitamin B2) plays an essential role in cellular metabolism, being the precursor of the coenzymes flavin mononucleotide (FMN) and flavin adenine dinucleotide (FAD) (Li et al., 2014). Microbial riboflavin biosynthesis has been extensively described in Bacillus subtilis (Perkins and Pero, 2002) and Escherichia coli (Bacher et al., 1996, 2000). Vitamin B2 is contained in dairy products, bananas, popcorn, green beans and asparagus.

Cobalamin (vitamin B12) is the only vitamin that is exclusively produced by microorganisms, particularly by anaerobes (Roth et al., 1996; Martens et al., 2002; Smith et al., 2007).

Besides, the production of niacin and pyridoxine appears to be generated by Lactobacilli used in yogurt, cheese, and fermented foods (Shahani and Chandan, 1979; Alm, 1982).

Pantothenate (vitamin B5) is a coenzyme A (CoA) precursor and it can be synthesized de novo from 2-dihydropantoate and balanine. Bacteroidetes and several species of Proteobacteria and Actinobacteria have been demonstrated to be CoA producers (Magnúsdóttir et al., 2015). This vitamin is also contained in meat, broccoli, avocado. Detection of this vitamin is mostly obtained using LC-MS and <sup>1</sup>H-NMR spectroscopy platforms.

Furthermore, Vitamin K operates as a co-factor for the enzymatic conversion of specific protein glutamyl to γcarboxyglutamyl residues. The daily requirement of vitamin K is satisfied by dietary intake of phylloquinone and by the production of polyisoprenyl-containing compounds synthesized by the human gut microbiota (Suttie, 1995; Davidson et al., 1998; Martens et al., 2002). Green leafy vegetables such as spinach, egg yolks, and liver contain vitamin K.

### Bacterial-Transformed Compounds Bile Salts

The metabolism of bile salts is a well-known and basic skill of the gut microbiota metabolism, particularly associated to the genera of Bacteroides, Clostridium, Lactobacillus, Bifidobacterium, Enterobacter, Eubacterium, and Escherichia (Ridlon et al., 2006; Nicholson et al., 2012). Bacteria also contribute to the recovery of bile salts escaping from active transport in the distal ileum (Begley et al., 2006). The gut microbiota chemically modifies bile acids through a wide range of reactions, resulting in the formation of secondary and tertiary bile acids (Bortolini et al., 1997). Bile salts contribute to the absorption of dietary fats and lipid-soluble vitamins, facilitate lipid assimilation, maintain gut barrier function and regulate triglycerides, cholesterol and glucose by endocrine functions and energy homeostasis (Groh et al., 1993; Ridlon et al., 2006; Dawson et al., 2009).

However, bacterial bile salt hydrolysis has recently been considered as a risk factor for the development of colon cancer because it causes the formation of harmful secondary bile salts after an initial deconjugation step (De Boever et al., 2000). The secondary free bile acids can diffuse through the lipid bilayer of the membrane, thus being much more inhibitory for the cells than the conjugated forms (Mayo and van Sinderen, 2010). De Boever et al. (2000) have speculated a plausible mechanism for the protective properties of probiotic Lactobacillus reuteri, which could precipitate the deconjugated bile salts by a physical binding, making the harmful bile salts less bioavailable. Bile salts have antimicrobial activity on gut microbes with inhibitory effects on Bacteroidetes and Actinobacteria microbial population (Islam et al., 2011), but high levels of these biomarkers in serum and urine are correlated with liver diseases (Bathena et al., 2015). The LC-MS and <sup>1</sup>H-NMR spectroscopy platforms are the main techniques used to detect bile acids.

### Polyphenols

Polyphenols are considerably bioactive components in the diet (Manach et al., 2004). Hydroxycinnamic acids and flavonoids are the two major classes of polyphenols. Fruits commonly contain caffeic acid, representing the most abundant hydroxycinnamic acid (Clifford, 2004). In particular, the chemically derived chlorogenic acid is commonly present in apples, berries and kiwifruit, in vegetables such as potatoes (Manach et al., 2004) and, in high concentrations, in coffee (Clifford, 2004).

Recent studies have demonstrated that gut bacteria, including strains of Lactobacillus and Bifidobacterium, can metabolize chlorogenic acid to form caffeic acid and quinic acid (Couteau et al., 2001; Taverniti and Guglielmetti, 2012; Amaretti et al., 2015; Marín et al., 2015; Raimondi et al., 2015), while caffeic acid is further metabolized to form the µ-coumaric acid (3-hydroxycinnamic acid), 3-hydroxylphenylacetic acid and dihydroxyphenylpropionic acid (Konishi and Kobayashi, 2004). The 3,4- dihydroxyphenylacetic acid also derives from the colonic catabolism of rutin (Jaganath et al., 2009). Conversely, phenolic acid metabolites of rutin are not produced in germ-free mice, implying that ring-fission products are generated only by intestinal bacteria (Selma et al., 2009; Parkar et al., 2013). The polyphenol detection is performed using LC-MS, GC-MS, and <sup>1</sup>H-NMR spectroscopy platforms.

### Lipids

Significant amounts of glycerol derive from daily dietary intake and/or from in situ microbial production, or from enterocyte desquamation. Some gut bacteria may anaerobically reduce glycerol to 1,3-propanediol, with the production of the intermediate 3-hydroxypropanal. The accumulation of this metabolite leads to the formation of reuterin, which is known for its antimicrobial properties (De Weirdt et al., 2010). Lipids are also involved in intestinal permeability, in the regulation of glucose homeostasis via intestine-brain-liver-neural axis, in the promotion of chronic systemic inflammation by LPS, in the improvement of hyperinsulinemia by conjugated fatty acids (FAs), in the enhancement of the immune system, and in the alteration of lipoprotein profiles. Bifidobacterium, Roseburia, Lactobacillus, Klebsiella, Enterobacter, Citrobacter, Clostridium genera have been recognized as the main actors in lipid metabolism (Nicholson et al., 2012). The platform constituted of LC-MS, GC-MS, and <sup>1</sup>H-NMR spectroscopy is mainly used to perform lipid detection.

### Amino Acids

The bacterial fermentation of proteins, occurring in the distal colon, leads to AAs fermentation products having some relevance for health (Macfarlane and Macfarlane, 1995). For instance, AAs deamination produces ammonia, whereas decarboxylation produces amines, which may have toxicological effects (Silla Santos, 1996). High ammonia concentrations have been found to act as tumor promoters (Clinton et al., 1988). Bacterial degradation of AAs cysteine and methionine leads to the formation of H2S, which is toxic and has also been reported to be responsible for inhibition of butyrate oxidation in colonocytes (Roediger et al., 1993). Furthermore, the anaerobic fermentation of the aromatic AAs tyrosine and tryptophan by colonic bacteria produces phenols and indoles respectively, which are eventually excreted in the urine (Macfarlane and Macfarlane, 1997). Phenols, such as p-cresol, have been proposed to act as procarcinogens in colon cancer (Bone et al., 1976). Interestingly, gut bacterial production of p-cresol is significantly related to autism (Clayton, 2012; De Angelis et al., 2013), and C. difficile appears to be a significant p-cresol producer (Sivsammye and Sims, 1990).

Finally, certain species of Clostridium (Moss et al., 1970) and Peptostreptococcus anaerobius (Lambert and Moss, 1980) can convert phenylalanine to benzoic acid in a multistep process with phenylpropionic acid (toxic metabolic product) acting as intermediate (Macfarlane and Macfarlane, 1997; Smith and Macfarlane, 1997). The AAs are mainly detected using LC-MS and <sup>1</sup>H-NMR spectroscopy techniques, and also using FT-ICR-MS.

### BIOLOGICAL ACTION OF THE GUT MICROBIOTA IN HEALTHY AND DISEASED SUBJECTS

In recent years, growing attention has been targeted to the role of the gut microbiota in the pathogenesis of gastrointestinal (GI) diseases (Lozupone et al., 2012). The alteration of the interplay between host and microbes at the gut level stimulates perturbation of the homeostasis and leads to the development of disorders.

The microbial ecosystem undergoes changes when the equilibrium is broken, which leads to the modification of the bacterial metabolic activity, and/or to transfers in the distribution of local bacterial communities. In fact, the phylotype complexity regulates the equilibrium between pathogenic and commensal taxa at the GI interface (Prakash et al., 2011).

The intestinal gut dysbiosis is associated with a plethora of children and adult diseases, including genetic (i.e., cystic fibrosis [CF]), inflammatory (i.e., inflammatory bowel diseases and syndrome [IBDs, IBS], Chron's [CD], ulcerative colitis [UC], and celiac disease), metabolic (i.e., diabetes, obesity and non-alcoholic fatty liver disease [NAFLD]), and allergic (i.e., atopic dermatitis, food allergies) disorders (Del Chierico et al., 2012), and neuropathologies (i.e., autism) (**Figure 2**).

Indeed, metabolomics is an approach allowing to perform a careful diagnosis of diseases, since metabolite profiles have a high resolution power, which enable to separate the groups based on microbial community profiles (Dicksved et al., 2008). Moreover, metabolites represent the terminal enzymatic process signature occurring in the gut, and the molecules within the pathways range allow to distinguish healthy from diseased subjects, as well as among disease phenotypes (Jansson et al., 2009; **Table 3**).

### Inflammatory Bowel Disease/Inflammatory Bowel Syndrome (IBD/IBS)

As concerns inflammatory diseases at GI tract, it is well known that the microbiota results to be abnormal both in IBD and IBS, showing decreased levels of Actinobacteria and Firmicutes, and high levels of Proteobacteria compared to healthy subjects (Kinross et al., 2011; Carroll et al., 2012; Mukhopadhya et al., 2012; De Preter et al., 2013).

Irregular microbial fermentation leads to a high production of hydrogen (in IBS), indole, phenols and others (Kumar et al., 2010). In fact, bacteria release volatile organic compounds (VOCs), determined by SPME-GC-MS, as by-products of metabolism. Hence, the rising acceptance of the gut microbiota involvement in the pathogenesis of IBD has led to the use of fecal matrix as a sample to determine metabolite profiling (Walton et al., 2013). Indeed, specific microbial VOCs profiles can provide specific biomarker candidates for diagnostic purposes (Schöller et al., 1997; Lechner and Rieder, 2007; Bunge et al., 2008).

Walton et al. (2013) observed differences among patient categories (IBD, UC, and CD) based on compounds detected in fecal samples, such as SCFAs and their corresponding alcohols, esters, and molecules, such as indoles and phenols, acetone and sulfur compounds.

The concentrations of propanoic and butanoic acids, revealed by using GC-MS, represent a source of energy affecting colonic mucosal growth, and these concentrations were found to be higher in CD subjects, compared to healthy controls (Best and Laposata, 2003), while acids such as oleic, stearic, palmitic, linoleic and arachidonic were higher in the ileum of CD patients (Jansson et al., 2009). Ahmed et al. observed an increase of esters in diarrhea predominant IBS patients using the SPME-GC-MS technique (Ahmed et al., 2013). On the other side, Walton et al. (2013) detected high levels of indole, phenol and p-cresol, generally considered to be toxic for the gut, in CD and UC groups compared to controls.

Moreover, Jansson et al. (2009), using FT-ICR-MS, detected several masses related to metabolites within the tyrosine metabolic pathway, which differentiated CD from healthy controls. In particular, dopaquinone (a dopa oxidation product and intermediate in the melanin formation from tyrosine) was significantly elevated in CD patients compared to healthy subjects. The authors also indicated that tryptophan and phenyalanine were related to the ileum CD phenotype. It was also observed, using <sup>1</sup>H-NMR spectroscopy, that AAs in CD patients with active disease showed a different profile (i.e., alanine, isoleucine, leucine, and lysine) compared to CD patients in remission (Marchesi et al., 2007). Furthermore, metabolites related to bile acids pathways (i.e., glycocholate) were found in CD patients in remission (Jansson et al., 2009).

Other studies (Marchesi et al., 2007; Bjerrum et al., 2015; De Preter, 2015; De Preter et al., 2015) using <sup>1</sup>H-NMR and GC-MS showed a depletion of bacterial products, such as SCFAs, branched chain FAs, dimethylamine and trimethylamine,


3|Correlatedmicrobesandmetabolitestodiseasesandtherelativemetabolomic

and high levels of AAs, suggesting a breakdown of the normal bacterial ecology that induces dysbiosis, as a cause or a consequence of the disease in IBD. Le Gall et al. (2011), using Denaturing Gradient Gel Electrophoresis (DGGE), discriminated UC patients from controls, and also found a correlation between the gut microbiota composition and the metabolite composition (high levels of cadaverine and taurine). Besides, UC patients with high levels of AAs showed low fecal concentrations of Faecalibacterium prausnitzii and therefore a small amount of butyric acid, since this bacteria is an SCFA producer (De Preter et al., 2015).

The correction of some microbial fermentations by antibiotics or diet could improve symptoms and abnormal fermentation, which, in some IBS cases, is believed to underlie food intolerances (Nanda et al., 1989). Furthermore, using NMR spectroscopy analysis of urine, it is possible to discriminate IBD patients from controls based on metabolite content of hippurate (Stephens et al., 2013). The levels of hippurate were found to be lower in IBD patients compared to controls, suggesting that hippurate is a biomarker of IBD. In particular, hippurate or Nbenzoylglycine is a mammalian-microbial co-metabolite deriving from the microbial fermentation of dietary aromatic compounds (polyphenols, purines, or aromatic AAs) to benzoic acid, further conjugated to glycine in the liver (Williams et al., 2010).

### Obesity

There is evidence that obese and healthy children (HC) show a different gut microbiota profile. Several mechanisms are involved in the energy metabolism regulation and represent the link between the gut microbiota and the metabolic disease pathophysiology, such as energy harvesting from the diet, fat storage regulation, and energy homeostasis of peptide synthesis (Cani and Delzenne, 2009; Krajmalnik-Brown et al., 2012).

In a study on mouse models, Zhang et al. (2011), using LC-ESI-Q-TOF/NMR, found in urine increased excretion of hippurate, 4-hydroxylphenylacetic acid and phenylacetylglycine, and decreased excretion of acetate and lactate, related to body weight gain and to an alteration of gut microbial changes (Veselkov et al., 2009). Moreover, Paul et al. (2016) identified, with the support of <sup>1</sup>H-NMR metabolomics, a maternal metabolic signature that may be related to programming offspring obesity risk in rats. In particular, pregnant rats showed high levels of circulating ketone bodies and free FA (FFA), especially associated with gestational diabetes (Catalano, 2010). The branched chains AA (BCAA) have also found to be related to increased insulin resistance (Scholtens et al., 2014). In humans, FFA are transferred through the placenta from mother to child and are used for lipogenesis. FFA, together with the circulating ketons, may play an important role in the early deposition of excess of body fats in offspring. The production of these molecules in serum metabolome is normalized when it is associated to diet enriched with oligrofructosaccharides (FOS) and to gut microbiota modulation (Paul et al., 2016). Respondek et al. (2013) studied the effects of FOS on the composition of the fecal microbiota and the metabolic parameters in animal models of diet-induced obesity (Respondek et al., 2013). The authors firstly found that the strains particularly stimulated by FOS were Clostridium coccoides, Ruminococcus torques, and Dorea longicatena, and the fecal metabolites modulated by the supplementation, analyzed by using LC-ESI-TOF-MS, were primary bile acids (i.e., cholic and chenodeoxycholic acid) and secondary bile acids (i.e., lithocholic acid).

## Cystic Fibrosis (CF)

Regarding CF disease, the majority of studies on the microbiota metabolome concern lung and the upper airways. It was observed that in CF the production of isoprene is associated to either Gram+ and Gram- species (Kuzma et al., 1995), while the production of hydrogen cyanide is prevalently associated to Pseudomonas aeruginosa, (Carterson et al., 2004; Cody et al., 2009). Robroeks et al. (2010) analyzed exhaled breath samples by GC-TOF-MS, and they managed to discriminate between CF and healthy controls, mainly based on the presence of C5– C16 hydrocarbons and N-methyl-2-methylpropylamine. Barker et al. (2006) reported a significantly lower level of dimethyl sulfide (probably associated to microbial metabolism), detected by GC-MS in CF patients compared to controls. On the other side, Montuschi et al. (2012), using <sup>1</sup>H-NMR forexhaled breath condensate (EBC) analysis, detected significantly higher values of ethanol, acetate, 2-propanol and acetone in CF patients, which differentiated them from controls, whereas acetate, ethanol, 2 propanol and methanol were found to be relevant metabolites for distinguishing between patients with stable CF and patients with unstable CF. Moreover, 2-Propanol, which represents an enzyme-mediated product of acetone reduction, was detected in a breath sample of CF colonized by P. aeruginosa (Wang et al., 2006). The high level of ethanol in EBC samples of CF, detected by MS based metabolomics, could also be related to the decreased capacity of P. aeruginosa to oxidize ethanol to acetate (Wang et al., 2006). On the contrary, elevated acetate concentrations in healthy subjects may reflect resident bacteria in the oral cavity, such as Streptococcus mutans, debasing pyruvate into metabolic end products, such as acetate and lactate (Korithoski et al., 2008).

### Non-alcoholic Fatty Liver Disease (NAFLD)

Several VOCs, including ethanol, seem to be produced by colonic bacteria and may have toxic effects on the host after intestinal absorption and delivery to the liver via the portal vein (Raman et al., 2013). Moreover, Raman et al. (2013), using SPME-GC-MS, identified esters (i.e., ethyl propanoate, butyl butanoate, methyl pentanoate, methyl acetate) in fecal samples of obese NAFLD patients more frequently than in healthy controls.

The bacterial production of SCFAs and ethanol by several gut microbes is well known, but very little is known about bacteria and biochemical pathways that may be involved in ester production in the intestinal microbiota, even though most of the esters linked to NAFLD are derivatives of short chain aliphatic alcohols and carboxylic acids (Raman et al., 2013). Also Del Chierico et al. (2016) evaluated the gut microbiota profiling of NAFLD and obese patients. The authors evidenced, with the support of multivariate analysis, that OTUs such as Oscillospira, Ricknellaceae, Parabacteroides, Bacteroides fragilis, Sutterella, and Lachnospiraceae, and metabolites such as 4- Methyl-2-pentanone, 1-butanol and 2-butanone (detected with SPME-GC-MS), discriminated NAFLD from healthy subjects (Del Chierico et al., 2016).

### Celiac Disease

As regards studies on celiac disease, it was clearly shown that metabolic differences between controls and celiac patients exist (Calabrò et al., 2014).

The main differences detected coupling MS-and NMR- based metabolomics approaches in celiac patients compared to controls were lower levels of several AAs, as asparagine, isoleucine, methionine, proline, and valine, and also methylamine, pyruvate, creatinine, choline, methyl glutarate, lactate, lipids, and glycoproteins, and higher levels of glucose and 3-hydroxybutyric acid in serum and acetoacetate in the urine of celiac patients (Calabrò et al., 2014). The same authors also found higher levels of some metabolites related to the gut microbiota in the urine, such as: indoxyl sulfate, meta-[hydroxyphenyl] propionic acid (m-HPPA), and phenylacetylglycine. In fact, M-HPPA in urine mostly originates from the gut microbiota, being one of the many products of the microbial mediated breakdown of plant phenolic compounds, such as caffeic acid and its conjugate chlorogenic acids (Phipps et al., 1998). Besides, Di Cagno et al. (2011) analyzed, using SPME-GC-MS and <sup>1</sup>H-NMR, VOCs, and AAs of fecal and urine samples of treated (gluten free diet) celiac children. The samples showed higher levels of free AAs (proline, methionine, histidine, and tryptophan) and lower levels of SCFAs, as butyric, isocaproic, and propanoic acids compared to controls. In this study, it was also found that the levels of some alcohols, such as 1-octen-3-ol, ethanol and 1-propanol were higher in treated celiac children compared to controls and it was hypothesized that when alcohol production is correlated with intestinal bacteria synthesis this may also induce endotoxemia (Cani et al., 2008).

In another study, saliva samples have been analyzed, using SPME-GC-MS, revealing high levels of non-anal, 4-methyl-2 hexanone, and ethyl-acetate in treated celiac children (Francavilla et al., 2014). These findings suggest the presence of microbial metabolic activities at the oral cavity level (by Firmicutes, Actinobacteria and Bacteroidetes) that may also affect the synthesis of VOCs (Kusano et al., 2013).

### Food Allergies

Finally, the gut microbiota is believed to be associated with food allergies. In particular, the prevalence of atopic diseases, including eczema and asthma, suggests that the modulation of the immune response mechanisms in the gut can directly affect the development of allergic diseases and the development of tolerance (Watanabe et al., 2003; Penders et al., 2007). Moreover, the advent of dysbiosis during the early post-natal period may further pre-dispose individuals to later inflammatory, immune, and allergic disorders (Francavilla et al., 2012).

However, there is still little scientific evidence on the relation between the gut microbiota metabolome and food allergy. A study by Francavilla et al. (2012) describing the metabolome of infants with cow's milk allergy, was conducted using the combinate of SPME-GC-MS and <sup>1</sup>H-NMR techniques on a group of children fed with hydrolyzed formula with no lactose (CMA-NL), and a group of children fed with lactose-containing (CMA-L) formula compared with controls. The authors found that the addition of lactose to the formula resulted in a significant increase of Bifidobacteria and LAB counts, and a decrease of Bacteroides/Clostridia. Consequently, the levels of SCFAs increased, especially for acetic and butyric acids, in controls and CMA-L compared to CMA-NL infants. The same trend was found for lactic acid and threonine.

### Neuropathology

The composition of the intestinal microbiota plays a key role in neuro-gastroenterology, which deals with the interactions between the central nervous system and the gut (gut–brain axis). Numerous neuropathological diseases, such as autism spectrum disorder (ASD), are probably associated with the gut microbiota and thus the possibility to influence this connection is alluring (Holmes et al., 2011), even if to date there are still few studies investigating this field. De Angelis et al. (2013) used SPME-GC-MS and <sup>1</sup>H-NMR to study the fecal microbiota and the metabolome of children with Pervasive Developmental Disorder Not Otherwise Specified (PDD-NOS), and children with ASD compared to healthy controls. The authors found an altered composition of the microbiota and VOCs, which were partially different between children with PDD-NOS and ASD. The main biological significance of this work was related to the increased levels of Clostridum in PDD-NOS and ASD and the decreased levels of some health promoting bacteria (i.e., Bifidobacterium) and metabolites, such as free AA and SCFAs in PDD-NOS, in ASD children compared to controls (De Angelis et al., 2013). Furthermore, Kidd (2002) found that subjects with ASD together with their non-ASD siblings, presented with a deep alteration in the tryptophan–nicotinic acid and sulfur metabolic pathways (Kidd, 2002; Oresic et al., 2008). Important metabolic phenotype differences were observed between ASD and controls with perturbations in the relative patterns of urinary metabolites related to the gut microbiota (Kidd, 2002).

Hsiao et al. (2013) used GC- and LC-MS platforms to study the oral treatment with human commensal Bacteroides fragilis (that corrects gut permeability) to modulate the microbial composition and the related defects in communicative behaviors in mouse models with maternal immune activation (MIA), showing GI barrier defects and microbiota alterations in displaying features of ASD. The authors detected the presence of indolepyruvate 4-ethylphenylsulfate and p-cresol in mice serum metabolome, presumably deriving from microbial metabolism. Furthermore, since B. fragilis improves intestinal health, it could also have a role in regulating intestinal permeability and metabolic homeostasis (Nicholson et al., 2012).

### CONCLUSIONS AND PERSPECTIVE

The challenge of systems medicine is to interpret the body structure as a whole system and not as a sum of single parts (Moco et al., 2013). To pursue this aim, the wide range of top-down systems biology analyses should be used to interpret the metabolic interactions between the host and its gut microbiota, and to comprehend how these interactions affect the physiological and pathological conditions. Furthermore, the combination of these techniques with genome analysis may lead to a holistic view of the metabolic pathways, which can also be backed up using mathematical models and statistical assessment of data. In fact, by managing data it is possible to achieve a higher level of biological understanding. Therefore, novel algorithms and statistical analysis need to be improved to integrate the "omics" data, and a stochastic model of metabolic networks needs to be introduced to lead to a novel knowledge of co-regulation in biochemical networks.

The metabolomics approach may identify physiological and clinical biomarkers that are not obtainable using targeted methods (Weckwerth and Morgenthal, 2005).

In conclusion, the generation of new gut microbiota biomarkers will offer the chance to associate complex metabolic pathways with the etiology of different diseases, in order to evaluate the causal relationship between metabolites and pathogenesis. Moreover, these novel biomarkers could lead to the

### REFERENCES


development of mechanistic hypotheses that could be targeted to the development of nutritional and personalized therapy tools in early disease prediction in asymptomatic conditions, and enable a more accurate prognosis of the disease progress.

### AUTHOR CONTRIBUTIONS

PV, conceived and wrote the manuscript. FD, participated in the writing and produced tables and figures. LP, supervised and reviewed the manuscript.

### ACKNOWLEDGMENTS

This work was supported by the Ministry of Health, Ricerca Corrente RC201402G003251 and 201502P003534 assigned to LP, Children's Hospital Bambino Gesù, IRCCS. The authors thank Paola Giovanna Volpi for her careful English revision of the manuscript.


production in Pseudomonas aeruginosa. J. Bacteriol. 186, 6837–6844. doi: 10.1128/JB.186.20.6837-6844.2004


antioxidant capacity of its catabolites. Free Radic. Biol. Med. 47, 1180–1189. doi: 10.1016/j.freeradbiomed.2009.07.031


metabolic parameters of humanized gnotobiotic diet induced obesity mice. PLoS ONE 8:e71026. doi: 10.1371/journal.pone.0071026


safety and shelf-life of minimally processed sliced apples and lamb's lettuce. Food Microbiol. 47, 74–84. doi: 10.1016/j.fm.2014.11.008


**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 Vernocchi, Del Chierico and Putignani. 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.

# Molecular Players Involved in the Interaction Between Beneficial Bacteria and the Immune System

#### Arancha Hevia, Susana Delgado, Borja Sánchez \* and Abelardo Margolles

Instituto de Productos Lácteos de Asturias, Consejo Superior de Investigaciones Científicas, Villaviciosa, Spain

The human gastrointestinal tract is a very complex ecosystem, in which there is a continuous interaction between nutrients, host cells, and microorganisms. The gut microbiota comprises trillions of microbes that have been selected during evolution on the basis of their functionality and capacity to survive in, and adapt to, the intestinal environment. Host bacteria and our immune system constantly sense and react to one another. In this regard, commensal microbes contribute to gut homeostasis, whereas the necessary responses are triggered against enteropathogens. Some representatives of our gut microbiota have beneficial effects on human health. Some of the most important roles of these microbes are to help to maintain the integrity of the mucosal barrier, to provide nutrients such as vitamins, or to protect against pathogens. In addition, the interaction between commensal microbiota and the mucosal immune system is crucial for proper immune function. This process is mainly performed via the pattern recognition receptors of epithelial cells, such as Toll-like or Nod-like receptors, which are able to recognize the molecular effectors that are produced by intestinal microbes. These effectors mediate processes that can ameliorate certain inflammatory gut disorders, discriminate between beneficial and pathogenic bacteria, or increase the number of immune cells or their pattern recognition receptors (PRRs). This review intends to summarize the molecular players produced by probiotic bacteria, notably Lactobacillus and Bifidobacterium strains, but also other very promising potential probiotics, which affect the human immune system.

Keywords: Bifidobacterium, immunomodulation, Lactobacillus, molecular players, probiotic

### THE HUMAN GUT MICROBIOTA AND THE IMMUNE SYSTEM

From the early stages of life, one of the most important roles of the gut microbiota is to contribute to the development of a proper immune system. Normally, humans live in a homeostatic symbiosis with their gastrointestinal microbes, providing them with nutrients and a friendly environment, whereas the microbiota aids in the appropriate development and maintenance of the host's gut mucosa. Epithelial function is influenced by direct host/microbiota interactions and microbial metabolism. The large intestine acts as an anaerobic bioreactor for the enteric bacterial community, which is fueled by host diet components that cannot be processed in the small intestine, as well as by endogenous nutrients, such as host glycans from mucus and cell debris released from epithelial cells. Additionally, the microbiota synthesizes essential amino acids, vitamins, and short chain fatty acids (SCFAs) by degrading a variety of proteins and otherwise non-digestible polysaccharides (Sekirov et al., 2010).

Edited by:

Enrica Pessione, University of Turin, Italy

#### Reviewed by:

Hailan Piao, Washington State University in Tri-Cities, USA Edoardo Puglisi, Università Cattolica del Sacro Cuore, Italy

> \*Correspondence: Borja Sánchez borja.sanchez@csic.es

#### Specialty section:

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

Received: 10 August 2015 Accepted: 31 October 2015 Published: 18 November 2015

#### Citation:

Hevia A, Delgado S, Sánchez B and Margolles A (2015) Molecular Players Involved in the Interaction Between Beneficial Bacteria and the Immune System. Front. Microbiol. 6:1285. doi: 10.3389/fmicb.2015.01285

The commensal microbiota ensures the mechanical integrity of the mucosal barrier, thereby offering protection against harmful pathogenic microbes (**Figure 1**). Commensal bacteria can adhere to the intestinal mucus and competitively inhibit the adhesion of enteropathogens; they also produce bacteriocins and SCFAs, compounds that are able to inhibit the growth of other microorganisms. Additionally, some antimicrobial metabolites, such as the defensins secreted in the intestine, contribute to the host's control of these microorganisms (Salzman et al., 2007). Further protection of the host is provided by inducing the mucosal immune system to produce immunoglobulin A, which is released in the intestinal lumen in large amounts and limits local bacterial colonization, thereby preventing bacteria from penetrating the epithelium (Salzman et al., 2007).

A major issue is how the intestine distinguishes between the abundant, normal microbiota and rare pathogens. The immune system fights pathogenic bacteria, but tolerates the presence of commensal species, even though their cellular structures are quite similar and they have common mechanisms of interacting with host immune cell receptors; this phenomenon is called immune tolerance. In this way, our immune cells differentiate between commensals and pathogens. This is carried out by our innate immune system through pattern recognition receptors (PRRs) (**Figure 1**), including Toll-like receptors (TLRs), transmembrane receptors that scan the external milieu of the intestinal lumen, and Nod-like intracellular receptors (NODLR), which guard the cytoplasmic space (Claes et al., 2015; Sellge and Kufer, 2015). Other PRRs have also been described, such as C-type lectin receptors, formylated peptide receptors, retinoic acidinducible (RIG)-like helicases, and intracellular interleukin-1 (IL-1)-converting enzyme protease-activating factor (Denes et al., 2012; Bufe et al., 2015; Dambuza and Brown, 2015; Yao et al., 2015). PRRs are able to specifically recognize and bind different microbial macromolecular ligands, which are designated as microbial-associated molecular patterns (MAMPs), such as lipopolysaccharide, flagellin and other proteins, bacterial peptidoglycan, viral RNAs, and fungal carbohydrates. As a result, the T cell subset involved in regulating the immune balance is finely tuned by the host and the microbes with which it interacts, and disequilibrium between effector T helper (Th) and regulatory T cells (Treg) leads to impaired immune responses (Noack and Miossec, 2014; Nyirenda et al., 2015; Yousefi et al., 2015). Effector Th cells are derived from progenitor naïve CD4+ T cells via maturational processes

FIGURE 1 | Schematic representation of the interactions established between the intestinal microbiota and the host immune system. (A) General overview of the epithelium in contact with multiple species of microorganisms that constitute the intestinal microbiota: (1) enterocytes; (2) M cells; (3) Goblet cells; antigen presentation cells (APC): (4) dendritic cells and (5) macrophages; (6) defensins, bacteriocins, and secreted IgA (sIgA) also play an important role in controlling the levels of the different populations of microorganisms. A fine-tuned balance of Tcell maturation toward Treg or Thelper cells must be established to assure the tolerogenic response of the host immune system. (B) Examples of molecular interactions between microbial antigens and host cells through Pattern Recognition Receptors (PRRs). LPS, lipopolysaccharide; PGN, peptidoglycan; dsRNA, double-strand RNA; TLR, toll-like receptor; NODLR, nucleotide-binding oligomerization domain-like receptors; RIG-like helicases, retinoic acid-inducible gene 1 like helicases.

that are induced by antigenic stimulation. Their function depends on complex interactions with antigen-presenting cells (APCs) in a permissive environment, which is characterized by the antigen type and load, costimulatory molecules, and cytokine signaling. CD4+ T cells may differentiate into different Th phenotypes (mainly Th1, Th2, and Th17) that produce distinct cytokines with different biological functions, or they may evolve into the inducible Treg lineage, which performs immunomodulatory functions (Sakaguchi et al., 2010; Wing and Sakaguchi, 2014). The Th1 subgroup recognizes intracellular pathogens and mainly produces IL-2, interferon (IFN), and tumor necrosis factor alpha (TNFα), thereby supporting typical cellular immunity. Th2 cells, which are essential for eliminating extracellular pathogens such as helminths, express IL-4, IL-5, IL-10, and IL-13, which aid humoral immunity. The Th17 subset, which is involved in fighting Gram-negative bacteria, fungi, and some protozoa, secretes IL-17, which has strong pro-inflammatory effects. Overall, Th responses are accurately balanced to avoid both self-antigen reactivity and excessive reactions to antigens. In fact, dysregulated Th1 responses drive cell-mediated autoimmune disorders, and enhanced Th2 activity is involved in atopy, whereas Th17 cells are probably responsible for chronic tissue inflammation. In contrast, skewing the response away from Treg cells may lead to the onset and/or progression of autoimmune diseases in humans (Eisenstein and Williams, 2009).

### PROBIOTICS AND THE IMMUNE SYSTEM

During the last few years, it has been proposed that the intestinal microbiota can be positively modulated by the administration of bacteria or bacterial substrates, and it is likely that, to some extent, this might lead to a significant modulation of the immune system (Dongarrà et al., 2013; Sánchez et al., 2015; Scott et al., 2015). To this end, substantial research efforts are concentrated on using probiotics as potential modulators of gut microbial community. Probiotics are commensal microorganisms that are present in the intestinal tract and in many fermented foods, and they are defined as "live microorganisms that, when administered in adequate amounts, confer a health benefit on the host" (Hill et al., 2014). The vast majority of probiotic bacteria are Gram-positive strains, mainly species of the Lactobacillus and Bifidobacterium genera, although some non-pathogenic strains of Escherichia coli and certain yeasts are also considered to be probiotics. Currently, there is an increasing interest in considering some common colonizers of the human gut to be novel probiotics, because of their potential health properties; they are called emerging probiotics (Hill et al., 2014; Rodríguez et al., 2015). Some examples are Faecalibacterium prausnitzii and Akkermansia muciniphila.

Probiotics can exert their beneficial properties in a wide range of ways, including direct cell-to-cell contact in the human gut, by secreting diverse molecules that act as the final mediators of probiotic crosstalk, or through cross-feeding mechanisms. The chemical composition of the molecular effectors is very diverse and consists of proteins that are secreted into the extracellular milieu or localized on the surface of the bacteria, low molecular weight peptides, amino acids, cellwall polysaccharides or components, bacterial DNA, or SCFAs (Macpherson and Harris, 2004; Turroni et al., 2013). Given the different molecular natures of these molecular effectors, their mechanisms of action are very diverse. Therefore, this review includes only a summary of the molecular bases underlying the immunomodulatory properties of probiotic bacteria (**Figure 2**). In addition, we must consider that genetic differences in the expression of host receptors, the variable composition of the autochthonous microbiota in different individuals, and other host factors that contribute to the response to bacterial signals are likely to explain the variability in responses to probiotics in responding and non-responding individuals (Salonen et al., 2014).

### IMMUNOMODULATORY EFFECTORS

A significant number of relevant studies have highlighted the immunomodulatory effects that Lactobacillus and Bifidobacterium strains exert on the host immune system. For instance, there is evidence that Bifidobacterium bifidum LMG13195 and Bifidobacterium breve IPLA 20004 enhance intestinal barrier function and preferentially elicit Treg cell differentiation, which induces the expression of antiinflammatory cytokines, when co-cultured with the human colorectal adenocarcinoma cell line HT29 (López et al., 2012). Lactobacillus rhamnosus GG interacts with macrophages in such a way that activated macrophages can discriminate between probiotic and pathogenic bacteria by INF-mediated TLR gene regulation (Miettinen et al., 2008), and the interaction between Lactobacillus casei CRL 431 and gut-associated immune cells induces an increase in the number of CD-206 and TLR2 receptors (Aragón et al., 2014).

The mediators of these interactions are largely unknown, although surface and cell-envelope molecules have been identified as some of the main players. Among them, we can distinguish between proteins and other components, such as peptidoglycan (PG), exopolysaccharides (EPS), teichoic acids (TA), and lipoteichoic acids (LTA). Known molecular effectors that mediate immunomodulatory mechanisms are listed in **Table 1**.

### Surface Proteins

Cell surface proteins include the S-layer proteins (Slp), which constitute the major surface proteins of some lactobacilli. In Lactobacillus helveticus fb213, Lactobacillus acidophilus fb116, and L. acidophilus fb214, Slp are well studied, and it is likely that they are necessary for lactobacilli survival in the gastrointestinal tract, as they can bind to components of the extracellular matrix, such as collagen and fibronectin, of intestinal cells (Meng et al., 2014; Yadav et al., 2015). Konstantinov and colleagues used an slpA knockout mutant of L. acidophilus to show that the interaction occurs via the recognition of SlpA by a specific receptor of dendritic cells, denominated DC-SIGN (Konstantinov et al., 2008). Additionally, proteins from structures that are included in the PG layer, such as pili, fimbria, and flagella, are recognized by the host immune

system. Recently, it has been reported that bacterial SpaCBA pilus fibers in L. rhamnosus GG may be responsible for its wellknown adhesion properties and confer the ability to contact host cells (Reunanen et al., 2012). B. bifidum PRL2010 pili have been shown to induce TNF-α production and decrease IL-10 production in the mouse mucosa, as well as to adhere TABLE 1 | Examples of immunomodulatory effectors produced by classic/emerging probiotics.


\*Synonym of B. animalis subsp. lactis.

to diverse extracellular matrix proteins (Turroni et al., 2013), while B. breve UCC2003 pili are essential for host colonization (O'Connell Motherway et al., 2011). In another recent work, gene complementation studies were used to show that the fimbriae of the probiotic strain E. coli Nissle 1917 were involved in the adhesion to porcine intestinal cells, thereby helping to prevent infection with enteropathogenic E. coli (EPEC) (Kleta et al., 2014).

### Cell Wall Non-proteinaceous Components

Non-proteinaceous cell wall components have different roles in microbe-host crosstalk. It has been shown that the EPS from Lactobacillus and Bifidobacterium strains can have a modulator role in preventing pathogen invasion, even though the EPS of pathogenic bacteria have been classically viewed as possible virulent factors. Examples of immunomodulatory EPS are those from B. breve and Bifidobacterium animalissubsp. lactis (Fanning et al., 2012; Hidalgo-Cantabrana et al., 2014) or Lactobacillus plantarum strains (Murofushi et al., 2015). TAs are linear polymers of ribitol phosphate or glycerol phosphate that are covalently bound to D-alanine, monosaccharides, or amino sugars, and they are attached either to PG (wall TAs) or to the cytoplasmic membrane (membrane TAs or lipoteichoic acids; LTAs). TAs from L. plantarum were shown to display antiinflammatory properties, as shown by the different cytokine production profiles of peripheral blood mononuclear cells (PBMCs) and monocytes exposed to this molecule (Grangette et al., 2005). In addition, mice fed a diet supplemented with L. plantarum LTAs or with an LTA-producing strain showed better scores in a colitis model compared with the control group and mice that were fed a L. plantarum LTA-deficient strain (Grangette et al., 2005). Although there have been a few promising results, this topic requires further research to clarify the mechanisms of action of the cell wall components of probiotics on the human gut microbiota.

### Soluble Compounds

Soluble components that are produced by probiotic bacteria can also affect the bacterial-host interplay. In Bifidobacterium longum the secretion of serpin, a serine protease inhibitor, which specifically binds and inactivates human neutrophil and pancreatic elastase, was shown to contribute to gut homeostasis (Ivanov et al., 2006). Additionally, it has been observed that some proteins with characteristic biochemical motifs that are produced by both commensal and pathogenic bacteria can elicit specific functions and affect immune cells of the intestinal lumen. This is the case for a family of serinethreonine rich proteins, which was described in species of Lactobacillus and Bifidobacterium, with a recently described kinase function (Zakharevich et al., 2012; Nezametdinova et al., 2014). In lactobacilli, a serine-threonine peptide, STp, which is contained in a protein secreted by L. plantarum, was shown to be involved in bacterial aggregation (Hevia et al., 2013). Additionally, this peptide can modulate the dendritic cell phenotype of ulcerative colitis (UC) patients (Bernardo et al., 2012; Al-Hassi et al., 2014). It was also demonstrated that the immunomodulatory effect of Lactobacillus paracasei is mediated, at least in part, by the secreted protease lactocepin, which selectively degrades the chemokine IFN-γ-inducible protein 10 (IP-10) that functions in lymphocyte recruitment (von Schillde et al., 2012). There are other examples of non-proteinaceous compounds that can exert certain effects on the host. Some species of Bifidobacterium possess unmethylated CpG motifs in their DNA that were able to induce TLR9 activation, which is known to trigger a Th1 orientation of the immune system (Ménard et al., 2010). In contrast, in other studies, it was shown that intragastric and subcutaneous administration of DNA from a probiotic mix ameliorated the severity of colitis in a murine experimental colitis model, whereas a methylated probiotic DNA had no effect (Rachmilewitz et al., 2004).

### EMERGING PROBIOTICS, A NOVEL SOURCE OF IMMUNOMODULATORY EFFECTORS

In addition to Lactobacillus and Bifidobacterium, other microorganisms have received substantial interest among researchers as potentially new, beneficial gut bacteria. Most of them are common colonizers of the human gut under normal conditions. Some of these microbial types are considered to be markers of dysbiosis in intestinal inflammatory diseases, such as UC and Crohn's disease (Manichanh et al., 2006; El Aidy et al., 2013). In these conditions, a loss of microbial diversity and a significant reduction of members of Clostridium clusters IV and XIVa have been reported, particularly in bacteria involved in butyrate and propionate metabolism, such as Ruminococcus, Eubacterium, Roseburia, and Faecalibacterium. In this section, we will highlight current research on F. prausnitzii and A. muciniphila, two bacteria that have received much attention during the last few years because of their potential immunomodulatory properties.

F. prausnitzii is a "novel" intestinal bacterium whose immunomodulatory properties have been well characterized in vitro and in vivo. This anaerobic, Gram-positive bacterium seems to play a role in the maintenance of gut homeostasis, and its population is normally reduced in intestinal inflammatory diseases (Sokol et al., 2008; Cao et al., 2014; Machiels et al., 2014). In 2008, Sokol and colleagues studied the effects of whole bacteria, a cell culture supernatant, bacterial DNA, or membranederived fractions in vitro using the Caco-2 epithelial colorectal adenocarcinoma cell line and PBMCs, as well as in vivo in a mouse model of 2,4,6-trinitrobenzene sulfonic acid (TNBS) induced colitis (Sokol et al., 2008). The results showed that F. prausnitzii cells exerted anti-inflammatory effects in PBMCs. Furthermore, its culture supernatant reduced IL-8 secretion and abolished the activation of nuclear factor kappa-light-chainenhancer of activated B cells (NF-κB) in Caco-2 cells. Moreover, although no significant improvement of the disease was detected in mice, partial disease scores significantly improved in colitic mice receiving the cell culture supernatants, compared with the non-receiving group. Accordingly, the authors hypothesized that the beneficial effects of F. prausnitzii must be executed by a soluble compound that is secreted by the bacteria. In relation to this, recent work showed that F. prausnitzii secreted a 15 kDa protein with anti-inflammatory properties. This protein was able to inhibit the NF-κB pathway in intestinal epithelial cells, and it prevented colitis in an animal model (Quévrain et al., 2015). Additional research showed that this bacterium restored physiological parameters and downregulated cytokine profiles in mice with colitis, as well as increased the Treg population to a greater degree than other commensals such as B. longum (Qiu et al., 2013; Martín et al., 2015). UC patients have fewer butyrate-producing Roseburia hominis and F. prausnitzii (Machiels et al., 2014). It is likely that a significant part of their anti-inflammatory action results from the effect of SCFAs in colonocytes, as acetate, propionate, and butyrate modulate the inflammatory responses of immune cells through receptors such as Gpr43 and Gpr41 (Maslowski et al., 2009). However, despite all the information that has recently been discovered about these bacterial groups in healthy and diseased states, and besides butyrate seeming to be the key homeostasis promoter, additional work is required to elucidate the molecular mechanisms through which F. prausnitzii interacts with the host gut environment.

A. muciniphila is another common member of the healthy gut microbiota in humans at all stages of age (Collado et al., 2007; Belzer and de Vos, 2012). A. muciniphila is a Gram-negative, strictly anaerobic, mucin-degrading microorganism member of the Verrucomicrobia phylum, and it was one of the first bacteria shown to utilize mucin, the glycosylated protein layer that covers the gut epithelium, as its sole carbon, nitrogen, and sulfur source. The products derived from mucin degradation are mainly SCFAs that feed colonocyte metabolism and confer health properties to the host. By degrading the mucin of the external mucus layer, A. muciniphila helps with the continuous renovation of the protective cover of the mucosae, and it maintains a healthy protective barrier that prevents the entrance of enteropathogens into the epithelium (Lukovac et al., 2014). In addition, when A. muciniphila was administered to mice, there were increased intestinal levels of endocannabinoids that control inflammation, the gut barrier, and gut peptide secretion, suggesting an immunomodulatory role for this bacterium (Everard et al., 2013).

## CONCLUDING REMARKS

In conclusion, even though much effort has been put into probiotic research during recent decades, the mechanisms underlying the immunomodulatory effects of beneficial intestinal bacteria have scarcely been elucidated. There is compelling evidence that novel bacterial players, other than Lactobacillus and Bifidobacterium, could play a role in these processes and are much more important than previously thought; however, difficulties in growing some of these bacteria on laboratory- and industrial-scales, and the lack of molecular tools needed to perform functional genomic analyses, seriously hamper the characterization of novel strains. Further research is needed to overcome these culturing and functional characterization difficulties to perform well-designed pre-clinical and intervention studies that shed new light on the mechanisms responsible for the beneficial effects attributed to these bacteria.

### AUTHOR CONTRIBUTIONS

AH, SD, BS, and AM contributed to the conception and design of the work, and to the acquisition, analysis, and interpretation of the data. All authors contributed to the drafting of the manuscript and approved the final version to be published.

### REFERENCES


### ACKNOWLEDGMENTS

This research was funded by grants AGL2013-44039-R and AGL2013-44761-P from the Spanish "Plan Estatal de I+D+I." BS and AH were recipients of a Ramón y Cajal postdoctoral contract and a Formación de Personal Investigador—Research Personnel Training (FPI) grant, respectively, from the Spanish Ministry of Economy and Competitiveness.

mutant synthesizing modified teichoic acids. Proc. Natl. Acad. Sci. U.S.A. 102, 10321–10326. doi: 10.1073/pnas.0504084102


**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 Hevia, Delgado, Sánchez and Margolles. 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.

# Bioactive Molecules Released From Cells Infected with the Human Cytomegalovirus

#### Anna Luganini, Maria E. Terlizzi and Giorgio Gribaudo\*

Laboratory of Microbiology and Virology, Department of Life Sciences and Systems Biology, University of Turin, Turin, Italy

Following primary infection in humans, the human cytomegalovirus (HCMV) persists in a latent state throughout the host's lifetime despite a strong and efficient immune response. If the host experiences some form of immune dysregulation, such as immunosuppression or immunodeficiency, HCMV reactivates, thereby emerging from latency. Thus, in the absence of effective functional immune responses, as occurs in immunocompromised or immunoimmature individuals, both HCMV primary infections and reactivations from latency can cause significant morbidity and mortality. However, even in immunocompetent hosts, HCMV represents a relevant risk factor for the development of several chronic inflammatory diseases and certain forms of neoplasia. HCMV infection may shift between the lytic and latent state, regulated by a delicate and intricate balance between virus-mediated immunomodulation and host immune defenses. Indeed, HCMV is a master in manipulating innate and adaptive host defense pathways, and a large portion of its genome is devoted to encoding immunomodulatory proteins; such proteins may thus represent important virulence determinants. However, the pathogenesis of HCMV-related diseases is strengthened by the activities of bioactive molecules, of both viral and cellular origin, that are secreted from infected cells and collectively named as the secretome. Here, we review the state of knowledge on the composition and functions of HCMV-derived secretomes. In lytic infections of fibroblasts and different types of endothelial cells, the majority of HCMV-induced secreted proteins act in a paracrine fashion to stimulate the generation of an inflammatory microenvironment around infected cells; this may lead to vascular inflammation and angiogenesis that, in turn, foster HCMV replication and its dissemination through host tissues. Conversely, the HCMV secretome derived from latently infected hematopoietic progenitor cells induces an immunosuppressive extracellular environment that interferes with immune recognition and elimination of latently infected cells, thereby promoting viral persistence. Characterization of the composition and biological activities of HCMV secretomes from different types of infected cells will lay the foundation for future advances in our knowledge about the pathogenesis HCMV diseases and may provide targets for the development of novel antiviral intervention strategies.

Keywords: human cytomegalovirus, lytic infection, latency, bioactive molecules, secretome, chronic disease, angiogenesis, immunoevasion

Edited by:

Katharina Riedel, University of Greifswald, Germany

#### Reviewed by:

Paul Richard Himes, University of Louisville, USA Donatella De Pascale, The Institute of Protein Biochemistry-CNR, Italy

\*Correspondence: Giorgio Gribaudo giorgio.gribaudo@unito.it

#### Specialty section:

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

Received: 18 January 2016 Accepted: 29 April 2016 Published: 13 May 2016

#### Citation:

Luganini A, Terlizzi ME and Gribaudo G (2016) Bioactive Molecules Released From Cells Infected with the Human Cytomegalovirus. Front. Microbiol. 7:715. doi: 10.3389/fmicb.2016.00715

### THE HUMAN CYTOMEGALOVIRUS

Human cytomegalovirus (HCMV) is an opportunistic beta-Herpesvirus that infects more than 90% of people worldwide with an infection rate that increases with age (Landolfo et al., 2003; Britt, 2008; Mocarski et al., 2014). HCMV infection, as for the other human Herpesviruses, is characterized by two phases: lytic and latent. Following primary infection, in which the virus productively replicates in a broad range of different cell types (Sinzger et al., 2008), HCMV establishes latency in cells of myeloid lineage through a still poorly understood mechanism (Britt, 2008; Crough and Khanna, 2009; Mocarski et al., 2014). In the normal immunocompentent host, primary infection is usually asymptomatic or mildly symptomatic and persistent (Britt, 2008; Mocarski et al., 2014). Nevertheless, HCMV infection is considered, even in immunocompetent hosts, a risk factor for the development of various vascular diseases, immunosenescence, and tumor development (Britt, 2008; Mocarski et al., 2014; Nogalski et al., 2014). In contrast, in individuals in which the ability to develop an appropriate cellular immune response is compromised, such as immunosuppressed patients and the immunoimmature fetus during pregnancy, HCMV infection is one of the major causes of morbidity and mortality. Indeed, primary infection or reactivation from latency causes overt diseases in immunosuppressed hosts, such as transplant recipients taking immunosuppressive drugs and AIDS patients. In these settings, HCMV can be responsible for a wide range of clinical conditions, for example retinitis, pneumonia, colitis, hepatitis, and several chronic inflammatory diseases, such as atherosclerosis, transplant vascular sclerosis (TSV), and chronic allograft rejection (CR; Britt, 2008; Crough and Khanna, 2009; Mocarski et al., 2014). Congenital HCMV infections, on the other hand, represent the most prominent viral cause of birth defects such as malformations, hearing loss and learning disabilities (Kenneson and Cannon, 2007; Britt, 2008; Mocarski et al., 2014), and the incidence of HCMV transmission to the fetus is strictly related to maternal seroprevalence. Primary maternal infection carries a risk of transmission between 14.2 and 52.4% (Kenneson and Cannon, 2007), while a transmission rate of 1.4% has been reported in relation to maternal reactivated infections. However, the severity of congenital HCMV disease is similar for both primary and non-primary infections (Ahlfors et al., 2001; Rahav et al., 2007). Globally, the incidence of HCMV congenital infection is between 0.3 and 2.3% of all live births in developing countries. In congenitally infected fetuses, about 10–15% develops evident symptoms after birth, with an incidence of perinatal mortality of about 3–20 newborns per 100,000 live births, and of 24–160 cases of neurologic diseases (hearing loss and mental retardation) per 100,000 neonates. In the remaining 85–90% of HCMV-infected newborns, no symptoms are displayed at birth, but late signs of infection, such as hearing defects, are evident in 30–200 newborns per 100,000 live births. The impact of congenital HCMV infection on public health is thus significant, given that hearing loss, the most common long-term sequelae of infants with congenital HCMV, it is the leading cause of non-genetic deafness in children.

The broad range of clinical manifestations of HCMV diseases reflects the capacity of the virus to productively infect an extremely wide range of cell types in the host, such as skin and lung fibroblasts, epithelial and endothelial cells (ECs), vascular smooth muscle cells, hepatocytes, monocyte-derived macrophages, neuronal, and glial cells, thus determining its potential to spread to all areas of the body (Sinzger, 2008; Sinzger et al., 2008).

The HCMV genome is the largest among the Herpesviruses; its 235 kb double-stranded DNA is structured into a unique long (UL) and a unique short (US) region, both of which are flanked by terminal and internal inverted repeats (TRL/S and IRL/S, respectively; Landolfo et al., 2003; Mocarski et al., 2014). Even though its annotation remains provisional and its coding capacity has recently been proposed to be much greater than originally thought (Stern-Ginossar et al., 2012), it is generally accepted that the HCMV genome encodes at least 170 canonical proteins (Mocarski et al., 2014). Whole-genome functional profiling of two HCMV laboratory strains revealed that a set of just 50 herpesvirus-common proteins, encoded by genes mainly located in the central region of the U<sup>L</sup> domain, is required for productive viral replication in primary fibroblasts (Murphy and Shenk, 2008; Mocarski et al., 2014). The remaining two-thirds of canonical HCMV protein-coding genes, mostly betaherpesvirus- or CMVspecific, are confined within the terminal regions of the genome and are not essential in cultured fibroblasts. Although, specific functions have yet to be assigned to many of these non-essential genes, they are mainly thought to be involved in regulating virus cell tropism, dissemination, and viral persistence and latency within the host, as well as the modulation of intrinsic, innate, and acquired host immune responses, thus contributing to viral pathogenesis in a variety of ways (Mocarski et al., 2014).

During lytic infection, viral gene expression occurs in three phases, named immediate early (IE or α), early (E or β), and late (L or γ) in relation to their temporal kinetic profiles, and leads to genome replication, assembly and release of mature infectious viral particles. In brief, HCMV gene expression begins with de novo expression of IE genes, predominantly IE1-72 and IE2-86, that activate the expression of E genes, required for replication of the viral genome and the subsequent transcription of L (primarily structural) genes. IE2-86 protein autoregulates its own expression by negatively acting on the Major IE Promoter (MIEP) of HCMV (**Figure 1**; Stinski and Petrik, 2008); it binds to the cis-repressive sequence (CRS) of MIEP, resulting in a decrease in MIEP transcription (Stinski and Petrik, 2008). The activities of IE proteins, in turn, determine the expression of E genes, which are divided in two subgroups, β1 (E) and β2 (E-L), according to their time of expression; their functions are mainly related to HCMV replication machinery, viral DNA replication factors, repair enzymes, and immune evasion (Landolfo et al., 2003; Mocarski et al., 2014). L proteins are then expressed according to two different kinetics profiles (γ1 and γ2), shaped by distinct times of expression and their sensitivity to inhibitors of viral DNA replication. L proteins functions are associated to the assembly, maturation and egress of newly formed viral particles from the host cell.

In contrast to lytic replication, only a small subset of HCMV genes are expressed during latency (e.g., transcripts for the latency-associated viral IL-10/UL111A, LUNA [Latency Unique Natural Antigen], US28, UL138, and UL144) without any detectable production of infectious virus (Avdic et al., 2011; Reeves, 2011; Poole et al., 2014). HCMV latency is restricted to primary myeloid progenitors, such as granulocyte macrophage progenitors (GMPs), CD34<sup>+</sup> hematopoietic progenitor cells, and CD14<sup>+</sup> monocytes (Hahn et al., 1998; Scrivano et al., 2011; Rossetto et al., 2013). In these infected cell-types the carriage of viral DNA occurs in the absence of the expression of lytic genes (Mocarski et al., 2014). Only upon the terminal differentiation of these cells into macrophages or dendritic cells (DCs), HCMV lytic genes are expressed: IE gene products activate the expression of other genes, leading to viral DNA replication and de novo virus production (Taylor-Wiedeman et al., 1994; Hahn et al., 1998; Soderberg-Nauclér et al., 2001; Reeves et al., 2005; Britt, 2008; Huang et al., 2012; Mocarski et al., 2014). Since the number of cells carrying the latent viral genome is extremely low in vivo (about 0.001–0.01% of monocytes), it has been necessary to develop complex experimental latency models in order to study the related mechanisms. Thus, experimental HCMV latency and reactivation analyses have been performed using in vitro models that exploit primary myeloid progenitors cells: namely GMP cells, CD34<sup>+</sup> hematopoietic cells, and CD14<sup>+</sup> monocytes (Reeves et al., 2005; Cheung et al., 2006; Reeves and Sinclair, 2010; Liu et al., 2013). All of these studies have highlighted that a crucial aspect of HCMV latency is the repression of MIEP following its association with repressive chromatin markers (Bain et al., 2003; Wright et al., 2005; Sinclair and Sissons, 2006), involving the recruitment of histones and transcriptional silencing factors to the MIEP (Reeves, 2011). In addition to several binding sites for cellular transcription factors (e.g., NF-κB, CREB/ATF, AP1, SRF, Elk-1) that stimulate its transcriptional activity (**Figure 1**; Caposio et al., 2007a, 2010; Lashmit et al., 2009; Isern et al., 2011; Mocarski et al., 2014), the MIEP contains a series of multiple binding sites for transcription factors that may lead to its repression (YY1, ERF and Gfi-1; **Figure 1**; Reeves, 2011). In fact, during HCMV latency, YY1 and ERF bind to MIEP, and recruit histone deacetylases and methyltransferases than then target histones associated with MIEP (Wright et al., 2005). The methylated histones then become targets for the recruitment of heterochromatin protein 1 (HP-1), which augments MIEP repression and contributes to the establishment of latency (Liu et al., 2010). Therefore, permissive or latent infection may be determined by the balance between activating and repressive transcription factors that control MIEP activity (**Figure 1**; Reeves and Sinclair, 2013). However, this balance is thought to be mainly under the control of host inflammatory responses, either in immunocompetent or immunosuppressed individuals (Liu et al., 2013). Various studies have indeed shown that HCMV reactivation, during the allogeneic response to a transplanted

Luganini et al. HCMV Secretome Regulates Viral Infections

organ, is mediated by the expression of inflammatory cytokines (i.e., TNF, LPS, and IL-6) that may, in turn, activate MIEPinteracting cellular transcription factors (e.g., NF-κB and AP-1), thus promoting MIEP activation and lytic HCMV gene expression (Hummel and Abecassis, 2002). However, in other experimental systems, the induction of IE gene expression has been associated with the state of cellular differentiation, as observed in non-permissive monocytes upon their differentiation into permissive macrophages in response to tumor necrosis factor (TNF) and IFN-γ (Soderberg-Nauclér et al., 1997).

### MODULATION OF HOST IMMUNE RESPONSES BY HCMV

Among the HCMV genes not essential for growth in cultured fibroblasts, about 40 have been shown to be involved in the modulation of host immune responses to viral infection by interfering with innate immune mechanisms and counteracting adaptive antibody and T-cell mediated immunity (Mocarski et al., 2014). The ability of HCMV to establish persistent infection, in spite of a robust T-cell and neutralizing antiviral antibody response, is mainly ensured by the activity of many virus-encoded immunomodulatory proteins. HCMV immunomodulation is therefore of pathogenetic importance in the establishment of virus persistence in the host. In this regard, HCMV is considered a master of immune evasion strategies, as highlighted by the results of several functional genomic, transcriptomic, and proteomic analyses that have defined many immunomodulatory functions of several HCMV proteins. In particular, these viral proteins have been observed to: (1) interfere with antigen presentation by major histocompatibility complex-I (MHC-I) molecules (e.g., the viral gene products US2, US3, US6, US10, US11, and UL82); (2) permit immune evasion through interference with and T and NK cells functions (e.g., the proteins encoded by genes UL16, UL18, UL40, UL83, UL141, UL142, US18, and US20); and (3) mimic the activities of cellular cytokines/chemokines (e.g., proteins UL111A, UL128, UL146, and UL147) or act as fake host cytokine/chemokine receptors (as observed for proteins UL144, UL21.5, US27, US28, UL33, and UL78; McSharry et al., 2012; Van Damme and Van Loock, 2014).

The best-characterized mechanisms of HCMV immunoevasion can be summarized as follows:

(1) Interference with antigen presentation by MHC-I molecules. A remarkable example is provided by glycoproteins US2, US3, US6, and US11, characterized by a single transmembrane (TM)-spanning and an immunoglobulin (Ig) domain, which most likely arose by duplication of a single viral ancestor gene (Gewurz et al., 2001). These HCMV glycoproteins, when expressed in fibroblasts, endothelial/epithelial cells, or DCs, induce, through different mechanisms, a decrease in the cell-surface expression of MHC-proteins required for the activation of CD8<sup>+</sup> T lymphocytes (Mocarski, 2002). In particular, both US2 and US11 proteins are able to redirect nascent MHC class I proteins from the ER into the cytosol until ubiquitin-dependent proteasome-mediated degradation occurs. The US2 glycoprotein has also been observed to inhibit MHC class II protein translocation, thus inducing a substantial block of CD4<sup>+</sup> T cell expansion. On the other hand, the US3 gene product prevents the egress of MHC class I proteins from the endoplasmic reticulum (ER) to the Golgi apparatus, whereas the US6 protein is able to bind the transporter of antigen processing (TAP) in the lumen of the ER, preventing ATP-dependent peptide loading on the cell membrane (Mocarski, 2002). In conclusion, this type of immunoevasion prevents the immune recognition of HCMV-infected cells.


host counterparts. In this regard, an important example of a HCMV-encoded chemokine receptor homolog is represented by the multifunctional US28 protein. US28 is considered a putative immunoevasion molecule due to its capability to bind and internalize a range of cellular chemokines, including a broad spectrum of CC and CX3C chemokines, thus limiting the ability of host chemotactic molecules to elicit their effects in immune cells stimulation (Mocarski et al., 2014). However, US28 makes use of both the extracellular chemokine milieu and different intracellular G-protein pathways to produce a wide variety of pathophysiological cell responses (Beisser et al., 2008). In fact, upon stimulation with some chemokines, US28 specifically promotes smooth muscle cells and macrophage migration, thus contributing to both virus dissemination and pathogenesis of HCMV-associated vascular diseases (Vomaske et al., 2009). On the other hand, constitutive expression of US28 induces, in a ligand-independent manner, cyclooxygenase-2 (COX-2) expression via NF-κB activation, leading to the production of VEGF, one of the most abundant angiogenic factors (Maussang et al., 2009).

Moreover, the release of virus-encoded cytokines/chemokines homologs (see HCMV-Induced Secreted Cellular Proteins) permits the virus to modulate specific host immune defense mechanisms in infected tissues, since these bioactive soluble molecules act in a paracrine fashion.

### BIOACTIVE MOLECULES RELEASED FROM HCMV-INFECTED CELLS

The pathogenesis of acute HCMV diseases is related to endorgan damage that results from both lytic virus replication and host immune responses, whereas diseases associated to persistent infections in both immunocompetent and immunocompromised patients (transplant recipients and AIDS patients) are related to chronic inflammation (Britt, 2008). However, besides the direct cytopathic effect of virus replication on host tissues, the pathogenesis of HCMV diseases may be influenced by the activity of virus-induced molecules secreted from virus-infected cells. These secreted factors, of both viral and cellular origin, by acting in a paracrine fashion alter and modify the local microenvironment, thus contributing to the development of HCMV-related diseases (Streblow et al., 2008).

## HCMV-Induced Secreted Cellular Proteins

Early in vitro studies on the ability of HCMV to stimulate the secretion of bioactive cellular factors identified several cytokines and growth factors released from different types of infected cells. First, Almeida et al. (1994) observed that HCMV infection of HUVECs (Human Umbilical Vein Endothelial Cells) greatly increased the expression of IL-6 mRNA and the secretion of this cytokine into the supernatant of infected cells. Dengler et al. (2000) went on to demonstrate the paracrine activity of IL-1β released from HCMV-infected cells, which resulted in an upregulation of pro-inflammatory adhesion molecules on noninfected neighboring cells. In fact, by analyzing the expression of plasma membrane proteins on the surface of both HCMVinfected HUVECs and human vascular smooth muscle cells (hvSMC), they observed a 200-fold overexpression of the Inter Cellular Adhesion Molecule-1 (ICAM-1) and de novo induction of both Vascular Cell Adhesion Molecule 1 (VCAM-1) and E-selectin (Dengler et al., 2000). In regard to the EC model, we observed an upregulation of ICAM-1, IL-8, CCL5/RANTES, CXCL10/IP-10, CXCL11/I-TAC, and COX-2 gene expression in HUVECs cells infected with a low-passage HCMV strain, thus indicating a direct involvement of HCMV in the modification of the extracellular vascular environment through the release of inflammatory mediators (Caposio et al., 2007b).

Using astrocyte and microglia cell models, Maxim et al. (2001) observed that HCMV stimulated the release of CCL2/MCP-1 and IL-8 in the supernatants of infected astrocytes, whereas infection of microglia cells led to an increased secretion of TNFα, IL-6, CCL2/MCP-1, IL-8, CCL5/RANTES, and CCL3/MIP-1α. These findings thus suggest the ability of infected astrocytes to recruit microglia cells, through the release of the chemoattractant CCL2/MCP-1 (Maxim et al., 2001).

However, a different picture emerged from the investigation of the HCMV-mediated modulation of chemokine gene expression in the context of HCMV retinitis – a virus-induced inflammation of the retina characterized by vasculitis and retina degeneration that leads to retinal detachment in immunocompromised hosts. In the in vitro cell model provided by human retinal pigment epithelial cells, Momma et al. (2003) observed, by means of RT-PCR and ELISA assays, that HCMV infection caused an upregulation of IL-8, whereas CCL2/MCP-1 and CCL7/MCP-3 levels were downregulated. The authors thus suggested that the altered secretion of IL-8, CCL2/MCP-1 and CCL7/MCP-3 by retinal cells may be involved in the initiation and development of the inflammatory process responsible for the pathogenesis of HCMV retinitis. In this scenario, the increased secretion of IL-8 may stimulate the recruitment and trafficking of leukocytes to the site of infection within the retina; on the other hand, the decrease in CCL2/MCP-1 and CCL7/MCP-3 levels in the extracellular environment may limit the migration of leukocytes, thus contributing to the virus escape of the host innate immune response (Momma et al., 2003).

Together, these earlier studies demonstrated how HCMV is able to alter in a cell-type specific manner the extracellular microenvironment surrounding infected cells by stimulating the secretion of cellular bioactive factors.

However, only after the introduction of high-throughput large-scale proteomic approaches including antibody-based arrays and liquid chromatography-mass spectrometry (MS/LC), was it possible to drawn up an in-depth depiction of the complexity of the bioactive molecules released upon HCMV infection (Dumortier et al., 2008; Botto et al., 2011; Caposio et al., 2011, 2013; Fiorentini et al., 2011; Mason et al., 2012; MacManiman et al., 2014; Noriega et al., 2014; Gustafsson et al., 2015). These global biochemical characterizations confirmed the cell-type specific patterns of secreted cellular factors from HCMV-infected cells and most of the alterations of types and

quantities of bioactive molecules already observed in earlier studies.

Thanks to proteomics studies, the entire array of bioactive proteins of present in supernatants from different types of HCMV-infected cells, has now been defined, qualitatively and quantitatively, and designated the HCMV secretome (Streblow et al., 2008).

In the first of these studies, Dumortier et al. (2008) analyzed virus-free supernatants from lytically infected fibroblasts (NHDF) to investigate the complexity and heterogeneity of the HCMV-induced secretome. Using gel-free liquid chromatography (LCQ)-MS-MS, they identified more than 1,200 proteins in the secretome of HCMV-infected cells. Among them, several factors involved in angiogenesis (AG) and wound healing (WH) were further confirmed using a wide range human cytokine antibody array (RayBio G Series 2000 arrays). A conspicuous number of cytokines (IL-5, IL-6, IL-1α, IL-1β, GM-CSF, osteoprotegerin, TNF-α, TNF-RI and -RII), growth factors (angiopoietin, angiotensinogen, FGF, GDNF, HGF, IGF-BP, osteopontin, PDGF, PIGF, SPARC, and VEGF), extracellular matrix proteins, chemokines (IL-8/CXCL8 and CXCL1/GRO-α), enzymes, and adhesion molecules were identified (**Table 1**). The biological effects of HCMV secretomederived proteins on AG and WH were then assessed using a modified matrigel in vitro tubule formation assay and an electric WH assay, respectively (Dumortier et al., 2008). However, many of the identified factors in the NHDF-derived secretome are also involved in the inflammatory response and EC activation, thus leading to cell proliferation, adhesion, and inflammatory response, all of which are indeed linked to TVS pathogenesis (**Figure 2**; Dumortier et al., 2008; Streblow et al., 2008). Thus, it was suggested that HCMV infection alters the local microenvironment through the secretion of cellular factors from infected cells that, in turn, by acting in a paracrine fashion, may stimulate AG and WH, both associated to the development of vascular diseases, such as TVS (Streblow et al., 2008).

Subsequent work by Botto et al. (2011) analyzed the secretome of infected HUVECs and identified by means of a cytokine/growth factors antibody array, 29 soluble factors, including IL-6, GM-CSF, and IL-8/CXCL8. A comparison of the HCMV secretomes derived from NHDF and HUVEC, highlighted that the presence in both secretomes of abundant cytokines/chemokines involved in AG, such as IL-6, GM-CSF, IL-8/CXCL8, CCL4/MIP-1β, CCL7/MCP-3, CCL20/MIP-3α, and CXCL11/I-TAC (Dumortier et al., 2008; Botto et al., 2011; Caposio et al., 2011, 2013). Among the EC-derived factors, IL-6 was identified as the major bioactive molecule involved in neovessel formation since its selective depletion severely decreased tubule formation (Botto et al., 2011). IL-6 is an inflammatory cytokine involved in different pathophysiological contexts, such as inflammation, lymphocytes differentiation, cell proliferation, and the inhibition of apoptotic signals (Scheller et al., 2011). IL-6 binding to its receptor (gp130) activates three different intracellular transduction pathways: (i) the JAK/STAT cascade, involved in growth regulation, survival, differentiation, and long-term inflammation-promoting effects (Horvath, 2004); (ii) the ERK1/2-MAPK pathway, and (iii) the phosphoinositol 3-kinase (PI3K)/Akt pathway, both of which are implicated in the regulation of cell growth and differentiation (Wegiel et al., 2008; Rose-John, 2012). In particular, IL-6 binding to its receptor leads to the phosphorylation of STAT3, which subsequently prevents apoptosis by blocking the activation of caspase 3 and 7, and most importantly, by increasing the expression of an apoptosis inhibitor, such as survivin (Hodge et al., 2005). Addition of the secretome derived from HCMV-infected ECs to uninfected ECs was found to stimulate expression of survivin, while antibody neutralization of IL-6 in the same secretome, abolished its capability to induce survivin or activate caspase 3 or 7 (Botto et al., 2011). Therefore, it was concluded that ECderived HCMV secretome promotes EC survival through the expression of survivin via activation of the IL-6 pathway (Botto et al., 2011).

At the same time, in a study aimed at investigating the susceptibility of lymphatic endothelial cells (LECs) to HCMV infection, we characterized in detail the secretome of this particular type of EC (Fiorentini et al., 2011). Although, early reports have shown the presence of HCMV antigens and DNA in lymphoid tissues, the ability of HCMV to infect LECs remained unaddressed due to the lack of a suitable in vitro system. In this study, we observed that a clinical isolate of HCMV productively infected purified lymph nodederived LECs and dysregulated the expression of several LEC genes involved in the inflammatory response to viral infection (Fiorentini et al., 2011). Qualitative and quantitative cytokine antibody array analysis of virus-free supernatants from HCMVinfected LECs revealed the virus-induced secretion of several cytokines, chemokines, and growth factors that may be involved in the regulation of EC physiological properties. Among the released bioactive cellular proteins, the 20 most abundant included several cytokines (IL-1α, IL-3, IL-5, IL-6, IL-13, IL-15, GM-CSF, and TNF-β), several chemokines (CCL1/I-309, CCL7/MCP-3, CCL20/MIP-3α, CXCL1/GROα, CXCL5/ENA-78, CXCL9/MIG, and CXCL11/I-TAC), receptors (ICAM-1 and TNF-R1), and growth factors (TGF-β1 and bFGF). Functional assays then allowed us to establish that the secretome produced by HCMV-infected LECs indeed stimulated AG in both LECs and blood vessel ECs. Furthermore, neutralization of either IL-6 or GM-CSF in the secretome brought about the loss of its angiogenic properties. Involvement of IL-6 and GM-CSF in the HCMV-mediated lymphoangiogenesis was further supported by the finding that recombinant IL-6 and GM-CSF reproduce the same angiogenic effects as seen for the whole LECderived secretome. Thus, these results suggest that HCMV can stimulate both hemangiogenesis and lymphangiogenesis through an indirect mechanism that relies on the stimulation of IL-6 and GM-CSF secretion from virus-infected cells (Fiorentini et al., 2011).

A comparison of the HCMV-secretomes from different types of ECs (LECs and HUVEC) and fibroblasts (NHDF) highlights the presence of several common factors involved in both AG and WH processes, such as IL-6, GM-CSF, CCL7/MCP-3, CCL20/MIP-3α, CCL4/MIP-1β, and CXCL11/I-TAC (**Table 1**; Caposio et al., 2013).

#### TABLE 1 | The most abundant bioactive factors determined by Ray Biotech protein assay analysis in the secretome of lytic HCMV-infected cells.


Taking into consideration the results of the aforementioned studies, the overall effect of HCMV-derived secretomes on AG and WH in ECs is very likely related to the combination of experimental conditions and different bioactive factors, which may cooperate in a synergetic manner. In this regard, Gustafsson et al. (2015) recently observed the influence of different experimental settings, such as the cell type from which the HCMV-secretome is derived, the culture conditions, the time points of supernatant collection, and the concentration of relevant bioactive molecules, in promoting AG and WH in cultured ECs.

Whilst a number of studies have provided evidence for functional roles of HCMV-derived secretomes produced by lytically infected fibroblasts and ECs (Dumortier et al., 2008; Botto et al., 2011; Fiorentini et al., 2011; MacManiman et al., 2014), little is known about the activities of secretomes released from latently HCMV-infected cells. However, exploitation of a new experimental in vitro model of HCMV latency allowed Mason et al. (2012) to obtain some first insights into how latent infection is able to stimulate the release of bioactive molecules. This experimental model involves the infection of CD34<sup>+</sup> hematopoietic progenitor cells cultivated in the absence of stimuli that may promote their differentiation into mature DCs or macrophages, that are fully permissive to lytic HCMV infection. These experimentally latently infected CD34<sup>+</sup> cells showed long-term carriage of viral genomes in which the MIEP remained associated with transcriptionally repressive chromatin, thus hampering further HCMV lytic gene expression. These observations correlate with those made ex vivo of natural HCMV latent infections (Reeves et al., 2005), and provide evidence in support of the use of experimentally latently infected CD34<sup>+</sup> cells as an effective model for the study of HCMV latent infection and reactivation (Reeves et al., 2005; Reeves and Sinclair, 2010). Indeed, analysis of supernatants from HCMV-infected CD34<sup>+</sup> cells by large-scale cytokine antibody array has revealed the presence of many bioactive cellular proteins known to be involved in both regulation of the immune response and chemoattraction of T cells (**Table 2**). The most highly abundant cellular factors identified in this latent HCMV-secretome included: several cytokines/chemokines (XCL1/LYMPHOTACTIN, CXCL9/MIG, CXCL12/SDF-1, CXCL13/BLC, CCL1/I-309, CCL2/MCP-1, CCL8/MCP-2, CCL13/MCP-4, CCL15/MIP-1δ, CCL17/TARC, CCL20/MIP-3α, CCL27/CTACK, IL-8, IL-10/CSIF, IL-13, IL-15, LEPTIN, LTA/TNF-β, and TNF-α), growth factors (BMP-4, BMP-6, CSF-1/M-CSF, IGF-1/MGF, IGFBP3/IBP3; NT-3/NTF3, TGF-β1, and TGF-β3), a receptor (ICAM-1), and the angiogenic RNase angiogenin. However, an increase in secreted CCL2 was also observed from latently infected GMPs, and related to the stimulation of CD14<sup>+</sup> monocytes migration toward the site of latency (Stern and Slobedman, 2008); whereas secretion of both CCL2 and CCL8 was measured during a short-term experimental latent infection of CD14<sup>+</sup> monocytes, and linked to the recruitment of CD14<sup>+</sup> monocytes and CD4<sup>+</sup> Th1 cells to latently infected cells (Noriega et al., 2014). Functional analysis of the HCMV-secretome derived from latently infected CD34<sup>+</sup> cells, however, highlighted that chemoattraction of CD4<sup>+</sup> T cells exclusively depended on the binding of CCL8 to CC chemokine receptors CCR3 and CCR5 expressed on CD4<sup>+</sup> T cells. In fact, the depletion of CD4<sup>+</sup> T cells bearing CCR3 or CCR5 receptors, as well as treatment of the secretome with a neutralizing anti-CCL8 antibody, resulted in a substantially blockade of CD4<sup>+</sup>

#### TABLE 2 | Main bioactive proteins assessed in the HCMV secretome of latently infected CD34+ cells.


T-cell migration toward latently infected cells (Mason et al., 2012). Moreover, in the same HCMV-secretome, high levels of TGF-β and cIL-10, both involved in the induction of immune tolerance by suppressing proliferation and T-cells functions were detected (Bettini and Vignali, 2009). This observation, led to the hypothesis that the secretome of latently infected CD34<sup>+</sup> cells may create an immunosuppressive microenvironment, thus interfering with the immune recognition of latently infected CD34<sup>+</sup> and promoting viral persistence. This hypothesis was indeed supported by the finding that TGF-β and cIL-10 in the latent secretome inhibited both the CD4<sup>+</sup> Th1-mediated cytotoxicity and the production of antiviral Th1-type cytokines, such as IFN-γ, TNF-α, and TNF-β. Furthermore, treatment of latent secretome with neutralizing antibodies specific for cIL-10 and TGF-β significantly restored IFN-γ production and the cytotoxic functions of CD4<sup>+</sup> cells, thus confirming that the observed inhibition of CD4<sup>+</sup> T cells functions strictly depended on the secretion of TGF-β and cIL-10 by latently infected CD34<sup>+</sup> cells (Mason et al., 2012). In addition, cIL-10 may prevent cell death and promote survival of latently infected cells (Weber-Nordt et al., 1996; Poole et al., 2011), since it was observed that cIL-10 is required for upregulation of the anti-apoptotic protein PEA-15 – which abolishes FAS-induced apoptosis (Poole et al., 2015). As described in Section "HCMV-Induced Secreted Cellular Proteins" above, the viral homolog LAvIL10 also stimulates the release of cIL-10 in latently infected CD34<sup>+</sup> cells. Subsequently, cIL-10 may downregulate the expression of cellular microRNA hsa-miR-92a, thus inducing the secretion of the cellular CCL8 (Poole et al., 2014). Together these findings suggest that TGF-β and cIL-10 in the latent HCMV secretome by acting in a paracrine manner on uninfected bystander CD34<sup>+</sup> cells, create and further expand a microenvironment conducive to the latent carriage of HCMV genomes. In fact, it was observed that uninfected CD34<sup>+</sup> cells synthetized TGF-β and cIL-10, when cultured in the presence of the latent secretome (Mason et al., 2012). Overall, these findings support the existence of a complex virus strategy during latency aimed at the generation of a heavily immunosuppressive microenvironment around latently infected CD34<sup>+</sup> that allows them to evade host immune recognition (**Figure 3**). In fact, even if HCMV infections leads to the recruitment of CD4<sup>+</sup> T cells to site of latency through the release of CCL8, it then inhibits their antiviral functions by generating the strong immunosuppressive microenvironment, thereby avoiding the elimination of latently infected cells (Sinclair and Reeves, 2013). Furthermore, the upregulation of cIL-10 protects latently infected CD34<sup>+</sup> from apoptosis, thus

promoting their survival in order to maintain latent viral genome carriage (Poole et al., 2015).

Although, HCMV lytic and latent host-cell interactions involve different target cell types and viral genetic programs, and lead to different outcomes, many bioactive factors are common between the HCMV-induced secretome derived from lytically and latently infected cells. In particular, various cytokines/chemokines, such as CXCL9/MIG, CCL1/I-309, CCL8/MCP-2, CCL15/MIP-1δ, CCL20/MIP-3α, IL-8, IL-10, IL-13, IL-15, LTA/TNF-β, and TNF-α, the growth factor TGF-β, the receptor ICAM-1, and the enzyme ANG are present in both lytically- and latently-infected cells secretomes. However, the presence of a set of common bioactive factors between the two functionally different secretomes is of great interest considering the opposite outcomes of both forms of infections; indeed, we might hypothesize that some of the pathophysiological alterations of the microenvironment surrounding HCMVinfected cells that are induced by the bioactive molecules that the two secretomes have in common, may occur during both lytic and latent infection.

### Secreted HCMV Proteins

Many viral-encoded secreted bioactive proteins have been identified in the supernatants of herpesvirus-infected cells. Among them, cytokine-like proteins, such as the viral counterparts of cellular IL-10, IL-17, and IL-6 were found in the supernatants from cells infected with Epstein–Barr virus (EBV; Ryon et al., 1993; Swaminathan et al., 1993), herpesvirus saimiri (Yao et al., 1995), or human herpesvirus-8 (HHV-8; Neipel et al., 1997), respectively. In this section, we review the main biochemical and functional features of the HCMV-encoded proteins that have identified to date within the secretome of infected cells.

### UL21.5

This was the first protein of viral origin identified as a secreted product in the supernatant of HCMV-infected cells (Müllberg et al., 1999). UL21.5 mRNA is abundantly expressed at the late stage of HCMV infection (Rawlinson and Barrell, 1993) and is transcribed by two exons; the first encodes a signal peptide of 20 amino acids followed by an additional sequence of eight amino acids, while the second exon codes the remaining 75 residues. Intriguingly, UL21.5 mRNA is also packaged into virions, thus suggesting a role of the encoded protein even before the infecting viral genome reaches the nucleus and becomes transcriptionally active (Bresnahan and Shenk, 2000).

Later, it was demonstrated that pUL21.5 is a soluble CC chemokine receptor able to selectively bind CCL5/RANTES with high affinity and thus preventing its interaction with cellular receptors (Wang et al., 2004). pUL21.5 by acting as a secreted receptor decoy even at a distance from infected cells, may therefore counteract chemokine-mediated host antiviral response (Wang et al., 2004).

### UL146 and UL147

fmicb-07-00715 May 12, 2016 Time: 17:53 # 11

UL146 and UL147 proteins are considered as potential viral encoded CXC-chemokines due to the presence of putative signal sequences, cysteine spacing, and (in pUL146 only) an ELR-CXC sequence (glutamic acid -E-, leucine -L-, arginine -R-, nominated the ELR sequence, upstream of the CXC motif). Such features are common to cellular CXC cytokines, such as IL-8 and Groα (Penfold et al., 1999; McSharry et al., 2012). Sequence analysis of UL146 and UL147 proteins from the Toledo strain of HCMV reveals 24 and 16% amino acid identity to IL-8 (McSharry et al., 2012), respectively; accordingly, UL146 protein is also known as vCXCL-1, and UL147 protein as vCXCL-2.

While the functions of UL147 have not yet been characterized, light has been shed on those of UL146 (Miller-Kittrell et al., 2007; Heo et al., 2015). vCXCL-1 (UL146) is a 117 amino acids glycoprotein secreted into the culture medium with late kinetics. The UL146 gene product is highly polymorphic among low-passage strains (Heo et al., 2008) and absent in laboratory strains of the virus, such as AD169 and Towne (Cha et al., 1996). Recently, it was observed that vCXCL-1 is able to: (i) induce calcium flux via CXCR2 (Heo et al., 2015); (ii) selectively stimulate the expression of beta2 integrin (CD11b and CD11c), responsible for the adherence of leukocytes to the vascular endothelium through an interaction with ICAM-1 and for modulating the life span of neutrophils (Mayadas and Cullere, 2005; Heo et al., 2015); (iii) induce the migration of neutrophils in vitro (Heo et al., 2015); and (iv) upregulate CCL22, a chemokine involved in chemotaxis in monocytes, dendritic and NK cells, and chronically activated T lymphocytes (Heo et al., 2015).

Together, the capability of vCXCL-1 to alter the trafficking of HCMV-infected neutrophils, as potential HCMV carriers, may facilitate viral dissemination and promote the spread of HCMV through the host (McSharry et al., 2012).

### UL128

The UL128 protein is part of the pentameric complex composed of gH (pUL115), gL (pUL75), pUL130, and pUL131A that is required for HCMV entry into epithelial, endothelial and DCs (Revello and Gerna, 2010). However, pUL128 contains four conserved cysteine amino acids at its N-terminus, similar to those found in the CC-chemokines (Akter et al., 2003). To this regard, it was observed that UL128 is able to induce the migration of human PBMCs (peripheral blood mononuclear cells) to levels comparable to those induced by the human CCL3/MIP-1α chemokine (Zheng et al., 2012). In addition, treatment with soluble pUL128 increases the expression of both IL-6 and TNF-α in PBMCs, and stimulates PBMCs proliferation through activation of the MAPK pathway (Zheng et al., 2012). Alteration in PBMCs trafficking and cytokines secretion by soluble pUL128 may thus facilitate HCMV dissemination through the recruitment of infected carrier cells.

### IL-10

Among the HCMV-encoded cytokine-like molecules, the latency-associated viral cellular IL-10 homologu (LAcmvIL-10), encoded by HCMV UL111A gene is worthy of note (Ouyang et al., 2014). This HCMV gene encodes two IL-10 homologs generated by alternative splicing (Kotenko et al., 2000; Jenkins et al., 2004): cmvIL-10 of 175 amino acids, expressed during productive infection with late kinetics (Spencer et al., 2002; Chang et al., 2004), and LAcmvIL-10, of 139 amino acids, the result of a C-terminal truncation, reported to be expressed with early kinetics during both productive and latent infection (Jenkins et al., 2004).

As the cellular IL-10 (cIL10), cmvIL-10, and LAcmvIL-10 downregulate the expression of MHC-II molecules in latently infected GMPs (Spencer et al., 2008). However, cmvIL-10 shares more functions in common with cIL-10 than LAcmvIL-10 (Jenkins et al., 2008b). In fact, cmvIL-10, but not LAcmvIL-10, increases the expression of the IgG (FCγ) receptors CD32 and CD64, and increases FCγ receptor-mediated phagocytosis (Jaworowski et al., 2009). Similar to cIL-10, cmvIL-10 then inhibits the expression of proinflammatory cytokines in LPSstimulated MDDCs (Monocyte-Derived Dendritic Cells; Jenkins et al., 2008a,b). Concerning the role of LAcmvIL-10, it was recently observed that its expression in latently infected CD34<sup>+</sup> cells resulted in the downregulation of cellular microRNA hsamiR-92a, which upregulates the myeloid transcription factor GATA2 (Poole et al., 2013, 2014). GATA2, in turn, increases the transcription of the latency-associated HCMV genes LUNA and UL144 (Reeves, 2011; Poole et al., 2013), as well as the transcription of the cellular IL-10 (Poole et al., 2011). Since cIL-10 is one of the most abundant cytokines in the secretome produced by cells latently infected with HCMV, and believed to create an immune suppressive environment and to suppress apoptosis (Mason et al., 2012; Poole and Sinclair, 2015), LAcmvIL-10 is thought to reduce the ability of CD4<sup>+</sup> cells to recognize HCMVinfected cells during latent infection (Mason et al., 2012; Poole and Sinclair, 2015).

### UL7

More recently, a novel HCMV-encoded secreted molecule, the protein encoded by UL7 gene, has been identified as a critical component of the HCMV-secretome responsible for vascular dysregulation associated with persistent HCMV infection (MacManiman et al., 2014). The UL7 gene belongs to the RL11 family, located in the left end of the viral genome, and characterized by early-late-phase kinetics during lytic HCMV infection (Engel et al., 2011). Structure prediction of UL7 protein identified a 222-amino-acid type I glycoprotein characterized by a putative leader peptide (35 amino acids), an extracellular immunoglobulin superfamily domain (Ig-like domain; 102 amino acids), a mucine-like stalk region (55 amino acids), a hydrophobic transmembrane sequence (22 amino acids) and a short cytoplasmic tail (eight amino acids; Engel et al., 2011). Engel et al. (2011) observed that the UL7 protein is proteolytically cleaved in correspondence with the stalk region, resulting in a heavily glycosylated ectodomain that is released from infected cells. Interestingly, the UL7 Ig-like domain shares significant amino acid identity with both CD229 (member of SLAM family), which is involved in T-cell signaling (Engel et al., 2011), and the carcinoembryonic antigen-related cell adhesion molecule (CEACAM) protein, that is highly expressed during

vasculogenesis (MacManiman et al., 2014). However, UL7 is not able to bind directly CD229 or any other member of the SLAM family, but it can contact a putative UL7 ligand on the surface of monocyte-derived DCs, thus interfering with pro-inflammatory responses (Engel et al., 2011). In particular, UL7 has been observed to downregulate the production of pro-inflammatory cytokines, such as TNF, IL-6, and IL-8 in primary human monocyte-derived DCs and in the PMAinduced myeloid cell lines, U937 and THP-1 (Engel et al., 2011). These findings have thus suggested UL7 as a novel HCMV-encoded product able to contribute to the inhibition of the host antiviral defense in favor of the establishment of persistent infection (Engel et al., 2011). Moreover, the activity of CEACAM-1 in promoting vasculogenesis led MacManiman et al. (2014) to hypothesize a potential role of UL7 in this process. In support of this, using in vitro AG assays, they found that the secretome generated by an UL7-deficient HCMV produced a 50% reduction in tubulogenesis compared with that for the secretome from wild-type virus. Furthermore, following adenovirus-mediated transduction of human aortic endothelial cells (HEACs) with either the full length UL7 or the UL7 ectodomain, they observed the appearance of a robust network of interconnecting tubules, thus sustaining the involvement of UL7, and in particular its ectodomain, in the promotion of ECs differentiation (MacManiman et al., 2014). Moreover, the observation of an increased secretion of IL-6 and the activation of STATs and MAPK pathways in ECs overexpressing UL7 further substantiates the role of UL7 in angiogenesis (MacManiman et al., 2014). Taken together, these findings support a role of secreted UL7 protein in regulating some physiological properties of ECs and in the stimulation of angiogenesis.

### The Importance of HCMV-Induced Bioactive Molecules in Virus Pathogenesis

Human cytomegalovirus infections are associated with the acceleration of various long-term vascular diseases, especially in immunocompromised hosts (Griffiths et al., 2015). In these settings, HCMV has been correlated to the acceleration of a number of vascular diseases, such as atherosclerosis, restenosis, and TVS, which are all determinants of chronic rejection (CR), one of the most important long-term conditions leading to graft failure and re-transplantation (Grahame-Clarke, 2005; Streblow et al., 2007). TVS is a complex phenomenon in which the leading event is a diffuse and concentric intimal proliferation, which, in turn, determines vessel occlusion due to perivascular inflammation, ECs dysfunction, and hvSMC proliferation with extracellular matrix deposition. In addition, the progression of TVS is promoted by the same mechanisms that lead to AG and WH (Streblow et al., 2007, 2008). In early TVS lesions, macrophages, T cells, B cells, and NK cells are present, while progression to late lesions is associated with a thinning intima containing SMC and macrophages (Streblow et al., 2005). Strong correlations between HCMV reactivation and the acceleration of vascular diseases development have been made on the basis of: (i) the high efficiency rate of infection for those cell types involved in TVS, including EC, SMC, and monocyte-derived macrophages (Lemström et al., 1993); (ii) the delay in the TSV progression and a prolonged graft survival in transplant recipients after treatment with ganciclovir, an approved anti-HCMV drug (Lemström et al., 1997); and (iii) in vivo studies on CMV animal models (Lemström et al., 1997; Streblow et al., 2003). In particular, in a rat model of aortic allografts, it was observed that infection with rat cytomegalovirus (RCMV) accelerates atherosclerosis, leading to an overall rearrangement of both the structural and functional architecture of vessel cells (Lemström et al., 1993; Streblow et al., 2003), the reversal of these effects through the use of an anti-CMV agent, such as ganciclovir, further support a role of the virus in the pathogenesis of experimental atherosclerosis (Lemström et al., 1994, 1997). RCMV infection has also been associated with an altered profile of released cytokines, in which higher levels of IL-2 and IL-4 were evident compared with controls (Zhou et al., 1999).

Furthermore, the observation that the HCMV-secretomes from both fibroblasts and ECs stimulate AG and WH in in vitro models suggests that HCMV infection of the allograft in transplant recipients and the subsequent release of a virus-altered secretome may promote AG and WH in the host, thus resulting in the acceleration of TVS (Streblow et al., 2008; Caposio et al., 2011, 2013). On the other hand, the stimulation of "de novo" lymphatic and blood vessels sprouting may represent a strategy adopted by the virus to enhance its replication and dissemination from the initial infection site to other host tissues.

In contrast, the available data about the activities of the HCMV-derived secretome from latently infected cells, indicate that the generation of a secretome-induced immunosuppressive microenvironment around latently infected cells may help the virus escape recognition by the hosts's immune system, thus favoring cell survival, viral persistence and potential reactivations.

Clearly, in both lytic and latent infections, the activities of the wide range of bioactive molecules in HCMV-secretomes may have severe pathogenetic consequences since they facilitate both virus dissemination and persistence in the host in spite of a robust antiviral response.

## CONCLUSION

One of the main intriguing features of HCMV biology is the exceptionally wide arsenal of virus-encoded proteins that have been observed to be capable of counteracting host innate and adaptive immune defenses. Persistence of the virus, even in the immunocompetent host, and its ability to avoid immune clearance may be due in part to the activities of the many viral immunomodulatory proteins (Mocarski et al., 2014). In this review, we have examined what is known about the bioactive molecules, of both viral and cellular origin, secreted from HCMV-infected cells. These proteins are able to modify the extracellular environment of host infected tissues, thus promoting both virus dissemination and persistence, as well as some pathophysiological processes, such as AG and WH, that may drive the development of associated vascular diseases.

However, more work is needed to fully appreciate the importance of HCMV-derived secretomes in the pathogenesis of HCMV diseases.

To date, ECs, fibroblasts, and hematopoietic precursor cells have been mainly used to investigate the composition and functions of HCMV-derived secretomes; but, in the future, a deeper understanding of the importance of the HCMV secretome may come from the analysis of secretomes derived from other cell types. For example, little is known about the impact of HCMV infection on the secretion of bioactive molecules from epithelial cells, which represent one of the main cell targets for productive infection in the natural host. Does the HCMV-induced secretome of epithelial cells able to alter the host's immune recognition of infected cells or to catalyze organ damage in relation to lytic infection as occurs in the retina (Momma et al., 2003)?

The findings that we have reviewed here have led to a better comprehension of both the viral and cellular bioactive molecules involved in the host-HCMV relationship. Little is known, however, about the molecular mechanisms underlying the modulation and secretion of such cellular bioactive proteins. For example, the HCMV mechanisms that interfere directly or indirectly with the cellular secretion pathways remain to be identified. Knowledge of these molecular dynamics would be

### REFERENCES


useful for the development of strategies aimed at attenuating or blocking the ability of HCMV to interfere with the host immune responses that contributes to the pathological outcome. Finally, it will also be necessary to confirm the in vitro findings in in vivo models, such as a humanized mouse model. The confirmation of experimental data and hypotheses would support the rationale to deepen our understanding of HCMV immunomodulation in order to identify targets for novel therapeutic strategies to prevent or control HCMV diseases.

### AUTHOR CONTRIBUTIONS

AL wrote the manuscript; MT wrote the manuscript; and GG conceived of the review and wrote the manuscript.

### ACKNOWLEDGMENTS

The authors would like to acknowledge support from the Italian Ministry for Universities and Scientific Research (Research Programs of Significant National Interest, PRIN 2010-11, grant no. 2010PHT9NF) to GG, and Ex-60% to AL, MT, and GG.

in quiescent cells and compensate for inactivation of the NF-κB sites in proliferating cells. J. Virol. 84, 4481–4493. doi: 10.1128/JVI.02141-09


that induce angiogenesis and wound healing. J. Virol. 82, 6524–6535. doi: 10.1128/JVI.00502-08




**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 Luganini, Terlizzi and Gribaudo. 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 Neuro-endocrinological Role of Microbial Glutamate and GABA Signaling

### Roberto Mazzoli\* and Enrica Pessione

Laboratory of Biochemistry, Proteomics and Metabolic Engineering of Prokaryotes, Department of Life Sciences and Systems Biology, University of Torino, Torino, Italy

Gut microbiota provides the host with multiple functions (e.g., by contributing to food digestion, vitamin supplementation, and defense against pathogenic strains) and interacts with the host organism through both direct contact (e.g., through surface antigens) and soluble molecules, which are produced by the microbial metabolism. The existence of the so-called gut–brain axis of bi-directional communication between the gastrointestinal tract and the central nervous system (CNS) also supports a communication pathway between the gut microbiota and neural circuits of the host, including the CNS. An increasing body of evidence has shown that gut microbiota is able to modulate gut and brain functions, including the mood, cognitive functions, and behavior of humans. Nonetheless, given the extreme complexity of this communication network, its comprehension is still at its early stage. The present contribution will attempt to provide a state-of-the art description of the mechanisms by which gut microbiota can affect the gut–brain axis and the multiple cellular and molecular communication circuits (i.e., neural, immune, and humoral). In this context, special attention will be paid to the microbial strains that produce bioactive compounds and display ascertained or potential probiotic activity. Several neuroactive molecules (e.g., catecholamines, histamine, serotonin, and trace amines) will be considered, with special focus on Glu and GABA circuits, receptors, and signaling. From the basic science viewpoint, "microbial endocrinology" deals with those theories in which neurochemicals, produced by both multicellular organisms and prokaryotes (e.g., serotonin, GABA, glutamate), are considered as a common shared language that enables interkingdom communication. With regards to its application, research in this area opens the way toward the possibility of the future use of neuroactive molecule-producing probiotics as therapeutic agents for the treatment of neurogastroenteric and/or psychiatric disorders.

Keywords: inter-kingdom communication, gut–brain axis, gut microbiota, enteric nervous system, enteroendocrine cells, neuroactive molecules, mood, anxiety

### COMMUNICATION SIGNALS THROUGH DIFFERENT BIOLOGICAL KINGDOMS

Living organisms exchange information through systems that are based upon signal-receptor interactions. The specific sense organs of higher animals have evolved to perceive signals, but the whole living world, including protozoans, plants, fungi and bacteria, efficiently communicates by exchanging information at a molecular level. Some universally sensed signals and widespread receptors can be found in different kingdoms.

#### Edited by:

Janice Lorraine Strap, University of Ontario Institute of Technology, Canada

#### Reviewed by:

Kevin Graeme Murphy, Imperial College London, UK Attila Szabo, University of Oslo, Norway

> \*Correspondence: Roberto Mazzoli roberto.mazzoli@unito.it

#### Specialty section:

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

Received: 20 May 2016 Accepted: 17 November 2016 Published: 30 November 2016

#### Citation:

Mazzoli R and Pessione E (2016) The Neuro-endocrinological Role of Microbial Glutamate and GABA Signaling. Front. Microbiol. 7:1934. doi: 10.3389/fmicb.2016.01934

It has been suggested that ghrelin-like molecules, i.e., those structurally related to the appetite-stimulating lipopeptide hormone which was first isolated from the gastrointestinal tract of rats (Kojima et al., 1999), are ubiquitously present in living organisms (Aydin et al., 2006). These molecules have been found in microorganisms (i.e., viruses, archaea, and phototrophic bacteria), animals and plants, where they fulfill different functions (e.g., inhibition of apoptosis), especially those connected with food intake and cell proliferation (Aydin, 2007). For instance, a structural homolog (N-octanoyl homoserine lactone) of ghrelin is produced by Gram-negative bacteria as a quorum sensing (QS) autoinducer and cell-to-cell communication molecule, which is also involved in food-searching (Aydin et al., 2006).

Bacteria produce a huge range of compounds that are involved in inter-microbial and host–microbe relationships. It is wellknown that QS peptides, once interpreted as specific bacteria– bacteria communication signals (and regulating phenotypes such as competence, sporulation, bioluminescence, biofilm formation, bacteriocin/toxin production) also have metal (e.g., iron) binding and antibiotic properties, that is, additional functions which can promote the improvement of human health (Schertzer et al., 2009). It has been established that antimicrobial pigments, such as Serratia spp. prodigionines (active against fungal and protozoan infections), can also control cancer and immunity (Williamson et al., 2006). So, a single signal can be detected by different cellular systems. "Dark side" examples also exist, as in the case of farnesol, a QS-like molecule synthesized by the yeast Candida albicans. Farnesol controls yeast/mycelium transition, but also triggers the disruption of human erythrocytes, and interferes with cytokine expression (Hornby et al., 2001; Navarathna et al., 2007). The extremely long coevolution of host and colonizing microbes can account for these inter-kingdom effects.

The neurochemicals and their related receptors that are found in mammals are far from being an exclusive to these organisms, and are widely dispersed throughout nature, even in microorganisms and plants that do not have nervous systems. For instance, the stress-related neuro-endocrine hormone family of catecholamines has also been demonstrated in fish, insects, plants and bacteria (Lyte, 2011). Microorganisms are able to produce a wide variety of neurochemicals. This feature has been studied extensively by scientists engaged in food safety, since the presence of neuroactive molecules in food has been found to be responsible for cases of poisoning, such as the "cheese reaction" and the "scombroid syndrome" (related to food containing large amounts of tyramine and histamine, respectively) (Pessione et al., 2005, 2009). It has been suggested that at least some molecular neurotransmitters and/or neuromodulators and/or neurohormones have been conserved or shared during co-evolution as the "words" of a common language, thus allowing communication between phylogenetically very distant organisms (Lyte, 2011). It has been hypothesized that such a common language contributes to the homeostatic regulation of gut microbiota and possibly to the functioning of the brain and behavior (Lyte, 2011).

### MICROBIAL NEURO-IMMUNO-ENDOCRINOLOGY: HOW GUT MICROBIOTA AND THE HOST CROSS-TALK

Microbial neuro-endocrinology is a branch of microbiology which has recently arisen, and it refers to the bidirectional communication that exists between commensal or parasite microbiota and the host. Although, some reports refer to the ability of bacteria to control the level of neurotransmitters through Toll-like receptors (TLRs) and heat-shock proteins (Sivamaruthi et al., 2015), this form of communication may occur directly by means of neurochemicals.

The first evidence that bacteria could respond to mammalian neuro-endocrine hormones dates back to 1929. These studies reported the effect of neuro-hormones, and in particular those belonging to the catecholamine family, on enhancing the pathogenic phenotype of several bacterial strains, such as Clostridium perfringens, during in vivo infections (Lyte, 2011). At that time, these observations were interpreted to mean that neurochemicals suppressed local immunity and thus favored the rapid unimpeded growth of infectious microbes (Lyte, 2011). However, evidence that prokaryotes can directly sense and respond to neurochemicals (e.g., altering their growth and/or virulence potential) has started to accumulate since the early 1990s. As an example, it is known that some intestinal molecules, such as serotonin [5-hydroxy tryptamin (5-HT)], can modulate the pathogenic potential of Pseudomonas fluorescens by affecting its motility and pyoverdin production, but without affecting its growth (Biaggini et al., 2015). Currently, this feature is not considered rare, but is widespread over a high number of diverse microorganisms (Lyte, 2011).

These findings have gained a great deal of attention in light of the parallel emerging evidence of the so-called microbiota-gutbrain axis, which allows bidirectional communication between the gut microbiota and the central nervous system (CNS) of animals, including humans (Cryan and Dinan, 2012). It has been reported that gut microbiota can control the tryptophan metabolism of the host by enhancing the fraction of tryptophan available for the kynurenine route and decreasing the amount available for 5-HT synthesis (O'Mahony et al., 2015). These authors suggested that the deficiencies observed in the serotonergic system of the elderly could be due to quantitative and qualitative modifications of gut microbiota during aging (O'Mahony et al., 2015). The cognitive and behavioral functions of germ-free mice can be altered following colonization with microbiota from a different mouse strain, and this can lead to a similar increased exploratory behavior to that of the donor mouse strain (Bercik et al., 2011b).

The observation of psychiatric co-morbidities in various chronic inflammatory intestinal disorders has been interpreted by some authors as further evidence of the influence of gut microbiota on the CNS (Bercik et al., 2012; Cryan and Dinan, 2012). Altered gut microbiota has been observed in individuals affected by severe psychiatric disorders, such as autism (Adams et al., 2011; Williams et al., 2011). Studies on

animals infected by pathogenic bacteria, or treated with probiotic or antibiotic agents have suggested that gut microbiota may play a role in the genesis of multiple sclerosis, anxiety and depression (Berer et al., 2011; Cryan and Dinan, 2012). However, many psychiatric co-morbidities are thought to be the result of perinatal infections, and should be interpreted within the context of altered predisposition rather than microbiota-CNS communication (Buka et al., 2008; Eßlinger et al., 2016; Gilman et al., 2016). Since clear evidence of the role of gut microbiota in the development of these syndromes is still lacking, this subject is still the topic of debate.

However, all these reports have contributed to the concept of the microbiota-gut-brain axis emerging as a "circular" interactive network consisting of multiple (immune, neural, and endocrine) interaction mechanisms. On one hand, the role of the psychological status in the pathogenesis of/recovery from infectious disease is recognized within the medical community (Lyte, 2011). On the other hand, these studies have revolutionized the traditional conception of human–microbiota relationships, and hopefully opened fascinating perspectives as regards the balancing of microbiota through the application of probiotics as adjuvant agents in both neuro-enteric and psychiatric disorders (Mazzoli, 2014). However, it should be pointed out that, due to their extreme complexity and several confounding factors, the understanding of the molecular mechanisms by which intestinal microbiota can influence gut–brain communication or respond to human-derived signals is still at its very beginning. And it is also for this reason that the efficacy of probiotics as therapeutic agents in psychiatric disorders is still a subject of massive debate within the circle of experts (Bron et al., 2011; Kelly et al., 2015; Santocchi et al., 2016).

### The Gut–Brain Axis

The gut–brain axis defines a bidirectional communication system that connects the GI tract with the CNS through multiple pathways involving neural, endocrine, and immune cells. The gut brain axis allows the CNS to regulate GI functions, including motility and secretion, and the GI tract to signal sensations such as hunger, pain, or discomfort to the CNS (Julio-Pieper et al., 2013). However, new complex emotional (affective mood), cognitive (memory formation) and behavioral functions (food intake) are beginning to be discovered (Berntson et al., 2003), and bacteria could play a role in this scenario (Cani and Knauf, 2016). Direct evidence that gut stimuli can affect emotional states includes the fact that the intragastric infusion of fatty acids in humans has been shown to reduce the brain response to experimentally induced sad emotions (Van Oudenhove et al., 2011).

The reciprocal influence between the GI tract and the CNS is sustained by the enteric nervous system (ENS) (Furness, 2006; Cryan and Dinan, 2012), which is considered as the third branch of the autonomic nervous system and consists of about 200– 600 million neurons. Owing to its high number of neurons, complexity, and similarity in signaling molecules with the brain, ENS has also been referred to as the "second brain" (Gershon, 1998). ENS interfaces with the gut-associated lymphoid tissue (GALT, which contains more than two-thirds of the body's immune cells) and with 1000s of entero-endocrine cells (EECs, which contain more than 20 identified hormones). Furthermore, the GI tract is directly innervated by the CNS through spinal and vagal afferents (Mayer, 2011). For this reason, gut-to-brain communications can occur through three main routes: (i) direct perception of stimuli by primary afferent neurons (belonging to the ENS or the CNS); (ii) immune system mediated connections; (iii) EEC mediated connections.

### Direct Neuronal Perception of Gut Chemical Stimuli

It is generally believed that the afferent terminals that innervate the gut cannot sense luminal chemosignals directly, but can through intermediate cells (e.g., EECs, immune cells) (Mayer, 2011). Under physiological conditions, protein-sized molecules cannot pass the intestinal epithelium through the paracellular pathway. However, lipophilic and small hydrophilic compounds of up to 600 Da can cross the intestinal barrier through multiple routes, e.g., through transcellular (namely, passive diffusion into the lipid bilayer and/or small aqueous pores) and paracellular routes (Keita and Söderholm, 2010). It can therefore be hypothesized that small neuroactive compounds may also use the latter pathways to diffuse in the lamina propria, that is, in contact with intrinsic and/or extrinsic neural afferents, or into the portal circulation, and therefore possibly exert extraintestinal effects. Some experimental evidence can in fact support this hypothesis (Psichas et al., 2015). For instance, FFAR3 receptors for short chain fatty acids (SCFAs) have been detected in submucosal and myenteric ganglia (Nøhr et al., 2013), and the responsiveness of enteric neurons to glucose, amino acids and fatty acids has been demonstrated (Liu et al., 1999; Furness et al., 2013; Neunlist and Schemann, 2014). Furthermore, pattern recognition receptors that are able to bind a variety of microbial antigens have been identified in neural cells, including enteric neurons, which could possibly enable a rapid activation without the intermediation of immune cells (Lim et al., 2016) (for more details, see Bacteria-Immune-Endocrine-Vagal Connections).

### Neuroimmune Connections

About 70% of the body's immune cells are present within the GALT (Mayer, 2011). The existence of close connections between afferent nerve terminals and immune cells (e.g., plasma cells, eosinophils, and mast cells) within the gut mucosa has been well-established (Keita and Söderholm, 2010; Mayer, 2011). For example, the terminals of some vagal afferents can respond to a variety of neuroactive compounds secreted by lymphocytes and mast cells, such as histamine, 5-HT, prostaglandins and various cytokines (Keita and Söderholm, 2010). Furthermore, neuropeptide receptors have been identified in mast cells, thus suggesting bi-directional communication between the nervous and immune systems (Keita and Söderholm, 2010).

### Neuro-endocrine Connections

Entero-endocrine cells account for less than 1% of epithelial cells in the gut; nonetheless, they constitute the largest endocrine organ of the body (Mayer, 2011). More than 20 different types of EECs have been identified, and these differ according to the type(s) of regulatory peptides (e.g., glucagon-like peptides,

GLPs, pancreatic peptide YY, PYY, cholecystokinin, CCK, secretin) or bioactive molecules that they secrete. EECs regulate digestive functions through ENS circuits, and communicate with the CNS (e.g., with the hypothalamus), either directly, i.e., through endocrine pathways, or through paracrine signaling to vagal afferents (Raybould, 2010). Although, EECs and afferent neurons have long been thought to communicate indirectly, i.e., through neuropeptides released by EECs, a recent study has also demonstrated direct synaptic contacts between EECs and neurons (Bohórquez et al., 2015). Experimental evidence has reported that suggests EECs could act as both pre-synaptic (to communicate gut feelings to the nervous system) and postsynaptic (e.g., for the possible modulation of the responsiveness of EECs by efferent neurons) elements.

### Immune-Endocrine Connections

Immune and enteroendocrine pathways are not distinct routes of the gut–brain axis, but can influence each other to some extent. Indeed, the release of interleukin-4 and 13 from CD4+ T cells, in a mouse gut inflammation model has been reported to increase cholecystokinin secretion in EECs (McDermott et al., 2006). Enterochromaffin cells (ECs) have been described as being able to modulate gut inflammation through 5-HT signaling (Mawe et al., 2009).

## Bacteria-Immune-Endocrine-Vagal Connections

Studies in germ-free animals have underlined the importance of gut microbiota in the early-life development of the gut– brain axis, with particular reference to the hypothalamicpituitary-adrenal axis (Sudo et al., 2004). Pattern recognition receptors such as TLRs, which are transmembrane proteins that are able to recognize bacterial envelope components, such as lipopolysaccharide (LPS) and lipoteichoic acids (Medzhitov, 2007), have been detected in a number of neuronal cells which could potentially mediate the direct neuronal sensing of microbial antigens (Lim et al., 2016). TLR4, which can be activated by bacterial LPS, has been detected in the nodose ganglion of the vagus nerve of rats (Hosoi et al., 2005). TLRs 3, 7, and 4 are expressed in the myenteric and submucosal plexus of the murine intestine, the human ileum and the lower dorsal root ganglia (Barajon et al., 2009). Similar results were also obtained by Qi et al. (2011), who detected the expression of TLRs 3, 7, and 9 in both human and murine dorsal root ganglion neurons. High mRNA levels of TLRs 1, 4, 5, and 6 have been found in the colonic dorsal root ganglion neurons of mice (Ochoa-Cortes et al., 2010). Furthermore, a number of studies have reported the direct excitation of neurons by LPS (Lim et al., 2016). However, in yet other studies the stimulation of pattern recognition receptors of neurons did not cause any direct excitation, but instead sensitized them and potentiated their excitability (Lim et al., 2016). Nonetheless, these researches have highlighted the existence of other possible pathways for bacteria sensing and communication pathways between the nervous and the immune systems.

Microbial exocellular polysaccharides (EPSs) are essential for protecting bacteria from the host immune response, but also for interacting with intestinal mucosal cells, including epithelial cells and EECs, thus inducing the release of molecular messengers that are able to modulate neural signaling or directly act on primary afferent axons (Forsythe and Kunze, 2013).

Mucosal immune cells are able to distinguish commensal from pathogenic bacteria (Artis, 2008) through TLRs, which are able to detect microbial antigens. Similarly, dendritic immune cells (extending their dendrites between epithelial cell tight junctions so as to directly sense the luminal environment) and some EECs can respond to the presence and activity of intraluminal microbial organisms through TLRs (Mayer, 2011). The intestinal infusion of Escherichia coli proteins has been reported to increase the secretion of anorexigenic GLP-1 and PYY, thus providing direct evidence of the ability of the gut microbiota to control appetite (Tennoune et al., 2014; Breton et al., 2016). In addition, EECs possess G-protein-coupled receptors that are able to sense microbial metabolites, such as amino acids (for example glutamate) (Bezençon et al., 2007), protein hydrolysates (Choi et al., 2007) and both long and short fatty acids (Tanaka et al., 2008; Nøhr et al., 2013). The expression of GPR41 receptor of SCFAs has been detected in several types of EECs (e.g., those secreting CCK, ghrelin, gastrin, GLP-1, PYY, neurotensin, secretin) throughout the GI tract of mice, from the stomach to the colon (Nøhr et al., 2013). Both the GPR41 and 43 receptors of SCFAs are expressed by enteroendocrine L-cells (Nøhr et al., 2013). Their stimulation promotes the secretion of GLP-1 and PYY (Samuel et al., 2008; Tolhurst et al., 2012; Nøhr et al., 2013). Furthermore, gut microbiota indirectly regulates the release of GLP-1, GLP-2 and PYY by modulating the differentiation of stem cells into EECs, and thus modifying the number of GLPs and PYY secreting L-cells (Everard et al., 2011). Another group of gut microbiota metabolites, namely indole and tryptamine derived from tryptophan, are known to modulate gut hormone release. Indole triggers GLP-1 secretion by enteroendocrine L-cells (Chimerel et al., 2014). Tryptamine induces ECs to secrete 5-HT (Takaki et al., 1985). Gut microbiota also contributes to regulating sensitivity to leptin, the so-called "satiety hormone" (Schéle et al., 2013).

Several microbial metabolites that display neuroactive properties have been described: (i) gaseous molecules, such as carbon monoxide, hydrogen sulfide and nitric oxide (Bienenstock and Collins, 2010); (ii) SCFAs such as n-butyrate, propionate, and acetate (Engelstoft et al., 2008; Nicholson et al., 2012); (iii) amines, such as putrescine, spermidine, spermine and cadaverine, which have been shown to be involved in CNS responses to stress (Bienenstock and Collins, 2010).

Curiously, bacteria can produce a wide range of molecules which mimic human hormones. For instance, muramyl dipeptide (similar to serotonin) (Masek and Kadlec, 1983) and indole (similar to melatonin) (Norris et al., 2013) can both cause sleep and drowsiness, while LPS from Gram-negative bacteria can directly act on thyroid cells, via type 4 TLRs, and up-regulate thyroglobulin gene expression (Nicola et al., 2009).

On the other hand, the direct response of bacteria to some of the regulatory peptides/neuroactive molecules that are secreted by EECs and/or the human nervous system has been demonstrated, which indicates the presence of

neuromodulator/neurotrasmitter receptors in the bacterial envelope (Lyte, 2011). For example, the stress hormones epinephrine and norepinephrine increase the in vitro growth of E. coli by more than four orders of magnitude (Freestone et al., 2002) and the Clostridium/Bacteroides ratio in the human gut (Bailey et al., 2011). A recent report has demonstrated that Vibrio cholerae can respond to epinephrine and norepinephrine (enhancing the growth rate, swimming motility, and production of virulence factors such as iron sequestrating phenotype) by means of specific sensor proteins (Halang et al., 2015).

Actually, the most forefront hypotheses consider that brain and gut commensal bacteria communicate with each other through shared chemical mediators, and this is part of the homeostasis mechanisms that help to maintain gut microbiota stability and possibly brain functions and behavior (Bienenstock and Collins, 2010; Lyte, 2011). This neurochemical-mediated "two-way street" (**Figure 1**) is one of the principles that supports the microbial endocrinology construct (Lyte, 2011).

### BACTERIAL PRODUCTION OF NEUROACTIVE MOLECULES

The molecules that have assessed neuroactive properties produced by bacteria are reported in **Table 1**. This list is far from being exhaustive, and will likely include a progressively higher number of strains and molecules as the results of a greater number of studies become available. The bacteria that are able to produce catecholamines include Bacillus spp. (producing dopamine), Escherichia spp., and Bacillus spp. (producing noradrenalin) (Lyte, 2011). The presence of the complete biosynthetic pathway for catecholamines in these organisms has raised the hypothesis that cell-to-cell signaling in vertebrates may be the result of a late horizontal gene transfer from bacteria (Lyte, 2011).

Strains belonging to Streptococcus spp., Escherichia spp. and lactic acid bacteria (LAB) are known to synthesize 5-HT (Lyte, 2011; Matur and Eraslan, 2012). It should be underlined that, unlike what is generally believed, the gut (and not the brain) is the main producer of 5-HT in mammals. The ECs of the gut mucosa are the predominant site for the synthesis and storage of 5-HT in humans (Gershon and Tack, 2007). The release of 5-HT occurs in response to chemical stimuli, e.g., the presence of nutrients (e.g., glutamate, glucose), or food-related/bacterial toxins in the intestinal lumen (Mayer, 2011; Kitamura et al., 2012). However, 5-HT produced by gut microbiota could possibly contribute to the overall 5-HT pool in the gut. It is known that 5-HT is involved in the regulation of gut peristalsis, in vagal circuits associated with nausea and vomiting and in the perception of visceral discomfort and pain through spinal afferents (Costedio et al., 2007). For these reasons, 5-HT has been the target of several treatments of gastrointestinal and gut–brain associated disorders such as inflammatory bowel disease, irritable bowel syndrome, postinfectious irritable bowel syndrome, and idiopathic constipation (Costedio et al., 2007).

Several LAB strains are able to synthesize β-phenylethylamine and/or tyramine and/or tryptamine that belong to a group of structurally related neuroactive compounds which are commonly known as trace amines (Mazzoli, 2014). Trace amines (which also include octopamine) are physiologically present in the human nervous system, although in low amounts compared to 5-HT and catecholamines, with which they display close metabolic and neurophysiological relationships (Burchett and Hicks, 2006). Tryptamine production has been detected in Lactobacillus bulgaricus (Pessione et al., 2009), while Leuconostoc and Enterococcus species have been reported to produce both tyramine and β-phenylethylamine (Pessione et al., 2009). It is worth noting that tryptophan decarboxylases have been found in at least 10% of the samples tested from the NIH Human Microbiome Project (Williams et al., 2014). Despite their low amounts, trace amines, such as β-phenylethylamine, could play a significant role as the co-transmitters and neuromodulators (Burchett and Hicks, 2006) that are involved in mood control, appetite/satiety circuits, and attention deficit/hyperactive disorders (Shimazu and Miklya, 2004). Alterations in the trace amine network also seem to be involved in psychiatric syndromes such as bipolar disorder (BD), parkinsonism, and hepatic encephalopathy (Burchett and Hicks, 2006).

Lactic acid bacteria and several other bacteria, such as those that can contaminate fish or shellfish products, produce large amounts of histamine, which is widely known as a mediator of allergy and anaphylaxis but also plays a role as a neurotransmitter in both the CNS and the ENS (Lyte, 2011).

Molecules that control QS communication between bacteria (mainly peptides or acyl-homoserine lactones) have also been shown to be involved in neuronal functioning (Hughes and Sperandio, 2008).

### Bacterial Production of Glutamate and γ-Aminobutyric Acid (GABA)

L-Glutamate (Glu) and γ-aminobutyric acid (GABA) are mainly known for their role as the main neuro-transmitters in the mammalian CNS, with excitatory and inhibitory roles, respectively (Hyland and Cryan, 2010; Julio-Pieper et al., 2013). However, these molecules are widespread in nature and have multiple functions, including plant signaling and communication between bacteria (Dagorn et al., 2013).

Several bacterial strains are able to produce Glu. Coryneform bacteria, such as Corynebacterium glutamicum, Brevibacterium lactofermentum and Brevibacterium flavum, have been used extensively for the industrial fermentative production of Glu (Sano, 2009). LAB strains belonging to Lactobacillus plantarum, Lactobacillus paracasei, and Lactococcus lactis are also able to synthesize Glu (Tanous et al., 2005). A recent study has revealed that about 15% of the LAB strains isolated from Asian fermented foods are Glu producers (Zareian et al., 2012).

Both prokaryotes and eukarya synthesize GABA through the decarboxylation of Glu by glutamate decarboxylase (GAD). GAD has been found in both Gram-positive and Gram-negative bacteria, where it is associated with systems that are involved in pH homeostasis and the generation of metabolic energy (i.e., proton motive force) (Pessione, 2012; Tsai et al., 2013). Marine microorganisms (Morse et al., 1980), E. coli (Richard and Foster, 2003) and Pseudomonas (Chou et al., 2008) can synthesize GABA.

Among those microorganisms that are generally recognized as safe or health-promoting, several LAB (e.g., strains belonging to Lactobacillus, Lactococcus, and Streptococcus genera) and Bifidobacterium strains have been reported to biosynthesize GABA (Siragusa et al., 2007; Li and Cao, 2010; Mazzoli et al., 2010; Lyte, 2011). Recently, one Lactobacillus strain and four strains of Bifidobacterium isolated from the human intestine have been reported to being able to produce GABA (Barrett et al., 2012). Furthermore, a very recent analysis on metagenomic data from the Human Microbiome Project suggests that genes encoding GAD could be present in a significant proportion of human gut microbiota (Pokusaeva et al., 2016). Lactobacilli

#### TABLE 1 | Main neuroactive amines and amino acids released by bacteria.


include the strains with the highest GABA production, although this metabolic ability is more likely strain- rather than genusrelated (Li and Cao, 2010).

### Glu/GABA: RECEPTORS, SIGNALING, AND TRANSPORT

### Bacterial Targets: Glu and GABA Receptors in Prokaryotes

The role of GABA in the communication between bacteria is in line with the identification of GABA-binding proteins in different prokaryotes. These proteins consist not only of transporters, such as GabP in E. coli or Bacillus subtilis) (Brechtel and King, 1998; Hu and King, 1998) and Bra in Agrobacterium tumefaciens (Chevrot et al., 2006), but also of possible specific receptors (Guthrie and Nicholson-Guthrie, 1989). It has been reported that metabotropic Glu and GABA<sup>B</sup> receptors, as well as bacterial periplasmic amino acid binding proteins, may have evolved from a common ancestor (Cao et al., 2009). In addition, as reported in the previous section, many bacteria (including those colonizing the human gut) can synthesize GABA, thus suggesting that, as in the case of eukaryotes, GABA might be a conserved and ubiquitous communication molecule.

The fact that GABA has an effect on the physiology of Pseudomonas can be supported by several observations. A periplasmic protein showing high affinity for GABA and displaying similar biochemical features to a subunit of the mammalian GABA<sup>A</sup> ionotropic receptor was identified in an environmental strain of P. fluorescens (Guthrie and Nicholson-Guthrie, 1989; Guthrie et al., 2000). Specific receptors for sensing GABA have been found in other Pseudomonads including: the PctC of P. aeruginosa and the McpG of P. putida, which are involved in GABA chemoreception (Reyes-Darias et al., 2015). Another study has demonstrated that GABA increases the cytotoxicity and virulence of P. aeruginosa through a sophisticated modulation of the protein expression (Dagorn et al., 2013). A functional proteomic analysis revealed that six proteins were differentially expressed in P. aeruginosa when exposed to GABA (Lyte, 2011). Hence, some bacteria are able to respond to GABA, of both prokaryote and mammalian origin.

A structurally similar ion channel to eukaryotic ionotropic GABA<sup>A</sup> receptors was identified in the plant pathogen Erwinia chrysanthemi (Zimmermann and Dutzler, 2011). The prokaryotic protein is activated by different amines, including GABA, and is modulated by benzodiazepines in a similar manner to its eukaryotic homolog (Zimmermann and Dutzler, 2011; Spurny et al., 2012).

Less information is available on prokaryotic Glu receptors. As far as we know, only a single case of a Glu-activated potassium channel in a bacterium (i.e., GluR0 from Synechocystis PCC 6803) has been reported to date (Chen et al., 1999). However, 100 prokaryotic channel proteins with putative Glu binding domains have recently been identified through a bioinformatic study (Ger et al., 2010). Among them, 22 proteins have been found to be homologs of vertebrate ionotropic Glu receptors (Ger et al., 2010).

Based on these findings, it is possible to hypothesize that members of the human gut microbiota can also sense and respond to GABA/Glu, although this has yet to be confirmed through dedicated studies. Furthermore, human-derived or dietary Glu can affect the balance of gut microbial populations by stimulating the growth of certain bacterial species at the expense of others. It is known, for instance, that LAB decarboxylate Glu to GABA in order to create a protonic gradient and enhance metabolic energy. Hence, Glu can constitute a privileged energy source that can be used to expand LAB populations.

### Host-Targets: the Glutamatergic/GABAergic System in the CNS

Glu and GABA are the major excitatory and inhibitory neurotransmitters found in the human CNS, respectively (Julio-Pieper et al., 2013). These compounds are the molecular effectors of a homeostatic neuronal circuit that is characterized by extreme complexity and sophistication at both a physiological and a

biochemical level. This system also includes glutamine (Gln), which plays a key role as a non-neuroactive intermediate in the recycling of neurotrasmitters in the brain, mainly Glu and GABA (Soeiro-de-Souza et al., 2015).

At a physiological level, this network consists of feed forward (FF) and feedback (FB) neuronal connections in which the main role of inhibitory (i.e., GABAergic) neurons is currently considered as that of balancing the activity of excitatory (i.e., Glutamatergic) neurons (Roth and Draguhn, 2012; **Figure 2A**). Both Glutamatergic and GABAergic neurotransmissions are complex systems in which glial cells (mainly astrocytes), pre- and post-synaptic neurons and a large spectrum of different receptors, transporters and enzymes are involved (Cherlyn et al., 2010; Femenía et al., 2012; **Figure 2B**).

The biochemical relationship between Glu and GABA goes far beyond this reciprocally modulating activity, since Glu and GABA are easily interconverted. Actually, the biochemical reactions involved in GABA-to-Glu-to-Gln conversion are integral features of such a regulatory network, which also involves the central energy metabolism of neurons (Soeiro-de-Souza et al., 2015; **Figure 2B**).

Glu is synthesized in situ in the CNS since it cannot pass the blood–brain barrier (BBB) (Hawkins, 2009). One of the main Glu biosynthetic pathways consists of the conversion of glucose (through the Embden-Meyerhof-Parnas pathway and the Krebs cycle) to α-ketoglutarate (α-KG), which is subsequently transaminated (Cherlyn et al., 2010). Glu is accumulated in vesicles and, after suitable stimulation, released to the synaptic space, where it can bind to a panel of Glu receptors (GluRs) that may be found in post-synaptic neurons. GluRs differ on the basis of their mechanism of action (e.g., ionotropic, iGluR, or metabotropic, mGluR, Glu receptors) and pharmacological responsiveness [e.g., N-methyl D-aspartate, NMDA, or α-amino-3-hydroxy-5-methylisoxazole-4-proprinoic acid (AMPA) or kainate sensitive receptors]. Ligand-gated ionotropic receptors mediate fast synaptic responses by directly regulating the ion influx. Metabotropic receptors are coupled to G-proteins and modulate slower signal transduction cascades

(Hyland and Cryan, 2010; Julio-Pieper et al., 2013). Any excess Glu present in the synaptic cleft is taken up by glial cells or neurons to prevent the prolonged excitation of post-synaptic neurons, which can lead to cell death (Cherlyn et al., 2010). Glu in the glia is either oxidatively deaminated to α-KG [by Glu dehydrogenase (GluDH)], or converted to Gln by Gln synthetase. Gln is released from glial cells, uptaken by pre-synaptic neurons [via Gln transporters (GlnT)], and is then converted back to Glu by glutaminase. Apart from its role in neurotransmission, Glu is also involved in other neural processes, such as neuronal development and synaptic plasticity (Cherlyn et al., 2010).

Inhibitory pre-synaptic neurons convert Glu to GABA through GAD and pack it into vesicles via vesicular GABA transporters (vGATs). Again in this case, Gln is an alternative Glu source to α-KG. The multiple GABA receptors that are present in post-synaptic neurons include ionotropic (GABAA, GABAC) and metabotropic (GABAB) receptors. Apart from GABA, these receptors are possibly modulated by a number of other compounds, e.g., barbiturates, baclofen, benzodiazepines and steroids (Simeone et al., 2003; Goetz et al., 2007). Any excess unbound GABA is cleared by glial cells through GABA transporters (GAT-1). GABA is then converted by GABA transaminase (GABA-T) into succinic semialdehyde, with concomitant Glu biosynthesis (Sivilotti and Nistri, 1991; Michels and Moss, 2007).

### Glu/GABA: From the Gut to the Brain

As described above, several bacteria, including probiotic bacteria and bacteria that colonize the human GI tract, are able to produce Glu or GABA (Barrett et al., 2012). As far as we know, only one study concerning the modulation of host Glutamate/GABAergic systems by GABA/Glu producing gut-colonizing bacteria is available in the literature (Bravo et al., 2011). A recent study has demonstrated that a GABA-producing Bifidobacterium dentium is able to attenuate sensitivity of dorsal root ganglia neurons in a rat model of visceral pain (Pokusaeva et al., 2016). It is worth noting that these few examples refer to the limited number of studies on interaction between gut microbiota and the host nervous system which is currently available. Experimental evidence is likely to grow as a higher number of reports become available. To date, most studies on the Glu/GABA effects on humans concern dietary Glu/GABA.

Glu and GABA can be found in food as natural components or as food supplements. Glu is among the most abundant amino acids (8–10%) found in dietary proteins. Furthermore, monosodium glutamate is employed extensively as a flavor additive in food (Brosnan et al., 2014). In Germany, Glu average intake (comprinsing both free Glu and Glu as protein constituent in foods) of about 10 g/day has been estimated (Beyreuther et al., 2007). In asian consumers this value is likely higher since free Glu intake can reach up to 4 g/day in these countries (Beyreuther et al., 2007). GABA is relatively abundant in plant food, such as brown rice germs and sprouts, spinach, barley and bean sprouts, where GABA concentrations of between 300 and 720 nmol/g dry weight have been detected (Oh et al., 2003), while higher levels are found in fermented foods (Abdou et al., 2006). In recent years, the utilization of GABA as a food supplement has progressively increased in both Asian and Western (USA and Europe) countries (Boonstra et al., 2015). GABA concentration of few micrograms per gram of cecal content has been detected in mice (Pokusaeva et al., 2016).

Hundreds of people have reported the benefits of using GABAsupplemented food, for instance in alleviating anxiety and/or improving sleep quality (Boonstra et al., 2015). Moreover, at least five different studies have reported that the oral administration of GABA or GABA-supplemented food/beverages (corresponding GABA amounts of about 50–100 mg) has had positive effects on human health. These effects include: (i) the reduction of psychological stress in people who performed arithmetic tasks (Nakamura et al., 2009; Kanehira et al., 2011; Yoto et al., 2012); (ii) the reduction of stress in acrophobic subjects exposed to heights (Abdou et al., 2006); (iii) an increased ability to perform prioritized planned actions (Steenbergen et al., 2015).

Multiple pathways by which gut luminal Glu/GABA may affect the CNS can be hypothesized, including their transport across intestinal and BBB and their sensing through afferent terminals that innervate the GI tract (**Figure 3**). In the next sections, evidence possibly supporting these mechanisms will be illustrated.

### Direct Influence of the Luminal Glu/GABA in the Gut on the CNS

Multiple transporters that mediate Glu absorption have been found in the apical membrane of GI epithelial cells, mainly in the small intestine, but also in the stomach (Burrin and Stoll, 2009), while little or no transport of amino acids from the lumen to portal blood occurs in the colon (Julio-Pieper et al., 2013). Studies on pigs and rodents have indicated the excitatory amino acid carrier 1 is the most abundant Glu transporter in the mucosa of the small intestine (Burrin and Stoll, 2009). A study performed in infant pigs has indicated that rate of Glu absorption and transport to the portal circulation linearly increases with increasing intraduodenal Glu intake (Janeczko et al., 2007). Portal absorption rate increased from 3.80 µmol kg−<sup>1</sup> h −1 for normal Glu intake (i.e., 510 µmol kg−<sup>1</sup> h −1 ) to 20.46 3.80 µmol kg−<sup>1</sup> h −1 for piglets infused with 3.5-fold increased Glu intake (Janeczko et al., 2007). Glu is one of the main nutrients for enterocytes. Several studies on different animal models, including pre-term infants and human adults, have agreed on that most Glu present in the GI lumen is oxidized to CO<sup>2</sup> or, secondarily, converted to other amino acids by the gut mucosa (Burrin and Stoll, 2009). Only a small percentage (between 5 and 17%, depending on the studies) of the ingested Glu is transported into the portal circulation but this does not generally affect the Glu concentration in the plasma to any great extent (Hays et al., 2007; Brosnan et al., 2014). Furthermore, an increased plasma Glu concentration (following Glu intake) does not necessarily affect the Glu concentration in brain tissues, since it is generally recognized that Glu cannot pass the BBB (Janeczko et al., 2007). A 20-fold (or more) increase in plasma Glu concentration was found necessary to access the brain tissues in rodents (Brosnan et al., 2014). The reaching of such a high Glu concentration in plasma after dietary intake (or Glu biosynthesis by gut microbiota) seems rather unlikely.

As far as GABA transporters are concerned, they are mostly found in the CNS (Gadea and López-Colomé, 2001). However, H+/GABA symport across the apical membrane of human intestinal epithelial (Caco-2) cells has been demonstrated (Thwaites et al., 2000). In fact, the hPAT1 H+/amino acid symporter, which is thought to mediate the uptake of luminal GABA, is present throughout the human GI tract, from the stomach to the discending colon, with maximal expression in the small intestine (Chen et al., 2003). Furthermore, GABA carriers have been detected in the basolateral membrane of

Caco-2 cells (Nielsen et al., 2012). These findings suggest that luminal GABA should be able to cross the intestinal barrier and possibly reach extra-intestinal targets. In fact, orally administered GABA in rats was reported to increased GABA concentration in their blood, with a peak after 30 min (Abdou et al., 2006). Nonetheless, GABA ability to cross the BBB is still controversial (Boonstra et al., 2015). Most early studies, performed between 1950 and 1980, agreed on the impermeability of BBB to GABA, but more recent investigations have reported that GABA may actually cross this barrier, although in small amounts (Boonstra et al., 2015). These discrepancies have been ascribed to the different method of administration (e.g., oral versus injection), or animal model employed. Some studies have also suggested that BBB permeability to GABA could diminish with increasing age (Boonstra et al., 2015). Interestingly, GABA-transporters have been found in the BBB of mice (Kakee et al., 2001). Direct evidence of GABA facilitated transport across the BBB of mice has been reported, although the efflux rate from the brain was 17 times higher than the influx rate (Kakee et al., 2001). Unfortunately, no information about the possible presence of similar transporters in the BBB of humans is currently available.

All together, these findings suggest that a direct action of luminal gut Glu and/or GABA on the CNS cannot be excluded, although further in vivo studies are required to confirm this hypothesis. These investigations should also take into account that gut and/or BBB permeability are affected by several factors, including stress, diet and gut microbiota (Braniste et al., 2014; Kelly et al., 2015).

### Sensing of Luminal Glu/GABA in the Gut

Multiple Glu receptor types (including ionotropic, types 1 and 4 metabotropic receptors and heterodimeric, TAS1R1 + TAS1R3, L-Glu taste receptors) have been detected in GI epithelial cells and/or enteric neurons in the stomach, small intestine and colon (Kondoh et al., 2009; San Gabriel et al., 2009; Kitamura et al., 2012). More in detail, mGlu<sup>4</sup> receptors have been detected in the mucosa of both the gastric antrum and duodenum (Akiba et al., 2009), while both mGlu<sup>4</sup> and mGlu<sup>7</sup> receptors have been identified in the colon epithelium (Chang et al., 2005; Julio-Pieper et al., 2010). TAS1R3, a subunit of umami taste receptors, has been found in enteroendocrine L-cells and X/A-like cells (Jang et al., 2007; Hass et al., 2010). This evidence suggests a possible role in luminal Glu sensing. In fact, small intestine luminal L-Glu seems to play a role in the defense mechanisms of duodenal mucosa in rats, by increasing the intracellular pH and mucus gel thickness (Akiba et al., 2009). A role of mGluRs in the human colon in the control of colon peristalsis and electrolyte transport has been described (Julio-Pieper et al., 2013). High levels of mGlu<sup>7</sup> and mGlu<sup>8</sup> have been detected in myenteric neurons, where they are possibly involved in the regulation of gut motility (Julio-Pieper et al., 2013). However, results on the ability of gut luminal Glu to directly excite enteric neurons still appear contradictory (Kirchgessner, 2001; Wang et al., 2014). On the other hand, an indirect activation of vagal afferents (i.e., via the production and release of nitric oxide and 5-HT) through the administration of Glu to the gastric lumen has been

demonstrated (Uneyama et al., 2006). These data strongly suggest a role of ECs in the mediation of luminal Glu sensing. This luminal Glu sensing mechanism stimulates different areas in the CNS, including the cerebral cortex, basal ganglia, limbic system, and hypothalamus (for a comprehensive review, refer to Kondoh et al., 2009), and induces flavor preference in rats (Kitamura et al., 2012).

GABA<sup>B</sup> receptors are abundantly expressed in the GI tract (Hyland and Cryan, 2010). GABA and its ionotropic and metabotropic receptors are widely distributed throughout the ENS, in both submucosal and myenteric neurons, from the stomach to the ileum (Auteri et al., 2015). In addition, Nakajima et al. (1996) reported the expression of GABA<sup>B</sup> receptors in cells morphologically similar to EECs, where they co-localized with cells containing somatostatin (in the stomach) or serotonin (in the duodenum). However, a later study failed to detect any positive EEC for GABA<sup>B</sup> receptors (Casanova et al., 2009). Schwörer et al. (1989) showed that the release of 5-HT by ECs in the small intestine of guinea-pig is modulated by GABA<sup>A</sup> and GABA<sup>B</sup> receptors. Involvement of GABA<sup>B</sup> receptors in modulation of sensitivity of vagal and spinal afferents has been reported (Hyland and Cryan, 2010). Actually, the GABA<sup>B</sup> metabotropic receptors in the GI tract are thought to regulate several functions, including gut motility and gut-to-brain signaling (Hyland and Cryan, 2010). Experimental evidence of GABAB/µ-type opiod receptor interaction, which mediates the synergistic potentiation of the anti-nociceptive action of GABA/morphine co-administration, has been reported (Hara et al., 1999). A similar GABA<sup>B</sup> receptor-mediated potentiation effect has also been described for drugs used to reduce visceral pain (Hara et al., 2004).

Since GABA receptors have been found in a wide range of immune cells, such as dendritic cells, mast cells and T-cells (Auteri et al., 2015), and are involved in regulating immunological processes, such as the down-regulation of proinflammatory cytokine release (Bjurstöm et al., 2008), their role in neuro-immune dialog in the gut has been proposed (Auteri et al., 2015).

It is worth reminding that GABA is produced through Glu decarboxylation in both eukarya and prokaryotes. Hence, GABAproducing microorganisms that are present in the gut may significantly affect luminal Glu/GABA ratio and, therefore, gut signaling.

## PROBIOTICS FOR THE TREATMENT OF PSYCHIATRIC DISORDERS?

As far as its application to the field of probiotics is concerned, microbial neuro-endocrinology addresses the ability of probiotics to both synthesize and respond to neuroactive compounds, and thus to affect the host neurological processes and sometimes the overall physiology of the host. A microbial endocrinology-based approach may be relevant to help understand how probiotics can influence the health of a host and how host can select commensal microbial populations.

A number of studies have provided evidence on the ability of probiotic strains to modulate the mood and stress responses of humans, and to reduce anxiety and depression (Logan and Katzman, 2005; Messaoudi et al., 2011; Arseneault-Breard et al., 2012). As an example, the administration of Bifidobacteria and LAB as probiotic supplements to humans can increase the levels of morning salivary melatonin, thus reducing the symptoms of irritable bowel syndrome (Wong et al., 2015). Actually, most of the currently recognized probiotic strains belong to Bifidobacteria and LAB, with lactobacilli being among the most abundant members of such health-promoting strains (Vaughan et al., 2011). A direct modulation of myenteric neuron activity by unidentified Bifidobacterium longum fermentation product(s) displaying anxiolytic effects has also been reported (Bercik et al., 2011a).

Alterations in the Glu/ Gln/ GABA circuits in the CNS have been reported for generalized anxiety disorders, as well as in other psychiatric conditions, such as the major depressive disorder (MDD), BD (also known as manic depressive disorder) and schizophrenia (SZ) (Cherlyn et al., 2010; Femenía et al., 2012; Soeiro-de-Souza et al., 2015). It is worth noting that these complex neurobehavioral disorders affect a significant portion of the overall population throughout the world (e.g., individuals affected by BD or SZ account for about 2–3% of the global population) (Cherlyn et al., 2010). Different classes of drugs addressed to different steps of Glutamate/GABAergic neurotransmission (e.g., reducing Glu release, increasing Glu re-uptake, or increasing GABA levels by enhancing GAD or inhibiting GABA transaminase and succinic semialdehyde dehydrogenase) or serendipitously discovered (e.g., valproic acid) have been used in the treatment of such mood disorders (Terbach and Williams, 2009; Femenía et al., 2012). However, drugs frequently have negative secondary effects. For instance, most antidepressants (such as tricyclic antidepressants), while having a positive effect on the mood of humans, do not alleviate and can even aggravate cognitive symptoms, or impair vigilance (e.g., selective 5-HT reuptake inhibitors), and may cause dependence (e.g., benzodiazepines) (Femenía et al., 2012). Finding alternative or adjuvant therapies is therefore a research priority and, in this respect, the use of neuromodulating bacteria or so-called psychobiotics is a fascinating perspective.

Interestingly, the anxiolytic and anti-depressant-like effects of Lactobacillus rhamnosus JB-1 ingestion in mice involve a reduction in the plasma corticosterone levels and alterations of both GABA<sup>A</sup> and GABA<sup>B</sup> receptor expression patterns in specific brain areas (Bravo et al., 2011). Furthermore, increases in the Glx (i.e., Glu + Gln) and GABA levels in the CNS have been observed after 2 and 4 weeks of L. rhamnosus JB-1 treatment, respectively (Janik et al., 2016). These effects were mediated by gut-to-brain communication through the vagus nerve (Bravo et al., 2011). However, whether L. rhamnosus stimulates the vagus nerve with its own GABA or induces an endogenous GABA production still has to be clarified.

Modulation of the glutamatergic system in the central amygdala, cortex and hippocampus of mice by gut microbiota has been described (Sudo et al., 2004; Neufeld et al., 2011).

Although, GABA production has been speculated as the key factor in the ability of a Lactobacillus helveticus–B. longum mixture to reduce anxiety-like behavior(s) (Lyte, 2011), the observed probiotic effects on the psychological health of a host are generally likely the result of multiple interactions with the host (ENS, EEC, GALT). On the other hand, anorexigenic and anxiolytic soymorphin peptides (produced by LAB proteolytic action toward soy betaconglycinine) act via the GABA<sup>B</sup> receptor and, in this overall action, multiple circuits (including 5-HT and dopamine receptors) are involved (Kaneko et al., 2010). A recent investigation have suggested that a GABA-producing B. dentium is able to reduce nociception in a rat model of visceral pain (Pokusaeva et al., 2016).

Because of the wide range of potential neuro-active compounds (orexant, anorexant, opioid, opioid antagonist) produced by bacteria and the complex network that exists between them, a deep characterization of the proteome and metabolome of probiotic strains is one key aspect that would help to understand the different effects exerted by each strain on human health. Since not only Glu/GABA receptors but also other circuits can be potential targets for bacterial-derived molecules, it is tempting to hypothesize the use of probiotics that produce bioactive compounds (e.g., GABA, 5-HT, opioid peptides) as a novel frontline strategy for the adjuvant treatment of gastrointestinal, neuroenteric, neurological and psychiatric conditions (Hyland and Cryan, 2010; Lyte, 2011; Reid, 2011).

Together with this optimistic viewpoint and the enthusiasm that these findings have generated in the scientific domain, it is worth stating that a number of issues suggests caution in application of psychobiotics. The comprehension of the mechanisms by which gut microbiota and the nervous system communicate is still at its beginning. Multiple levels of complexity hinder the detailed understanding of this network and these include (but are not limited to) the fact that: the gut–brain axis comprises multiple (i.e., neural, immune, and endocrine) pathways with several reciprocal interactions; multiple molecular mediators, such as GABA and Glu, are involved, which play roles in a number of different biological processes; gut microbiota and even single commensal microorganisms interact with the host through multiple pathways and chemical mediators (e.g., cell envelope components, production/metabolism of neurotransmitters or neuroactive molecules). Therefore, it may be difficult to specifically address health problems with such microbial agents without the risk of unbalancing other pathways. As for other potential therapeutic agents, rigorous evaluation of the potential beneficial effects together with possible negative side effects should be performed for each probiotic/psychobiotic agent.

### CONCLUSION

The increasing knowledge about gut bacteria–brain bidirectional communication has provided scientific evidence that can explain the popular statement that somebody has

made a decision based his/her on gut feeling, but also the saying "we are what we eat," as it is supposed that diet can affect the microbiota composition of the gut. It is now evident that gut microbiota is able to interfere, through its cellular components or by secreting bioactive compounds, with the gut– brain axis through immune, neural and humoral pathways. As the knowledge of this domain increases, new unexpected and exciting perspectives for both basic and applied science will likely emerge. At the same time, the observed effects of human-derived neuro-active compounds on gut microbiota would seem to suggest that some neurochemicals share a common language that enables interkingdom communication, as proposed by Lyte and Freestone (2010). The ability of a microbial strain to affect the overall patho-physiology of a host thus appears to be the result of a long and sophisticated synergic evolution between gut bacteria and the host.

### REFERENCES


### AUTHOR CONTRIBUTIONS

Authors equally contributed to writing the manuscript. EP mainly contributed to describe bioactive molecules produced by bacteria and involved in interkingdom signaling, with particular attention to microbiota-gut-brain axis. RM focused his contribution mainly on Glutamate- and GABA-mediated communication, including Glu and GABA biosynthesis and receptors in both bacteria and humans, gut-to-brain transport and sensing and Glu-ergic/GABA-ergic neural transmission in the CNS.

### ACKNOWLEDGMENT

EP benefited from a MIUR-University of Torino "Fondi ricerca locale (ex-60%)" grant.


central GABA receptor expression in a mouse via the vagus nerve. Proc. Natl. Acad. Sci. U.S.A. 108, 16050–16055. doi: 10.1073/pnas.1102999108


Furness, J. B. (2006). The Enteric Nervous System. Oxford: Blackwell.




Tn3 transposon carried by a large plasmid. J. Bacteriol. 187, 5019–5022. doi: 10.1128/JB.187.14.5019-5022.2005


of gut microbiota decarboxylases that can produce the neurotransmitter tryptamine. Cell Host Microbe 16, 495–503. doi: 10.1016/j.chom.2014.09.001


**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 Mazzoli and Pessione. 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.

# Melatonin and Other Tryptophan Metabolites Produced by Yeasts: Implications in Cardiovascular and Neurodegenerative Diseases

*Ruth Hornedo-Ortega1, Ana B. Cerezo1, Ana M. Troncoso1, M. Carmen Garcia-Parrilla1 and Albert Mas2\**

*<sup>1</sup> Facultad de Farmacia, Universidad de Sevilla, Sevilla, Spain, <sup>2</sup> Facultad de Enología, Universitat Rovira i Virgili, Tarragona, Spain*

Yeast metabolism produces compounds derived from tryptophan, which are found in fermented beverages, such as wine and beer. In particular, melatonin and serotonin, may be relevant due to their bioactivity in humans. Indeed, the former is a neurohormone related to circadian rhythms, which also has a putative protective effect against degenerative diseases. Moreover, serotonin is a neurotransmitter itself, in addition to being a precursor of melatonin synthesis. This paper summarizes data reported on fermented beverages, to evaluate dietary intake. Additionally, the article reviews observed effects of yeast amino acid metabolites on the prevention of neurodegenerative diseases (Alzheimer's and Parkinson's) and angiogenesis, focusing on evidence of the molecular mechanism involved and identification of molecular targets.

#### *Edited by:*

*Enrica Pessione, University of Turin, Italy*

## *Reviewed by:*

*Robin Anderson, United States Department of Agriculture/Agricultural Research Service, USA Paul Richard Himes, University of Louisville, USA*

> *\*Correspondence: Albert Mas albert.mas@urv.cat*

#### *Specialty section:*

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

*Received: 20 October 2015 Accepted: 27 December 2015 Published: 19 January 2016*

#### *Citation:*

*Hornedo-Ortega R, Cerezo AB, Troncoso AM, Garcia-Parrilla MC and Mas A (2016) Melatonin and Other Tryptophan Metabolites Produced by Yeasts: Implications in Cardiovascular and Neurodegenerative Diseases. Front. Microbiol. 6:1565. doi: 10.3389/fmicb.2015.01565*

Keywords: wine, beer, VEGF, **β**-amyloid, **α**-synuclein, tryptophol, serotonin

### ORIGIN, OCCURRENCE, AND DIETARY INTAKE

The presence of bioactive compounds in fermented beverages has long been observed and they have been studied with great interest. A large body of research has focused on polyphenols, in particular, since these bioactive compounds are already present in plants and released into fermented products. Yeast also transforms certain other molecules into biologically active compounds. Among these, the case of amino acid tryptophan is of interest, since it is the precursor of at least three biologically active compounds: melatonin, serotonin, and tryptophol (Mas et al., 2014).

Tryptophol is an alcohol produced by the Ehrlich pathway and it has long been detected in appreciable concentrations in wines and beers in the mg/L range (Bartolomé et al., 2000; Monagas et al., 2007). Therefore, its occurrence in beverages is widely recognized. Moreover, tryptophol has also been indicated as a quorum sensing molecule for yeast (Sprague and Winans, 2006).

Just a few years ago, melatonin was detected in wines in much lower levels: within the ng/L range. Not only was it evidenced in wines, but also in other fermented foods, as summarized in **Table 1**. Furthermore, Rodriguez-Naranjo et al. (2011) highlighted that melatonin was produced after alcoholic fermentation, pinpointing the role *Saccharomyces* plays. Indeed, different strains synthetized melatonin at different levels (Rodriguez-Naranjo et al., 2012).

The synthetic pathway of melatonin in yeast is not completely elucidated, yet it seems the formation of serotonin might be an intermediate in the pathway (Mas et al., 2014). In addition, serotonin has been detected at mg/L levels in red wine following malolactic fermentation (Wang et al., 2014). Further research is required to explore the roles of yeast and bacteria in the occurrence of these bioactive compounds in fermented products.

One of the characteristics of bioactive compounds is the minimal concentration required for them to act. The reported concentrations in wine and beer would mean that someone consuming these beverages would obtain a low daily intake of these compounds. According to WHO, the daily intake of ethanol should not exceed 30 g and 20 g for men and women, respectively. That is to say that daily intake for a man of a wine can provide up to 0.00005–0.13 mg of melatonin. In a comprehensive review summarizing the results of human intervention studies, Harpsøe et al. (2015) concluded that the bioavailability of melatonin was 15%. In our example, its bioavailable concentration should result in 1.5–4000 pg/mL of melatonin in blood. Physiological values for day plasma melatonin are very low, accounting for several pg/mL (5– 10 pg/mL in human plasma) (de Almeida et al., 2011). Thus, pg/mL in plasma might be expected after dietetic intake of wine or beer, considering the values displayed in **Table 1**. Indeed, Maldonado et al. (2009) determined an increase in the concentration of plasmatic melatonin after the ingestion of a moderate dose of beer (330 mL for women volunteers, 660 mL for men). To the best of our knowledge, there is no published data on the bioavailability of serotonin after food or beverage intake.

### BIOLOGICAL EFFECTS AND PREVENTION OF CHRONIC DISEASES

Literature on the biological effects of these compounds is extensive and encompasses circadian rhythm, antioxidant properties, and reproductive function. Due to the length of this mini-review, we will focus on more recent findings on the prevention of the most prevalent degenerative diseases, such as cancer, and cardiovascular and neurodegenerative diseases.

### IMPLICATIONS FOR CANCER AND CARDIOVASCULAR DISEASE: THE ROLE OF ANGIOGENESIS

Angiogenesis, which consists of the formation of new blood vessels from pre-existing ones, is crucial for organ growth during embryonic development and after birth. However, in adulthood, angiogenesis plays an essential role in the pathogenesis of diverse chronic diseases, such as cancer and cardiovascular disease, involving the progression and development of the tumor, and development and destabilization of atherosclerotic plaques (Celletti et al., 2001; Bergers and Benjamin, 2003).

Angiogenesis occurs when there is an imbalance between pro-angiogenic (e.g., vascular endothelial growth factor (VEGF), basic fibroblast growth factor, alfa tumor necrotic factor, etc.) and anti-angiogenic (e.g., angiostatin and endostatin) factors. VEGF is the most active endogenous pro-angiogenic factor in humans (Giles, 2001; Dulak, 2005; Cebe-Suarez et al., 2006; Cook and Figg, 2010). It exerts its angiogenic effect by stimulating VEGF receptor 2 (VEGFR-2), which is critical for promoting the proliferation and differentiation of endothelial cells (Giles, 2001; Ferrara and Kerbel, 2005). It has been demonstrated that VEGF promotes atherosclerotic plaque progression (Celletti et al., 2001; Khurana et al., 2005) and tumor angiogenesis (Senger et al., 1993). Indeed, VEGF is a target for drug therapies that aim to

TABLE 1 | Concentration of melatonin and other tryptophan metabolites in fermented products.


inhibit VEGF signaling (Ferrara and Kerbel, 2005). Anti-VEGF antibodies, aptamers and small molecule VEGFR tyrosine kinase inhibitors have been developed and given regulatory approval for the treatment of colon, lung, breast, kidney, and liver cancer, in addition to neovascular age-related macular degeneration (Giles, 2001; Ferrara and Kerbel, 2005). However, serious side effects, such as hypertension, have been reported with prolonged use of anti-VEGF therapies (Zhu et al., 2007; Wu et al., 2008; Kappers et al., 2010). The use of natural products in reducing VEGFinduced angiogenesis may prove to be more beneficial than the current anti-VEGF drugs available (Moyle et al., 2015).

Melatonin has been associated with a decline in VEGF secretion levels in the serum of advanced cancer patients (Lissoni et al., 2001), in addition to markedly reducing the expression of VEGF in HUVEC and culture cancer cells at 1 μM and 1 mM (Dai et al., 2008; Cui et al., 2012; Álvarez-García et al., 2013; Gonçalves et al., 2014). Melatonin has also been proven to reduce endothelial cell proliferation, invasion, migration, and tube formation, through downregulation of VEGF at 1 mM (Álvarez-García et al., 2013). The possible cell signaling pathway when melatonin inhibits HUVEC proliferation has been related to the following pathway: melatonin receptors/ERK/PI3K/Akt/PKC/NF-kB (Cui et al., 2008). Additionally, Sohn et al. (2015) have recently demonstrated that melatonin (1 mM) upregulates miRNA3195 and miRNA374b, whose overexpression synergistically reduced VEGF production in hypoxic PC-3 prostate cancer cells, indicating the important role of miRNA3195 and miRNA374b in melatonin induced antiangiogenic activity. Melatonin (40 mg/kg) has also shown an antitumor effect on mammary tumor growth in mice after 21 days of treatment; the mice displayed significantly smaller tumor volume and tumor regression (Jardim-Perassi et al., 2014). Additionally, in the same study, a lower expression of VEGFR2 was observed in the melatonin-treated tumors compared to the vehicle-treated tumors. More research is consequently needed to focus on determining the molecular mechanism by which melatonin exerts its angiogenic effect and the molecular target involved.

### NEURODEGENERATIVE DISEASES

Alzheimer's disease (AD) and Parkinson's disease (PD) are the most common human neurodegenerative diseases. In both cases, their incidence increases with age. The aggregation of proteins that results in different fibrillar structures is responsible for these disorders. Specifically, they are owing to the abnormal pathological assembly of amyloid- beta (Aβ), tau and α- synuclein (αS).

Indeed, several studies have demonstrated that protofibrils and oligomers of αS and Aβ are more neurotoxic than fibrils (Pike et al., 1993; Lashuel et al., 2002; Volles and Lansbury, 2003; Outeiro et al., 2008).

This review focus on the evidence of certain bioactives which can present in fermented products. However, to give a fair balance, alcohol effects on neurodegeneration have to be highlighted as it is a major component formed by yeast in alcoholic beverages. It is well-Known that alcohol intake crosses the BBB (Blood–Brain Barrier) easily producing the excessive release of neurotransmitters, oxidative stress and inflammatory response which turns out in neurotoxicity and finally cell death. (Persidsky and Potula, 2014).

### ALZHEIMER'S DISEASE

Alzheimer's disease is a progressive and irreversible neurodegenerative disorder characterized by loss of memory and cognition, abstract thinking, and personality alteration. The etiology of AD is unknown in more than 90% of cases. In the pathogenesis of AD there are two principal hallmarks: neurofibrillary tangles (NFTs) and amyloid plaques. NFT are formed by the intracellular accumulation of phosphorylated tau protein and amyloid plaques, by extracellular accumulation of amyloid β peptides (Hardy and Selkoe, 2002). The amyloid beta peptide is formed via cleavage of the amyloid precursor protein (APP). In the non-amyloidogenic pathway (normal state), APP is cleaved by α-secretase, to generate sAPP (soluble N- terminal fragment), which is neuroprotective as it is involved in the enhancement of synaptogenesis, neurite outgrowth, and neuronal survival. Conversely, in the disease state, APP is cleaved by β and γ secretase, resulting in insoluble beta amyloid peptide, which has high potential for assembly and formation of toxic aggregates (Gandy, 2005).

Several mechanisms have been proposed to explain βA neurotoxicity, such as oxidative stress and loss of endogenous antioxidants (Behl et al., 1994; Abramov and Duchen, 2005; Hamel et al., 2008); mitochondrial damage, depolarization, and mitochondrial permeability transition pore opening (Moreira et al., 2001, 2010; Abramov et al., 2004, 2007); destabilization of intracellular calcium homeostasis in neurons (Bezprozvanny and Mattson, 2008); glial cells (Abramov et al., 2003, 2004), and neuroinflammation (McNaull et al., 2010).

Levels of melatonin and its precursors (serotonin and tryptophan) are significantly decreased in elderly AD individuals and are associated with the emergence of AD (Zhou et al., 2003; Greilberger et al., 2010). A growing body of evidence supports the protective role of melatonin in several molecular mechanisms implicated in the development of AD.

Among these mechanisms, the most significant one is that melatonin prevents amyloid aggregation and overproduction. This neurohormone has a great affinity for Aβ peptide, preventing amyloid fibril formation (Masilamoni et al., 2008), as determined by circular dichroism (CD) spectroscopy, electron microscopy, nuclear magnetic resonance (NMR) and electrospray ionization-mass spectrometry (ESI-MS). In particular, a hydrophobic interaction has been observed between melatonin and Aβ, specifically on the 29–40 residues of the Aβ segment (Skribanek et al., 2001). Additionally, melatonin has inhibitory effects on the formation of secondary β-sheet structures through the disruption of the histidine (His+) and aspartate (Asp−) salt bridges in Aβ peptide that promote fibril dissolution (Fraser et al., 1991; Huang et al., 1997; Pappolla et al., 1998).

Melatonin presents a great capacity to regulate the synthesis and maturation of APP at different levels by: decreasing its mRNA encoding β-APP (Song and Lahiri, 1997; Lahiri, 1999); blocking cAMP production, which is involved in activating the APP gene promoter, (Husson et al., 2002); and inactivating GSK-3, which promotes α-secretase mediated cleavage of APP, favoring the non-amyloidogenic pathway (McArthur et al., 1997; Zhu et al., 2001; Hoppe et al., 2010). *In vivo* studies with transgenic mice over-expressing APP (in 9–10 months they develop senile plaques) and fed with 0.5 mg/mL of melatonin in their drinking water (3 mL/day) found a reduction in important markers of the disease, including Aβ levels in the brain, and that some animals survived (Matsubara et al., 2003). The amount given to rodents are within the pharmacological dose and out of the range of the dose that can be achieved with moderate consumption of wine. Therefore it cannot be concluded that these effect will be observed in humans after wine intake. Further research is required to obtain the evidence at dietary doses.

Furthermore, melatonin exhibits a protective effect on the cholinergic system. In AD patients, a dramatic decrease of acetylcholine has been observed (Francis et al., 1999), which was related to a decrease in enzyme choline acetyltransferase (ChAT) activity and an increase in acetylcholinesterase (AChE) activity (Bieschke et al., 2005). Indeed, AChE inhibitors increase the synaptic levels of acetylcholine, which is why they are used as a treatment for mild to moderate AD. *In vivo* administration of melatonin in rats (50 mg/kg body weight) has led to significantly reduced AChE activity, with maintenance of calcium levels under conditions of oxidative stress (Masilamoni et al., 2008). Regarding ChAT, melatonin increased its activity, after 4 months of melatonin administration in rats (Feng et al., 2004).

Finally, melatonin reduces Aβ-induced oxidative stress related to reactive oxygen species (ROS) and proinflamatory cytokines, such as IL6 and IL1-β in *in vivo* studies (Masilamoni et al., 2008). As a result of these effects, melatonin protects brain neurons from damage and death by increasing viability in hippocampal neurons and glial cells following treatment with Aβ1–40, Aβ25–40, and Aβ1–28. Moreover, melatonin prevents the death of murine N2a neuroblastoma and PC12 cells by using Aβ25–35 (Pappolla et al., 1997; Ionov et al., 2011).

There is scarce literature available in relation to the activity of other tryptophan metabolites, with indole 3-acetic acid and tryptophol being the only bioactive molecules reported so far. Morshedi et al. (2007) proved the inhibitory effect of these indole derivatives on the amyloid fibrillation of hen eggwhite lysozyme, which is another model for exploring the amyloidogenic mechanism.

### PARKINSON'S DISEASE

Parkinson's disease is the second most common neurodegenerative disorder. Its diagnosis is based on motor abnormalities, such as resting tremor, bradykinesia, and rigidity (Duvoisin, 1992). Indeed, patients present other non-motor symptoms, such as depression, anxiety, and sleep disorders (Jenner et al., 2013). Only 10% of patients have a genetic basis, with 90% being considered sporadic cases. PD is characterized by the degeneration of the subcortical structure of the brain. Specifically, there are significant losses of dopaminergic neurons in the substantia nigra pars compacta (SNpc; Forno, 1996), although other cell populations are also susceptible to the neurodegeneration process.

α-Synuclein (αS) is a 140 amino acid and a highly abundant neuronal protein. It is found as a soluble cytoplasmatic protein associated with synaptic vesicles (Iwai et al., 1995). It is thought that it plays a role in neurotransmission and cognitive function. Although its physiological function is uncertain, the pathology is associated with the accumulation of αS aggregates, which are the main component of Lewy bodies (LBs; Spillantini et al., 1997). LBs are spherical inclusions formed by αS aggregate (99%) and other proteins.

Despite the main risk factor being aging, other possible risk factors include mutation in the SNCA (alpha-synuclein gene) and exposure to environmental toxins. The latter are also linked to metabolic abnormalities involving neurotransmitter systems (dopamine, serotonin, GABA, and glutamate), fatty acids, such as arachidonic acid-cascade, oxidative stress and mitochondrial function (Henchcliffe and Beal, 2008; Quinones and Kaddurah-Daouk, 2009; Kaidanovich-Beilin et al., 2012; Lei and Powers, 2013).

Furthermore, several studies suggest that αS oligomers and protofibrils are an important factor in neurotoxicity in PD. αS protofibrils cause membrane permeabilization, which alters cellular homeostasis and may activate an apoptotic process (Volles and Lansbury, 2003). Indeed, there is evidence to support the capacity of αS to inhibit proteosomal activity, which would prevent elimination of misfolded proteins (Giasson and Lee, 2003).

Substantial evidence also suggests that a significant factor in dopaminergic neuronal loss in the PD brain are ROS, which result from dopamine metabolism, low glutathione concentration and high levels of iron and calcium in the SNpc (Jenner and Olanow, 2006). Additionally, the brain contains high concentrations of polyunsaturated fatty acids, which, under oxidative stress, result in lipid peroxidation and generation of toxic products (Liu et al., 2008).

No treatment is currently available for the prevention or cure of PD. However, a combination of L-DOPA and antioxidants has been recommended to reduce the rate of progression of the disease, due to the decrease in dopamine levels and significant increase of oxidative stress commonly concomitant to this type of disorder (Zhu et al., 2004).

Concerning the role of melatonin in PD, several works have reported different mechanisms of action. Lin et al. (2007) demonstrated that melatonin attenuates arsenite-induced apoptosis by reducing aggregated αS levels in rat brains, by means of Western blot analysis. Additionally, Ishido (2007) showed that melatonin inhibits αS assembly, using immunostaining in rat pheochromocytoma cells.

It is also important to highlight that melatonin dosedependently inhibits all steps of the αS assembly process. Ono et al. (2012) observed a reduction in the number of fibrils and the corresponding increase of the number of short fibrils and amorphous aggregates (25–250 μM) using electron microscopy and thioflavin S experiments. Indeed, melatonin presents a significant destabilization effect (also dosedependently), suggesting a decrease in beta-sheet levels. In the same study, the authors performed experiments with primary cultures of mesencephalon and neostriatum with MTT: (3- (4,5-Dimethylthiazol-2-yl)-2,5-Diphenyltetrazolium Bromide) a colorimetric assay for assessing cell metabolic activity. The results showed that melatonin reduced the toxic effects of αS after pretreatment (2–6 days) with an increase in cell viability of between 56 and 97%.

In addition to this, it is well known that melatonin exhibits antioxidant properties (Reiter et al., 1997; Kotler et al., 1998). Cellular injury cause by αS-mediated perturbation of cellular redox reactions is an important mechanism proposed for PD (George et al., 2009). Melatonin has been suggested as a potential therapeutic agent in diseases where oxidative stress is thought to be a major pathogenic factor. Mayo et al. (1998) observed that this hormone was an effective free radical scavenger and that it prevented apoptosis in neuronal cells. Moreover, *in vitro* studies on MPTD-induced (1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine) PD in mice have shown that melatonin protects against excitotoxicity by reducing the autoxidation of dopamine. The administration of melatonin leads

### REFERENCES


to normalization of complex I activity and oxidative status in mitochondria (Escames et al., 2010).

In conclusion, and based on the preceding evidence, we should consider that melatonin presents strong inhibitory effects on protofibril formation and peptide oligomerization.

### AUTHOR CONTRIBUTIONS

RH-O and AC Literature search and first draft. AT, MG-P, and AM thorough revision and discussion and final document.

### FUNDING

This work was supported by the Ministry of Economy and Competitiveness, Spain (grant no. AGL2013-47300-C1 and AGL2013-47300-C3).

### ACKNOWLEDGMENTS

We acknowledge the Universidad de Sevilla for the contracts of AC and RH-O.


angiogenesis in xenograft model of breast Cancer. *PLoS ONE* 9:e85311. doi: 10.1371/journal.pone.0085311


amyloid beta peptide injected mice. *Free Radic. Res.* 42, 661–673. doi: 10.1080/10715760802277388


**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 Hornedo-Ortega, Cerezo, Troncoso, Garcia-Parrilla and Mas. 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 broad-spectrum antibiotic, zeamine, kills the nematode worm Caenorhabditis elegans

### *Josephine E. E. U. Hellberg1, Miguel A. Matilla1,2 \* and George P. C. Salmond1\**

<sup>1</sup> Department of Biochemistry, University of Cambridge, Cambridge, UK

<sup>2</sup> Department of Environmental Protection, Estación Experimental del Zaidín – Consejo Superior de Investigaciones Científicas, Granada, Spain

#### *Edited by:*

Enrica Pessione, University of Torino, Italy

#### *Reviewed by:*

Linda Thomashow, United States Department of Agriculture – Agricultural Research Service, USA Paul Alan Hoskisson, University of Strathclyde, UK

#### *\*Correspondence:*

George P. C. Salmond, Department of Biochemistry, University of Cambridge, Tennis Court Road, Cambridge CB2 1QW, UK e-mail: gpcs2@cam.ac.uk; Miguel A. Matilla, Department of Environmental Protection, Estación Experimental del Zaidín – Consejo Superior de Investigaciones Científicas, Profesor Albareda 1, Granada 18008, Spain e-mail: miguel.matilla@eez.csic.es

Soil bacteria can be prolific producers of secondary metabolites and other biologically active compounds of economic and clinical importance. These natural products are often synthesized by large multi-enzyme complexes such as polyketide synthases (PKSs) or non-ribosomal peptide synthases (NRPSs).The plant-associated Gram-negative bacterium, Serratia plymuthica A153, produces several secondary metabolites and is capable of killing the nematode worm Caenorhabditis elegans; a commonly used model for the study of bacterial virulence. In this study, we show that disruption of the hybrid PKS/NRPS zeamine (zmn) gene cluster results in the attenuation of "fast-killing" of C. elegans, indicating that zeamine has nematicidal activity. C. elegans also exhibits age-dependent susceptibility to zeamine, with younger worms being most sensitive to the bioactive molecule. The zmn gene cluster is widely distributed within Serratia and phytopathogenic Dickeya species and investigation of strains harboring the zmn gene cluster showed that several of them are highly virulent in C. elegans. Zeamine was described previously as a phytotoxin and broadspectrum antibacterial compound. In addition to its nematicidal properties, we show here that zeamine can also kill Saccharomyces cerevisiae and Schizosaccharomyces pombe.The expression of the zmn gene cluster and regulation of zeamine production were also investigated.Transcription of the cluster was growth phase-dependent, and was modulated by the post-transcriptional RNA chaperone, Hfq.The results of this study show that zeamine is a highly toxic molecule with little, or no, apparent host specificity in very diverse biological systems. In its current form, zeamine(s) may be useful as a lead compound suitable for chemical modification and structure-activity assays. However, because of widespread nonselective toxicity in multiple bioassays, unmodified zeamine(s) is unlikely to be suitable as a therapeutic antibiotic.

**Keywords:** *Serratia plymuthica***,** *Dickeya***, secondary metabolite, PKS, NRPS, antibiotic,** *Caenorhabditis elegans***, zeamine**

### **INTRODUCTION**

Bacteria belonging to the Gram-negative genus *Serratia* are members of the γ-proteobacterial family *Enterobacteriaceae*. The best characterized member of the genus, *Serratia marcescens*, was first described in 1819 as a pigmented microbial isolate (Grimont and Grimont, 1978). *Serratia* species are ecologically diverse although strains of *Serratia plymuthica* are commonly isolated from soil as well as from the rhizosphere of wheat (Åström and Gerhardson, 1988), oilseed rape (Neupane et al., 2012a,b,c), melon (Kamensky et al., 2003), and pea (Gould et al., 2008), the anthosphere of oil pumpkin (Fürnkranz et al., 2012) and from rotting potato tissue (Czajkowski et al., 2012). Many strains of *S. plymuthica* are capable of producing compounds with antibiotic activity, such as the antifungal and anti-oomycete haterumalide, oocydin A (Levenfors et al., 2004), and the antifungal pyrrolnitrin (Liu et al., 2007). Several *S. plymuthica* strains were shown to be effective as biocontrol agents (De Vleesschauwer and Höfte, 2003), for example being capable of controlling the phytopathogenic gray mold *Botrytis cinerea* and white mold, *Sclerotina sclerotiorum,* on greenhousegrown melon (Kamensky et al., 2003), as well as suppressing the growth of *Penicillium*blue and green mold on citrus fruit (Meziane et al., 2006). The strain used in this study, *S. plymuthica* A153, was isolated from the rhizosphere of wheat (Åström and Gerhardson, 1988), despite producing phytopathogen-antagonistic compounds (Thaning et al., 2001; Levenfors et al., 2004). Recently, the strain has also been shown to be virulent in the *Caenorhabditis elegans* animal infection model (Matilla et al., 2012).

*Caenorhabditis elegans* is a free-living nematode that reaches an adult length of 1–2 mm. Although first developed as a genetic model in the 1970s (Brenner, 1974), *C. elegans* has undergone a renaissance over the past decade as a model system for assaying and understanding bacterial pathogenesis (Sifri et al., 2005). The importance of this has been two-fold. Firstly, not only are some soil-dwelling nematodes agricultural pests that attack economically important crops (reviewed by Jones et al., 2013) but they are also hard to control with traditional pesticides, and that makes nematode-pathogenic bacteria attractive as potential biocontrol agents (Chitwood, 2002). Secondly, although nematodes and mammals are separated by over 900 million years of evolution (Hedges et al., 2006), there are a surprising number of

bacterial virulence factors affecting both taxa (Rahme et al., 2000). Altogether, *C. elegans* has become a useful model for the identification of novel virulence factors and the characterization of the relationships between pathogens and genetically amenable hosts (Sifri et al., 2005).

Bacteria are capable of antagonizing *C. elegans* by several mechanisms (Sifri et al., 2005). The most common of these is bacterial colonization of the nematode intestine, where bacteria accumulate in the intestinal lumen (causing it to expand) and interfere with its normal function (Sifri et al., 2005). This is the mechanism by which many human pathogens – such as *Salmonella typhimurium* (Aballay et al., 2000), *Pseudomonas aeruginosa* (Mahajan-Miklos et al., 1999), and *S. marcescens* (Kurz et al., 2003) – infect *C. elegans*. This type of 'worm killing,' also referred to as 'slow killing,' takes place over the span of several days and correlates with bacterial proliferation in the intestine (Sifri et al., 2005; Portal-Celhay et al., 2012). On the other hand, some bacteria are capable of killing worms over much shorter time-spans. This 'fast killing' is usually mediated by toxins that are produced by the bacteria (Sifri et al., 2005). Although toxic proteins have been described (Wei et al., 2003), the toxins are typically secondary metabolites. For example, *P. aeruginosa* PA14 is capable of killing *C. elegans* through oxidative stress by using the phenazine compound pyocyanin as a virulence factor (Mahajan-Miklos et al., 1999; Cezairliyan et al., 2013).

Secondary metabolites are traditionally considered to be nonessential organic molecules that are synthesized by cells during the later stages of growth, without playing any direct role in growth or development (Price-Whelan et al., 2006). Although the true roles of secondary metabolites remain elusive, some of them can confer fitness advantages to producing bacteria. Thus, secondary metabolites with antibiotic activity are thought to increase the fitness of bacteria in complex natural environments by antagonizing microbial competitors (Mazzola et al., 1992) or by deterring predation (Pradel et al., 2007). Additionally, since soil is a stressful environment where nutrients are limited (Challis and Hopwood, 2003), it has been speculated that soil bacteria are enriched for the production of secondary metabolites performing primary functions acting, for example, as synergistic siderophores (Price-Whelan et al., 2006). Many secondary metabolites are synthesized by large multidomain proteins such as non-ribosomal peptide synthases (NRPSs) or polyketide synthases (PKSs) (Sattely et al., 2008). Frequently, the genes encoding these enzymes are carried on the biosynthetic gene clusters together with the genes for tailoring enzymes responsible for the modification of the final structure of the molecule (Osbourn, 2010). The modular and mobile nature of biosynthetic gene clusters can enable the mixed assembly of biosynthetic genes from different sources to form hybrid gene clusters containing both NRPS- and PKS-encoding genes.

Preliminary work in this laboratory showed that *S. plymuthica* A153 is a nematode-pathogen capable of killing *C. elegans* rapidly (Matilla et al., 2012)*,* suggesting that it produced a nematicide toxin. In this study we investigated the relationship between A153 and *C. elegans* to characterize the pathogen-worm interaction, with particular emphasis on identifying the A153

nematicide, genes involved in its biosynthesis and the regulation thereof.

### **MATERIALS AND METHODS**

#### **STRAINS, PLASMIDS, PHAGES, CULTURE MEDIA, AND GROWTH CONDITIONS**

Bacterial strains, plasmids, and phages used in this study are listed in **Table 1**. *Serratia, Dickeya*, and derived strains were grown at 30◦C, unless otherwise indicated, in L broth (LB, per liter: 5 g yeast extract, 10 g tryptone, 5 g NaCl), potato dextrose (16 g of potato dextrose broth l−1), minimal medium [MM: 0.1% (w/v) (NH4)2SO4, 0.41 mM MgSO4, 15 mM carbon source, 40 mM K2HPO4, 14.7 mM KH2PO4, pH 6.9–7.1], optimized minimal medium [OMM: 0.2% (w/v) (NH4)2SO4, 1.66 mM MgSO4, 0.2% (w/v) mannitol, 0.2% (w/v) glycerol, 60.3 mM K2HPO4, 33.1 mM KH2PO4, 15.9 μM MnCl2, 90.1 μM CaCl2, 32.9 μM FeSO4], 1-carbon OMM [1C-OMM: 0.2% (w/v) (NH4)2SO4, 1.66 mM MgSO4, 15 mM carbon source, 60.3 mM K2HPO4, 33.1 mM KH2PO4, 15.9 μM MnCl2, 90.1 μM CaCl2, 32.9 μM FeSO4]. *Escherichia coli* strains were grown at 37◦C in LB. *E. coli* DH5α was used for gene cloning. Media for propagation of *E. coli* β2163 was supplemented with 300 μM 2,6-diaminopimelic acid. Where appropriate, antibiotics were used at the following final concentrations (in μg mL−1): ampicillin, 100; kanamycin, 25 (*E. coli*), 75 (*Serratia*); streptomycin, 50; tetracycline, 10. Sucrose was added to a final concentration of 10% (w/v) to select for derivatives of a second crossover event during marker exchange mutagenesis. *C. elegans* was maintained at 15◦C using standard methods (Brenner, 1974). *Saccharomyces cerevisiae* and *Schizosaccharomyces pombe* were grown at 30◦C in yeast peptone dextrose (YPD, per liter: 10 g yeast extract, 20 g peptone, 20 g glucose).

### **DNA TECHNIQUES, DNA MANIPULATION, AND BIOINFORMATICS ANALYSES**

Plasmid DNA was isolated using the Anachem Keyprep plasmid DNA kit. Manufacturer's instructions were followed for DNA digestion (New England Biolabs). The Anachem gel recovery kit was used to recover DNA from agarose gels. Ligation reactions and total DNA extraction were performed using standard protocols (Sambrook et al., 1989). Competent cells were prepared using calcium chloride and transformations were performed using standard protocols (Sambrook et al., 1989). PCR fragments for cloning were amplified using *Phusion*® high fidelity DNA polymerase (New England Biolabs), and all sequences were confirmed. DNA sequencing was performed at the University of Cambridge DNA Sequencing Facility on an Applied Biosystems 3730xl DNA analyzer. Genome comparison analyses were performed using the wgVISTA on-line tool (Frazer et al., 2004). Open reading frames (ORFs) in the zeamine (*zmn*) gene cluster were predicted using Glimmer 3.0 (Delcher et al., 1999). BLAST was used for functional gene assignment. Protein domain organization was identified using the NCBI conserved domains database (Marchler-Bauer et al., 2011).

#### **Table 1 | Strains, phages, and plasmids used in this study.**


(Continued)

#### **Table 1 | Continued**


aAp, ampicillin; Km, kanamycin; Sm, streptomycin; Tc, tetracycline; Zmn, zeamine. bStrain A153L is considered the A153 wild-type precursor of derivatives used in this study and is referred to as such throughout. cTransposon insertions in these strains were in previously unlabelled genes, and have been named as follows: A153JH9, mannose-1-phosphate guanyltransferase (mpg); A153JH11, ferric aerobactin receptor (far).

#### *IN VITRO* **GROWTH KINETICS**

The kinetics of bacterial growth (OD600) were measured with a Unicam Heλios spectrophotometer at 600 nm, using bacterial cultures grown in LB or OMM at 215 r.p.m. at 25◦C, unless otherwise indicated. Three replicates were used for each condition and strain.

#### **RANDOM TRANSPOSON MUTAGENESIS, MUTANT SCREENING, AND GENERALIZED TRANSDUCTION**

Random transposon mutagenesis of A153 was performed by conjugation with *E. coli* β2163 harboring plasmids containing transposons Tn-KRCPN1 or mini-Tn*5*Sm/Sp, using a previously published protocol (Matilla et al., 2012). Mutant libraries for phenotypic screening were collected on antibiotic-selective plates before screening. Screening for mutants with reduced nematicidal activity was performed as described in Kurz et al. (2003), and screening for mutants with reduced antibacterial activity was performed as described by Zhou et al. (2011). To confirm that mutant phenotypes were associated with single insertions and to ensure association between mutation and phenotype, the mutations were transduced into the parent genetic background using the generalized transducing bacteriophage MAM1, as described in Matilla and Salmond (2014). Random-primed PCR and amplicon sequencing were used to identify transposon insertion points, as previously described Matilla et al. (2012). The genome of *S. plymuthica* A153 has been recently sequenced (Matilla and Salmond, unpublished results).

#### **PHENOTYPIC AND VIRULENCE ASSAYS**

Antibacterial and antifungal assays against *Bacillus subtilis* JH642 and *Verticillium dahliae* 5368, respectively, were performed as described in Matilla et al. (2012), at 25◦C unless stated otherwise. Assays for fungicidal activity against yeast were performed in the same manner as the antibacterial assays, but using overnight cultures *S. cerevisiae* and *S. pombe* grown in YPD. *C. elegans* virulence assays were performed as previously described Matilla et al. (2012), with the exception that synchronized worms were obtained by egg-lay, as described by Gems et al. (1998), with synchronized populations allowed to develop at 25◦C to induce the *fer-15* sterility phenotype. Large-scale phenotypic screens for nematicidal activity were performed using around 50 synchronized L1-stage worms per well in 24-well plates (Kurz et al., 2003), where each well was inoculated with a transposon mutant for testing. Plates were incubated at 25◦C and the survival of the worms was scored after 24 and 48 h. A Mantel-Cox log-rank test was used for statistical analysis of worm survival curves and was performed using Prism 5.0 (GraphPad Software). *P* values of 0.05 and below were considered statistically significant.

#### **MEASUREMENT OF** β**-GALACTOSIDASE ACTIVITY**

Transcriptional fusion assays to detect expression of the *lacZ* reporter gene were performed as described previously (Ramsay, 2013), using the fluorogenic substrate 4-methylumbelliferyl β-Dgalactoside (MUG2). Gene transcription values were expressed as relative fluorescence units (*r.f.u.)* s <sup>−</sup><sup>1</sup> OD600−1.

#### **QUANTIFICATION OF ZEAMINE ACTIVITY**

Quantification of A153 supernatant bioactivity was performed using cultures of the andrimid-negative mutant, A153A. Bacterial cells were grow at 25◦C, for 72 h, in MM supplemented with one of the 15 different carbon sources tested. Cultures were harvested after 72 h and culture supernatants were sterile-filtered (0.2 μm). Supernatant bioactivity (given as zeamine activity, ZA) was determined using a *B. subtilis* antibacterial bioassay (Matilla et al., 2012), and given as the fraction of the diameter of the resultant halo (H) and the diameter of the well (W) over the maximum OD600 attained by the bacteria in the relevant media.

#### **GENETIC COMPLEMENTATION OF hfq**

For single-copy complementation of the in-frame *hfq* deletion mutant A153H, an inducible plasmid construct was first generated, using wild-type *hfq*. For this, the A153 *hfq* ORF was amplified using primers hfq-BamHI (5 -TAAT TCCGCTAAGGGGCAATCTTTGCA-3 ) and hfq-PstI (5 -TAATC TGCAGCTCGCAACGCGCTTTATTC-3 ). The PCR product was digested with PstI and BamHI and inserted at the same sites in the vector pTRB30. The resulting plasmid, pJEEUH13, was introduced into A153H by electroporation. Gene expression was induced with 0.1 mM IPTG.

#### **RESULTS**

### **CHARACTERIZATION OF A153 FAST-KILLING OF** *Caenorhabditis elegans* It has been reported previously that *Serratia* spp. such as *S. marcescens* (Kurz and Ewbank, 2000) and *Serratia* sp. ATCC 39006 (Coulthurst et al., 2004) are capable of killing *C. elegans* over the span of 3–5 days by establishing an infection in the nematode intestine. We observed that *S. plymuthica* A153 is capable of killing worms quickly, within hours (**Figure 1**). To better understand this interaction, the A153 killing dynamics of *C. elegans* were subjected to an initial characterization, whereby *C. elegans* L4 larvae were transferred from the standard *C. elegans* food-source *E. coli* OP50 onto lawns of A153. After transfer, the worms succumbed in stages. Initially, they were strongly repulsed by the bacteria and showed strong avoidance of the bacterial lawn. The immediacy of this effect suggests that the bacteria may produce odorants or surfactants that deter nematode grazing (Pradel et al., 2007; Burlinson et al., 2013). Worms that remained on the bacterial lawn quickly become immobilized, within hours of transfer, and remained alive for a latency period of about 6 h, before they started dying. While immobilized, but before death, worms showed little to no spontaneous movement, although the pharynx and body wall muscles could be induced to contract by the gentle touch of an instrument. The majority of worms were dead within 24 h of transfer. The speed of A153 killing of *C. elegans* suggested that A153 produces a fast-acting nematicide that functions as a potent virulence factor in the nematode infection model.

#### *Caenorhabditis elegans* **SHOW AGE-DEPENDENT SENSITIVITY TO A153 FAST-KILLING**

The age of individual *C. elegans* has been shown previously to determine worm susceptibility to bacterial pathogens. Some fastkilling bacterial toxins such as pyocyanin have been shown to be more toxic to younger worms (Mahajan-Miklos et al., 1999). In contrast, under slow-killing infection models, older worms are generally more sensitive (Laws et al., 2004; Portal-Celhay et al., 2012). To determine if there were any age-dependent susceptibility effects to A153 fast-killing of *C. elegans*, worms from each of the four *C. elegans* larval stages (L1–L4) and day-1 and day-2 adults were transferred onto lawns of A153 and nematode survival was

assayed hourly. The results revealed that*C. elegans* show significant age-dependent differential susceptibility to the A153 nematicide, with early larvae being more sensitive than late larvae (**Figure 2**; L1 vs. L4: *P* < <0.05), and with larvae being more sensitive than adult worms (**Figure 2**; L4 vs. d1: *P* < < 0.05). This shows that A153 fast-killing and *C. elegans* susceptibility to the A153 nematicide is inversely correlated with the developmental stage and age of the worm.

#### **ISOLATION OF A153 MUTANTS WITH REDUCED VIRULENCE AGAINST** *C. elegans*

To identify the genes involved in A153 fast-killing of *C. elegans*, a mutant library was generated using random transposon mutagenesis. In an initial screen, the library was screened looking for mutants with reduced nematicidal activity against L1-stage larvae.

**FIGURE 2 |** *Caenorhabditis elegans shows age-dependent sensitivity to the A153 nematicide.* Survival of different larval stages and adult C. elegans worms when cultured on S. plymuthica A153. The results of a representative trial with at least 50 worms under each condition is shown. This approach yielded one mutant, A153Ce10, which showed significantly reduced'fast-killing' of worms. This mutant still showed wild-type antibacterial (**Figure 3A**) and antifungal (**Figure 3B**) activities, suggesting that A153Ce10 is specifically deficient in worm-killing. Random-primed PCR confirmed that the transposon in A153Ce10 was in the gene *zmn16*, encoding a putative thioester reductase, and forming part of a hybrid FAS/PKS/NRPS gene cluster responsiblefor the biosynthesis of the broad-spectrum antibacterial antibiotic, zeamine (zmn), which was first described by Masschelein et al. (2013).

Zeamine was first identified in the plant pathogen *Dickeya zeae* EC1 as a phytotoxic virulence factor with antibacterial properties (Wu et al., 2010; Zhou et al., 2011), and our results suggest that zeamine also has nematicidal activity. In A153, the

**FIGURE 3 |** *Serratia plymuthica A153 mutants that show reduced virulence in C. elegans are specifically attenuated for zeamine production.* Plate-based bioassays showing andrimid **(A)** oocydin A **(B)** and zeamine **(C)** production. **(A,B)** The upper culture spot shows the A153 wild-type whereas the lower culture spot shows the derived mutant A153Ce10. **(C)** The upper culture spot shows the A153 andR mutant A153A whereas the lower culture spot shows the derived mutant A153Ce10A, which is representative of all mutants characterized in this study. In all cases, plates were incubated at 25◦C either overnight **(A,C)** or for 5 days **(B)**.

antibacterial activity of zeamine was masked by the production of another potent antibiotic, andrimid (cf. **Figures 3A,C**; Matilla and Salmond, unpublished results). Using a mutant defective in the production of this antibacterial compound, A153A, a random transposon mutant library was constructed and screened for mutants lacking the small halo associated with the production of zeamine byA153 (**Figure 3C**). Six mutants showing no, or reduced, antibacterial activity (A153JH1, A135JH5, A153JH6, A153JH24, A153JH27, and A153JH28) and with transposon insertions in the *zmn* gene cluster (**Figure 4A**) were isolated. These mutants showed significantly reduced nematicidal activity compared to that of the A153A parent strain (*P* < < 0.05; **Figure 5A**, Figure S1), confirming that zeamine is indeed the A153 nematicide and responsible of the observed A153 fast-killing of *C. elegans*.

The A153 *zmn* gene cluster spans over 54 kb and contains 22 genes (*zmn1-22*), organized into three putative operons (**Figure 4A**; Table S1). As previously described in *S. plymuthica* RVH1, the A153 *zmn* gene cluster contains the genes for three multidomain PKSs (*zmn10, zmn11, zmn18*) and two multidomain NRPSs (*zmn16, zmn17*). In addition to these biosynthetic genes, the cluster also contains genes encoding modifying enzymes (*zmn3, zmn12, zmn14, zmn15, zmn22*) and transportrelated proteins (*zmn7, zmn8, zmn9, zmn20, zmn21*) – the latter with a proposed role in conferring innate resistance to the zeamine antibiotic (Masschelein et al., 2013). A putative integrase-encoding gene marks the downstream end of the cluster, suggesting that it could have been acquired by horizontal gene transfer.

#### **THE** *zmn* **GENE CLUSTER IS WIDELY DISPERSED WITHIN** *Serratia* **AND** *Dickeya* **GENERA**

Genome comparison analyses revealed that the *zmn* gene cluster is present in *S. plymuthica* strains AS9, AS12, AS13, A30, S13, and V4 (Table S2). Additionally, we also identified the biosynthetic gene cluster in several phytopathogenic strains belonging to the *Dickeya* genus, including *D. solani* MK10, MK16, IPO222; *D. zeae* DZ2Q and ZJU1202; and *Dickeya* spp. MK7 and NCPPB 3274 (Table S2). The *Serratia* and *Dickeya zmn* clusters span between 50.64- and 54.02-Kbp and are between 59.7 and 96.4%

*Dickeya* **strains.** Genetic organization of the zmn gene cluster sequence in S. plymuthica A153 **(A)**. Location of the Tn-KRCPN1 transposon insertions are indicated by red arrowheads. Inserts show mutant antibacterial phenotypes against B. subtilis after an overnight incubation

parent strain to the far right. DNA homology (%) between the zmn gene cluster of A153 and those of S. plymuthica RVH1 **(B)** and D. solani MK10 **(C)** is presented. Sequence comparisons were performed using wgVISTA and show regions of >50% homology.

#### **FIGURE 5 | Continued**

**Kinetics of the nematicidal properties of** *Serratia plymuthica* **A153 strains. (A)** Mutation of the zmn gene cluster attenuates S. plymuthica A153 fast killing of C. elegans. Compared with the A153A parent strain, mutants A153JH1, A153JH5, A153JH6 and A153Ce10A, show reduced virulence in C. elegans (P < < 0.05). **(B)** Mutation of the chaperone Hfq eliminates A153 virulence in C. elegans. The virulence could be partially restored by the in trans expression of hfq (using plasmid pJEEUH13). Gene expression was induced using 0.1 mM IPTG. Inserts show antibacterial phenotypes of the parent and complemented strains after an overnight incubation at 25◦C. **(C)** Mutation of the regulators rpoS, pigP, and csrB had no effect on A153 fast−killing of C. elegans, and neither did deletion of the A153 T6SS. For each experiment, worms were transferred at the L4 stage and the results of one representative trial with at least 50 worms under each condition is shown.

fast killing of C. elegans **(A)**. In contrast, D. solani strains MK10, MK16, IPO2222 and Dickeya spp. MK7 and NCPPB3274 showed no fast killing of C. elegans **(B)**. For each experiment, worms were transferred at the L4 stage and the results of a representative trial with at least 50 worms under each condition is shown.

identical at the DNA level with the A153 *zmn* gene cluster (Table S3).

*In silico* analyses showed that the A153 and RVH1 *zmn* gene clusters have the same gene and domain organization (**Figure 4B**).

However, we found that the A153 PKS Zmn10 contains a previously unidentified dehydratase (DH) domain. Based on the high sequence identity (around 95%) between *zmn* gene clusters in A153 and related *S. plymuthica* strains, we hypothesized that all these strains could possess high nematicidal activities*.* Thus, using L4-stage larvae, we showed that *S. plymuthica* strains AS9, AS12, and AS13 are highly virulent and capable of killing *C. elegans* at similar levels to those of A153, with 50% of worms dying within 18 h of transfer (**Figure 6A**).

#### **FIGURE 7 | Continued**

**The expression of the** *S. plymuthica* **A153** *zmn* **gene cluster is growth phase-dependent and post-transcriptionally regulated by the RNA chaperone Hfq. (A)** Transcription of the zmn gene cluster throughout growth in Serratia plymuthica A153. The dip in β−galactosidase activity post−peak suggests the enzyme is subject to proteolytic turnover, which has been observed previously in A153 (Matilla and Salmond, unpublished results). **(B)** Different carbon sources have different effects on the correlation between zmn gene transcription and the bioactivity of A153A−derived supernatants. β−galactosidase activity was measured in strain A153JH6 grown in 1C−OMM supplemented with 15 mM sorbitol, mannitol, sucrose, or gluconic acid. **(C)** Expression of the zmn gene cluster is regulated by the RNA chaperone Hfq. β−galactosidase activity was measured in strain A153JH6 in the presence (red) or absence (blue) of a chromosomal hfq gene deletion. **(A,B)** The strains were grown in LB at 25◦C and β-galactosidase activities were measured in strains expressing chromosomal zmn13::lacZ fusions. In all panels solid symbols represent growth of the corresponding strains whereas open symbols represent β−galactosidase activity. Data shown are the average values ±SD of at least three experiments.

Importantly, pairwise comparisons *in silico* indicate that genes *zmn1-4* are not present in strains of *Dickeya* and that the *zmn5* homolog constitutes the first gene of these *zmn* gene clusters

by zeamine.

(**Figure 4C**). In addition, whereas the *zmn* gene clusters in different strains of *Dickeya* show around 60% sequence identity with the A153 cluster, the putative permease Zmn9 is only about 50% conserved between the genera. The NCBI conserved domains database predicts that the *Serratia* Zmn9 contains a zinc-dependent phospholipase domain, whereas Zmn9 in *Dickeya* is predicted to contain a CDP-alcohol phosphatidyltransferase domain. To investigate if strains of *Dickeya* show similar nematicidal differences to *zmn*-carrying strains of *Serratia* despite these differences, a subset of *Dickeya* strains that carry the *zmn* gene cluster (MK7, MK10, MK16, IPO2222, and NCPPB 3274) were chosen for *C. elegans* virulence assays. Interestingly, although the tested *Dickeya* strains had variable pathogenic capacity against L4-stage *C. elegans* (**Figure 6B**), none of them was found to show the fast-killing phenotype observed in strains of *S. plymuthica* that carry the *zmn* gene cluster.

#### **THE** *zmn* **GENE CLUSTER IS TRANSCRIBED IN A GROWTH PHASE-DEPENDENT MANNER**

To investigate the transcription of the *zmn* gene cluster, βgalactosidase activity was measured in a chromosomal*zmn13*::*lacZ* fusion strain (strain A153JH6). Transcription of *zmn* biosynthetic genes started in mid-logarithmic phase of growth and reached an apparent maximum in early stationary phase of growth (**Figure 7A**). The sharp decrease in β-galactosidase levels beyond this point may be explained by proteolytic turnover of β-galactosidase (Matilla and Salmond, unpublished results).

#### **THE PRODUCTION OF ZEAMINE IS CARBON SOURCE-DEPENDENT**

Zeamine production has been shown previously to differ between growth media, being higher in *D. zeae* EC1 when grown in an OMM, compared to the standard LB culture medium (Zhou et al., 2011). Thus, we investigated zeamine production in strains grown in different carbon sources by determining the bioactivity of cell-free supernatants against *B. subtilis,* which is sensitive to zeamine (**Figure 3C**). Our results showed that the production of the bioactive molecule is carbon source-dependent, with some carbon sources favoring high levels of zeamine biosynthesis (e.g., sorbitol and mannitol) whereas others do not support the production of zeamine at all (e.g., gluconic acid; Table S4). Expression of the *zmn* gene cluster, reported using β-galactosidase assays, was examined in a modified OMM with different carbon sources (1C-OMM). Unexpectedly, no correlation between *zmn* gene transcription and zeamine production was observed (**Figure 7B**; Table S4).

#### **THE RNA CHAPERONE Hfq REGULATES THE PRODUCTION OF ZEAMINE AND THE EXPRESSION OF** *zmn* **BIOSYNTHETIC GENES**

The RNA chaperone Hfq acts as a regulator of gene expression by interacting with small regulatory RNAs to stabilize the interaction between these and their target mRNAs through the formation of regulatory RNA-RNA complexes (Vogel and Luisi, 2011). Mutants defective in *hfq* are highly pleiotropic and can be attenuated in both virulence and the production of secondary metabolites in *Serratia* sp. ATCC39006 (Wilf et al., 2012). To investigate whether Hfq is involved in regulating the production of zeamine in *S. plymuthica* A153, *C. elegans* virulence assays were performed. The results showed that deletion of *hfq* in strainA153H strongly attenuated virulence to *C. elegans* L4 larvae (*P* < < 0.05; **Figure 5B**). The virulence of A153H could be partially restored by expressing *hfq in trans* (**Figure 5B**). β-galactosidase assays showed that deletion of *hfq* abolished the transcription of the *zmn* gene cluster (**Figure 7C**), confirming that Hfq positively regulates the production of zeamine. It is well known that Hfq regulates the translation of the stationary-phase sigma factor RpoS (Vogel and Luisi, 2011). To investigate whether Hfq regulation is dependent on RpoS, we phenotypically characterized an *rpoS* mutant in A153. However, this mutant showed the same antibacterial and nematicidal activities as the wild-type strain (**Figure 5C**). Mutants defective in the non-coding small RNA *csrB* (Babitzke and Romeo, 2007) and the transcriptional regulator PigP (Fineran et al., 2005) were also unaffected in their virulence against *C. elegans* (**Figure 5C**).

#### **ZEAMINE IS TOXIC TO ASCOMYCETE YEASTS**

Zeamine shows some structural similarity to another family of hybrid polyamino-polyketides: fabclavines. These natural products have been shown to have broad-spectrum antibiotic activity against a diverse set of organisms including bacteria and ascomycete fungi (Fuchs et al., 2014). Using cell-free supernatants of A153A and A153JH6, we showed that zeamine is bioactive against the ascomycete yeasts *S. cerevisiae* and *S. pombe* (**Figure 8**).

#### **THE TYPE VI SECRETION SYSTEM OF** *S. plymuthica* **A153 IS NOT INVOLVED IN VIRULENCE**

The bacterial Type VI secretion system (T6SS) is the most recently described secretion system in Gram-negative bacteria and has been found to promote bacterial virulence against both prokaryotic competitors and eukaryote hosts (Coulthurst, 2013). *In silico* analyses revealed that a T6SS gene cluster highly homologous to that present in *Citrobacter rodentium* ICC168 is present in the genome of A153 (Matilla and Salmond, unpublished results). The role of the T6SS in virulence against nematodes remains largely uncharacterized but Sana et al. (2012) showed that it plays a role in *P. aeruginosa* "slow killing" of *C. elegans*. To investigate if the *S. plymuthica* A153 T6SS is involved in virulence, an A153 T6SS mutant was constructed (A153T6) and characterized. However, no difference in the virulence was observed between A153T6 and the A153 wild-type (**Figure 5C**).

### **DISCUSSION**

Although zeamine was first described as a phytotoxin with broadspectrum antibacterial properties (Wu et al., 2010; Zhou et al., 2011), this study showed that zeamine is also a potent nematicide. Furthermore, it is a characteristic of some bacterial toxins that younger hosts are more susceptible than older ones (Mahajan-Miklos et al., 1999), and consistent with this, *C. elegans* showed age-dependent sensitivity to the A153 nematicide. In addition, we have shown that zeamine is capable of killing *S. cerevisiae* and *S. pombe*, and therefore zeamine, by definition, is also a fungicide. This leads us to conclude that zeamine is a very broad-spectrum antibiotic that is capable of antagonizing a phylogenetically diverse set of organisms, making it unsuitable, in an unmodified form, for application as a therapeutic antibiotic.

Like the fabclavines (Fuchs et al., 2014), the zeamine molecule has a polyaminated fatty acid backbone that is derived from modified secondary lipid metabolism (Wu et al., 2010; Masschelein et al., 2013). Zeamine also shows some structural similarity with compounds such as the phytotoxin syringomycin, which has been shown to antagonize plant cells by forming ion channels in the plant cell membrane (Hutchison and Gross, 1997). Altogether, it is possible that zeamine has lipophilic properties and might interact with the lipids of cell membranes through a mechanism analogous to that of cationic antimicrobial peptides (Hancock, 2001). A membranal target would be entirely consistent with the observed strong susceptibility of very diverse organisms to zeamine.

Fatty acids have been shown previously to function as nematicides against various phytopathogenic nematodes, and have been hypothesized to disrupt plasma membranes to facilitate solubilization of the nematode cuticle or hypodermis (Anke et al., 1995; Davis et al., 1997). Similarly, cationic peptides are capable of interacting with and disrupting cell membranes owing to a three-dimensional amphiphilic structure (Joondan et al., 2014), and various classes thereof have been shown effective against fungi, protozoa and mammalian cells (Hancock, 2001). Intriguingly, cationic peptides share many of these properties with zeamine, and have been shown previously to function as nematicides (Colgrave et al., 2008), and are capable of causing severe damage to the intestine of lepidopteran larvae (Barbeta et al., 2008). It is therefore possible that the nematicidal cytotoxicity of zeamine involves damage to, and vacuolisation of, the cells that line the *C. elegans* intestine.

Our results raise questions about the role of zeamine in nature. The majority of sequenced *S. plymuthica* strains that carry the *zmn* gene cluster were isolated from agricultural contexts, such as the plant antho- and rhizosphere (e.g., Åström and Gerhardson, 1988; Fürnkranz et al., 2012; Neupane et al., 2012a,b,c). Plant root exudates are rich in sugars and other organic nutrients that favor root colonization by soil-borne bacteria (Bais et al., 2006), and this study has shown that different sugars have different effects on zeamine production by A153 – with some repressing the production of the antibiotic whilst others favor it. In this regard, it is notable that A153 was initially isolated on the basis that it antagonized plant growth (Åström and Gerhardson, 1988), and that zeamine has been found to be a potent phytotoxin, capable of antagonizing both shoot and root development in rice seedlings (Zhou et al.,2011). Together with our results showing that zeamine is also a potent nematicide, these observations raise the question: are there any possible large-scale effects of zeamine production by rhizosphere-associated soil bacteria?

Various *Dickeya* spp. are pathogens of plants, and as some of these have been found to contain the *zmn* gene cluster, their nematicidal properties were investigated. However, contrary to expectation, none of the selected strains were found to show fastkilling of *C. elegans*. The absence of genes *zmn1-4* in strains of *Dickeya* suggests that the cryptic *zmn* gene clusters of the assayed *Dickeya* isolates may not be effectively or functionally expressed, correlating with a lack of rapid nematicidal activity. Alternatively, and considering the low sequence conservation between *Serratia* and *Dickeya zmn9*, it is possible that the final

biosynthetic products of the *zmn* gene clusters of the *Dickeya* strains analyzed in this study do not possess the same biological properties as zeamine. For example, Masschelein et al. (2013) showed that *S. plymuthica* RVH1 is capable of producing three different zeamine molecules (zeamine, zeamine I, and zeamine II), of which zeamine I is the predominant molecule produced by *D. zeae* EC1 (Wu et al., 2010). The individual contributions of these molecules to the antibiotic activity of zeamine, *sensu lato*, is unclear, although further work on this topic might reveal differential toxic activities between different zeamines and derived molecules. If different zeamine species are found to show differential specific toxicities, this would raise the possibility that these may be used as more-specific antibiotics or pesticides. It is also possible that the *zmn* gene clusters of the *Dickeya* isolates tested in this study are simply cryptic under the conditions tested – as is known for gene clusters encoding other secondary metabolites (Osbourn, 2010). It is, however, clear that various *Dickeya* spp. are virulent in the *C. elegans* model, consistent with previous work that found that the plant pathogens *D. dadantii* 3937c, *Agrobacterium tumefaciens* CFBP2413, and *Pectobacterium carotovorum* CFBP 2141, are capable of killing *C. elegans* through infection (Couillault and Ewbank, 2002).

The expression pattern of the A153 *zmn* gene cluster is characteristic for secondary metabolite antibiotics, which are typically produced during conditions of nutrient limitation and reduced growth (Bibb, 2005). As is common with secondary metabolites, transcription of *zmn* biosynthetic genes is sensitive to environmental conditions. In contrast to findings in other bacteria such as *S. plymuthica* RVH1 (Masschelein et al., 2013), *zmn* gene transcription in A153 does not appear to be thermoregulated (Figure S2). Zeamine production does, however, show media-dependent effects in A153, as has previously been reported for *D. zeae* EC1 (Zhou et al., 2011). The basis of this effect appears to be post-transcriptional, as we did not observe a correlation between *zmn* gene transcription and zeamine bioactivity between different carbon sources (cf. **Figure 7B**; Table S4).

We observed a correlation between *zmn* gene transcription and the activity of the RNA chaperone Hfq – with the A153 *zmn* gene cluster being silent in a *hfq* background – consistent with its role as a regulator of secondary metabolism in other species of *Serratia.* Mutation of *hfq* has been shown previously to reduce the transcription of genes involved in iron uptake in *E. coli* (Veˇcerek et al., 2003), and intriguingly, as part of our mutagenesis program, we isolated a mutant (A153JH11) with a transposon insertion in the A153 homolog of the ferric aerobactin receptor (Figure S3A), which showed increased production of zeamine (Figure S3B), suggesting that production of the antibiotic is increased during conditions that might mimic iron limitation.

In summary, the work presented in this study has shown that *S. plymuthica* A153 produces the very broad-spectrum antibiotic, zeamine. Although zeamine was initially described as a phytotoxin with antibacterial activity, our results show that zeamine is also a potent nematicidal compound and antifungal. That zeamine antagonizes such a phylogenetically diverse set of organisms suggests that it targets a highly conserved cellular process, which

would make it unsuitable as a specific antibiotic. That target is likely to be the cell membranes of diverse hosts. However, the research presented here may help toward the development of zeamine analogs with enhanced host specificity in nematodes and fungi.

#### **ACKNOWLEDGMENTS**

We thank Ian Toth, Richard Cooper, Juan Mata, and Steve Oliver for generous donations of bacterial and fungal strains. The DH26 nematode strain was provided by the *Caenorhabditi*s Genetics Center. We would also like to thank Eric Miska and Peter Sarkies for advice on the *C. elegans* virulence assay and Alison Drew for technical support. MAMV was supported by the EU Marie-Curie Intra-European Fellowship for Career Development (FP7-PEOPLE-2011-IEF), grant number 298003. The Salmond laboratory is supported by funding through the Biotechnology and Biological Sciences Research Council (BBSRC; UK). Work with plant pathogens was carried out under DEFRA licence No. 50864/197900/1.

#### **SUPPLEMENTARY MATERIAL**

The Supplementary Material for this article can be found online at:http://www.frontiersin.org/journal/10.3389/fmicb.2015.00137/ abstract

#### **REFERENCES**


*Serratia* sp. ATCC39006. *Environ. Microbiol*. 13, 2649–2666. doi: 10.1111/j.1462- 2920.2011.02532.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.

*Received: 13 January 2015; paper pending published: 26 January 2015; accepted: 05 February 2015; published online: 26 February 2015.*

*Citation: Hellberg JEEU, Matilla MA and Salmond GPC (2015) The broad-spectrum antibiotic, zeamine, killsthe nematode worm Caenorhabditis elegans. Front. Microbiol. 6:137. doi: 10.3389/fmicb.2015.00137*

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

*Copyright © 2015 Hellberg, Matilla and Salmond. 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.*

# Prodigiosin Induces Autolysins in Actively Grown *Bacillus subtilis* Cells

*Tjaša Danevci ˇ cˇ 1, Maja Boric Vezjak ´ 1, Maja Tabor1, Maša Zorec2 and David Stopar1\**

*<sup>1</sup> Chair of Microbiology, Department of Food Science and Technology, Biotechnical Faculty, University of Ljubljana, Ljubljana, Slovenia, <sup>2</sup> Chair of Microbiology and Microbial Biotechnology, Department of Animal Science, Biotechnical Faculty, University of Ljubljana, Ljubljana, Slovenia*

Prodigiosin produced by marine bacterium *Vibrio ruber* DSM 14379 exhibits a potent antimicrobial activity against a broad range of Gram positive and Gram negative bacteria. The mechanism of prodigiosin antimicrobial action, however, is not known. In this work, the effect of prodigiosin on *Bacillus subtilis* growth, cell membrane leakage, and induction of autolysins was studied. Treating *B. subtilis* with prodigiosin resulted in rapid decline of optical density and increased cell membrane leakage measured by β-galactosidase activity. Cell lysis was initiated immediately after treatment with prodigiosin in the middle exponential phase and was completed within 2 h. Lytic activity of prodigiosin in mutant strains with impaired autolysin genes *lytABCD* decreased for 80% compared to the wild type strain, while in *lytABCDEF* mutant strain prodigiosin had no bacteriolytic but only bacteriostatic effect. Fast prodigiosin lytic activity on individual *B. subtilis* cells was confirmed by a modified comet assay. The results indicate that prodigiosin autolysin induction in *B. subtilis* is growth phase dependent.

#### *Edited by:*

*Enrica Pessione, University of Turin, Italy*

#### *Reviewed by:*

*Sylvie Garneau-Tsodikova, University of Michigan, USA Paul Richard Himes, University of Louisville, USA*

> *\*Correspondence: David Stopar david.stopar@bf.uni-lj.si*

#### *Specialty section:*

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

*Received: 28 August 2015 Accepted: 11 January 2016 Published: 28 January 2016*

#### *Citation:*

*Danevci ˇ c T, Bori ˇ c Vezjak M, Tabor M, ´ Zorec M and Stopar D (2016) Prodigiosin Induces Autolysins in Actively Grown Bacillus subtilis Cells. Front. Microbiol. 7:27. doi: 10.3389/fmicb.2016.00027*

Keywords: prodigiosin, autolysis, *Bacillus subtilis*, antimicrobial, mechanism, autolysin, lytic rate, comet assay

## INTRODUCTION

Prodigiosin, a secondary metabolite, is produced by several bacterial genera including *Serratia, Streptomyces, Vibrio*, *Hahella*, *Zooshikella,* and *Pseudoalteromonas* (Farmer et al., 1988; Sawabe et al., 1998; Shieh et al., 2003; Williamson et al., 2006; Kumar and Nair, 2007; Fehér et al., 2010; Staric et al., 2010 ˇ ; Lee et al., 2011; Stankovic et al., 2012). Prodigiosin provides numerous ecophysiological benefits for the producing cell and has a good potential for biotechnological as well as medical applications (Burger and Bennett, 1985; Hood et al., 1992; Ryazantseva et al., 1995; Seganish and Davis, 2005; Haddix et al., 2008; Staric et al., 2010 ˇ ; Boric et al., 2011 ´ ). Prodigiosin has been used in control of plant diseases caused by bacteria and fungi (Okamoto et al., 1998; Someya et al., 2001; Meschke et al., 2012). It has affinity to DNA (Melvin et al., 2000), but shows no *in vitro* or *in vivo* genotoxic effect on *Salmonella typhimurium* cells (Guryanov et al., 2013). It modulates H+/Cl− symport activity (Konno et al., 1998). When used as a coloring agent, it retains its antibiotic activity (Alihosseini et al., 2008). Prodigiosin has anticancer, antimalarial and immunosuppressant properties (Pérez-Tomás et al., 2003; Williamson et al., 2007). It inhibits growth of a wide range of Gram positive bacteria including *Bacillus subtilis* and *Staphylococcus aureus*, as well as Gram negative *Escherichia coli*, *Salmonella enterica,* and *Erwinia carotovora* (reviewed in Stankovic et al., 2014). The mechanism of its antibacterial action, however, is not understood on the molecular level. In this work, we study the molecular targets of prodigiosin in *B. subtilis*.

Many antibacterial agents target cell membrane and cell wall (Falk et al., 2010; Lacriola et al., 2013). In *B. subtilis* induction of autolysins and subsequent cell wall degradation is a major mechanism of antimicrobial action of several different compounds (Falk et al., 2010; Lacriola et al., 2013). Autolysins are enzymes involved in hydrolysis and remodeling of the peptidoglycan in the bacterial cell wall (Smith et al., 2000). It is absolutely critical to maintain the regulation of potentially suicidal activity of autolysins. The regulation of autolysin activity is mainly due to the local pH in respiring cells. In respiring cells, protons are extruded across the cytoplasmic membrane and bind to cell wall constituents (Jolliffe et al., 1981; Kemper et al., 1993; Calamita and Doyle, 2002). The protonation of the cell wall constituents create a wall matrix of low pH. Proton motive force sustains this local low pH and, in turn, autolytic activity is inhibited because *N*-acetylmuramyl-L-alanine amidase cannot function below pH 5.5. There are 36 autolysins present in the *B. subtilis* genome (Smith et al., 2000). Among different autolysins in *B. subtilis*, two autolysins of vegetative growth named LytC and LytD are responsible for about 95% of the bacterial cell autolysis during active growth (reviewed in Smith et al., 2000). It has been shown previously that β-galactosidase release correlates with activation of pathways leading to autolysis (Falk et al., 2007). Apart from β-galactosidase release several other methods (e.g., optical density, cell count) can be used to determine bacterial cell lysis. The comet assay is a method primarily used for the detection of DNA damage and cell lysis in a single cell (Ostling and Johanson, 1984; Singh et al., 1988). Recently, comet assay was applied to detect bacterial lysis after bacteriophage treatment (Khairnar et al., 2014). Using comet assay to detect single cell lysis provides an advantage over enzyme activity or optical density measurements of cell lysis which probe a population response.

In this work, the mechanism of antibacterial activity of prodigiosin on *B. subtilis* was studied. The ability of prodigiosin to affect bacterial population structure was determined in simple co-culture experiments. The effect of prodigiosin on the bacterial growth, cell wall integrity, autolysis, and DNA damage was determined on the wild type strain and several autolysin mutants. This study is the first attempt to determine the mechanism of antibacterial activity of prodigiosin in *B. subtilis in vivo*. The results indicate that prodigiosin activates rapid autolysis in *B. subtilis*.

## MATERIALS AND METHODS

## Bacterial Strains, Mutant Strain Construction and Growth Conditions

Bacterial strains used in this study were *B. amyloliquefaciens* FZB42, *B. licheniformis* ATCC9445A*, B. subtilis* NCIB3610 all obtained from BGSC, *B. subtilis* ATCC6051, *B. mycoides* DSM2048 obtained from DSMZ, *B. subtilis* PS-216 *wt* (Stefanic and Mandic-Mulec, 2009), *B. subtilis* PS-216 *amyE::*Phyperspank*mKate2 cat* (Stefanic et al., 2015), *Escherichia coli* MG1655 and *Vibrio ruber* DSM14379 (Stopar et al., 2004; Boric et al., 2011 ´ ). In addition mutant strains of *B. subtilis* PS-216 *lytABC::neo*, *B. subtilis* PS-216 *lytABC::neo lytD::tet*, *B. subtilis* PS-216 *lytABC::neo lytD::tet lytE::cam*, *B. subtilis* PS-216 *lytABC::neo lytD::tet lytE::cam lytF::spc* and *B. subtilis* PS-216 *srfA-lacZ neo* were prepared. The *lyt* mutants were constructed by transforming the *B. subtilis* PS-216 *wt* strain with the DNA isolated from *B. subtilis* L16611 and *B. subtilis* L16628 obtained from BGSC (Margot and Karamata, 1992). The *srfAlacZ* construct was made by transforming *B. subtilis* PS-216 *wt* strain with the DNA isolated from *B. subtilis* KTB308 (Bacon Schneider et al., 2002). *Bacillus* strains were grown at 37◦C (except *B. mycoides* at 28◦C) and 200 rpm in LB medium or CM medium (Albano et al., 1987) or at 28◦C and 200 rpm in PYE medium with 3% (w/V) NaCl (Danevci ˇ c et al., 2005 ˇ ).

### Prodigiosin Extraction

*Vibrio ruber* DSM14379 was cultured overnight in PYE medium with 3% (w/V) NaCl at 28◦C and 200 rpm (Danevci ˇ c et al., 2005 ˇ ). Overnight culture was diluted 100 times in 400 mL of fresh M9 medium supplied with 5 g/L glucose (Staric et al., 2010 ˇ ) and incubated for 16 h at the same growth conditions. Cells were harvested with centrifugation at 10397 *g* for 10 min. Prodigiosin was extracted from cells with an equal volume of acetone at 28◦C and 200 rpm for 2 h. After centrifugation for 15 min at 10397 *g* to remove cell debris, the solvent was evaporated. Dry biomass was resuspended in sterile 96% (V/V) ethanol and the pigment extract was filtered through 0.20 μm filter. Prodigiosin concentration in the extract was determined as described previously by Staricˇ et al. (2010). The prodigiosin purity in the extract was confirmed by HPLC using Kinetex C18 column (250 mm × 4.6 mm, 5 μm, Phenomenex, USA) according to the modified method described previously (Nakashima et al., 2006). The separation was achieved using 0.1% trifluoroacetic acid (A) and methanol (B) mobile phases, and a gradient elution program at 1 mL−<sup>1</sup> min with the following parameters: 0–5 min 100% A, 5–25 min 0– 100% B (linear gradient), 25–30 min 100% B (isocratic), and 30–50 min 100% A (isocratic) to re-equilibrate the column, monitored by UV-VIS detection. The purity was determined by HPLC at different wavelengths in the range from 400 to 600 nm. The HPLC elugram at 535 nm is given in supplementary materials (Supplementary Figure S1). In the tested range the extract contained more than 98% prodigiosin of both isomers α and β.

## Minimal Inhibitory Concentration Determination

Minimal inhibitory concentration (MIC) values were determined in 96-well microtiter plate assay according to CLSI standards (Clinical and Laboratory Standards Institute, 2009). Briefly, bacterial strains were grown overnight in LB medium at 28◦C or 37◦C. Inoculum was prepared by diluting overnight cultures in 0.9% NaCl to 0.5 McFarland (∼107 cells per mL). Wells of microtiter plates were filled with 0.1 mL of the appropriate concentration of the prodigiosin, ethanol or water and 0.01 mL of the inoculum to have the approximate number of cells in each inoculated well 105. The dilutions of prodigiosin, ethanol and water were made in LB medium. Prodigiosin was diluted in the final concentration range from 0.6 mg L−<sup>1</sup> to 367.1 mg L−1, while ethanol was diluted in the final concentration range from 0.02 to 12.03% (V/V). For positive controls LB medium was diluted with the same amounts of sterile distilled water as ethanol. Microtiter plates were incubated without shaking at 37◦C for 20 h and the optical density at 650 nm (OD650) was measured spectrophotometrically at the beginning and the end of incubation. The lowest prodigiosin concentration in the well with no growth after 20 h of incubation, was taken as the MIC value.

### Co-culture Experiments

*Bacillus subtilis* PS-216 *wt* was grown either in co-culture with *V. ruber* DSM14379 or in pure culture. For co-culture experiments, strains were separately grown in PYE medium with 3% (w/V) NaCl (*V. ruber*) or in LB medium (*B. subtilis*) until the stationary phase. Co-cultures were inoculated into PYE medium with 3% (w/V) NaCl with *V. ruber* DSM14379: *B. subtilis* PS-216 in the ratio 1:2 (V:V), the overall inoculum being 1% (V/V). CFU was determined at the beginning and 20 h after incubation at 28◦C and 200 rpm. CFU values were determined on PYE plates without NaCl (for *B. subtilis* PS-216 *wt*) and PYE plates with 3% (w/V) NaCl (for both strains). Due to the red pigmentation of *V. ruber* DSM14379 the colonies were clearly distinguished. Malthusian fitness was calculated based on CFU counts according to Ahn et al. (2006). Prodigiosin concentration in co-cultures was determined as described previously in Boric et al. (2011) ´ .

## Cell Treatment with Prodigiosin in Different Growth Phases

*Bacillus subtilis* PS-216 *wt* cells were cultured in liquid LB medium at 37◦C and 200 rpm overnight. Overnight culture was diluted 100 times in 20 mL of fresh LB medium and incubated further at the same growth conditions. OD650 was measured in regular intervals during the culture incubation. At appropriate OD650 in different growth phases [i.e., middle exponential (OD650 ∼ 0.7), late exponential (OD650 ∼ 1.1) and stationary phase OD650 ∼ 1.3], the strains were treated with 5.9 mg L−<sup>1</sup> of prodigiosin. As a control, an equivalent amount of sterile 96% (V/V) ethanol was added to the LB medium in the final concentration 0.19% (V/V).

### Treatment of Different Bacterial Strains with Prodigiosin

*Bacillus mycoides* DSM2048 cells were cultured in liquid LB medium at 28◦C and 200 rpm overnight, while *B. licheniformis* ATCC9445A cells were cultured in liquid LB medium at 37◦C and 200 rpm overnight. Overnight cultures were diluted 100 times in 20 mL of fresh LB medium and incubated further at the same growth conditions. Cultures were treated with prodigiosin (1.2 mg L−<sup>1</sup> *B. mycoides* or 6.9 mg L−<sup>1</sup> *B. licheniformis*) or ethanol as a control [0.04% (V/V) *B. mycoides* or 0.23 % (V/V) *B. licheniformis*] in the middle exponential phase at OD650 between 0.4 and 0.5. Treated cells were then incubated further and OD650 was measured in regular intervals during 6 h of incubation. To determine the influence of prodigiosin on different strains, the rate of lysis was calculated from measured decrease of OD650 after prodigiosin treatment.

### Cell Morphology and Viability

*Bacillus subtilis* PS-216 *wt* and *B. subtilis* PS-216 *amyE::*Phyperspank*-mKate2* strains were grown in LB medium as described above and treated with 5.9 mg L−<sup>1</sup> of prodigiosin or 0.19% (V/V) of ethanol in early exponential phase at OD650 between 0.4 and 0.5. Cell morphology and viability were inspected by Differential Interference Contrast (DIC) and fluorescence microscopy using appropriate filter set under the inverted microscope Axio Observer Z1 (Carl Zeiss, Germany) at 0, 1, and 5 h after the prodigiosin or ethanol treatment.

### Effect of Prodigiosin on Inactive Cells

*Bacillus subtilis* PS-216 *wt* cells were grown in LB medium as described above. When cells reached OD650 between 0.4 and 0.5, bacterial culture was autoclaved at 121◦C and 1.03 bar for 20 min to inactivate cells. After autoclaving, cell suspension was cooled down to room temperature and 5.9 mg L−<sup>1</sup> of prodigiosin was added. Cell suspension was incubated further at 37◦C and 200 rpm and OD650 was measured every 30 min for 5 h.

### Cytoplasmic Membrane Leakage

Cytoplasmic membrane leakage was determined as β-galactosidase activity according to the modified method by Mensa et al. (2011). Briefly, *B. subtilis* PS-216 *srfA-lacZ* (*neo)* cells were grown overnight in LB medium with appropriate antibiotic at 37◦C and 200 rpm and diluted 100 times in 20 mL of the fresh CM medium. Cells were incubated at 37◦C and 200 rpm until they reach OD650 ∼ 0.6. At that point 70 μL of cells were transferred in a microtiter plate and mixed with 20 μL of Z buffer (60 mM Na2HPO4, 40 mM NaH2PO4, 10 mM KCl, 10 mM MgSO4, pH 7) supplemented with 4 mg mL−<sup>1</sup> ONPG (*o*-nitrophenyl-β-galactoside) and 40 μM β-mercaptoethanol. Cells were then treated with 10 μL of prodigiosin to a final concentration of 2.9 mg L<sup>−</sup>1. Microtiter plates were incubated in a spectrometer for 2 h at 37◦C and absorbance at 420 nm was measured every 2 min to determine the rate of ONPG hydrolysis. The addition of 10 μL of 0.1% (w/V) sodium dodecyl sulfate (SDS) detergent [final concentration 0.01% (w/V)] represented a positive control, while the addition of 10 μL of 96% (V/V) ethanol [final concentration 9.6% (V/V)] represented a negative control. To calculate specific enzyme activity, the rate of ONPG hydrolysis was normalized to the cell biomass.

### Bacterial Response to Different Autolysin Inducing Agents

*Bacillus subtilis* PS-216 *wt* strain was grown in LB medium as described above and treated with 0.0125% (V/V) Triton X-100, 0.01% (w/V) SDS , 0.01% (w/V) cetyl trimethylammonium bromide (CTAB), 100 μg mL−<sup>1</sup> ampicillin or 100 μg mL−<sup>1</sup> kanamycin in the middle exponential phase at OD650 between 0.4 and 0.5. Ampicillin was used as a positive control known to induce cell lysis, whereas kanamycin was used as a negative control (Chung et al., 2009; Falk et al., 2010; Lacriola et al., 2013). Cell suspension was incubated further at 37◦C and 200 rpm and OD650 was measured every 30 min for 5 h. To determine the impact of different detergents on bacterial strain, the rate of lysis was calculated from measured decrease of OD650 after treatment with antimicrobial agents.

### Autolytic Response of Bacterial Strains to Prodigiosin

*Bacillus subtilis* strains PS-216 *wt*, PS-216 *lytABC*, PS-216 *lytABCD*, PS-216 *lytABCDE* and PS-216 *lytABCDEF* were grown in LB medium as described above and treated with 5.9 mg L−<sup>1</sup> of prodigiosin or 0.19 % (V/V) of ethanol as a negative control at OD650 between 0.4 and 0.7. Cells were then incubated further and OD650 was measured in regular intervals during 6 h of incubation at 37◦C and 200 rpm. To determine the influence of prodigiosin on different mutant strains, the rate of lysis was calculated from measured decrease of OD650 after prodigiosin treatment.

### Modified Comet Assay

*Bacillus subtilis* strains PS-216 *wt*, PS-216 *lytABC*, PS-216 *lytABCD*, PS-216 *lytABCDE* and PS-216 *lytABCDEF* were grown in LB medium as described above. At OD650 between 0.4 and 0.6 cells were treated with 5.9 mg L−<sup>1</sup> of prodigiosin or 0.19% (V/V) of ethanol. Cells were then incubated further for 1 h at 37◦C and 200 rpm to lyse. Minigels were prepared on frostedend microscopic slides. Slides were first dipped into 1% (w/V) normal melting point agarose (NMP) and one side was wiped clean. After drying on a hot plate (1 min, 100◦C), the second layer of 300 μL 1% (w/V) NMP agarose was applied and solidified at 4◦C for 5 min. For the third layer treated or non-treated cells were resuspended in 1% (w/V) low melting agarose (ratio 1:10) and left to solidify for 5 min. Minigels were immediately electrophoresed in TBE (Tris/borate/EDTA buffer, pH 8.3) at 1 V/cm and run for 30 min. Afterward minigels were neutralized in 400 mM Tris buffer (pH 7.4). Gels were stained with 30x solution of GelRedTM (Biotium, USA) and observed under epifluorescent microscope (Olympus BX-50, Japan) equipped with EMCCD camera (LucaEMr, Andor Technology Ltd., United Kingdom) and the appropriate filter set. Images were captured by Komet 7.0 software (Andor Technology Ltd., UK).

### Data Analysis

All the values presented are averages with standard deviations or standard errors. Results were statistically evaluated using oneway ANOVA. Samples with *p*-values equal or lower than 0.05 were taken as statistically significant.

## RESULTS AND DISCUSSION

Prodigiosin isolated from *V. ruber* DSM14379 inhibits growth of different bacterial species. The MIC value for *Escherichia coli* is 103.4 <sup>±</sup> 6.3 mg L<sup>−</sup>1, while different *Bacillus* sp. strains tested have significantly lower MIC values. The MIC values for different *Bacillus* sp. strains were in the range between 5 and 7 mg L−<sup>1</sup> (MIC values for *B. amyloliquefaciens* FZB42 was

FIGURE 1 | The effect of prodigiosin addition on *Bacillus subtilis* PS-216 *wt* growth in LB medium at 37**◦**C and 200 rpm at different growth phases. The arrows represent the time of addition of 5.9 mg L−<sup>1</sup> prodigiosin (opened squares) or 0.19% (V/V) ethanol in the control samples (filled squares). (A) cells treated in the middle of the exponential phase; (B) cells treated in the late exponential phase; (C) cells treated in the stationary phase. Data are presented as averages and standard deviations (*n* = 3).

6.1 ± 1.2 mg L−1, *B. licheniformis* ATTC9445A was 6.9 ± 2.2 mg L<sup>−</sup>1, *B. mycoides* was 1.2 mg L−1, *B. subtilis* NCIB3610 was 5.2 <sup>±</sup> 1.4 mg L−1, *B. subtilis* ATCC6051 was 5.9 mg L−<sup>1</sup> and *B. subtilis* PS-216 was 5.9 mg L<sup>−</sup>1) and were lower as previously reported (Stankovic et al., 2012; Priya et al., 2013). Approximately the same MIC values for different *Bacillus* species may suggest that prodigiosin has the same mode of action on all tested *Bacillus sp.* Only undomesticated natural isolate *B. subtilis* PS-216 *wt* was used for further analysis with MIC value 5.9 mg L<sup>−</sup>1. To test the *in vivo* activity of prodigiosin on *B. subtilis* PS-216 *wt* simple co-cultures with prodigiosin producing *V. ruber* DSM14379 were made. There was a significant decrease in the number of *B. subtilis* cells after incubation in co-culture. The Malthusian fitness of *B. subtilis* grown in a monoculture was 5.7 ± 0.2, but was significantly lower −6.8 ± 1 in co-culture experiments. The concentration of prodigiosin in co-culture at the end of the experiment was 16.3 <sup>±</sup> 0.4 mg L−<sup>1</sup> which is 2.7 fold higher than the MIC required for growth inhibition of *B. subtilis* PS-216 *wt*. The results indicate that prodigiosin has *in vivo* antimicrobial activity.

The effect of prodigiosin on *B. subtilis* was dependent on the growth phase as shown in **Figure 1**. Prodigiosin treatment in the middle of the exponential phase (**Figure 1A**) caused instantaneous termination of growth and immediate decrease of optical density. Cell lysis was completed in 2 h after prodigiosin addition. As shown in **Figure 2** cell number significantly

decreased after prodigiosin treatment. After 5 h of incubation with prodigiosin the bacterial solution was clear and only a few bacteria were observed. Cells treated with prodigiosin in the late exponential phase (**Figure 1B**) lysed as well. However, the rate of lysis was three times lower than in the middle of the exponential phase. After 5 h of prodigiosin treatment, optical density decreased for approximately 50%. When cells were treated with prodigiosin in the stationary phase (**Figure 1C**) their growth ceased, but cells did not lyse. When those cells were re-inoculated into a fresh medium and incubated at optimal growth conditions, they resumed grow immediately. This implies that the mode of action of prodigiosin on *B. subtilis* cells depends on the growth phase. During the exponential growth prodigiosin has a bacteriolytic activity, whereas in the stationary phase it has bacteriostatic activity. This finding is not surprising, as it is generally accepted that effectiveness of antibiotic is greatly dependent on bacterial growth rate (Hobby et al., 1942; Greulich et al., 2015). For example, β-lactam antibiotics that target cell wall have their maximum effect during fast bacterial growth and are not effective toward stationary cells (Eng et al., 1991). The requirement of cell growth for prodigiosin susceptibilty was further tested with inactivated cells from the middle of the exponential phase. There was no significant change in optical density during 5 h of the incubation with prodigiosin, suggesting that the inactive cells did not lyse (data not shown). After the addition of the control containing ethanol (**Figure 1**), the growth of *B. subtilis* continued. Although the culture was not significantly diluted (the total volume increased only by 0.2%), the added ethanol may have had a positive effect on the growth of *B. subtilis* as demonstrated previously (Gomaa, 2014).

For cell lysis one expects both cell wall and membrane disintegration. Prodigiosin is a small molecule that can diffuse through the cell wall and due to its amphiphilic character it can incorporate in the membrane (Diaz et al., 2007). Membrane loaded with prodigiosin may be compromised and allow leakage of cell constituents. To test cytoplasmic membrane permeability ONPG hydrolysis by β-galactosidase was used. ONPG can enter the cell and reach cytoplasmatic β-galactosidase only if cytoplasmic membrane is compromised. As given in **Figure 3** prodigiosin was able to permeabilize cytoplasmic membrane. Similar permeabilization was observed with SDS detergent, which is a known cytoplasmic membrane solubilization agent in *B. subtilis* and *Streptococcus faecalis* (Bishop et al., 1967; Cornett and Shockman, 1978). There were no statistically significant differences between prodigiosin and SDS treatment (*p* = 0.4). The results indicate that prodigiosin interacts with cytoplasmic membrane and increases its permeability. Increased cytoplasmic permeability was associated with autolysis in *B. subtilis* (Falk et al., 2007). It has been shown that proton-motive force influences regulation of autolysins in *B. subtilis* cells (Calamita and Doyle, 2002). Prodigiosin functions in an energy-spilling process as a tightly regulated uncoupler of proton transport and ATP synthesis by oxidative phosphorylation (Haddix et al., 2008).

When cells were treated with prodigiosin in the middle of the exponential phase, approximately half of the cells lysed in 45 min. This lytic rate of prodigiosin (0.37 ± 0.01 h−1) is comparable to cell lysis induced by Triton X-100 (0.36 <sup>±</sup> 0.03 h<sup>−</sup>1) and ampicillin (0.49 ± 0.06 h−1), which is a known autolysin induction agent (Lacriola et al., 2013). On the other hand, the other known agents that induce autolysins (i.e., SDS and CTAB) had lower lytic rates. SDS had lytic rate 0.13 <sup>±</sup> 0.01 h<sup>−</sup>1, while CTAB had lytic rate 0.15 <sup>±</sup> 0.01 h−1. Kanamycin known as a non-autolysin agent did not lyse *B. subtilis*, the lytic rate was – 0.12 <sup>±</sup> 0.03 h<sup>−</sup>1. The results indicate that there was a significant difference between lytic rate of prodigiosin and SDS, which is different from the <sup>β</sup>-galactosidase results (**Figure 3**). Whereas β-galactosidase release assay shows the cytopasmic membrane permeabilisation, the lytic rate shows autolysin induction. The

two are not necessarily coupled. For example, SDS is a typical solubilizing agent of cytoplasmic membrane proteins in *B. subtilis* (Bishop et al., 1967), but is usually not used as an autolysin inducer due to its low lytic effectiveness. It has been demonstrated in *Enterobacteriacea* that the SDS-grown cells underwent rapid lysis only when they ran out of energy (Kramer et al., 1980). The energy dependence reflects a requirement for ATP rather than for a proton gradient or membrane potential (Aspedon and Nickerson, 1994). Prodigiosin, on the other hand, acts as a potent lytic agent for *B. subtilis* that uncouples proton transport in the cytoplasmic membrane (Haddix et al., 2008), thereby activating autolysins. Prodigiosin lytic activity was not limited to *B. subtilis* cells only. For example, treating *B. mycoides* and *B. licheniformis* with prodigiosin at MIC concentrations induces cell lysis and the cell suspension cleared after 20 h of treatment. The corresponding lytic rates were 0.014 <sup>±</sup> 0.002 h−<sup>1</sup> for *B. mycoides* and 0.009 ± 0.001 h−<sup>1</sup> for *B. licheniformis*, which indicates that prodigiosin lytic activity could be a more general mechanism in *Bacillus* species.

To prove that prodigiosin induces autolysis several mutant strains of *B. subtilis* PS-216 deficient in major autolysis genes (*lyt* genes) were produced. The lytic rates after treatment with prodigiosin of mutant strains deficient to a different extent in autolytic response are given in **Figure 4**. In the *lytABC* deficient strain the cell lysis was significantly reduced and the lytic rate was approximately two fold lower compared to the wild type strain. When additionally *lytD* gen was deleted a further decrease in the lytic rate was observed. The *lytABCD* deficient strain

showed approximately 80% reduction of lysis compared to the wild type strain. This is consistent with previous reports of a key role of LytC and LytD in autolysis (Smith et al., 2000). In addition, it has been shown previously that *lytE* mutant may significantly reduce autolysis (Margot et al., 1998). Consistently in *lytABCDE* mutant there was almost no lysis after prodigiosin addition. Finally, in *lytABCDEF* mutant the lytic rate upon addition of prodigiosin was 50 times lower compared to the wild type strain. Although no lytic response was observed in *lytABCDEF* mutant the addition of prodigiosin in the middle of exponential phase had a bacteriostatic effect. The results obtained by different *lyt* deficient strains strongly suggest that prodigiosin induces autolysin dependent bacteriolysis in *B. subtilis*. For a complete lysis of *B. subtilis* by prodigiosin a complement of different autolysins is needed. Since autolysins can be induced also by other amphiphatic molecules such as short- and medium– chain fatty acids (Tsuchido et al., 1985) this further supports the hypothesis that prodigiosin is a potent autolysin modulator.

The release of nucleic material during cell lysis was followed with modified comet assay (**Figure 5**). Non-treated cells retained cell integrity through the experiment, displaying clearly defined shape without tail or halo (**Figure 5A**). Similary, in control cultures treated with 0.19% (V/V) ethanol, the same concentration as used for prodigiosin solubilization, cells were intact (**Figure 5B**). On the other hand, treatment with prodigiosin induced instant lysis in almost all examined cells (**Figure 5C**). Cells that lysed appeared blurred with rapid decline of fluorescence indicating DNA diffusing out of disrupted cells. The comet assay was also applied to *B. subtilis* mutant strains deficient in major autolysin genes. *Lyt* mutants treated with prodigiosin responded differently. The comet assay suggests that lysis was most intense in *lytABC* mutant. The degree of cell lysis decreased from *lytABC* to *lytABCDEF* deficient strain (**Figure 6**). Negligible cell lysis was observed in *lytABCDEF* mutant strain after prodigiosin treatment. The results obtained with the modified comet assay confirm that prodigiosin induces rapid autolysis in *B. subtilis* cells.

Finally, we note that upon entry into the stationary phase prodigiosin acts as a bacteriostatic agent. During transition to the stationary phase *B. subtilis* cells begin to lose their ability to move and start to produce biofilm. Biofilms and motility are mutually exclusive lifestyles of *B. subtilis* (Diethmaier et al., 2011) and are controlled by a common regulator, the transcription factor SinR. During biofilm formation SinR interacts with SinI and derepresses synthesis of extracellular matrix important in biofilm formation. One of the components of extracellular matrix EpsE interacts with flagellar motor switch protein FliG and prevents rotation of the flagellum (Blair et al., 2008). In addition, upon the entry into a stationary phase cells shut down synthesis of flagella components. The synthesis of flagella and autolysin are under the same regulation control. The synthesis of flagella proteins requires genes which are under the control of alternative RNA polymerase sigma factor, σ<sup>D</sup> (Shahbabian et al., 2009). σ<sup>D</sup> is a pleiotrophic transcription regulator during the exponential growth and has a positive effect on autolysin synthesis (Fein, 1979). Upon entry into the stationary phase σ<sup>H</sup> decreases the expression of σ<sup>D</sup> (Britton et al., 2002), which in turn downregulates autolysin production. In the absence of autolysins, as demonstrated in this work, the lytic activity of prodigiosin is abolished. However, as shown with *lytABCDEF* mutant prodigiosin may still act as a bacteriostatic agent preventing further growth of *B. subtilis*.

### CONCLUSION

The application of prodigiosin as an antibacterial agent is hampered by the lack of knowledge of its molecular targets. The results of this study demonstrate that prodigiosin exhibits a potent antimicrobial activity against *B. subtilis*. It acts as a bacteriolytic agent during the exponential growth and as a bacteriostatic agent during the stationary growth. Prodigiosin interferes with cytoplasmic membrane function and increases its permeability. The obtained results indicate that prodigiosin's

bacteriolytic activity is due to the induction of autolysins. Prodigiosin proves to be a strong autolysin inducer in *B. subtilis* cells comparable in its activity to Triton X-100. Since prodigiosin functions as an uncoupler of proton transport and ATP synthesis, it is ideally suited for autolysins induction and killing of different *Bacillus* species.

### AUTHOR CONTRIBUTIONS

Conceived and designed the experiments: TD, MV, and DS. Performed the experiments: TD, MV, MT, and MZ. Analyzed the data: TD, MV, MT, and MZ. Wrote the paper: TD, MV and DS.

### REFERENCES


### ACKNOWLEDGMENTS

We acknowledge financial support from the Slovenian Research Agency (ARRS no. 1000-09-310208). This work was also supported by the Slovenian Research Agency grant P4-0116. We are thankful to Polonca Stefanic for providing *B. subtilis* PS-216 *wt* and PS-216 *srfA-lacZ* (*neo*) strains.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.00027


causative agent of red spot disease of *Laminaria japonica*. *Int. J. Syst. Bacteriol.* 48, 769–774. doi: 10.1099/00207713-48-3-769


**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 Danevˇciˇc, Bori´c Vezjak, Tabor, Zorec and Stopar. 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.*

# **Bioengineering Lantibiotics for Therapeutic Success**

*Des Field <sup>1</sup> \*, Paul D. Cotter 2,3, Colin Hill 1,3 and R. P. Ross 2,3*

*<sup>1</sup> School of Microbiology, University College Cork, Cork, Ireland, <sup>2</sup> Teagasc Food Research Centre, Fermoy, Ireland, <sup>3</sup> APC Microbiome Institute, University College Cork, Cork, Ireland*

Several examples of highly modified antimicrobial peptides have been described. While many such peptides are non-ribosomally synthesized, ribosomally synthesized equivalents are being discovered with increased frequency. Of the latter group, the lantibiotics continue to attract most attention. In the present review, we discuss the implementation of *in vivo* and *in vitro* engineering systems to alter, and even enhance, the antimicrobial activity, antibacterial spectrum and physico-chemical properties, including heat stability, solubility, diffusion and protease resistance, of these compounds. Additionally, we discuss the potential applications of these lantibiotics for use as therapeutics.

**Keywords: antimicrobial peptide, nisin, mutagenesis, lantibiotic, post-translational modification, bacteriocin, bacterial resistance**

## **INTRODUCTION**

#### *Edited by:*

*Enrica Pessione, University of Torino, Italy*

#### *Reviewed by:*

*Dimitris Tsaltas, Cyprus University of Technology, Cyprus Shuji Tani, Osaka Prefecture University, Japan*

> *\*Correspondence: Des Field des.field@ucc.ie*

#### *Specialty section:*

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

*Received: 29 September 2015 Accepted: 17 November 2015 Published: 27 November 2015*

#### *Citation:*

*Field D, Cotter PD, Hill C and Ross RP (2015) Bioengineering Lantibiotics for Therapeutic Success. Front. Microbiol. 6:1363. doi: 10.3389/fmicb.2015.01363* Given that antibiotic resistance has now reached a crisis point, novel compounds and innovative methods are urgently required to arrest the spread and development of drug-resistant infections in both the nosocomial and community environments. Ideally, such novel substances should exhibit distinctly different mechanisms of action to currently used chemotherapeutics in order to decrease resistance development. Ribosomally synthesized antimicrobial peptides produced by bacteria (bacteriocins) constitute an emerging class of natural products that have attracted considerable interest as promising alternatives to existing antibiotics (Sahl and Bierbaum, 2008). Within this diverse group of peptides, the lantibiotics, i.e., class I bacteriocins which contain the posttranslationally modified amino acids lanthionine and methyllanthionine, have become the focus of many biomedical and pharmaceutical research groups due to their demonstrable high potency *in vitro*, multiple modes of action and ability to destroy target cells rapidly (Cotter et al., 2005; Cavera et al., 2015). In general, lantibiotics exhibit activity against Gram positive bacteria. Importantly, this includes many drug resistant targets including methicillin resistant *Staphylococcus aureus* (MRSA), vancomycin intermediate *S. aureus* (VISA), vancomycin resistant enterococci (VRE), *Streptococcus pneumoniae* and *Clostridium difficile*, amongst others (Cotter et al., 2013). Furthermore, several lantibiotic peptides have demonstrated excellent *in vivo* activities and have progressed toward clinical evaluation for the treatment of life-threatening diseases (Dawson and Scott, 2012; Sandiford, 2015). Indeed, these and a range of other desirable features make them suitable for use in human and veterinary medicine and also in the pharmaceutical industry (Dischinger et al., 2014). However, despite these promising attributes, there are a number of limitations that has prevented their more widespread use, including instability and/or insolubility at physiological pH, low production levels and susceptibility to proteolytic digestion. The implementation of multiple technologies, including genome mining as well as high-throughput screening strategies in combination with *in vivo* and *in vitro* expression systems has provided a wealth of information relating to the widespread existence, structural diversity and functionality of lantibiotics while facilitating the identification

**FIGURE 1 | Representative structures of various single and two-component lantibiotics.** Post translational modifications are indicated as follows: Abu: 2-aminobutyric acid, Ala-S-Ala: lanthionine, Abu-S-Ala: 3-methyllanthionine, Dha: dehydroalanine, Dhb: dehydrobutyrine. D-Ala residues (in lacticin 3147) are shaded gray.

of structural regions that can be targeted to enhance their biological and physicochemical properties. The present review will focus on recent developments with regard to these achievements.

### **LANTIBIOTICS: THE CASE FOR THERAPEUTIC USE (***IN VITRO* **AND** *IN VIVO* **POTENCY)**

New antimicrobials that possess novel modes of action, particularly against drug resistant organisms so that they can be specifically targeted for clinical applications, are required as a matter of urgency. In this regard lantibiotics hold considerable potential as a consequence of their unusual structure, unique mechanisms of action and their potency against multi-drug resistant bacteria. Today, close to 100 of these bioactive peptides have been described, the majority of which are produced by Gram-positive bacteria (Dischinger et al., 2014). The common feature that links all lantibiotics is the presence of a number of distinctive amino acids which result from enzymatically mediated post-translational modifications, including dehydration and cyclisation, leading to the formation of the eponymous (methyl)lanthionine bridges. These bridges convert the linear peptide chain into a polycyclic form giving structure and function to the peptide. It should be noted that only those peptides that display antimicrobial activity within the larger family of lanthionine-containing peptides or lanthipeptides are termed lantibiotics.

Many lantibiotics exert their antimicrobial action through complexation with lipid II, an essential precursor of the bacterial cell wall, either by inhibiting cell wall synthesis through sequestration of lipid II and/or by disruption of membrane integrity and pore formation (Breukink and de Kruijff, 2006). Indeed, the prototypical and best studied lantibiotic nisin performs both of these functions as a consequence of two distinct structural domains located at the N- and C-termini (**Figure 1**). It has been established that the A, B, and C rings form a "cage-like" enclosure that facilitates binding of the pyrophosphate moiety of lipid II, thus inhibiting cell wall synthesis (Hsu et al., 2004) This binding enhances the ability of the C-terminal segment, containing rings D and E, to form pores in the cell membrane, resulting in the rapid efflux of ions and cytoplasmic solutes (Wiedemann et al., 2001). This mechanism of action is not common to all lantibiotics, and some of them lack the



ability to elicit pores or to bind lipid II or both, but can still exhibit antimicrobial activity (Pag and Sahl, 2002). The poor activity of lantibiotics toward Gram negative bacteria is due to the outer membrane (OM) of the Gram negative cell wall which acts as a barrier for the cell, restricting the access of the peptides to the cytoplasmic membrane (Nikaido and Vaara, 1985).

Lantibiotics have been classified on the basis of their biosynthetic pathways (Willey and van der Donk, 2007). According to this scheme, class I lantibiotics are those modified by two separate enzymes, a LanB (dehydratase) and LanC (cyclase); class II are modified by a single LanM enzyme with both dehydratase and cyclase activity. The third and fourth classes of lanthipeptides are also modified by a single enzyme (general nomenclature LanKC for class III and LanL for class IV; van der Donk and Nair, 2014). Most of the class III lanthipeptides reported thus far have no or weak antimicrobial activities, but some have been shown to possess anti-allodynic/antinociceptive activity (Meindl et al., 2010; Iorio et al., 2014), antiviral activity (Férir et al., 2013) or morphogenetic activities (Willey and Gaskell, 2011). The designation Lan is used generically to refer to proteins associated with the biosynthesis of, or immunity to, lantibiotics. A typical lantibiotic operon will also contain genes encoding enzymes to carry out transport/processing (LanT), immunity (LanI and LanFEG), proteolytic processing (LanP) as well as the structural gene (LanA). Other enzymes, responsible for the formation of less common residues may also be present. Importantly, individual components of the lantibiotic biosynthetic machinery show even greater flexibility as demonstrated by their activity *in vitro* (Li et al., 2009).

## **ACTIVITY OF LANTIBIOTICS** *IN VITRO*

Although lantibiotics such as nisin have been in use for decades as safe and natural food preservatives (Delves-Broughton, 2005), the continued escalation of multi-drug resistant bacterial infections has led to a re-appraisal of their capacity for use against lifethreatening infections. A multitude of studies have highlighted the *in vitro* potency of lantibiotics against nosocomial pathogens (the reader is directed to a comprehensive review: Piper et al., 2009a). Many lantibiotics, including lacticin 3147, mutacins B-Ny266 and 1140, nisin, mersacidin, epidermin, Pep5, and planosporicin exhibit activity against clinically-relevant targets (**Table 1**) such as MRSA, VRE, *Propionibacterium acne*, *Streptococcus mutans*, *Streptococcus pyogenes*, *S. pneumoniae*, *C. difficile*, *Listeria*, and *Bacillus* species (Severina et al., 1998; Galvin et al., 1999; Mota-Meira et al., 2000; Brumfitt et al., 2002; Rea et al., 2007; Ghobrial et al., 2009; Piper et al., 2009b). Notably, both Pep 5 and epidermin successfully inhibit the adhesion of staphylococcal cells to the surfaces of siliconized catheters (Fontana et al., 2006). Although it is the general view that lantibiotics exhibit less potential as chemotherapeutics to combat infections with Gramnegative organisms, lantibiotics including mutacin B-Ny266 are selectively active against a few strains of *Neisseria* and *Helicobacter*

(Mota-Meira et al., 2000), while purified nisin displays activity against *Escherichia coli* (Kuwano et al., 2005).

Nisin has also been shown to effectively inhibit spore outgrowth including spores of *Bacillus anthracis*(Gut et al., 2008) and those of *C. difficile* (Nerandzic and Donskey, 2010). Additionally, studies have revealed that the lantibiotic gallidermin efficiently prevents biofilm formation in both *S. aureus* and *S. epidermidis* species (Saising et al., 2012).

Recently, an intriguing and novel (sub-)class of lantibiotics termed pinensins were found to be highly active against many filamentous fungi and yeasts but displayed only weak antibacterial activity. Not only do pinensin A and pinensin B represent the first examples of a lantibiotic fungicide, they are also the first lantibiotics to be isolated from a Gram-negative native producer (Mohr et al., 2015).

## **LANTIBIOTICS DEMONSTRATE** *IN VIVO* **POTENCY**

While the *in vitro* success of a chemotherapeutic agent does not always necessarily translate to *in vivo* efficacy, there have been a number of encouraging studies to suggest that this may not be a major shortcoming of lantibiotics. For instance, mutacin B-Ny266 was shown to be as active as vancomycin against MRSA *in vivo* (Mota-Meira et al., 2005), mersacidin was able to effectively eradicate an MRSA infection in a mouse rhinitis model (Kruszewska et al., 2004) and Nisin F, a natural nisin variant, was also found to successfully control *S. aureus* infection in rats (De Kwaadsteniet et al., 2009). Similarly, microbsporicin (**Figure 1**) (NAI-107) was evaluated for its therapeutic potential in nosocomial infection and demonstrated efficacy against MRSA in a rat endocarditis model (Jabes et al., 2011). The efficacy of MU1140 (mutacin 1140) has also been investigated *in vivo* (Ghobrial et al., 2009) and is currently in pre-clinical development for the treatment of Gram positive infections. NVB302, a derivative of deoxyactagardine B, is currently undergoing phase I clinical trials as a therapeutic for the treatment of *C. difficile* infections due to its selective targeting of this organism over the predominantly Gram negative normal gut flora (Dawson and Scott, 2012). Investigations into the use of lantibiotics to control the microorganisms responsible for dental plaque, halitosis and "sore throat" infections have also yielded promising results (Hillman, 2002; Burton et al., 2006; Dierksen et al., 2007).

Some lantibiotics possess additional bioactivities that hold promise for therapeutic application. A smaller subcategory of lantibiotics, such as cinnamycin and duramycin, have been found to influence eukaryotic metabolic functions by binding phosphatidylethanolamine in cell membranes and, in turn, inhibiting the enzyme phospholipase A2 (Marki et al., 1991). In addition to this activity, duramycin demonstrated efficacy in the treatment of cystic fibrosis by inhalation (Grasemann et al., 2007) as a result of its ability to stimulate chloride secretion in bronchial epithelial cells (Oliynyk et al., 2010).

Remarkably, the first instance of a lantibiotic, or indeed any bacteriocin, to prevent the growth of cancer cells has been confirmed. In a study by Joo and coworkers, nisin Z was shown to be effective in the treatment of head and neck squamous cell carcinoma (HNSCC; Joo et al., 2012). In subsequent mouse trials involving a highly purified form of nisin Z, reduced tumorigenesis *in vivo* was observed and long-term treatment with nisin Z extended survival. In addition, nisin treated mice exhibited normal organ histology with no evidence of inflammation, fibrosis or necrosis (Kamarajan et al., 2015).

## **BIOENGINEERING AND SYNTHETIC BIOLOGY- GENERATING MORE EFFECTIVE LANTIBIOTICS**

Bioengineering (engineering within the cell) and the use of synthetic biology-based (*in vitro* engineering) approaches have been important for advancing our understanding of the fundamentals of bacteriocin activity and structure–function relationships (these approaches are the subject of a number of recent comprehensive reviews: Tabor, 2014; Escano and Smith, 2015). However, there is also a steadily growing number of engineered lantibiotic peptides that demonstrate enhanced functionalities (activity and/or stability) which make them more attractive from a clinical perspective (Cotter et al., 2013). The following provides some recent examples of bioengineered lantibiotics exhibiting enhanced pharmacological and physicochemical properties as well as developments in genetic systems to increase peptide yields.

Several bioengineered variants of the prototypical lantibiotic nisin have been generated that provide excellent examples of how lantibiotic functionality can be modulated by as little as one residue change. The nisin Z derivatives N20K, M21K, N27K, H31K generated by protein engineering displayed improved solubility, particularly at alkaline pH values where the solubility of the parent nisin is particularly reduced (Rollema et al., 1995; Yuan et al., 2004). Furthermore, the consequences of effecting single residue alterations at distinct locations in nisin has generated variants that exhibit not only improved antimicrobial activity against strains of clinical relevance (MRSA, VRE, VISA, MRSP, and *C. difficile*) but has also brought about the widening of its antimicrobial spectrum to include some Gram negative bacteria (Field et al., 2008, 2012, 2015; Molloy et al., 2013). More dramatic substitutions at the location of rings A and B at the N-terminal end of nisin A revealed that the various activities of nisin can be altered by changing the amino acid arrangement in this region of the peptide (Rink et al., 2007). The hinge-region of nisin has also been the subject of mutagenesis resulting in variants with enhanced antimicrobial activity (Field et al., 2008; Healy et al., 2013) as well as derivatives with an enhanced ability to diffuse through complex polymers (Rouse et al., 2012). In both mutacin 1140 and nukacin ISK-1 peptides, single residue changes brought about a significant increase in activity against several Gram positive strains(Islam et al., 2009; Chen et al., 2013). Similarly, mutagenesis of the mersacidin gene was ultimately successful in that several variants were identified which exhibited enhanced activity against a range of different targets including clinically relevant MRSA, VRE and *S. pneumoniae* (Appleyard et al., 2009).

Generating enhanced variants of two-component lantibiotics presents an even greater challenge given that two peptides are required to work jointly in synergy. However, a lacticin 3147 derivative with enhanced activity against a pathogenic strain of *S. aureus* was recently identified (Field et al., 2013), the first occasion such an increase in antibacterial properties has been observed for bioengineered two-component lantibiotics.

Synthetic biology approaches are another promising means to provide insights into structure-stability relationships and generate novel derivatives with improved function. Chemical synthesis enables the limitations of the modification machinery to be bypassed, extending the range of analogs that can be produced. For example, deoxyactagardine B is a single peptide lantibiotic that is rigid, compact and globular and differs from actagardine (**Figure 1**) by two amino acids and the absence of a sulfoxide bond (Boakes et al., 2010). A synthetically introduced C-terminal modification (1, 7 diaminoheptane) produced a variant, NVB302, that displayed greater solubility and activity compared to the parent molecule. NVB302 is now in phase I clinical trials for the treatment of *C. difficile* infections (Dawson and Scott, 2012). It has also been established that lantibiotics are susceptible to oxidation of the sulfur-containing lanthionine and this can lead to sharp decreases in antimicrobial activity. In the case of lactocin S, lanthionines were replaced with diaminopimelate to produce several analogs, one of which revealed greater stability whilst still retaining 100% biological activity (Ross et al., 2012).

Chemical synthesis methods were employed to produce enhanced analogs of the lantibiotic epilancin 15X (Knerr and van der Donk, 2012). A novel approach termed *in vitro* mutasynthesis has produced improved variants of the class II lantibiotic lacticin 481. Here, non-standard amino acids were introduced into the structural peptide by organic synthesis, and subsequently modified *in vitro* with purified LctM to generate derivatives with superior specific activity against a target strain (Levengood et al., 2009). Notably, synthetic chemistry approaches were employed to generate hybrids of nisin and vancomycin that demonstrated a 40-fold increase in potency compared to each of the components separately (Arnusch et al., 2008). Similarly, the nisin N-terminus (1–12) was synthetically modified by the coupling of simple membrane-active lipids to create biologically active and proteolytically stable hybrids (Koopmans et al., 2015).

Regardless of these bioengineering successes, one concern that remains to be tackled is that of production. Indeed, the discovery, study and application of lantibiotics is often compromised by limited, or the absence of, production of these peptides by the native producer, a problem which is further compounded when working with bioengineered derivatives. However, a number of instances have demonstrated that quite the opposite effect can be achieved in terms of production. In the case of mutacin 1140 and nukacin ISK-1, single residue alterations did not increase specific activity but instead increased peptide production by up to fourfold (Islam et al., 2009; Chen et al., 2013). Importantly, a recent study involving synthetic biology approaches describes the development of a genetic system that facilitates significant overproduction of nisin (Kong and Lu, 2014). Although heterologous expression of lantibiotic peptides (and their bioengineered derivatives) has been demonstrated in the Gram negative host *E. coli* on several occasions (Nagao et al., 2007; Caetano et al., 2011, 2014; Shi et al., 2012; Basi-Chipalu et al., 2015), a recent study describes a multigene assembly strategy for the overexpression of the two-component lantibiotic lichenicidin in *E. coli* (Kuthning et al., 2015). Such systems may also help in attaining higher yields to simplify isolation of and improve cost-efficiency of novel derivatives that are often compromised by limited production.

A major drawback that has yet to be overcome with respect to therapeutic use is the sensitivity of lantibiotics to proteolytic cleavage by intestinal enzymes. For example, nisin, pep5 and epidermin have been shown to be susceptible to the proteases trypsin and chymotrypsin (Jarvis and Mahoney, 1969; Bierbaum et al., 1996). Bioengineering strategies could be employed to replace the residues that serve as recognition sites by these and other digestive enzymes and potentially overcome the issue of vulnerability to proteolytic breakdown in the gastrointestinal tract. Indeed, the recent discovery of the class II lantibiotic pseudomycoicidin (which was found to be resistant to trypsin) provides the perfect example for this approach. A trypsin cleavage site which is located in the conserved lipid II binding motif, is protected by the presence of at least one thioether ring structure. This was confirmed by experiments with site-directed mutant peptides where the removal of thioether forming Cys residues resulted in the establishment of protease sensitivity (Basi-Chipalu et al., 2015).

Lastly, it should be remarked that the efficacy of individual lantibiotics could be further boosted through combination with other antimicrobials or membrane-active substances. For example, nisin displayed synergistic activity with the antibiotics colistin and clarithromycin against *Pseudomonas aeruginosa* (Giacometti et al., 2000) and with ramoplanin and other nonβ-lactam antibiotics against many strains of MRSA and VRE (Brumfitt et al., 2002). Similarly, nisin-ceftriaxone and nisincefotaxime were found to be highly synergistic against clinical isolates of *Salmonella enterica* serovar Typhimurium as evident by checkerboard test and time-kill assay (Rishi et al., 2014).

### **CONCLUSION**

Lantibiotics possess many of the attributes essential for the treatment of infections caused by multi-drug resistant bacteria and their potential for use as alternatives to traditional antibiotic therapies has been mooted for decades. While greater than 100 lantibiotic peptides have been described, not all of these have been characterized in great depth and so many may possess traits of commercial value. Indeed, as the number of microbial genome sequences has increased dramatically, an even larger collection of new lantibiotic biosynthetic gene clusters has been revealed. These clusters can be applied directly or, the information gained from their analysis, can be used indirectly to guide the bioengineering of new and existing peptide structures.

Finally, although nisin remains the only lantibiotic that is extensively exploited, its full use as a therapeutic entity has not yet been fulfilled, in part due to its low solubility and stability at physiological pH. It is thus notable that a broad range of technologies have been developed for the engineering of lantibiotics and the past decade has seen several bioengineering studies describe the generation of peptide derivatives including nisin with enhanced functionality in terms of specific activity, spectrum of activity, solubility and/or temperature and pH stability. Critically, genetic systems are in continuous development to increase yields of peptide that may aid commercial viability. The further application of these systems to enhance nisin and other lantibiotics has the potential to lead to the development of novel derivatives for therapeutic use. Additionally, bioengineering in combination with semi-synthesis will expand structural diversity still further. It is thus likely that these peptides will be only the first of many generations of bioengineered lantibiotic and lantibiotic-like peptides. Given these recent developments and the fact that several lantibiotics are currently in clinical and preclinical trials reinforces our belief that bioengineered lantibiotics can contribute to a solution to antibiotic resistance across a broad range of bacterial pathogens.

### **REFERENCES**


### **AUTHOR CONTRIBUTIONS**

DF drafted the manuscript. PC, CH, and RR revised and approved the final manuscript.

### **ACKNOWLEDGMENTS**

DF, CH, PC, RR are supported by the Irish Government under the National Development Plan, through a Science Foundation Ireland (SFI) Technology and Innovation Development Award (TIDA 14/TIDA/2286) to DF, a SFI Investigator awards to CH and RR (10/IN.1/B3027), SFI-PI funding (11/PI/1137) to PDC and the Alimentary Pharmabiotic Centre under Grant Number SFI/12/RC/2273.


**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 Field, Cotter, Hill and Ross. 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.*