# NK CELL-BASED CANCER IMMUNOTHERAPY

EDITED BY: Francisco Borrego, Susana Larrucea, Rafael Solana and Raquel Tarazona PUBLISHED IN: Frontiers in Immunology

#### *Frontiers Copyright Statement*

*© Copyright 2007-2016 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 978-2-88919-934-1 DOI 10.3389/978-2-88919-934-1

#### 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

# **NK CELL-BASED CANCER IMMUNOTHERAPY**

#### Topic Editors:

**Francisco Borrego,** BioCruces Health Research Institute, Cruces University Hospital & Ikerbasque, Basque Foundation for Science, Spain **Susana Larrucea,** BioCruces Health Research Institute & Cruces University Hospital, Spain **Rafael Solana,** Maimonides Institute for Biomedical Research (IMIBIC) & Reina Sofía University Hospital, Spain **Raquel Tarazona,** University of Extremadura, Spain

**Citation:** Borrego, F., Larrucea, S., Solana, R., Tarazona, R., eds. (2016). NK Cell-Based Cancer Immunotherapy. Lausanne: Frontiers Media. doi: 10.3389/978-2-88919-934-1

# Table of Contents


Max P. Jung and Gabriele Multhoff

#### *129 Heat shock protein 70 (Hsp70) peptide activated Natural Killer (NK) cells for the treatment of patients with non-small cell lung cancer (NSCLC) after radiochemotherapy (RCTx) – from preclinical studies to a clinical phase II trial*

Hanno M. Specht, Norbert Ahrens, Christiane Blankenstein, Thomas Duell, Rainer Fietkau, Udo S. Gaipl, Christine Günther, Sophie Gunther, Gregor Habl, Hubert Hautmann, Matthias Hautmann, Rudolf Maria Huber, Michael Molls, Robert Offner, Claus Rödel, Franz Rödel, Martin Schütz, Stephanie E. Combs and Gabriele Multhoff

#### *138 NK cell-mediated antibody-dependent cellular cytotoxicity in cancer immunotherapy*

Wei Wang, Amy K. Erbe, Jacquelyn A. Hank, Zachary S. Morris and Paul M. Sondel

*153 Cetuximab reconstitutes pro-inflammatory cytokine secretions and tumor-infiltrating capabilities of sMICA-inhibited NK cells in HNSCC tumor spheroids*

Stephan Klöss, Nicole Chambron, Tanja Gardlowski, Sandra Weil, Joachim Koch, Ruth Esser, Elke Pogge von Strandmann, Michael A. Morgan, Lubomir Arseniev, Oliver Seitz and Ulrike Köhl

*171 NK Cell Subgroups, Phenotype, and Functions After Autologous Stem Cell Transplantation*

Benedikt Jacobs, Sara Tognarelli, Kerstin Poller, Peter Bader, Andreas Mackensen and Evelyn Ullrich


Loredana Ruggeri, Sarah Parisi, Elena Urbani and Antonio Curti

*197 Effector functions of natural killer cell subsets in the control of hematological malignancies*

Angela Gismondi, Helena Stabile, Paolo Nisti and Angela Santoni

*204 Increased NK cell maturation in patients with acute myeloid leukemia*

Anne-Sophie Chretien, Samuel Granjeaud, Françoise Gondois-Rey, Samia Harbi, Florence Orlanducci, Didier Blaise, Norbert Vey, Christine Arnoulet, Cyril Fauriat and Daniel Olive

*213 Natural Killer Cell Recognition of Melanoma: New Clues for a More Effective Immunotherapy*

Raquel Tarazona, Esther Duran and Rafael Solana

# Editorial: NK Cell-Based Cancer Immunotherapy

#### *Francisco Borrego1,2 \*, Susana Larrucea1 \*, Rafael Solana3 \* and Raquel Tarazona4 \**

*1BioCruces Health Research Institute, Cruces University Hospital, Barakaldo, Spain, 2 Ikerbasque, Basque Foundation for Science, Bilbao, Spain, 3 Instituto Maimónides de Investigación Biomédica de Córdoba (IMIBIC), Reina Sofía University Hospital, University of Córdoba, Córdoba, Spain, 4University of Extremadura, Cáceres, Spain*

Keywords: NK cells, adoptive cell therapy, cancer immunotherapy, ADCC, cytokines, CAR, NK-92

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

#### **NK cell-based cancer immunotherapy**

*Edited and Reviewed by:* 

*Yenan Bryceson, Karolinska Institutet, Sweden*

#### *\*Correspondence:*

*Francisco Borrego francisco.borregorabasco@ osakidetza.eus; Susana Larrucea susana.larruceabilbao@ osakidetza.eus; Rafael Solana rsolana@uco.es; Raquel Tarazona rtarazon@unex.es*

#### *Specialty section:*

*This article was submitted to NK Cell Biology, a section of the journal Frontiers in Immunology*

*Received: 22 May 2016 Accepted: 13 June 2016 Published: 27 June 2016*

#### *Citation:*

*Borrego F, Larrucea S, Solana R and Tarazona R (2016) Editorial: NK Cell-Based Cancer Immunotherapy. Front. Immunol. 7:249. doi: 10.3389/fimmu.2016.00249*

Innate and adaptive immunity cooperate to eliminate tumors. However, when infrequent cancer cell variants are not destroyed, tumor growth and immunosurveillance enter into a dynamic equilibrium until cancer cells evade the immune system, at which point malignancies appear clinically as a consequence. Therapies designed to induce potent antitumor responses by harnessing the power of the immune system are an appealing strategy to control tumor growth. Natural killer (NK) cells are innate lymphocytes that play a pivotal role in host immunity against cancer. The activity of NK cells is finely tuned by the balance between the signals that emanate from inhibitory and activating receptors. Inhibitory receptors, such as killer-cell immunoglobulin-like receptor (KIR) and CD94/ NK Group 2 member A (NKG2A), recognize human leukocyte (HLA) class I molecules whose expression is often altered on tumor cells. NK cells recognize tumor cells by activating receptors, such as natural cytotoxicity receptors (NCRs) and NKG2D, which sense the changed expression of their ligands on the cancer cell surface. Providing important insights, the past 15 years have witnessed an explosion of research into the biology and clinical applications of NK cells. Current NK cell-based cancer immunotherapy aims to reverse the tumor-induced NK cell dysfunction that is observed in patients with cancer and to increase and sustain NK cell effector functions. Therapies involving NK cells may either activate endogenous NK cells or involve transfer of exogenous cells by hematopoietic stem cell transplantation (HSCT) or adoptive cell therapy.

In this research topic, we have collected several articles that highlight the exciting potential that NK cells exhibit as an effective tool in cancer immunotherapy. We open the research topic with two articles that describe NK cell surface receptors involved in the recognition of tumor target cells. Chester et al. briefly describe NK-cell–tumor interactions and the three most important mechanisms of how NK cells kill target cells, i.e., natural killing, antibody-dependent cell-mediated cytotoxicity (ADCC), and death receptor-induced apoptosis, and follow with a description of the best studied activating and inhibitory receptors involved in tumor cell recognition, along with the ability of agonistic monoclonal antibodies (mAbs) specific for costimulatory molecules, such as CD137 and OX40. Horton and Mathew focus their review on NCRs, with a special attention to NKp44 and its dual activating and inhibitory function following recognition of different ligands. They suggest that NCRs serve as receptors for damage-associated molecular pattern (DAMP) molecules in association with HLA class I molecules, heparan sulfate proteoglycans (HSPGs), or other coligands and, therefore, regulate NK cell activity. A better understanding of the tumor immunosuppressive microenvironment is very important to design efficient NK cell-based therapies. Hasmim et al. review the cell types within the tumor that are involved in the suppression of NK cells, including M2-polarized macrophages, myeloid-derived suppressor cells (MDSC), regulatory T cells (Treg), and fibroblasts, and how the microenvironmental hypoxia that is characteristic of solid tumors inhibit NK cell functions.

Next, there is a group of articles that provide a general vision of how to obtain and harness NK cells to fight tumors. Dahlberg et al., Domogala et al., Pittari et al., and Rezvani and Rouce have reviewed the different cellular sources and methods to isolate, differentiate, genetically engineer, expand, and activate *ex vivo* and *in vivo* NK cells, including autologous and allogeneic NK cells and NK cell lines. In addition to NK cell lines, sources to generate NK cell products include NK cells from peripheral blood or from cord blood and NK cells differentiated from CD34<sup>+</sup> hematopoietic precursors or pluripotent stem cells (embryonic stem cells or induced pluripotent stem cells). In general, adoptive NK cell-based therapy has been more successful in the treatment of hematological tumors than in patients with solid tumors, and the use of tools aimed to reverse the immunosuppressive tumor microenvironment significantly will improve the efficacy of this type of therapy. A cell product termed cytokine-induced killer (CIK) cells, which possess phenotypic and functional features of both NK cells and T cells, is also described by Pittari et al.

Similar to T cells, genetic manipulation of NK cells is emerging as a promising tool to increase their antitumor activity. Chimeric antigen receptors (CARs) consist of an extracellular domain, generally a small chain variable fragment, specific for a tumor antigen that is linked with one or more intracellular domains able to induce activation signals. In this study, Hermanson and Kaufman extensively review the CAR constructs with different intracellular activation domains that have been described, to date, in NK cells from several sources, suggesting that depending on the tumor type and/or target antigen, different CAR constructs may be required for optimal activation of NK cells. Carlsten and Childs review the advantages and challenges of methods to genetically modify NK cells and give an overview of different strategies to reprogram NK cells with the objective to improve the persistence and expansion of infused cells, to enhance the migration to the tumors, and to improve their cytotoxicity. Cell lines, such as NK-92, could be an alternative to NK cells from patients or allogeneic donors. They could easily be expanded in culture, genetically manipulated, and may represent an off-theshelf product ready for use. Klingemann et al. review that NK-92 is the only cell line that has been studied in clinical trials with clinically significant responses and minimal adverse reactions.

In contrast to normal cells, many tumor cells express heatshock protein 70 (Hsp70) at the cell surface and get released into the circulation. Membrane Hsp70 (mHsp70) correlates with high aggressiveness of the tumors. In this research topic, Gunther et al. describe that patients with squamous cell and adeno non-small cell lung cancer (NSCLC) exhibited high levels of serum Hsp70. Furthermore, they found a positive correlation between serum levels of Hsp70 with gross tumor volume and an inverse correlation with CD69<sup>+</sup>/CD94<sup>+</sup> NK cells in squamous NSCLC, suggesting that activated NK cells somehow may be involved in the control of tumor growth. The same group has previously found in preclinical studies that NK cells activated with a naturally occurring Hsp70 peptide (TKD) and IL-2 are able of specifically kill mHsp70-expressing tumors, but not mHsp70 negative ones. Here, they summarize a Phase I clinical trial of TKD/IL-2-stimulated autologous NK cells with NSCLC and describe an ongoing Phase II clinical trial of TKD/IL-2 stimulated NK cells for the treatment of patients with NSCLC, following radiochemotherapy (Specht et al.).

The mechanism of action of many therapeutic mAbs for cancer treatment involves, at least partially, ADCC through FcγRIIIA/CD16a. Many studies have shown that the clinical outcome after treatment of patients with mAbs is correlated with polymorphisms at the *FCGR3A* gene, which encodes for FcγRIIIA/CD16a, that affect the binding affinity to mAbs. Wang et al. review some of the current therapeutic mAbs that are being used in the clinic and strategies that increase their ADCC, such as modifying the glycosylation patterns of the mAbs, combining them with other mAbs, radiation therapy, matrix metalloproteases inhibitors or cytokines, and by designing new molecular entities such as immunocytokines and bi-specific antibodies. Cetuximab, an anti-epidermal growth factor receptor (EGFR) mAb, exerts ADCC against EGFR<sup>+</sup> target cells. Kloss et al. show that patients with head and neck squamous cell carcinomas (HNSCCs) have elevated levels of soluble major histocompatibility complex class I chain-related peptide A (sMICA) and transforming growth factor beta 1 (TGF-β1) in serum, which are responsible for the impaired NK cell effector functions and decreased NKG2D expression. They show that cetuximab restores the NK cell-mediated killing of sMICA-inhibited patient NK cells against HNSCC cells *via* ADCC and enhances tumor infiltration of NK cells in HNSCC tumor spheroids.

Autologous hematopoietic stem cell transplantation (autoHSCT) is a therapeutic indication for multiple myeloma or malignant lymphoma, and it has been shown that the reconstitution levels of the NK cell pool after autoHSCT has a prognostic value. Jacobs et al. have studied the phenotype and function of NK cells after autoHSCT. They found that CD56++ NK cells were the major subset 1–2 days after leukocyte regeneration (>1000 leukocytes/μl) and that is characterized by a high expression of CD57 and KIRs, which is age dependent, and that are able to degranulate and produce cytokines after tumor interaction. On the other hand, preclinical and clinical data have demonstrated that, in the context of haploidentical T-cell-depleted HSCT, alloreactive NK cells are able to exert a very important antitumor activity, with no increased incidence of GVDH, and that mature alloreactive NK cells can be safely infused into patients. The KIR–HLA class I mismatch between donor and recipient in the graft versus leukemia (GVL) direction has demonstrated to enhance the antitumor activity of NK cells. In this research topic, two articles by Lim et al. and Ruggeri et al. review the present and future of alloreactive NK cells for tumor treatment, mostly acute myeloid leukemia (AML), in the context of allogeneic HSCT and infusions of alloreactive purified NK cells. These are emerging as safe and potent effectors against tumors.

Based on their expressed pattern of cell surface receptors, NK cells are divided in subsets that are able to mediate different effector functions and are characterized by distinct homing properties. Gismondi et al. have reviewed this issue, suggesting that, for improved and more efficient NK cell-based therapies, it is necessary to identify, isolate, expand, and administer NK cell subsets that exhibit increased effector functions and have the adequate homing capabilities to reach the tumors. Chretien et al., with an automated procedure using the FLOCK algorithm and a panel of three markers (CD56, CD57, and KIRs), define five maturation stages of NK cells from human peripheral blood. By analyzing a cohort of healthy volunteers and another cohort of patients with AML, they found that the latter displayed marked differences compared with healthy donors. Moreover, within the AML cohort, it was possible to define three distinct groups of patients according to their maturation profiles, which might be useful for prognostic purposes. Tarazona et al. have reviewed the current knowledge about the role of NK cells on the recognition and elimination of melanoma cells and the strategies against melanoma based on NK cells. *In vitro* experiments, *in vivo* data from murine models, and observations from melanoma patients indicate that NK cells have a role in the immune response against melanoma. NK cell-based therapies against melanoma include, among others, modulation of NK cell responses by administration of cytokines, treatment with checkpoint inhibitors and bi-specific antibodies, and by adoptive NK cell therapy with

**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 Borrego, Larrucea, Solana and Tarazona. This is an open-access article distributed under the terms of the Creative Commons Attribution License*  autologous or allogeneic NK cells and NK cell lines, genetically modified or not.

Finally, we want to express our gratitude to all the authors who have contributed to this research topic and to the reviewers for their magnificent job. We hope that the reader will find this research topic motivating and helpful. We invite you to read the following articles and immerse yourself in the interesting world of NK cell-based cancer immunotherapy.

# AUTHOR CONTRIBUTIONS

All authors listed, have made substantial, direct and intellectual contribution to the work, and approved it for publication.

# FUNDING

This work was supported by grants BIO13/CI/011 (to FB) from BIOEF-EiTB Maratoia Pediatric Cancer, BIO14/TP/001 (to SL) from BIOEF-EiTB Maratoia Transplantation, SAF2013-46161-R (to RT) from the Ministry of Economy and Competitiveness of Spain, and PI13/02691 (to RS) from Spanish Ministry of Health.

*<sup>(</sup>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.*

# Natural Killer Cell Immunomodulation: Targeting Activating, Inhibitory, and Co-stimulatory Receptor Signaling for Cancer Immunotherapy

#### *Cariad Chester1,2 , Katherine Fritsch1 and Holbrook E. Kohrt1 \**

*1Division of Oncology, Department of Medicine, Stanford University, Stanford, CA, USA, 2 Institute for Immunity, Transplantation and Infection, Stanford University School of Medicine, Stanford, CA, USA*

#### *Edited by:*

*Raquel Tarazona, University of Extremadura, Spain*

#### *Reviewed by:*

*Francisco Borrego, BioCruces Health Research Institute and Cruces University Hospital, Spain John T. Vaage, Oslo University Hospital, Rikshospitalet and University of Oslo, Norway*

> *\*Correspondence: Holbrook E. Kohrt kohrt@stanford.edu*

#### *Specialty section:*

*This article was submitted to NK Cell Biology, a section of the journal Frontiers in Immunology*

*Received: 02 May 2015 Accepted: 09 November 2015 Published: 02 December 2015*

#### *Citation:*

*Chester C, Fritsch K and Kohrt HE (2015) Natural Killer Cell Immunomodulation: Targeting Activating, Inhibitory, and Costimulatory Receptor Signaling for Cancer Immunotherapy. Front. Immunol. 6:601. doi: 10.3389/fimmu.2015.00601*

There is compelling clinical and experimental evidence to suggest that natural killer (NK) cells play a critical role in the recognition and eradication of tumors. Efforts at using NK cells as antitumor agents began over two decades ago, but recent advances in elucidating NK cell biology have accelerated the development of NK cell-targeting therapeutics. NK cell activation and the triggering of effector functions is governed by a complex set of activating and inhibitory receptors. In the early phases of cancer immune surveillance, NK cells directly identify and lyse cancer cells. Nascent transformed cells elicit NK cell activation and are eliminated. However, as tumors progress, cancerous cells develop immunosuppressive mechanisms that circumvent NK cell-mediated killing, allowing for tumor escape and proliferation. Therapeutic intervention aims to reverse tumor-induced NK cell suppression and sustain NK cells' tumorlytic capacities. Here, we review tumor– NK cell interactions, discuss the mechanisms by which NK cells generate an antitumor immune response, and discuss NK cell-based therapeutic strategies targeting activating, inhibitory, and co-stimulatory receptors.

Keywords: natural killer cells, immunotherapy, adoptive cell therapy, monoclonal antibody, cancer vaccines, checkpoint blockade

# INTRODUCTION

The recent FDA approvals of the programmed cell death protein 1 (PD-1)-targeted checkpoint inhibitors pembrolizumab and nivolumab mark the latest successes in the rapidly expanding field of cancer immunotherapies. Immunotherapy represents a paradigm shift in cancer treatment; instead of targeting tumor cells, the goal of immunotherapy is to augment and expand the immune system's intrinsic antitumor response. To date, diverse immunotherapeutic modalities have been accepted as viable strategies for eliminating cancerous cells. Cytokines, cancer vaccines, adoptive cell transfers, and especially checkpoint inhibitors constitute valuable elements in the immunotherapeutic armamentarium. However, a class of important immune-modulators is conspicuously absent: agents that utilize the power of innate immune cells to eradicate tumors. An important class of innate immune cells that play a critical role in mediating the antitumor immune response is the natural killer (NK) cell.

First described in 1975, NK cells were initially identified as a distinct sub-population of lymphocytes by their capacity to spontaneously lyse tumor cells (1). NK cells are now accepted to play an important role in both the adaptive and innate immune responses that govern infection, autoimmunity, and tumor immunosurveillance (2, 3). Human NK cells are phenotypically characterized by the expression of CD56 and the absence of CD3 and can be further subdivided into a CD56bright population and a CD56dim population. The CD56bright population produces immunoregulatory cytokines, including interferon-γ (IFN-γ), tumor necrosis factor-beta (TNF-B), tumor necrosis factor-α (TNF-α), granulocyte macrophage-colony stimulating factor (GMCSF), IL-10, and IL-13 (4). The CD56dim subset is the terminally differentiated successor of the CD56bright population and is primarily responsible for exerting cytolytic functions (5, 6). However, CD56dim NK cells can produce cytokines, specifically IFN-γ, after cell triggering via NKp46 of NKp30 activating receptors or after stimulation with combinations of IL-2, IL-12, and IL-15 (7).

The defining functional feature of NK cells remains their intrinsic ability to conduct "natural killing" of cellular targets without prior sensitization. The antitumor effect provided by natural killing has been observed in tumors of hematopoietic and non-hematopoietic origins and reported in diverse *in vivo* models and clinical series (8). NK cell infiltration into tumor tissue is associated with better disease prognosis in colorectal cancer, clear cell renal cell carcinoma, and lung carcinomas (9–11). Additionally, a 11-year prospective cohort study of Japanese inhabitants linked low peripheral-blood NK cell cytotoxicity with increased cancer risk (12). The combination of compelling preclinical evidence and early clinical success has established NK cell immunotherapy as a promising therapeutic strategy in cancer. Here, we review the current understanding of the NK cell mechanisms underpinning antitumor immunity and discuss immunomodulatory targets for augmenting NK cell-mediated tumor clearance.

#### Natural Killing

The initial hypothesis for the mechanism of NK cell-mediated killing postulated that the absence or altered expression of major histocompatibility complex (MHC) class I molecules would render target cells susceptible to NK cell attack (13). The "missingself " hypothesis was the result of observations that NK cells can directly reject MHC class I-deficient tumors (14). Later *in vivo* experiments in murine and human systems confirmed that NK cytotoxicity was directly related to the absence of MHC class I expression on target cells (15, 16). However, the contemporary understanding of NK cell activation suggests that the transition of the NK cell from quiescence to activation is mediated by a network of activating and inhibitory receptors (17). While NK cells do express inhibitory receptors that detect the presence of MHC Class I molecules, it is the integration of multiple activating and inhibitory signals that determines if the NK cell becomes cytotoxic.

Natural killer cell cytotoxicity can be demonstrated in several related ways. The primary mechanism of cytotoxicity is based on granule exocytosis upon formation of an immunological synapse. NK cells contain preformed cytoplasmic granules that resemble secretory lysosomes and contain perforin and granzymes (18). Perforin is a membrane-disrupting protein that perforates the target cell membrane, while granzymes are a family of serine proteases that trigger cell apoptosis (19, 20). Upon activation, NK cells rapidly polarize the granules and reposition the microtubule organizing center toward the synapse with the target cell (21). The granule membrane then fuses with the plasma membrane, externalizes, and releases the cytotoxic granule contents, triggering target cell apoptosis (22).

NK cells can also contribute to target cell death indirectly by secreting pro-inflammatory cytokines. Two of the primary cytokines released by activated NK cells are IFN-γ and TNF-α. IFN-γ is a type II interferon that plays a critical role in promoting host resistance to microbial infection and protecting against tumor development (4). In the tumor microenvironment (TME), the IFN-γ released by NK cells stimulates CD4<sup>+</sup> T cells to polarize toward a Th1 subset and accelerates the development of activated macrophages and cytotoxic, tumor-targeting CD8<sup>+</sup> T cells (23). TNF-α is a multifunctional cytokine that can cause direct tumor necrosis by inflicting tumor-associated capillary injury, but also generates an adaptive immune response (24). TNF-α can enhance B cell proliferation and also promote monocyte and macrophage differentiation (25, 26). Together IFN-γ and TNF-α help to activate both innate and adaptive immune cells in the TME and generate a sustained antitumor immune response.

# Antibody-Dependent Cell-Mediated Cytotoxicity

Another granule-mediated mechanism of NK cell targeted killing is antibody-dependent cell-mediated cytotoxicity (ADCC). ADCC is thought to play an important role in mediating the antitumor effects of many of the monoclonal antibody (mAb) therapies used today as standard of care treatments for both solid tumors and hematologic malignancies (27). In ADCC, the Fc receptor expressed by NK cells (FcγRIII or CD16) binds to the Fc portion of the therapeutic antibody, which in turn is bound to tumor-associated antigen (TAA) on the tumor surface. The effectiveness of ADCC depends on the FcγRIII ligation on the NK cell. Patients with a FcγRIIIa polymorphism, resulting in high-affinity binding of FcγRIII to IgG1, demonstrate enhanced clinical benefit. This effect has been seen in patients treated with rituximab, trastuzumab, and cetuximab (28–30). ADCC was initially described as the release of cytotoxic perforin and granzyme by NK cells following ligation of FcγRIII by IgG target cells. However, ADCC is now recognized as a multi-tiered process that involves a network of coordinated immune cells and an adaptive immune response (31). For example, FcγR ligation on NK cells can induce the secretion of pro-inflammatory cytokines like IFN-γ, which can accelerate dendritic cell (DC) maturation (32). Mature DCs enhance antigen presentation and train tumor-specific lymphocytes, producing an immunological memory response (33).

## Death Receptor-Induced Apoptosis

Death receptor-induced apoptosis is a perforin-independent mechanism by which NK cells lyse target cells (34). This cytotoxic pathway relies on target cell expression of tumor necrosis factor (TNF) receptor superfamily members. The two main TNF receptors used in apoptotic induction are Fas (CD95) and TNF-related apoptosis-inducing ligand (TRAIL) (35). Fas is expressed on a wide variety of tissues, but Fas ligand (FasL) expression is restricted to activated NK cells and cytotoxic T lymphocytes (CTLs) (36). Fas cross-linking induces nuclear condensation, membrane blebbing, and caspase activation (37). The initial optimism surrounding the Fas–FasL pathway as a means of tumor control has decreased following the observations that Fas is downregulated in a variety of cancers during tumor progression (38).

TNF-related apoptosis-inducing ligand-mediated signaling is another death receptor-induced mechanism NK cells employ to kill target cells. TRAIL is constitutively expressed on some populations of NK cells and TRAIL-mediated signaling can induce spontaneous cytotoxicity against TRAIL-sensitive tumor cells (39, 40). Binding of TRAIL to its receptor, TRAILR1 or TRAILR2, results in receptor oligomerization on the cell membrane and triggering of a pro-apoptotic signal through activation of caspases (41). Preclinically, recombinant forms of TRAIL and agonistic anti-TRAIL receptor antibodies can have single-agent activity against TRAIL-sensitive tumor cells *in vitro* and *in vivo* (42). Recently, artificial nanoparticles coated with bioactive TRAIL demonstrated cytotoxicity against primary leukemic cells from a patient with acute lymphocytic leukemia (ALL) (43). However, despite preclinical successes, clinical trials of TRAIL-based therapies have demonstrated little efficacy and tumors rapidly develop resistance mechanisms to TRAIL (44). A better understanding of how tumors evade targeting and removal by NK cells is needed to overcome immunosuppression in the TME.

#### NK–Tumor Interactions

Despite the diverse repertoire of killing strategies utilized by NK cells, the tumor cell often avoids attack by direct and indirect mechanisms (45). Direct mechanisms consist of shedding soluble ligands for NK cell-activating receptors, upregulation of HLA molecules, and release of inhibitory cytokines. Indirect mechanisms consist of activation of inhibitory regulatory T cells (Tregs), DC killing, and phagocyte-derived inhibitory cytokines. These immunosuppressive mechanisms collectively create a TME where NK cell cytotoxic functions are inhibited. By stifling NK-mediated tumor eradication, the tumor escapes immunosurveillance and is able to grow and develop. Restoring and augmenting NK cell cytotoxic functions in the TME is an important step in overcoming immunosuppression and eliminating tumor. In an attempt to generate potent tumor-lysing NK cells, therapeutics are being developed that target NK cell activating, inhibitory, and co-stimulatory receptors (**Figure 1**).

## ACTIVATING RECEPTORS

Activating receptors are a crucial element in regulating NK cell function. In the last decade, researchers have identified major signaling axes that control NK cell activation and suggested novel routes for therapeutic interventions (46). Some of the dominant activating receptors on NK cells are NKG2D, signaling lymphocytic activation molecule (SLAM) family molecule 2B4 (CD244), the DNAX accessory molecule (DNAM-1, CD226), and the NCRs: NKp30, NKp44, NKp46, and NKp80 (42). Recent work suggests that in NK cells, there is not a dominant receptor for activation, but instead receptors induce activation through combinatorial synergy (17). Only when multiple activating receptors are simultaneously engaged does the resulting signal surpass the requisite activation threshold and trigger cytokine secretion or direct cellular cytotoxicity. The requirement for activating receptor combinations helps prevent unrestrained activation of NK cells and provides flexibility in sensing and responding to environmental stimuli. What follows is a brief exploration of the dominant NK cell-activating receptors and summaries of attempts to target their tumorlytic capacity therapeutically.

#### NCRs

All NK cells express NKp30 and NKp46, whereas NKp44 is only expressed on activated NK cells (47, 48). The acquisition of NCR during NK cell maturation correlates with the acquisition of cytolytic activity against tumor target cells (49). Inversely, downregulation of NKp30, NKp44, and NKp46 correlates with low NK cytolytic activity (50). NKp80 is expressed by virtually all fresh NK cells and mAb-mediated cross-linking of NKp80 resulted in induction of cytolytic activity and Ca2<sup>+</sup> mobilization (51).

The NKp30 activating receptor has emerged as a promising therapeutic target in multiple cancer histologies. Downregulation is observed in patients with cervical cancer and high-grade squamous intraepithelial lesions (52). In lymphoma and leukemia models, ligation of NKp30 has been shown to activate human NK cells, trigger degranulation, and increase cytotoxicity (53). In patients with gastrointestinal sarcoma, the NKp30 isoform predicts the clinical outcome; patients with the immunostimulatory NKp30a and NKp30b isoforms have increased survival relative to patients with the immunosuppressive NKp30c isoform (54). Recently, the expression of distinct forms of NKp30 has been linked to 10-year progression free survival in patients with highrisk neuroblastoma (NB) (55). In NB patients with metastatic disease, the percentage of CD3<sup>−</sup>CD56<sup>+</sup> NK cells in the peripheral blood and bone marrow was significantly elevated relative to patients with localized tumors. Additionally, NKp30 expression in the bone marrow of patients with metastatic NB was lower than expression in patients with localized NB (55). The ligand for NKp30, B7-H6, was highly expressed in neuroblasts, and the serum soluble form of B7-H6 correlated with tumor load and disease dissemination. The authors conclude that NK cell modulating immunotherapeutics offer a promising strategy for treating NB patients and that antibodies neutralizing sB7-H6 serum molecules and antibodies targeting NKp30 are worth pursuing in future clinical development.

#### NKG2D

NKG2D, a homodimeric activating receptor and member of the C-type lectin superfamily, is expressed by all NK cells and subsets of T cells (56). NKG2D serves as a major recognition receptor for detection and elimination of infected and transformed cells (57). Ligands of the human NKG2D receptor are the MHC I-related molecules MICA/MICB, and the UL16 binding proteins (ULBP-1 to ULBP-6) (57). These ligands are rarely expressed in healthy tissues but induced by various

cellular stresses, such as DNA damage, heat shock, or cellular transformation. Primary tumors frequently express NKG2D ligands: NK cell killing of both an urothelial tumor cell line and a bladder cancer cell appeared to be triggered by NK cell detection of the NKG2D ligands MICA/MICB (58, 59). However, tumors have also developed mechanisms for NK cell evasion despite NKG2D ligand expression. One such mechanism is the systemic release of NKG2D ligands by tumors in cancer patients (60, 61). The secreted NKG2D ligand was believed to cause downregulation of NK cell-expressed NKG2D, thus, depriving the NK cell of an activating signal and facilitating tumor escape. Recently, evidence has emerged that demonstrates an activating, antitumor role for soluble NKG2D ligands. The highaffinity MULT1 mouse NKG2D ligand can stimulate NK cells and enhance antitumor activity (62). The NKG2D pathway is integral to immune surveillance and an active area of immunotherapy research.

## 2B4 and DNAM-1

One of the best-characterized NK cell activation receptors is 2B4, a member of the SLAM receptor family. The first data to suggest a role for 2B4 in regulating NK cell activation demonstrated that ligation of 2B4 by 2B4-specific antibodies induced IFN-y production *in vitro* and triggered NK cell-mediated cytotoxicity (63). Following the identification of the natural ligand for 2B4, CD48, researchers reported that target cell expression of CD48 augmented NK cell-mediated cytotoxicity (64). Researchers also reported significantly greater cytotoxic effects if 2B4 ligation was accompanied by ligation of DNAM-1 (65). DNAM-1 is an Ig-like family glycoprotein expressed on most human NK cells, monocytes, and T lymphocytes (66). Early support for DNAM-1 controlling NK cell activation was provided by Lanier and colleagues using DX11, an anti-DNAM-1 mAb (67). Blockade via DX11 inhibited the cytotoxicity of NK cells against an array of different tumor cell lines. CD112 and CD155, two nectin family proteins regulated by cellular stress, were soon identified as the ligands for DNAM-1 (68). CD155 and CD112 are expressed in a wide range of both solid and hematologic tumors (69). In patients with NB, expression levels of CD155 and CD112 correlate with tumor cells susceptibility to NK cell-mediated lysis (70). However, tumors have developed mechanisms for downregulating NK cell DNAM-1 and effecting NK cell immunosuppression (71). In the design of future NK cell-based immunotherapies, mechanisms for preserving activation receptor surface expression need to be considered. Additionally, combinations of synergistic activating receptor pairs, like DNAM-1 and 2B4, need to be taken into account.

# CHECKPOINT BLOCKADE IN NK CELLS

Immune checkpoint blockade strategies have proven a powerful approach to cancer immunotherapy. By blocking the receptors that transmit inhibitory signals to effector immune cells, checkpoint blockade aims to reverse immune suppression and generate robust antitumor immune responses. The successes of ipilimumab (anti-CTLA-4 mAb) and nivolumab and pembrolizumab (anti-PD-1 mAbs) demonstrate the potential of this therapeutic strategy. Ipilimumab (Yervoy, BMS) was approved in 2011 for the treatment of unresectable or metastatic melanoma, and blocks the CTLA-4-mediated signaling in T cells (72). CTLA-4 is an inhibitory receptor that upon ligation sends a negative regulatory signal to the T-cell receptor (TCR), limiting T cell activation (73). Nivolumab (Opdivo, BMS) and pembrolizumab (Keytruda, Merck) target programmed cell death protein-1 (PD-1). PD-1 is upregulated on T cells following activation and ligation of PD-1 transmits a negative regulatory signal (74). Histologically, diverse tumors upregulate the ligands of PD-1, PD-L1, and PD-L2 to take advantage of this immunosuppressive signaling pathway (75). Analogously to negative regulators of T cell activity, NK cells express surface receptors that can be targeted in checkpoint blockade strategies.

#### Killer Cell Immunoglobulin-Like Receptor

Within the signaling pathways that govern NK lytic capacity, the killer cell immunoglobulin-like receptor (KIR) family is a dominant group of negative regulators. KIR receptors bind to the self-MHC class I ligands (HLA-A, -B, -C) and upon ligation transmit signals that abrogate the effects of activating receptors (76). The prevalence of MHC I on healthy cells provides an inhibitory signal that prevents NK cells from inducing autoimmune responses. However, in acute myeloid leukemia (AML) patients following haploidentical stem cell transplantation from KIR mismatched donors, the absence of KIR–HLA class I interactions resulted in potent NK cell-mediated antitumor efficacy and increased survival (77, 78). The antitumor effect can also be obtained without undergoing stem cell transplant; mAb therapy provides a viable route for blocking KIR–HLA interactions. Preventing HLA ligation to KIRs with an anti-KIR mAb has been shown to increase NK cell degranulation, IFN-γ secretion, and tumor cell lysis as well as increasing overall survival in murine cancer models (79).

The development of a candidate anti-KIR antibody had to overcome significant challenges. The KIR gene content varies substantially from individual to individual depending on the inherited KIR haplotype and the KIR family is composed of several structurally different proteins, necessitating an antibody that has cross-reactivity between different KIRs (80). Despite these challenges, the anti-KIR mAb lirilumab (Innate Pharma) has entered clinical trials. The initial phase I safety trial reported safety and potential efficacy in patients with AML (81). A second phase I trial confirmed the early reports of safety and durable KIRblocking ability in patients with multiple myeloma (82). Recently, it has been reported that rituximab-mediated ADCC, a potent therapeutic mechanism of rituximab therapy, is reduced by KIR signaling (83). We have demonstrated that this KIR-mediated ADCC suppression can be overcome by combining rituximab with anti-KIR mAb therapy (84). Currently, multiple phase I and phase II clinical trials are ongoing, testing lirilumab (IPH2102/ BMS-986015), as a monotherapy or in combination with other

#### NKG2A

In addition to KIRs, the CD94/NKG2A heterodimer is another target for NK cell checkpoint blockade. The natural ligand of CD94/NKG2A is HLA-E, a non-classical HLA class I molecule that is expressed on the cell surface of most leukocytes and on transformed cells, including virus-infected cells and tumor cells (85, 86). Ligation of CD94/NKG2A by HLA-E transmits inhibitory signaling that suppresses the effector functions of NK cells, resulting in decreased cytotoxicity and cytokine secretion. HLA-E and CD94/NKG2A expression has been reported in multiple tumor histologies and is associated with poor prognosis. In colorectal cancer patients, tumor expression of HLA-E is associated with shorter disease-free survival time (87). In patients with head and neck squamous cell cancers (HNSCC), 78 to 86% of tumors express HLA-E (88). In patients with non-small cell lung cancer, intratumoral NK cells display higher expression levels of NKG2A mRNA relative to non-tumor NK cells (89). In breast cancer patients, expression of NKG2A by tumor infiltrating NK cells increases with cancer progression and correlates with impaired NK cell functions (90). Similarly to blocking KIR-mediated interactions, blockade of CD94/NKG2A-mediated signaling has the potential to restore and preserve NK cell cytotoxicity, leading to antitumor responses. A phase I/II trial testing an anti-NKG2A antibody (IPH2201, Innate Pharma) in HNSCC patients is ongoing (NCT02331875).

# CO-STIMULATORY SIGNALING VIA mAbs

Activating co-stimulatory pathways to potentiate antitumor immune responses is a promising approach for augmenting NK-mediated tumor clearance. Members of the tumor necrosis factor receptor superfamily (TNFRsf) include several co-stimulatory proteins with key roles in the regulation of the activation, proliferation, and apoptosis of lymphocytes, including NK cells.

# CD137

First identified in 1989, CD137 (or 4-1BB) is a co-stimulatory receptor and member of the TNF receptor superfamily (91). CD137 is expressed on T cells and DCs and is upregulated on NK cells following FcγRIIIa ligation (92). In a variety of different tumor models, agonistic anti-CD137 mAbs have demonstrated the capacity to amplify antitumor immune responses and eliminate established tumors (93). Despite the broad expression of CD137 and its multiple contributions to immune dynamics, the therapeutic efficacy of anti-CD137 relies on functional NK cells. In preclinical models, the selective depletion of NK cells via the anti-AsialoGM1 or anti-NK1.1 antibodies completely abrogated the antitumor effect of anti-CD137 mAb therapy (94). Simultaneously, anti-CD137 agonistic antibodies increase NK cell proliferation, degranulation, and IFN-γ secretion, leading to enhanced ADCC of tumor cells (95). Because of the potential to enhance ADCC-mediated tumor clearance, anti-CD137 antibodies are being tested in combination treatment strategies with FDA-approved mAbs. We have previously demonstrated that antibodies targeting CD137 synergize with rituximab and trastuzumab to clear tumors in murine xenograft models of lymphoma and breast cancer (27, 96). Recently, we combined cetuximab and anti-CD137 antibody therapy to obtain complete tumor resolution and prolonged survival in xenograft models of EGFR-expressing cancer cells (97). In all three disease models and combination treatment regimens, expression of CD137 on NK cells increases significantly when NK cells encounter mAbs bound to tumor cells. We believe the synergy between anti-CD137 treatment and established mAbs demonstrates a promising therapeutic strategy and warrants future investigation.

Anti-CD137 mAb therapy has also entered clinical testing. The anti-CD137 antibody, urelumab, is currently in clinical trials with rituximab for patients with non-Hodgkin's lymphoma (NCT01775631) and with cetuximab in patients with colorectal cancer or head and neck cancer (NCT02110082). In addition to urelumab, clinical trials of Pfizer's anti-CD137 mAb, PF–05082566, are also ongoing (NCT01307267). A recent presentation of the preliminary findings reports that 27 patients with mixed tumor types have been treated with PF-05082566; disease stabilization was the best overall response, observed in 22% (6/27) patients (98).

#### OX40

OX40, also known as CD134 or TNFRSF4, is a co-stimulatory molecule expressed primarily by activated T cells, but also expressed on natural killer T (NKT) cells and NKs (99). In NK cells, OX40 ligation appears to induce an activating signal and IFN-γ production (100). Engagement of the OX40 receptor *in vivo* in tumor-bearing mice enhanced antitumor immunity, resulting in increased survival in four separate murine tumor models of diverse histology and immunogenicity (101). The initial phase I/II trial of an anti-OX40 mAb demonstrated tolerability and regression of at least one metastatic lesion in 12 out of 30 study patients (102). Immunologically, treatment with agonistic anti-OX40 increased the proliferation of NK cells as well as CD4+ T cells (103). Additional trials of anti-OX40 are ongoing, including combination therapies with rituximab in patients with CLL and NHL (NCT01775631), with stereotactic body radiation in patients with metastatic breast cancer (NCT01862900), and with tremelimumab, an anti-CTLA-4 antibody, in patients with solid tumors (NCT02205333).

#### CD27

In addition to its co-stimulatory role on T cells, the expression of CD27, or TNFRSF7, differentiates the NK cell compartment into two functionally distinct subsets. Circulating CD27<sup>+</sup> NK have lower levels of perforin and granzyme B and demonstrate

## REFERENCES


lower levels of cytotoxicity relative to CD27- NK cells (104). The absence of CD27 expression in combination with the expression of CD11b is an indicator of cytolytic effector cells within human NK cell subsets. The natural ligand for CD27, CD70, induces downregulation of CD27 in a process controlled by the common γ-chain cytokine IL-15 (105). Signaling via CD27–CD70 interactions have been shown to accelerate NK-mediated tumor clearance while simultaneously stimulating cytokine secretion by NK cells that elicits an adaptive immune response (106).

The potential for CD27 ligation to generate an antitumor response has been confirmed in preclinical models. In a xenograft models of lymphoma, administration of the humanized anti-CD27 antibody, 1F5, significantly prolonged survival (107). The fully human 1F5 cannot bind to mouse CD27, therefore, any observed antitumor activity is attributed to effector mechanisms such as direct inhibition/apoptosis via CD27 signaling in tumors or ADCC. In syngeneic colorectal and lymphoma models with little to no expression of CD27, treatment with the 1F5 mAb also elicited antitumor activity and increased survival (108). By testing an aglycosylated version of the 1F5 mAb, the researchers demonstrated that FcR engagement was required for the antitumor effects of 1F5 therapy. An anti-CD27 mAb (Varlilumab or CDX-1127, Celldex Therapeutics) is currently being tested in a phase I trial in patients with solid tumors and hematologic cancers (NCT01460134). Preliminary findings report that of the 19 treated patients, 3 had stable disease and 1 had a complete readmission (109).

# CONCLUSION

In the future, immunotherapeutic agents that directly enhance NK cell-mediated tumor eradication will play a leading role in cancer treatment strategies. NK cells have novel mechanisms of participating in immune defense, making them uniquely appealing for cancer immunotherapy. Enhancing NK cell tumorlytic capacity is also a compelling combinatorial treatment strategy and would complement current standard of care treatments based on mAb therapy. The potential for NK-targeted agents to augment the antitumor effects of T cell checkpoint blockade is actively under consideration. As NK cell-based therapies move into the clinic, identifying prognostic biomarkers in the treatment populations will be crucial to the rational design of clinical studies. Concurrently, a greater effort must be made to profile the effects of novel immunotherapeutic agents, like checkpoint inhibitors, on NK cell function. The NK cell is now accepted as an integral part of the immunologic antitumor response. A number of promising NK-targeting therapeutics are in early-phase trials, and the results are eagerly awaited.


**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 Chester, Fritsch and Kohrt. 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.*

# NKp44 and natural cytotoxicity receptors as damage-associated molecular pattern recognition receptors

#### **Nathan C. Horton and Porunelloor A. Mathew \***

Department of Cell Biology and Immunology, Institute for Cancer Research, University of North Texas Health Science Center, Fort Worth, TX, USA

#### **Edited by:**

Rafael Solana, University of Cordoba, Spain

#### **Reviewed by:**

Hugh Thomson Reyburn, Spanish National Research Council, Spain Chiara Romagnani, Deutsches Rheuma-Forschungszentrum, Germany Alfonso Martin-Fontecha, Boehringer-Ingelheim, Germany

#### **\*Correspondence:**

Porunelloor A. Mathew, Department of Cell Biology and Immunology, University of North Texas Health Science Center at Fort Worth, 3500 Camp Bowie Blvd., Fort Worth, TX 76107-2699, USA e-mail: porunelloor.mathew@ unthsc.edu

Natural killer (NK) cells are a key constituent of the innate immune system, protecting against bacteria, virally infected cells, and cancer. Recognition and protective function against such cells are dictated by activating and inhibitory receptors on the surface of the NK cell, which bind to specific ligands on the surface of target cells. Among the activating receptors is a small class of specialized receptors termed the natural cytotoxicity receptors (NCRs) comprised of NKp30, NKp46, and NKp44. The NCRs are key receptors in the recognition and termination of virally infected and tumor cells. Since their discovery over 10 years ago, ligands corresponding to the NCRs have largely remained elusive. Recent identification of the cellular ligands for NKp44 and NKp30 as exosomal proliferating cell nuclear antigen (PCNA) and HLA-B-associated transcript 3 (BAT3), respectively, implicate that NCRs may function as receptors for damage-associated molecular pattern (DAMP) molecules. In this review, we focus on NKp44, which surprisingly recognizes two distinct ligands resulting in either activation or inhibition of NK cell effector responses in response to tumor cells. The inhibitory function of NKp44 requires further study as it may play a pivotal role in placentation in addition to being exploited by tumors as a mechanism to escape NK cell killing. Finally, we suggest that the NCRs are a class of pattern recognition receptors, which recognize signals of genomic instability and cellular stress via interaction with the c-terminus of DAMP molecules localized to the surface of target cells by various co-ligands.

#### **Keywords: NK cells, natural cytotoxicity receptors, NKp44, DAMPs, tumor ligands**

#### **INTRODUCTION**

Natural killer (NK) cells are a fundamental component of the innate immune system, capable of recognizing and destroying tumor cells as well as cells infected by viruses or bacteria (1, 2). NK cells also secrete cytokines such as interferon-γ (IFN-γ) and thus regulate the function of other immune cells. Furthermore, NK cells play an important role in adaptive immunity by modulating dendritic cell function and recent findings demonstrate that NK cells have memory (3, 4). The ability of NK cells to kill target cells and secrete cytokines is regulated by a delicate balance of activating and inhibitory signals received through distinct classes of receptors found on their cell surface. The balance of signals delivered by those receptors governs NK cell activation, proliferation, and effector functions (5–8). Traditionally, inhibitory killer cell immunoglobulin like receptors (KIRs) and killer cell lectin-like receptors (KLRs) bind cell surface human leukocyte antigen class I (HLA I) molecules expressed by healthy human cells and signal through immunoreceptor tyrosine-based inhibitory motifs (ITIMS) (9–11). When HLA I interacts with inhibitory receptors, dominant inhibitory signaling transmitted by ITIMS prevents activation and cytotoxic action by the NK cell against normal, healthy cells of the body. NK cells may also be inhibited by cytokines released by regulatory cells of the immune system,such as regulatory T cells and myeloid suppressor cells (12).

Activating receptors, including the natural cytotoxicity receptors (NCRs), NKG2D, and 2B4, bind ligands induced by cellular stress, infection, or tumor transformation (13–16). Activating signals are transmitted through immunoreceptor tyrosine-based activating motifs (ITAMs) located in the cytoplasmic tail of the receptor or through ITAMs in adaptor molecules, which associate with activating receptors at the cell surface (8, 17). Therefore, when a target cell lacks or under expresses HLA I and/or over expresses activating ligands, NK cells eliminate that target by releasing preformed cytotoxic granzymes and perforin stored as granules or activate apoptosis pathways in the target cell (8, 18).

#### **NATURAL CYTOTOXICITY RECEPTORS**

Among the activating receptors is a specialized group of receptors called the NCRs, which play a key role in recognition and killing of tumor and virally infected cells. Comprising the NCRs are the NKp44, NKp30, and NKp46 receptors. Binding of one or more of these receptors with a specific ligand induces strong NK cell activation and cytotoxicity (19). For optimal recognition and elimination of target cells, the NCRs work best as a team when identifying potential targets (20). This is evident through increased cytotoxicity when multiple NCRs are triggered versus an individual receptor, suggesting simultaneous NCR ligand expression on target cells (20, 21). Several studies have identified and characterized NCR ligands. NKp46 recognition of a ligand on tumor cells has been shown to play a role in prevention of tumor metastasis (22, 23). NKp30 is known to bind B7-H6, a member of the B7 family expressed exclusively on tumor cells (24). While many NCR ligands remain unidentified, they are believed not to be expressed by normal cells but induced by cellular stress or pathological conditions (14).

#### **NKp44**

NKp44 is unique and significant for several reasons. First, expression of the receptor is restricted to activated NK cells capable of initiating an immediate cytotoxic response (25). Second, NKp44 activating function is implicated in HIV-related T cell decline as expression of an activating ligand for NKp44 is induced in uninfected CD4 T cells by the gp41 envelope protein of HIV (26). Earlier studies have shown recognition of viral hemagglutinins of influenza virus by NKp44 enhanced killing of infected cells (27). Finally, NKp44 expression is responsible for a dramatic increase in killing of many tumor cell lines and cross linking the receptor results in the release of cytotoxic granules, IFN-γ, and TNF-α (25, 28–30). While only found on activated NK cells in circulation, NKp44 is constitutively expressed by a specialized subset of NK cells in the decidua, implicating a role for NKp44 during placentation (25, 31, 32). NKp44 is also expressed on a subset of interferon-producing cells located in human tonsils and ILC3 cells in mucosal-associated lymphoid tissues and human decidua (33–37). Recently, it has been shown that NKp44 is indeed functional in ILC3 and its engagement results in TNF but not in IL-22 production (38).

Crystallographic structure of NKp44 demonstrates a surface groove made by two facing β hairpin loops extending from the Ig fold core stabilized by a disulfide bridge between Cystine 37 and Cystine 45 (39). The Ig domain contains an arrangement of positively charged residues at the groove surface, suggesting that NKp44 ligands are anionic (39). Also, the groove appears wide enough to host a sialic acid or an elongated branched ligand. Interestingly, the cytoplasmic tail of NKp44 contains a tyrosine sequence resembling an ITIM (25, 40). Contrary to initial reports, this motif is functional and inhibits the release of cytotoxic agents and IFN-γ (25, 30, 40). NKp44 surface expression is dependent on its association with the ITAM containing DAP 12 accessory protein linked to NKp44 through Lysine 183 in the transmembrane domain (25). Upon recognition of activating ligands, signaling transduced through the ITAMs in Dap 12 result in release of cytotoxic agents, tumor necrosis factor-α, and IFN-γ (29, 40).

While NK cells utilize NKp44 to recognize and kill targets, tumors may also exploit NKp44 to escape NK cell recognition. By engaging NKp44, as well as the other NCRs, tumors can induce NK cell death via up regulation of Fas Ligand in the NK cell, inducing Fas-mediated apoptosis (41). Tumors may also downregulate NKp44 surface expression by shedding soluble MHC Class I chainrelated molecules or by releasing indoleamine 2,3-dioxygenase and prostaglandin E2 (42, 43). The latter two molecules are released by mesenchymal stem cells as well, inhibiting NKp44 expression in the tumor microenvironment (44). Additionally, tumors can regulate NKp44 ligand expression to escape NK cell killing, as is the case with acute myeloid leukemia (45). Finally, tumor cells may

induce expression of exosomal proliferating cell nuclear antigen (PCNA) when physically contacted by NKp44 expressing NK cells to inhibit NK cell effector function (30).

In addition to its role in immunity, NKp44 also has roles during pregnancy. Decidual NK cells (dNK) make up 50–90% of lymphocytes in the uterine mucosa during pregnancy and constitutively express NKp44 (36, 46, 47). Trophoblast cells and maternal stromal cells of the decidua both express unidentified NKp44 ligands (46). This ligand may be PCNA as the protein is over expressed in trophoblast cells during the first trimester (48). As an inhibitory ligand for NKp44, extracellular PCNA expression on trophoblast cells would help explain the diminished ability of dNK cells to lyse trophoblasts despite low levels of classical HLA I expression (47).

#### **NKp44 TUMOR LIGANDS**

NKp44 is implicated in recognition and killing of numerous types of cancer: neuralblastoma, choriocarcinoma, pancreatic, breast, lung adenocarcionma, colon, cervix, hepatocellular carcinoma, Burkitt lymphoma, diffuse B cell lymphoma, prostate (15, 21, 28). While most of these ligands have not been identified, they appear to be cell cycle regulated, with down regulation of expression during mitosis (28). Recognition of tumor cells is partially mediated through charged-based binding of NKp44 with heparan sulfate proteoglycans (HSPGs) on the surface of tumor cells (49–51). Of note, recognition of HSPG only evokes IFN-γ release by NK cells, not cellular cytotoxicity (49). Thus, HSPGs are believed to only be a co-ligand for NKp44 as well as the other NCRs, potentially facilitating binding with other cellular ligands.

Proliferating cell nuclear antigen is the inhibitory tumor ligand for NKp44 (15, 30). PCNA is a nuclear protein found in all replicating cells, which encircles DNA and increases processivity of DNA replication, but is also involved in DNA repair and cell cycle control (52). NKp44 recognizes PCNA expressed on exosomes shuttled to the surface of tumors cells when in contact with NK cells (15, 30). Recognition of cell surface PCNA colocalizing with HLA I on the cell surface inhibits NK cell cytotoxicity and IFN-γ release (15).

A truncated isoform of mixed-lineage leukemia-5 (MLL5) is an activating cellular ligand for NKp44 (53). This MLL5 isoform contains a specific exon encoding a C-terminus, which interacts with NKp44 (53). Typically located only in the nucleus, MLL5 is a lysine methyltransferase implicated in hematopoietic differentiation and control of the cell cycle (53). Contrary to normal MLL5, the isoform recognized by NKp44 is not found in the nucleus but in the cytoplasm and endoplasmic reticulum, destined to be expressed at the cell surface (53). While MLL5 is expressed in normal tissue, the isoform recognized by NKp44 is only present on tumor and transformed cells (53).

#### **NCR CO-LIGANDS**

Heparan sulfate proteoglycans have been identified as co-ligands involved in the recognition of tumor cells by the NCRs (49, 50, 54). HSPGs are complex glycoproteins found at the cell surface of mammalian cells or in the extracellular matrix (55, 56). Heparan sulfate is characterized by chains of disaccharide units of *N*acetyl-d-glucosamine linked to d-glucuronic acid (55, 57). Interestingly, each NCR recognizes distinct forms of heparan sulfate

epitopes on HSPGs, specifically highly sulfated microdomains on disaccharide units (58). 2-O-sulfation of iduronic acid and *N*-acetylation of glucosamine on HSPGs are important for interaction with NKp44 (50). NKp30 and NKp46 recognize HSPGs with 2-O-sulfation of iduronic acid and either 6-O-sulfation or 6- N-sulfation of glucosamine (50). Interactions between the NCRs and HSPGs are charge based as each NCR contains basic amino acid residues in their binding cleft and HSPGs are heavily charged molecules.

In addition to HSPGs, HLA I may also serve as a co-ligand for the NCRs. We have shown that HLA I and the NKp44 inhibitory ligand, PCNA, associate on the cell surface (15). In our search to identify a ligand for NKp44, several key pieces of evidence suggested that HLA I plays a role in ligand formation. First, HLA I has been demonstrated to coimmunoprecipitate with anti-NKp44 antibodies; reciprocally, NKp44 coimmunoprecipitates with antiβ-2-microglobulin antibodies (59). Additionally, the Nef protein of HIV prevents surface expression of NKp44 ligand isoform of MLL5 on CD4-infected T cells, which is also consistent with the ability of Nef to retain HLA I intracellularly (60, 61). Finally, the NKp30 ligand, Bat3, colocalizes with HLA I on the extracellular membrane of tumor cells, activating NK cell effector functions (62, 63). Interestingly, all 50 alleles of HLA Class A, B, and C molecules harbor an Asparagine at position 86, close to the residues on the α1 helix, which determine interactions with human NK receptors (64). This site allows for attachment of N-linked glycan structures, which could enable binding of other proteins (64). Electron density mapping of the HLA I glycan structure suggests that it is flexible and could serve as a ligand for other receptors or block access to HLA I molecules. Additionally, HLA-A and - B are almost exclusively disialylated, resulting in these molecules having a negative charge, a characteristic of NKp44 ligands (64). This negative charge combined with a protruding oligosaccharide could potentially facilitate interactions with NKp44.

As co-ligands, HSPGs and HLA I most likely facilitate binding of other proteins, which together form a complex ligand recognized by the NCRs. The most prevalent interaction between HSPGs and other proteins is charged based via clusters of positively charged amino acids on proteins forming ionic bonds with negatively charged sulfate and carboxyl groups on HSPGs (57). HSPGs may offer two mechanisms facilitating NCR ligand recognition. First, HSPGs may bind soluble proteins, which as a whole serve as ligands for the NCRs. Second, HSPGs could bind a soluble protein and an NCR separately, and then act as scaffolding to bring the NCRs into contact with a soluble protein. In the same manner, the protruding oligosaccharide of HLA I, or other regions, may enable assimilation of small proteins or DAMPs.

#### **DAMPs AND THE C TERMINUS**

Immune responses are initiated by pattern recognition receptors, which recognize microbial-derived products called pathogenassociated molecular pattern molecules (65). In a similar manner, pattern recognition receptors also recognize molecules released by dying or damaged cells, termed damage-associated molecular pattern (DAMP) molecules or alarmins (66, 67). Recognition of DAMPs contributes to the induction of inflammation, even in the absence of pathogens (68). Normally residing in the

nucleus, cytoplasm, or exosomes, DAMPs lack secretion signals but can be actively secreted by non-classical pathways or passively released by necrotic cells (67). DAMPs thus serve as endogenous danger signals when improperly released from damaged cells as well as tumors and activate innate immune cells (67). DAMPs are most often released after trauma, ischemia, or other tissue damage and initiate early inflammatory responses (67). By recruiting immune cells and promoting the release of proinflammatory mediators, DAMPs activate immune responses and initiate pathways leading to tissue repair and regeneration (67, 69). Heat shock proteins, high-mobility group box 1 (HMGB1), S100 proteins, hyaluronan, and heparan sulfate represent a few DAMPs known to date (68).

Like PAMPs, DAMPs are also recognized by pattern recognition receptors. In addition to binding PAMPs, the Toll-like receptors (TLRs) also recognize HMGB1 and a member of the S100 family (67). These two DAMPs are also recognized by the receptor for advanced glycation end products (RAGE) (67). Like the TLRs, RAGE is expressed on numerous immune cells and induces NFκB-mediated production of cytokines (67). Interestingly, DAMP molecules such as high-mobility group protein B1 and S100A8/9 have the ability to bind heparin sulfate and HSPGs, which are known to be co-ligands involved in NCR-dependent recognition of tumor cells resulting in secretion of IFN-γ but not cytotoxicity (49, 50, 67).

We postulate that DAMPs may serve as the missing link in NCRmediated recognition of tumor cells. The association of DAMPs with HSPGs, HLA I, or other potential co-ligands may form larger complex ligands for the members of the NCR family (**Figure 1**). Human leukocyte antigen-B-associated transcript 3 (Bat3), also known as BAG-6, could be considered a DAMP due to its release from tumor cells (62). Bat3 is typically located in the nucleus where it plays an essential role in controlling the acetylation of p53 in response to cellular DNA damage (62). However, upon non-lethal heat shock, nuclear Bat3 relocates to the cell membrane of tumors where it serves as a ligand for NKp30 (62, 63). Interestingly, this study found Bat3 colocalizes with HLA I, suggesting opposition to previous reports that NCRs do not associate with HLA I molecules (29, 63, 70).

Like Bat3, PCNA and MLL5 are located in the nucleus and cytoplasm. Additionally, all three molecules are intricately involved in processes regulating the cell cycle and/or DNA repair mechanisms. Thus, their presence on the cell surface may indicate intracellular stress related to DNA damage or improper cell cycle control, qualifying these molecules as DAMPs. This suggests that the NCRs may be pattern recognition receptors, which recognize DAMPs sequestered to the cell surface. Taken into context with other studies, the NCRs potentially have the ability to interact with HLA I, HSPGs, or other cell surface molecules as co-ligands in conjunction with soluble proteins, such as Bat3, PCNA, MLL5, or other DAMPs on the cell surface of tumors (62). Therefore, the NCRs may recognize DAMPs on the cell surface in association with a docking protein. Furthermore, the NCRs may be directly recognizing the C-terminal ends of DAMPs. NKp30 was recently shown to recognize the C terminus of Bat3 (71). In a similar manner, NKp44 also recognizes the C terminus of the MLL5 ligand (53). The molecular details of interaction between NKp44 and PCNA

have yet to be determined, but the above precedents suggest that interaction may occur at the C terminus of PCNA.

Natural cytotoxicity receptors ligands consisting of a DAMP and a co-ligand may add complexity in understanding how the NCRs regulate NK cell effector function. The NCRs were originally believed to be strictly activating NK cell receptors. However, NKp44 and NKp30 have recently been shown to exhibit both inhibitory and activating functions. NKp44 recognizes cell surface PCNA in an inhibitory manner while a soluble C-terminal fragment of Bat3 inhibits NK cell function via NKp30 (15, 71). However, recognition of MLL5 by NKp44- and NKp30-mediated recognition of Bat3 sequestered to the cell surface activates NK cell effector functions (53, 62, 63). Thus, modulation of NK cell activity via the NCRs could depend on the DAMP molecule, the co-ligand sequestering the DAMP, or the lack of a coligand and the soluble nature of the DAMP. NKp44 presents a more special case since it contains a functional ITIM-like sequence in its cytoplasmic tail. Due to the dual nature of NKp44 signaling, it will be of interest to determine if recognition of the DAMP, either PCNA or MLL5, the coligand, potentially an HSPG or HLA I, or the motif as a whole is responsible for inhibition or activation of NK cytotoxicity. Neither NKp30 nor NKp46 has been reported to contain an ITIM sequence. However, an immunosuppressive isoform of NKp30 resulting from a single-nucleotide polymorphism in the 3<sup>0</sup> -untranslatable region has been reported (72). Whether the divergence of NCR function depends on the individual DAMP molecule recognized or the binding of DAMP molecules to a specific coligand remains to be elucidated.

#### **CONCLUDING REMARKS**

Recent studies reveal a novel function for DAMP molecules, or proteins, which are located and function intracellular, but somehow localize to the extracellular membrane despite lacking a traditional secretory leader sequence. These proteins are released by cells, which have become injured in the absence of infection due to ischemia, hypoxia, transformation, chemotherapy, DNA damage, or other trauma. Analogous to TLRs recognizing pathogenassociated molecular patterns, the NCRs may represent a class of NK cell receptors that participate in pattern recognition of DAMP molecules, whose identities may reflect the intracellular health of a cell, particularly in regards to DNA damage or instability, in addition to the traditional method of HLA I presenting self peptide. In this manner, HLA I, HSPGs, or other co-ligands may present DAMP molecules for identification by the NCRs, which would then regulate NK cell function. Like other NK cell receptors, the NCRs undoubtedly recognize multiple ligands, which may be cell surface transmembrane proteins, like the recognition of B7-H6 molecule by NKp30 (73). Knowledge of the identities of NCR ligands and nature of DAMP molecules that bind to HLA I, HSPGs, or other cell surface molecules to form complex ligands for the NCRs will shed light on NK cell recognition of target cells under healthy and disease conditions and offer novel therapeutic targets.

#### **ACKNOWLEDGMENTS**

This work was supported by UNTHSC Seed Grant (Mathew).

#### **REFERENCES**


triggering partner in natural cytotoxicity. *Structure (Camb)* (2003) **11**:725–34. doi:10.1016/S0969-2126(03)00095-9


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

*Received: 10 November 2014; accepted: 15 January 2015; published online: 02 February 2015.*

*Citation: Horton NC and Mathew PA (2015) NKp44 and natural cytotoxicity receptors as damage-associated molecular pattern recognition receptors. Front. Immunol. 6:31. doi: 10.3389/fimmu.2015.00031*

*This article was submitted to NK Cell Biology, a section of the journal Frontiers in Immunology.*

*Copyright © 2015 Horton and Mathew. 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.*

# Critical role of tumor microenvironment in shaping NK cell functions: implication of hypoxic stress

*Meriem Hasmim1 , Yosra Messai1 , Linda Ziani1 , Jerome Thiery1 , Jean-Henri Bouhris1,2 , Muhammad Zaeem Noman1 and Salem Chouaib1 \**

*<sup>1</sup> INSERM U 1186, Equipe labellisée Ligue Contre le Cancer, Gustave Roussy Campus, Villejuif, France, 2Department of Hematology and Bone Marrow Transplantation, Gustave Roussy Campus, Villejuif, France*

#### *Edited by:*

*Susana Larrucea, BioCruces Health Research Institute, Spain*

#### *Reviewed by:*

*Stephan Gasser, National University of Singapore, Singapore Kamalakannan Rajasekaran, Blood Research Institute, USA*

#### *\*Correspondence:*

 *Salem Chouaib, INSERM U1186, Gustave Roussy Campus – PR1, 114 rue Edouard Vaillant, Villejuif 94805, France salem.chouaib@gustaveroussy.fr*

#### *Specialty section:*

*This article was submitted to NK Cell Biology, a section of the journal Frontiers in Immunology*

> *Received: 15 July 2015 Accepted: 04 September 2015 Published: 23 September 2015*

#### *Citation:*

*Hasmim M, Messai Y, Ziani L, Thiery J, Bouhris J-H, Noman MZ and Chouaib S (2015) Critical role of tumor microenvironment in shaping NK cell functions: implication of hypoxic stress. Front. Immunol. 6:482. doi: 10.3389/fimmu.2015.00482*

Blurring the boundary between innate and adaptive immune system, natural killer (NK) cells, a key component of the innate immunity, are recognized as potent anticancer mediators. Extensive studies have been detailed on how NK cells get activated and recognize cancer cells. In contrast, few studies have been focused on how tumor microenvironment-mediated immunosubversion and immunoselection of tumor-resistant variants may impair NK cell function. Accumulating evidences indicate that several cell subsets (macrophages, myeloid-derived suppressive cells, T regulatory cells, dendritic cells, cancer-associated fibroblasts, and tumor cells), their secreted factors, as well as metabolic components (i.e., hypoxia) have immunosuppressive roles in the tumor microenvironment and are able to condition NK cells to become anergic. In this review, we will describe how NK cells react with different stromal cells in the tumor microenvironment. This will be followed by a discussion on the role of hypoxic stress in the regulation of NK cell functions. The aim of this review is to provide a better understanding of how the tumor microenvironment impairs NK cell functions, thereby limiting the use of NK cell-based therapy, and we will attempt to suggest more efficient tools to establish a more favorable tumor microenvironment to boost NK cell cytotoxicity and control tumor progression.

#### Keywords: HIF, microenvironment, solid tumors, natural killer cells, immune suppression

#### Introduction

Natural killer (NK) cells are lymphoid cells that are considered to be major innate effector cells. They are endowed with a natural ability to kill tumor cells and infected cells (1). NK cell lytic functions are regulated by a balance of activating and inhibiting signals originating from membrane receptors (1). Despite their effective antitumor activity, their contribution in controlling solid tumor progression remains elusive. The immunosuppressive tumor microenvironment is undoubtedly involved in tumor evasion from NK cell-mediated killing through several cellular and metabolic factors. Immune and stromal cells as well as the hypoxic stress inside the tumor microenvironment are known to be negative regulators of NK cell infiltration into solid tumors and cytotoxicity (1). Tumor cells themselves develop several strategies to evade NK cell-mediated killing. In this regard, hypoxic stress through its ability to induce tumor resistance and to regulate the differentiation and function of immune-suppressive cells plays a determinant role in shaping the NK cell phenotype and function.

In this review, we propose an insight on how tumor microenvironment inhibits NK cell functions and how this may impact the therapeutic use of NK cells in anticancer treatments.

# Cell-Mediated Immune Suppression Toward NK Cells in the Tumor Microenvironment

The tumor microenvironment is a complex network of tumor cells, immune cells, stromal cells, and extracellular matrix accomplishing proliferation, migration, and dissemination of tumor cells. The immune cell subset comprises CD8<sup>+</sup> T cells, CD4<sup>+</sup> T cells, NK cells, and myeloid cells [dendritic cells (DCs), M2-macrophages, myeloid-derived suppressor cells (MDSCs)]. Despite the recognized role of NK cells in clearing circulating tumor cells (leukemia cells, metastatic cells) (2–4), the antitumor functions of NK cells in solid tumors are frequently mentioned due to the more favorable prognosis associated with higher NK cell infiltration in some type of cancers (5), the inverse correlation between natural cytotoxic activity and cancer incidence (6), or the faster tumor growth in NK cell-depleted mouse models (7–9).

Natural killer cells enter solid tumor site by extravasations through tumor vasculature (10). CXCR3 is a major chemokine receptor involved in NK cell migration toward tumor following a gradient of the tumor-derived chemokine (C-X-C motif) ligands CXCL9, 10, and 11 (11, 12). In particular, increased CXCL10 expression in melanoma tumors results in increased infiltration of adoptively transferred CXCR3-positive expanded NK cells, reflecting the role of CXCL10-induced chemoattraction (12). However, infiltrated NK cells often display a suppressed phenotype inside solid tumors. Accumulating evidence indicates that tumor-residing cells as well as a series of microenvironmental factors are endowed with suppressive properties that affect NK cell reactivity and inhibit their functions (**Figure 1**).

#### Macrophage Polarization Regulates NK Cell-Mediated Cytotoxicity

Within the tumoral tissue, macrophages and other myeloid cells constitute a major component of the immune infiltrate (13, 14). They differentiate into tumor-associated macrophages (TAM) with expression of TAM markers such as CD206 (15). Exposure of TAM to tumor-derived cytokines such as IL-4, IL-10, IL-13, and M-CSF is able to convert them into polarized type II or M2 macrophages with immune-suppressive activities resulting in tumor progression (15). M2-polarized macrophages appear to contribute to immune suppression through the production of immunosuppressive factors such as IL-10 and TGF-β (16). Recently, the role of myeloid cells including macrophages in immunosuppression of NK cells has been better understood by the involvement of A2AR receptors (17). Myeloid-selective deletion of A2ARs significantly activates macrophages by favoring M1 polarization, reduces lung metastasis, and increases CD44 expression on tumor-associated NK cells and T cells as well as numbers and activation of NK cells and antigen-specific CD8<sup>+</sup> T cells in lung infiltrates (17). In a xenografted lung carcinoma model, increased expression of surfactant protein-A (SP-A) was reported to be associated with reduced tumor growth and increased M1-TAM and NK cell recruitment and activation at the tumor site (18).

#### Myeloid-Derived Suppressor Cells Suppress NK Cell Activity

Myeloid-derived suppressor cells represent additional myeloid subsets involved in tumor-induced immunosuppression (19). MDSCs comprise immature macrophages, granulocytes, and DCs. Their expansion and immunosuppressive functions are well documented in tumor-bearing mice and cancer patients. As such, the NK cell activity was found to be inversely correlated with MDSC expansion (20, 21). In addition, MDSC-mediated inhibition of NK cells was found to be cell contact dependent via membrane-bound transforming growth factor-β (TGF-β) on MDSC (21) or inhibition of perforin and signal transducer and activator of transcription 5 (Stat5) activity in NK cells (20). MDSC from patients with hepatocarcinoma also show inhibitory effects on autologous NK cells after coculture (22). This inhibition was also found to be cell contact dependent and to involve blocking of the activating receptor NKp30 on NK cells (22).

#### CD4+CD25+ T Regulatory Cells Inhibit NK Cell Cytolytic Functions

T regulatory cells (Treg) are well described for their immunosuppressive functions (23). Studies performed by Trzonkowski et al. (24) and Xu et al. (25) report direct inhibitory effects of Treg on NK cell cytolytic functions and expression of the CD69 activation marker following *in vitro* cocultures. These studies indicate that the production of TGF-β by Treg is at least one mechanism of Treg-mediated NK cell inhibition. *In vivo*, Treg depletion was shown to increase NK cell proliferation by a mechanism involving IL-15Rα expression on DCs (26). In a murine model, such depletion was also shown to favor NKG2D-mediated tumor rejection (27).

#### Dendritic Cells Modulate NK Cell Cytotoxicity

The incrimination of TGF-β in the modulation of NK cell cytotoxicity is also reported when NK cells are cocultured with DCs. Signal transducer and activator of transcription 3 (STAT3) phosphorylation in DCs was reported to be associated with increased secretion of TGF-β, which inhibited NK cell activity, and inhibition of TGF-β restored NK cell functions (28). TGF-β production by DCs can be induced by coculture of immature DC with lung carcinoma cells (29) or by stimulation with LPS (30). Secretion of IL-6 and IL-10 by DC has also been incriminated in dendritic cell-mediated NK cell inhibition (31). Nevertheless, some reports show that DC can also activate NK cell functions. IL-15-stimulated DCs acquire the ability to increase surface expression of the NK cell-activating receptors NKp30 and NKp46, which is associated with an increased tumor target killing (32). This activation is cell contact dependent and required membrane-bound IL-15 on DC. DCs were also reported to induce NK cell proliferation and to activate NKp30 receptor-signaling in NK cells (33).

#### Cancer-Associated Fibroblasts Decrease NK Cell-Mediated Cytotoxicity

Among the stromal cells, modified/activated fibroblasts, often termed cancer-associated fibroblasts (CAFs), are considered to play a central role in the complex process of tumor–stroma interaction. CAFs, the prominent stromal cell population in most types of human carcinomas, are α-SMA (alpha-smooth muscle actin) positive, spindle-shaped cells, which closely resemble normal myofibroblasts but express specific markers [i.e., FAP (fibroblast-associated protein), FSP-1 (fibroblast specific protein 1), and PDGFR-β (platelet-derived growth factor)] together with vimentin (a mesenchymal marker) and the absence of epithelial (cytokeratin, E-cadherin) and fully differentiated smooth muscle (smoothelin) markers (34–36). CAFs differentiate in the tumor microenvironment in a TGF-β-dependent manner from other cell types such as resident fibroblasts, mesenchymal stem cells, and endothelial and epithelial cells (37, 38). In the tumor stroma, CAFs produce and secrete several factors such as extracellular matrix proteins (i.e., collagen I, III, IV), matrix metalloproteinases (MMPs), proteoglycans (i.e., laminin, fibronectin), chemokines (i.e., CXCL1, CXCL2, CXCL8, CXCL6, CXCL12/SDF1, CCL2, and CCL5), vascularization promoting factors (i.e., PDGF and VEGF), and other proteins that affect tumor cells' proliferation, invasiveness, and survival (i.e., TGF-β, EGF, HGF, and FGF) (39). Consequently, CAFs have been involved in tumor growth, angiogenesis, tissue invasion, and metastasis (40).

During the past few years, these activated tumor-associated fibroblasts have also been involved in the modulation of the antitumor immune response, especially by the secretion of soluble immunosuppressive factors in the tumor microenvironment (TGF-β, IL-1β, IL-6, and IL-10) (41). As such, CAFs can potentially affect both innate and adaptive antitumor immune response by increasing the recruitment to tumor of myeloid-derived suppressive cells (MDSC), by decreasing antigen presentation, by increasing the numbers of Tregs, by decreasing T-cell proliferation, cytotoxic T-cell (CTL) function and maturation, or by inhibiting B cell activation and differentiation (41, 42). Different studies also involved CAFs in the modulation of the NK cell functionality. Indeed, the secretion of TGF-β by CAFs could attenuate the expression of NK cell-activating receptors including NKG2D, NKp30, and NKp44 (43, 44). More recently, studies involving melanoma and hepatocellular and colorectal carcinoma-derived fibroblasts have shown that CAFs can decrease NKG2D expression on NK cells surface through the secretion of prostaglandin E2 (PGE2) and/or indoleamine-2,3-dioxygenase (IDO). In parallel, perforin and granzyme B expression (involved in NK cell-mediated killing of target cells) also seem to be decreased by coculture of NK cells with CAFs, affecting their lytic potential (45–47). Altogether, these findings highlight the direct and indirect action of CAFs on various levels of the antitumor immune response and on NK cells antitumor activity within the tumor microenvironment.

#### Tumor Cells May Develop Strategies to Evade NK Cell-Mediated Lysis

In many cancers, tumor cells down-regulate surface expression of MHC-I molecules in order to evade CD8-dependent T-cell killing, making them more susceptible to NK cell-dependent killing. Cancer cells can also up-regulate NKG2D ligands following activation of NFκB or Sp transcription factors (48). However, tumor cells also develop various strategies to inhibit NK cell-mediated cytotoxicity. Indeed, NK cells from multiple myeloma patients were shown to constitutively express the inhibitory receptor PD-1, as compared to NK cells from healthy donors, which contributes to NK cell inhibition by multiple myeloma cells (49). When the authors inhibit the PD-1/PD-L1 axis using lenalidomide or a blocking antibody, they restored NK cell lytic functions against tumor cells. In addition, tumor cells secrete a number of immunosuppressive cytokines such as TGF-β. In this regard, neuroblastoma cell-derived TGF-β has been reported to down-regulate the activating receptor NKp30 (50). Melanoma cells are also able to inhibit the expression of activating NK cell receptors including NKp30, NKp44, and NKG2D, resulting in impairment of NK cell-mediated cytotoxicity (51). This inhibitory effect is mediated via the production of IDO and PGE2 by melanoma cells. Tumor cells may release soluble NKG2D ligands through proteolytic cleavage, resulting in down-regulation of NKG2D and impairment of NK cell lytic functions (52, 53). The inhibitory consequences of releasing soluble NK cell receptor ligands may not be systematic. Indeed, NKp30 activation by tumor-released vesicles containing HLA-Bassociated transcript 3 (BAT3), a ligand for NKp30, was reported (54). Very recently, Deng et al. demonstrated that shedding of the NKG2D ligand MULT1 results in NK cell activation and increased surface expression of NKG2D (55). As suggested by the authors, the differential affinity of MULT1 (high-affinity NKG2D ligand) and MICA/B (low-affinity NKG2D ligand) for NKG2D may explain this discrepancy. Recently, Nanbakhsh et al. reported that the induction of c-myc in leukemic cells resistant to cytarabine resulted in up-regulation of NKG2D ligands (56). Since deregulated expression of c-myc is associated with many cancers in human, it raises the question of the expression of NK cell-activating ligands on c-myc-altered solid tumors. We and others have also provided evidence indicating a role of HIF factors in tumor resistance to NK cell-mediated lysis, which is further detailed in Section "Consequences of Hypoxia-Induced HIF Stabilization on NK Cell Functions."

# Consequences of Hypoxia-Induced HIF Stabilization on NK Cell Functions

Microenvironmental hypoxia is a prominent feature of solid tumors and is involved in fostering the neoplastic process and in the modulation of immune reactivity (57). It results from inadequacies between the tumor microcirculation and the oxygen demands of the growing tumor mass, which leads to a lowering of oxygen partial pressure and a metabolic switch toward glycolysis (58). Tumor hypoxia is a negative prognostic and predictive factor due to many effects on the selection of hypoxia-surviving clones (59), activation of the expression of genes involved in apoptosis inhibition (60), angiogenesis (61), invasiveness and metastasis (62), epithelial-to-mesenchymal transition (63), and loss of genomic stability (64). Accumulating evidences indicate that tumor hypoxia is also involved in the loss of immune reactivity either by decreasing tumor cell sensitivity to cytotoxic effectors or by promoting immunosuppressive mechanisms (57).

#### Cellular Adaptation to Hypoxia Through Hypoxia-Inducible Factors

Cells adapt to hypoxic microenvironment by regulation of hypoxia-inducible family of transcription factors (HIFs). This family comprises three members: HIF-1, HIF-2, and HIF-3. HIF-1 is a heterodimeric protein composed of a constitutively expressed β-subunit and an O2-regulated α-subunit. In the presence of O2, HIF-1α is hydroxylated on proline residue 402 and/or 564 by prolylhydroxylase domain protein 2 (PHD2), resulting in its interaction with the von Hippel-Lindau (VHL) tumor suppressor protein, which recruits an E3 ubiquitin-protein ligase that eventually catalyzes poly-ubiquitination of HIF-1α, thereby targeting it for proteasomal degradation (1). Under hypoxic conditions, hydroxylation is inhibited and HIF-1α rapidly accumulates, dimerizes with HIF-β, binds to the core DNA-binding sequence 50-RCGTG-30 [R being a purine base (adenine or guanine)] in the promoter region of target genes, recruits coactivators, and activates transcription (65). In addition, oxygen-dependent hydroxylation of asparagine-803 by factor inhibiting HIF-1 (FIH-1) blocks the interaction of HIF-1α with the coactivators P300/CBP under normoxic conditions, resulting in suppression of HIF transcriptional activity (66–68). Similar to HIF-1α, HIF-2α is also regulated by oxygen-dependent hydroxylation. HIF-1α and HIF-2α are structurally similar in their DNA-binding regions and dimerization domains but differ in their transactivation domains. Consistently, they share many target genes, but each one also regulates a unique set of genes (69). HIF-3α lacks the transactivation domain and may function as an inhibitor of HIF-1α and HIF-2α (70).

#### Hypoxic and Pseudo-Hypoxic Tumor Cells Are Resistant to NK Cell-Mediated Killing

It is now well established that the hypoxic tumor microenvironment favors the emergence of tumor variants with increased metastatic and invasive potential and alters immune reactivity as well (57).

Fink and colleagues reported the inhibition of NK cell cytotoxicity toward liver tumor cell lines under hypoxic conditions, suggesting for the first time that hypoxia is able to confer tumor resistance to NK cell-mediated cytotoxicity (71). Another study demonstrated that hypoxia decreased the expression of MICA (a NKG2D ligand) on osteosarcoma cell surface with a consistent decrease in the susceptibility of these cells to NK cell-mediated cytotoxicity (72). Consistently, HIF-1α knockdown using small interfering RNA increased the expression of cell surface MICA and concomitantly increased the level of soluble MICA. HIF-1α was also found to be inversely correlated with MIC gene expression, indicating that hypoxia was involved in the inhibition of NK cell reactivity toward tumor cells. Recently, we showed that hypoxiainduced autophagy in tumor cells mediated resistance to CTL (73). In this context, Baginska et al. demonstrated that hypoxia-induced autophagy in tumor cells was also involved in tumor resistance to NK cells via granzyme B degradation in autophagosomes (74).

In the particular context of renal cancer, hypoxic signaling is frequently constitutively active owing to the majority of renal cancers presenting with clear cell carcinoma (ccRCC) histology (75), which is usually associated with mutational or functional inactivation of the *VHL* gene (76). The VHL pathway targets the hypoxia-inducible factors (HIFs) family of transcription factors, in particular HIF-1α and HIF-2α, for ubiquitin-mediated degradation via the proteasome (77). Consequently, VHL inactivation leads to constitutive stabilization of HIFs, a process known as pseudo-hypoxia, and increased expression of HIF target genes. Our group has recently shown that, in VHL-mutated ccRCC cells, HIF-2 stabilization caused by mutated VHL induces upregulation of ITPR1 which is involved in ccRCC resistance to NK cells (78). NK cells were found to induce a contact-dependent autophagy in ccRCC cells that was dependent on ITPR1 expression in tumor cells. Blocking ITPR1 expression in ccRCC cells inhibited NK cell-induced autophagy and suppressed ccRCC resistance to NK cells.

On the contrary, in non-tumoral cells, Luo and colleagues demonstrated that HIF-1α overexpression in HK-2 cells induces MICA expression and enhances NK cell cytotoxicity toward target cells as well as IFNγ secretion by NK cells (79). Antibody blocking experiments using anti-MICA mAb were able to down-regulate NK cell-mediated killing and IFNγ secretion toward HIF-1α-overexpressing HK-2 cells confirming the involvement of MICA in the increased NK cell reactivity.

#### Hypoxia Inhibits NK Cell Functions via HIFs

The specific role of hypoxia and HIFs on NK cells is not well studied.

Balsamo and colleagues showed that NK cells adapt to a hypoxic environment by up-regulating HIF-1α. They demonstrated that, under hypoxia, NK cells lose their ability to upregulate the surface expression of the major activating NK-cell receptors (NKp46, NKp30, NKp44, and NKG2D) in response to IL-2 or other activating cytokines (including IL-15, IL-12, and IL-21). These altered phenotypic features correlated with reduced responses to activating signals, resulting in impaired capability of killing infected or tumor target cells. However, hypoxia does not significantly alter the surface density and the triggering function of the Fc-γ receptor CD16, thus allowing NK cells to maintain their capability of killing target cells via antibody-dependent cellular cytotoxicity (80).

Hypoxic primary tumors were shown to provide cytokines and growth factors capable of creating a pre-metastatic niche and a reduction of the cytotoxic functions of NK cells. In fact, Sceneay et al. reported that injection of mice with hypoxic mammary tumor cells resulted in increased CD11b<sup>+</sup>/Ly6Cmed/Ly6G<sup>+</sup> myeloid and CD3<sup>−</sup>/NK1.1<sup>+</sup> immune cell lineages infiltration into the lung and led to increased metastatic burden in mammary and melanoma experimental metastasis models (81). The cytotoxicity of NK cells was significantly decreased, resulting in a reduced antitumor response that allowed metastasis formation in secondary organs to an extent similar to that observed following depletion of NK cells. Sarkar and colleagues confirmed that hypoxia reduced NK cell killing of multiple myeloma cell lines (82). They showed that hypoxia significantly decreased expression of the activating receptor NKG2D by NK cells and of intracellular granzyme B and perforin. Whether HIF factors were able to directly regulate the expression of granzymes genes is not documented, but perforin has been reported not to be a direct target gene of HIF-1 (83).

Despite detailed description of the detrimental effects of hypoxia on NK-cell responses, the underlying molecular mechanisms remain unclear. In particular, whether HIF or other hypoxia-related factors are able to directly control NK cell receptor expression remain to be clarified.

#### Indirect Consequences of Hypoxic Stress on NK Cell Cytotoxic Functions

Despite the direct consequences of hypoxic stress on NK cells, intratumoral hypoxia is also involved in increased tumor infiltration by Treg and MDSC and in M2-polarization (57), which are cellular subsets that negatively regulate NK cell lytic functions (see Cell-Mediated Immune Suppression Toward NK Cells in the Tumor Microenvironment). Hypoxic stress is also involved in increased expression and secretion by tumor cells of NK cell-inhibiting cytokines such as TGF-β (84, 85). Of note, NK cell adhesion on hypoxic endothelial cells was reported to be not altered (86), but NK cell infiltration into hypoxic tumors has not been extensively studied.

## NK Cell: A Role in Tumor Immunoediting

The major focus of immunotherapy approaches has been enhancing the effectiveness of host antitumor immunity. However, while accumulating evidences indicate that tumor microenvironment might evade the innate host immune response to ensure tumor development and survival (58, 87), NK cells have also been reported to play a role in the selection of tumor-resistant and tumor-tolerant cells and therefore to shape tumor microenvironment (88). While the mechanisms of CTL-induced tumor editing are well known (89), only limited knowledge on how NK cells induce tumor editing is available. In this regard, the involvement of NK cells in immune editing has been studied in relation to NKG2D and DNAM1 (90, 91). Guillerey and Smyth have elegantly demonstrated the NK cell activity in the cancer immune editing process with particular emphasis on the elimination and escape phases (92). NK cells have been also shown to kill immature DC because of their low amount of surface human leukocyte antigen (HLA) class I molecules (33) and therefore impact the quality of adaptive immune response. In this regard, Ghadially et al. have reported data indicating that in the absence of NKp46, graft-versus-host disease (GVHD) is greatly exacerbated, resulting in rapid mortality of the transplanted animals (93). Furthermore, these authors have demonstrated that the exacerbated GVHD is the result of an altered ability of immune cells to respond to stimulation by immature DCs (94). Buchser et al. have shown that classic cytolytic cells, including NK cells, can often promote survival and autophagy in target cells (95). These authors provided evidence indicating that NK cells are a primary mediator of autophagy in tumor target cells by a mechanism involving cytokines (IL-10, IL-2, IFNγ) and that cell-to-cell contact strongly enhanced lymphocyte-mediated autophagy. The authors suggested that the NK cell-mediated autophagy promotes cancer cell survival and may represent an important target for development of novel therapies.

# NK Cell-Based Immunotherapies in the Context of Tumor Microenvironment Complexity and Heterogeneity

Natural killer cell-based immunotherapies have the advantage of circumventing antigen recognition restriction since NK cells do not need antigen recognition to kill tumor targets. In this context, NK cell infusion has been useful in leukemic patients, probably due to their primary location being the blood. Indeed, most of hematological malignancies display an autologous NK cell deficiency specifically in myeloid diseases. Autologous NK cells do not control acute myeloid leukemia (AML) blasts and several mechanisms have been hypothesized: down-regulation of the ligands for NK-cell activating receptors or up-regulation of NK cell inhibitory receptors (96). Allogeneic NK cells do not bear this deficiency and have demonstrated their ability to kill AML blasts targets. In this context, the killer Ig-like receptor (KIR)-ligand mismatch is considered fundamental for their antitumor effects (97, 98, 99). Modulating the immune reconstitution following allogeneic transplantation with NK cells is a potential powerful tool to increase the graft versus leukemia (GvL) effect against AML blasts and tumor cells (100). NK cells do recover early following allogeneic transplantation and exert cytotoxicity through MHC unrestricted killing. High numbers of allogeneic circulating NK cells improved remission duration in patients with leukemia and consolidate engraftment following haploidentical transplants (101).

On the other hand, the therapeutic potential of NK cells in solid tumors is not yet clearly established. However, preclinical studies support the antitumor activity of NK cells against solid tumors (102). Phase I and phase II clinical trials based on adoptive transfer of irradiated NK cell lines or allogeneic NK cells have been made in breast, ovarian, melanoma, and renal cancer patients (98, 103). These trials revealed mild and transient toxicities following NK-92 infusion and some severe syndromes following allogeneic NK cell administration. Further studies are still needed to increase NK cell persistence and expansion.

Moreover, producing sufficient amounts of allogeneic NK cells for clinical applications remain a technical challenge in cell therapy programs despite their useful and safe infusion 10 years ago (104). Dampening negative regulators of NK cell lytic functions should also be explored, in particular in the context of solid tumors. Strategies aimed at inhibiting NK cell suppressors such as TGF-β, expansion of immunosuppressive cells, and expression of inhibitory checkpoints should be considered. In particular, targeting HIF-1α by antisense plasmid in xenografted mice led to NK cell-dependent tumor rejection (105). Various anticancer drugs have been shown to inhibit HIFs (106, 107). We believe that pharmacologic manipulation of hypoxic signaling will result in increased target killing by effector cells and in general improving of antitumoral immunotherapy. Whether the suppression of hypoxia may be a promising strategy that is selective for facilitating immunotherapeutic efficacy in cancer patients is at present investigated. Nevertheless, a better understanding of functionally distinct KIR or NK cell receptor subsets within NK cell population is still needed for designing optimal immunotherapy based on NK cell administration or reactivation.

# Conclusion

During the last few years, cancer immunotherapy has emerged as a safe and effective alternative to cancers that do not respond to classical treatments including those types with high aggressiveness. New immune modulators like cytokines, blockers of CTLA4/CD28 and PD-1/PD-L1 interactions, or adoptive cell therapy have been developed and approved to treat solid tumors and hematological malignant diseases. In these scenarios, cytotoxic lymphocytes mainly CTLs and NK cells are the ultimate responsible for killing the cancer cells and eradicating the tumor. Many mechanisms have been proposed for the functional inactivation of tumor-associated NK cells. Thus, the definition of tumor microenvironment-related immunosuppressive factors, along with the identification of new classes of tissue-residing NK cell-like innate lymphoid cells, represents key issues to design effective NK-cell-based therapies for solid tumors.

# Acknowledgments

The authors are supported by the Ligue contre le Cancer.

# References


epithelial cells during hypoxia/reoxygenation. *BMC Cell Biol* (2010) **11**:91. doi:10.1186/1471-2121-11-91


**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 Hasmim, Messai, Ziani, Thiery, Bouhris, Noman and Chouaib. 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.*

# Natural Killer Cell-Based Therapies Targeting Cancer: Possible Strategies to Gain and Sustain Anti-Tumor Activity

#### *Carin I. M. Dahlberg1,2 , Dhifaf Sarhan3,4 , Michael Chrobok1,2 , Adil D. Duru1,2 and Evren Alici1,2,5\**

*1Cell Therapies Institute, Nova Southeastern University, Fort Lauderdale, FL, USA, 2Cell and Gene Therapy Group, Center for Hematology and Regenerative Medicine (HERM), Karolinska University Hospital Huddinge, NOVUM, Stockholm, Sweden, 3Oncology-Pathology, Cancer Center Karolinska, Karolinska Institutet, Stockholm, Sweden, 4Division of Hematology, Oncology and Transplantation, Masonic Cancer Research Center, University of Minnesota, Minnesota, MN, USA, 5Hematology Center, Karolinska University Hospital Huddinge, Stockholm, Sweden*

#### *Edited by:*

*Francisco Borrego, Cruces University Hospital, Spain*

#### *Reviewed by:*

*Subramaniam Malarkannan, Medical College of Wisconsin, USA Vincent Vieillard, Institut National de la Santé et de la Recherche Scientifique (INSERM), France*

> *\*Correspondence: Evren Alici evren.alici@ki.se*

#### *Specialty section:*

*This article was submitted to NK Cell Biology, a section of the journal Frontiers in Immunology*

> *Received: 02 September 2015 Accepted: 13 November 2015 Published: 30 November 2015*

#### *Citation:*

*Dahlberg CIM, Sarhan D, Chrobok M, Duru AD and Alici E (2015) Natural Killer Cell-Based Therapies Targeting Cancer: Possible Strategies to Gain and Sustain Anti-Tumor Activity. Front. Immunol. 6:605. doi: 10.3389/fimmu.2015.00605*

Natural killer (NK) cells were discovered 40 years ago, by their ability to recognize and kill tumor cells without the requirement of prior antigen exposure. Since then, NK cells have been seen as promising agents for cell-based cancer therapies. However, NK cells represent only a minor fraction of the human lymphocyte population. Their skewed phenotype and impaired functionality during cancer progression necessitates the development of clinical protocols to activate and expand to high numbers *ex vivo* to be able to infuse sufficient numbers of functional NK cells to the cancer patients. Initial NK cell-based clinical trials suggested that NK cell-infusion is safe and feasible with almost no NK cell-related toxicity, including graft-versus-host disease. Complete remission and increased disease-free survival is shown in a small number of patients with hematological malignances. Furthermore, successful adoptive NK cell-based therapies from haploidentical donors have been demonstrated. Disappointingly, only limited anti-tumor effects have been demonstrated following NK cell infusion in patients with solid tumors. While NK cells have great potential in targeting tumor cells, the efficiency of NK cell functions in the tumor microenvironment is yet unclear. The failure of immune surveillance may in part be due to sustained immunological pressure on tumor cells resulting in the development of tumor escape variants that are invisible to the immune system. Alternatively, this could be due to the complex network of immune-suppressive compartments in the tumor microenvironment, including myeloid-derived suppressor cells, tumor-associated macrophages, and regulatory T cells. Although the negative effect of the tumor microenvironment on NK cells can be transiently reverted by *ex vivo* expansion and long-term activation, the aforementioned NK cell/tumor microenvironment interactions upon reinfusion are not fully elucidated. Within this context, genetic modification of NK cells may provide new possibilities for developing effective cancer immunotherapies by improving NK cell responses and making them less susceptible to

the tumor microenvironment. Within this review, we will discuss clinical trials using NK cells with a specific reflection on novel potential strategies, such as genetic modification of NK cells and complementary therapies aimed at improving the clinical outcome of NK cell-based immune therapies.

Keywords: natural killer cells, adoptive cell therapy, immunotherapy, cancer, clinical trials, expansion, tumor microenvironment, genetic modifications

## INTRODUCTION

Natural killer (NK) cells are lymphocytes of the innate immune system. They are cytokine producing and have cytotoxic ability to kill both viral infected and tumor cells. Tumor-killing lymphocytes were first reported in 1968 by Hellström et al. (1). Kiessling and colleagues, in parallel with Ronald Herberman's research laboratory, defined a novel lymphocyte population named NK cells that are able to target tumor cells in 1975 (2–5). Unlike T cells and B cells, NK cell recognition is not governed by highresolution antigen specificity. Target cell recognition is mediated by the signals delivered through several activating and inhibitory receptors. The balance between activating and inhibitory signals decides the response of NK cells. When there is a mismatch between an inhibitory subgroup of killer immunoglobulin-like receptors (KIRs) on NK cells and self-human leukocyte antigen (HLA) class I proteins on the surface of target cells the NK cells can get activated due to lack of inhibitory signals leading to lysis of the host cell. This mismatch mediates alloreactivity and is the strategy behind the missing-self concept (6). KIRs can be divided into two haplotypes; the A haplotype with predominantly inhibitory KIRs plus only one activating KIR and the B haplotype containing inhibitory and activating receptors (7). During NK cell education, KIRs go through a random sequential acquisition process where they get functionally competent after they encounter self-MHC class I molecules. Consequently, mature NK cell function is inhibited by self-MHC class I and KIR interaction (8). When a NK cell confronts a target cell without expression of self-MHC class I molecules, the inhibitory signals are not active and the NK cell gets activated.

The majority of NK cells, as well as certain T cell subpopulations, may express the receptor family NKG2. One of the ligands for most NKG2 receptors is HLA-E, which is expressed on all nucleated cells. NKG2-family consists of seven members: NKG2A, B, C, D, E, F, and H in which NKG2A and B are inhibitory receptors. NK cells also express activation receptors on the surface, such as natural cytotoxicity receptors (NCRs), DNAM-1, and receptor members of the 2B4 family. NCRs, including NKp30, NKp44, and NKp46, are one of the main and initial groups of NK cellactivating receptors identified and they recognize viral ligands, heat shock-associated proteins, or tumor antigens (9). NK cells can also get activated by crosslinking of Fc receptor CD16 to target cell leading to antibody-dependent cellular cytotoxicity (ADCC) and lysis of the target cell (10, 11).

Natural killer cells perform their cytotoxic activity through granzyme B- and perforin-mediated apoptosis or by expression of death receptor ligands such as FasL and TNF-related apoptosisinducing ligand (TRAIL). While the release of cytolytic granules is one of the essential cytotoxic responses, perforin deficient NK cells can still kill tumor cells through Fas-mediated apoptosis (12). Moreover, TRAIL-TRAILR mediated cytotoxicity also plays an important role in eliminating the target cells. Various tumor cells express TRAIL death receptors, which could be upregulated by proteasome inhibitors such as bortezomib (13). Additionally, immunomodulatory drugs (IMiDs) such as lenalidomide upregulates TRAIL expression on NK cells that potentially enhance the TRAIL-mediated elimination of tumor cells (14, 15).

Natural killer cells are derived from hematopoietic stem cells (HSC) in the bone marrow. The differentiation from HSC can be divided into five stages based on surface markers [detailed review in Ref. (16)]. The stages can be identified by the following surface markers, CD34, CD117, CD94, and CD16 among the Lin<sup>−</sup> events, where stage 1 is CD34<sup>+</sup>CD117<sup>−</sup>CD94<sup>−</sup>CD16<sup>−</sup>. First at stage 2, the cells are able to respond to IL-15, which is necessary for NK cell development (17, 18). In the transition between stage 2 and 3, they lose their CD34 expression. At stage 4, the NK cells are CD56bright, produce IFNγ, and are capable of cytotoxic killing of K562 cells *in vitro* (19). NK cells in stage 5 are CD56dim and express CD16.

The majority of human NK cells are CD14<sup>−</sup>CD19<sup>−</sup>CD3 <sup>−</sup>CD56<sup>+</sup>. While most of the CD56<sup>+</sup> cells express lower levels of CD56 (~90% CD56dim), they are potent cytotoxic killers of target cells and secrete cytokines such as IFNγ. Approximately 10% of peripheral NK cells express high levels of CD56 (CD56bright), have low cytolytic activity, and have the capacity to produce high titers of immunoregulatory cytokines. The cell surface phenotypes of these two subpopulations also differ in respect to the receptors they express: the CD56bright population expresses the inhibitory receptor NKG2A that could also be expressed on CD56dim NK cells. While the CD56dim population expresses FcγRIIIa (CD16a) as well as the inhibitory receptors KIRs (20).

#### NK CELLS IN CANCER

Natural killer cells recognize tumor cells by the activating receptors like NCRs, which detect the altered expression of their ligands on the tumor cell surface. Additionally, downregulation or lack of MHC class I molecules on the cell surface of tumor cells can trigger NK cell activation since it diminishes the inhibitory signals transduced through KIR-MHC interactions. Moreover, since NK cells' target recognition and activation are mainly through NCRs and missing-self, this engagement could induce upregulation of FasL on the NK cell surface leading to an alternative pathway inducing apoptosis in tumor cells. Nevertheless, both IL-2 stimulation and NK cell activation through NCRs also upregulate Fas on NK cells that may initiate regulation of the NK cell activation and expansion (21, 22).

Many tumors have gained methods to evade the surveillance by NK cells and other members of the immune system. For example, 16 of 18 patients with acute myeloid leukemia (AML) had reduced NCR surface expression compared to healthy donor NK cells, resulting in reduced cytotoxic capacity against target cells (23). Another way for tumor cells to escape recognition by NK cells is upregulation of the non-classical MHC class I molecule HLA-G, which dampens NK cell responses (24, 25). In numerous malignancies, there are also abnormalities found in the NK cell population. Examples of this include defective expression of activating receptors found in hepatocellular carcinoma (26), metastatic melanoma (27), AML (23), chronic lymphocytic leukemia (CLL) (28), and multiple myeloma (29, 30) or defective NK cell proliferation in metastatic renal cell carcinoma (31) and chronic myelogenous leukemia (CML) (32).

In renal cell carcinoma, infiltrating NK cells have, compared to peripheral blood NK cells, increased expression of NKG2A receptor contributing to decreased NK cell activity (33). NKG2D is a well-studied activating receptor on NK cells. Membrane-bound NKG2D ligand has a stimulatory effect on immunity, while soluble NKG2D ligands have the opposite effect on immune system leading to metastatic cancer progression (34). Patients with colorectal cancer have increased serum titers of the soluble NKG2D ligand, MHC class I chain-related protein A (sMICA), compared to healthy controls, leading to downmodulation of activating and cytokine receptors on the NK cells (35). A potential way to reduce the risk of soluble NKG2D ligand is to give the patients neutralizing antibody treatment. Clinical observations demonstrate that patients treated with cytotoxic T lymphocyte-associated antigen 4 (CTLA-4) antibody blockade have reduced sMICA in a close correlation with increased titers of autoantibodies against MICA (36). Interestingly, a new report from Deng et al. shows that the soluble high-affinity ligand MUL1 causes NK cell activation and stimulates tumor rejection in mice, instead of inhibition of NK cells as earlier reported (37).

The potential benefits of NK cell-based cancer immunotherapy products have led to the design of *in vitro* methods aiming to cultivate NK cells in cGMP conditions. Some of these methods have already been tested in clinical trials, which will be discussed later in this review.

## CLINICAL-GRADE NK CELL PRODUCTS

It is possible to activate NK cells and increase their anti-tumor activity through short-term cytokine exposure *in vitro* prior to adoptive transfer (38). However, to achieve clinically relevant numbers of NK cells, there also needs to be development of longterm NK cell expansion protocols (**Table 1**; **Figure 1**) (39–47). Yet, there are concerns when expanding NK cells *in vitro*, such as potential phenotypic changes, selective expansion, and reduced cytotoxic killing. When expanded *in vitro* with IL-2, there is a chance of CD3<sup>+</sup> cell expansion as well (48, 49). Thus, there is still room for improvement to achieve optimum clinically relevant NK cell numbers, *in vivo* NK cell persistence and survival, and most importantly, anti-tumor activity. There are numerous parameters affecting the clinical-grade NK cell manufacturing such as source of the NK cells, cytokine stimulation, cell culture medium, and expansion platform. Here, in this section, we will address these parameters.

#### Source of the NK Cells

The majority of clinical NK cell products or pre-clinical research on efficient NK cell manufacturing platforms are making use of peripheral blood mononuclear cells (PBMC), umbilical cord blood (UCB), cell lines, and human embryonic stem cells (hESC), as well as induced pluripotent stem cells (iPSC) as a source of start material.

#### Peripheral Blood Mononuclear Cells

The majority of NK cell products are generated through utilization of PBMCs either by apheresis or ficoll separation under cGMP conditions. An advantage of using PBMCs is the ability to collect cells in a closed aseptic system. Although PBMC consists of 5–20% NK cells, it is not possible to achieve sufficient numbers of potent NK cells. Thus, various techniques to expand NK cells *ex vivo* have been developed. For example, we have designed a feeder-free NK cell expansion system where it is possible to expand and activate tumor-reactive NK cells in a clinically compatible manner (45). These cells have a high cytotoxic effect specifically against autologous and allogeneic tumors *in vitro* and *in vivo* (42, 45). We have also completed a first-in-man clinical trial using donor-derived *ex vivo* expanded NK cells in terminal cancer patients that had CLL, kidney cancer, colorectal cancer, and hepatocellular carcinoma with promising results (43). Having optimized the procedure for NK cell expansion in a closed-automated bioreactor using clinical-grade GMPcompliant components, we have initiated a first-in-man phase I/ II clinical trial to expand and restore the function of patients' own NK cells (45, 62). To our knowledge, this is the first advanced therapy investigational medicinal product trial performed using autologous NK cells in Sweden.

Sakamoto et al. have established another similar approach that generates large numbers of activated NK cells from peripheral blood without prior purification of the cells. The PBMCs are cultured with autologous plasma, IL-2, OK-432, and γ-irradiated autologous FN-CH296 stimulated T cells, reaching up to a median purity of 90.96% of NK cells at day 21 or 22. Many of the NK expansion protocols are based on enrichment of NK cells either prior to NK cell activation and expansion through cell selection or sorting in order to achieve pure cell therapy product and avoid unwanted side effects stemming from T cells especially in allogeneic NK cell transfusions.

One of the main methods of enriching the purity and the number of initial NK cells is the clinical-grade immuno-magnetic depletion of other lymphocyte subsets such as T cells and/or B cells as well as myeloid cells (60). Depletion of CD3<sup>+</sup> cells followed by CD56<sup>+</sup> cell enrichment can lead to highly pure NK cells which could be supplemented by CD19<sup>+</sup> cell depletion before infusion in order to prevent passenger lymphocyte syndrome in allogeneic transplantation (63). Nguyen et al. have shown that a partial depletion of T cells could get a more beneficial clinical outcome compared to a complete T cell depletion after hematopoietic stem cell transplantation, suggesting that T cells may have a positive role in *in vivo* NK cell function (64).

#### TABLE 1 | Clinical-grade NK cell products.


*PBMC, peripheral blood mononuclear cells; HS, human serum; FBS, fetal bovine serum; HP, human plasma; HAS, human serum albumin.*

Additionally, direct enrichment of CD56<sup>+</sup> NK cells through immuno-magnetic selection is an option to achieve high purity initial NK cell product. Nevertheless, NK cells might require physical and cytokine-dependent communication with other cells such as monocytes (65) in order to activate and expand. Thus, it is essential to fine-tune the enrichment of NK cells by making use of feeder cells and/or optimizing the cytokine cocktail used in *ex vivo* NK cell expansion protocols.

Furthermore, using feeder cells and cell lines is another approach in expanding NK cells *ex vivo* since feeder cells can provide essential stimulatory signals for NK cells proliferation. Monocytes, irradiated PBMC, feeder cell lines, and engineered feeder cell lines are the most commonly used sources for stimulation of NK cell expansion through humoral signals and cell-to-cell contact. Example of feeder cells that have been used in clinical trials are irradiated autologous PBMCs (60, 66), irradiated Epstein–Barr virus-transformed lymphoblastoid cells (61), and K562 engineered cells expressing 4-1BB ligand (67) or membrane-bound IL-21 (68, 69) on cell surface.

#### Stem Cells

While PBMC is one of the major sources for achieving clinically relevant doses of tumor-reactive NK cells, HSC and potentially hESC as well as iPSC are likewise essential sources for achieving clinically relevant doses of NK cells.

One of the potential sources to accomplish clinically relevant doses of tumor-reactive NK cells is making use of HSC (CD34<sup>+</sup>) through differentiation and expansion of CD34<sup>+</sup> cells isolated from bone marrow, peripheral blood, or UCB into functional NK cells. It was recently demonstrated that it is possible to expand activated, tumor cytotoxic and pure NK cells by differentiating UCB CD34<sup>+</sup> HSC under cGMP condition (46). Furthermore, NK cells derived from CD34<sup>+</sup> UCB cells lack expression of KIRs such as KIR2DL1 (CD158a), KIR2DL2/DL3 (CD158b), and NKB1, as well as diminished CD16 expression in the CD56dim population (70). Even though NK cells derived from UCB have reduced cytotoxicity, this could be restored by *ex vivo* cytokine stimulation such as IL-2, IL-12, and IL-15 (50, 71–73). Infusion of UCB-derived NK cells supplemented with IL-15 has shown to inhibit growth of human bone marrow resident leukemia cells *in vivo* (74). Recently, it was demonstrated that frozen UCB CD34<sup>+</sup> cells differentiate into NK cells with better expansion than freshly isolated UCB CD34<sup>+</sup> cells, and more importantly, UCB CD34<sup>+</sup> cells gave more NK cell product than peripheral blood HSC without jeopardizing NK cell functionality (75). Thus, UCB CD34<sup>+</sup> cells are one of the essential sources for manufacturing NK cell therapy protocols, providing an option to create NK cell biobanks.

Another potential source of NK cells is hESC and iPSC, with the advantage of potential usage of iPSCs in autologous settings with reduced risk of immune rejection. The first step is to generate CD34<sup>+</sup> hematopoietic precursor cells from the hESCs and iPSCs and then differentiate these cells into NK cells, which could be efficiently achieved through growing hESCs and iPSCs on murine stromal cells (76, 77). Yet, the involvement of xenogeneic cells could limit the potential clinical usage of hESCs and iPSCs. Addressing this potential problem, Knorr et al. developed a two-stage culture method where hESCs and iPSCs are first differentiated to CD34<sup>+</sup> hematopoietic cells by spin-EB system in xeno-free and serum-free conditions followed by stroma-free NK cell differentiation, which enables generation of cytotoxic NK cells without involvement of xenogeneic cells taking a step forward toward clinical-scale production (78). Since IL-2-activated NK cells are potent killers of both allogeneic and autologous iPSCs (79), it is possible to manufacture a pure NK cell therapy product. This sticks out as one of the advantages of using *in vitro* NK cell differentiation from iPSCs.

#### Cell Lines

Cell lines derived from NK cells with similar biological functions (NK-92, NKL, KYHG-1, and NKG) are potential candidates for NK cell-based products enabling design and development of off-the-shelf anti-cancer cell therapy products. Furthermore, it is more feasible to generate genetically modified NK cell lines expressing intracellular IL-2 for activation or cell surface molecules such as CD16, NCRs, and chimeric antigen receptors (CARs). To our knowledge, the NK-92 cell line is the most clinically studied one. The IL-2-dependent NK-92 cell line is cytotoxic to a wide range of malignant cells (80–83). It has also been used as a source of NK cells for cGMP-grade cellular therapy products (51) as well as in clinical trials (52, 84). The NK-92 cell line expresses several activating receptors but lacks most of the inhibitory KIRs, NKp44, and CD16 (80, 85). NK-92 cells require irradiation to prevent proliferation prior to being used effectively in immunotherapeutic approaches without compromising hematopoietic cell function. For example, recently, clinical-grade NK-92 cells have been manufactured and were safely used as anti-tumor therapy for patients with a variety of tumors (84) with promising results (52). As of today's date, two phase I clinical trials (NCT00900809 and NCT00990717) are recruiting patients with hematological malignancies for treatment with NK-92 cells. The first clinical phase II study (NCT02465957) with NK-92 cells has recently been initiated.

KHYG-1 is the first NK cell line derived from NK leukemia and has higher cytotoxicity than NK-92 cell line (86). Likewise NK-92 cells, these cells can also be irradiated to inhibit proliferation and can still efficiently kill tumor targets. Furthermore, NKL cell line, which is the most biologically and functionally similar to primary NK cells, is more cytotoxic to certain tumor cells than NK-92 cell line and, additionally, it has the ADCC capacity whereas NK-92 cells lack CD16 expression. Thus, both KHYG-1 and NKL cell lines have the potential to be used as anti-cancer NK cell products.

Additionally, one of the advantages of using such master cell bank is an appealing opportunity in the manufacture of cellular therapy products since it is possible to establish a comprehensive standardization and characterization of the cell source. It is also possible to genetically modify these cell lines to exert more tumor specificity and cytotoxicity. For example, NK-92 cell lines are dependent on external IL-2 stimulation, which increases manufacturing costs as well as potentially reducing the long-term cytotoxic capacity of these cells unless they are supported by IL-2 infusions. Thus, constitutive expression of IL-2 in NK-92 cells through genetic modification leads to auto-activated and -proliferating cells, which reduces the manufacturing costs as well as potentially increases the *in vivo* tumor reactivity (87, 88).

#### Cytokines

*Ex vivo* manufacturing of NK cell-based products is dependent on extensive use of cytokines to stimulate, differentiate, activate, and expand NK cells in order to get clinically relevant doses and enhanced anti-tumor reactivity. Historically, one of the most popular cytokines in NK cell research is IL-2 since it was the first cytokine to be injected to patients to treat metastatic melanoma (89). Thirty years ago, Rosenberg et al. published the first report where they treated 25 metastatic cancer patients, who did not respond to standard therapy, with autologous lymphokineactivated killer (LAK) cells together with recombinant-derived IL-2. LAK cells are generated from mononuclear cells collected from IL-2 injected patients. In 11 patients, the cancer regression was observed with >50% of tumor volume (90). This adoptive immunotherapy was followed by a larger scale study, where 157 patients with advanced metastatic cancer were treated with successful results (91). In the same year, it was shown that it was the NK cells that mediated the cytotoxic activity in response to systemic administered recombinant IL-2 (92). These reports were followed by many years of IL-2 and NK cell research. In a dose-dependent manner, IL-2 is important for NK cell infiltration and killing of the tumor. For example, in the bone marrow, there are hypoxic regions leading to reduced NK cell killing of plasma cells in multiple myeloma. IL-2-activated NK cells *ex vivo* have increased NKG2D expression resulting in increased targeting of multiple myeloma upon infusion (93). Cytokine-activated NK cells *in vitro* are dependent on constant stimulation both *in vitro* and *in vivo*. Basse et al. reported that when no exogenous IL-2 is present the amount of injected NK cells found in tumors were very low (94). The half-life of IL-2 in serum is not more than 10 min, which makes the administration of IL-2-dependent cells difficult (95). By transducing NK cells to produce IL-2 prior to transplantation, the activated NK cells would have a constant source of IL-2 *in vivo* (87, 96). One of the disadvantages of using IL-2 to activate NK cell *in vivo* is the competition over IL-2 by regulatory T cells, which express high levels of the high-affinity receptor for IL-2, IL-2Rα (CD25). By treating patients with lympho-depleting agents (fludarabine and cyclophosphamide) followed by NK cell infusion and IL-2 fused with diphtheria toxin (IL-2DT), CD25<sup>+</sup> cells are selectively depleted, leading to increased NK cell expansion and complete remission rate for patients with AML compared to regular IL-2 treatment (97). Overall, the majority of cGMP-grade NK cell therapy protocols include IL-2 as a main cytokine to stimulate NK cell activation and proliferation.

Another important cytokine is IL-15 which is required for both NK cell maturation and survival (98). IL-2 and IL-15 share the same receptor components: IL-2/15Rβ and common γ chain (also shared with IL-4, IL-7, IL-9, and IL-21). Recent advances in the production of cGMP quality cytokines enabled further optimization of cytokine supplementation during NK cell expansion. For example, use of IL-15 in combination with IL-2 has a synergetic effect on product viability and NK cell proliferation (66). This highlights the necessity of other cytokines to achieve NK cell product potency especially when it comes to the NK cell expansion protocols that are not using feeder cell support. Additionally, IL-21, primarily described in 2000 (99), has significant homology with IL-2 and IL-15. Compared to IL-2 and IL-15, IL-21 promotes maturation and survival but does not promote proliferation of NK cells alone. However, IL-21 does have synergetic effects with IL-2 and IL-15 (100). Interestingly, it has been suggested that IL-21 does not drive proliferation of regulatory T cells *in vivo* and might be a good candidate to substitute for IL-2 in CLL (101).

#### Other Factors

Besides NK cell source, feeder support, and cytokine stimulation, other parameters such as expansion platform, cell culture media, and serum supplementation are also very important in achieving clinically relevant cell numbers, viability, and tumor cytotoxicity. More specifically, we have recently investigated the importance of the culture vessels on the quality and efficacy of the NK cell product. Briefly, PBMCs from healthy donors and myeloma patients were cultured for 21 days using flasks, cell culture bags, and bioreactors. Even though we have achieved high yield in NK cell expansions in all systems, NK cells expanded in the bioreactor displayed significantly higher cytotoxic capacity. These results demonstrate that highly active NK cells can be produced in a closed, automated, large-scale bioreactor under feeder-free current GMP conditions facilitating adoptive immunotherapy clinical trials (45).

Additionally, cell culture media is another important factor to consider in the manufacturing of cellular therapy products. There are very few cGMP quality medias that work optimally for *ex vivo* NK cell expansion protocols. The most commonly preferred media in the generation of NK cell products are stem cell growth medium (SCGM; CellGenix, Freiburg, Germany), X-VIVO serum-free media (BioWhittaker, Verviers, Belgium), or AIM V (Life Technologies, Grand Island, NY, USA) (49, 102, 103). Generally, medium is supplemented by human AB serum or fetal bovine serum from certified sources.

Finally, there are numerous variables that may impact quality and quantity of NK cell products. Future pre-clinical research and results from more clinical trials will evaluate the contribution of each factor to the product purity, potency, and safety, as well as assist in acquiring NK cell products that can be manufactured reproducibly with the optimal safety and anti-tumor responses.

# CLINICAL USE OF NK CELL-BASED ANTICANCER PRODUCTS

#### Autologous NK Cells

Several clinical studies have been performed with adoptive autologous NK cells in an attempt to target tumors, such as breast cancer, lymphoma, glioma renal cell carcinoma, non-small cell lung cancer, and adenocarcinoma (**Table 2**) (39, 40, 55, 103–107). In general, autologous NK cell trials are safe with no toxic side effects (39, 40, 55, 105). For example, *ex vivo* activated autologous peripheral blood lymphocytes get enhanced cytolytic activity against heat shock protein 70 (Hsp70) membrane-positive tumors *in vivo* if pre-incubated with Hsp70 peptide and IL-2 (105, 107). However, some clinical trials with autologous NK cells have only partial effect on tumors, such as glioma (55). While other tumors, such as metastatic carcinoma or relapsed lymphoma, do not demonstrate any improvement (103, 104, 108). Moreover, a recent clinical trial used *ex vivo* FN-CH296 stimulated T cells and OK-432 expanded, autologous NK cells with enrolled patients diagnosed with rectal, esophageal, gastric, or colon cancer that was either recurrent or at metastatic disease stage. The NK cell therapy in these patients was well tolerated with no severe adverse events and the cytotoxicity of peripheral blood was elevated approximately twofold up to 4 weeks post the last transfer (47).

#### Allogeneic NK Cells

Allogeneic NK cell products have been used in the treatment of a range of malignancies, such as leukemia, renal cell carcinoma, leukemia, colorectal cancer, hepatocellular cancer, lymphoma, and melanoma (**Table 3**) (38, 109–113). The major risk with allogeneic NK cell transplantation is the development of graftversus-host disease (GvHD). Several precautions can be taken to reduce the risk of GvHD, for example, immunosuppression,


*NSCLC, non-small cell lung cancer; PBMC, peripheral blood mononuclear cell; RCC, renal cell carcinoma.*

#### TABLE 3 | Clinical trials with infusion of allogeneic NK cells.


*(Continued)* Natural Killer Cell-Based Cancer Therapies


infusion of CD3 depleted high purity NK cells and if available, selecting the donor that matches the host HLA (44, 114, 115).

In the first phase I clinical trial using the feeder-free *ex vivo* expansion platform, adoptive transfer of NK cells from HLA identical siblings into patients with leukemia or carcinoma was well tolerated and safe alongside *in vivo* NK cell expansion, with only some infusion-related complications (43).

If no HLA identical donor is available, host cells from a receptor–ligand-mismatched donor can be used. If the donor is HLA matched, it is preferentially better if the donor cells are KIR B haplotype. Also, to further improve the outcome, T cell depletion is performed (120). In haploidentical transplantation, at least one KIR ligand is not expressed on the host cells leading to reduced inhibition of donor NK cells. Less inhibited NK cells could lead to better prognosis and might be the best treatment for a good clinical outcome if GvHD can be avoided (38, 121). When haploidentical transplantation is performed, it is strictly necessary to make extensive T cell depletion to avoid GvHD. In most clinical trials, NK cells are collected from leukapheresis followed by a two-step purification procedure, with depletion of CD3<sup>+</sup> T cells followed by enrichment of CD56<sup>+</sup> cell (109, 110, 117, 118).

Completed clinical trials with haploidentical donors are safe with only a few reports of infusion-related complications such as dyspnea, nausea, hypertension, stroke, febrile reaction, and vomiting (38, 115). So far, allogeneic NK cell transplantations derived from PBMCs or CD34<sup>+</sup> cells have shown promising results with engraftment, *in vivo* expansion of NK cells, complete remission, and a 100% 2-year event-free survival in one clinical trial by Rubnitz et al. (109, 112–114, 116).

#### IMMUNE SUPPRESSION OF NK CELLS IN THE TUMOR MICROENVIRONMENT

Natural killer cells can recognize and kill tumor cells *in vitro*. However, their efficiency in targeting solid tumors has not yet been fully acknowledged in the clinical setting even though endogenous and adoptively transferred activated NK cells can be detected in various tumors (122, 123). Nevertheless, not all tumors are equally well infiltrated by NK cells, and many of the infiltrating cells are dysfunctional (124–127). The failure of immune surveillance may in part be due to sustained immunological selection pressure on tumor cells resulting in the development of tumor escape variants that are in fact invisible to the immune system (**Figure 2**). In addition, cytotoxic function of immune effector cells is also largely suppressed in the tumor microenvironment (128), which could be explained by suppressive tumor-secreted factors as well as suppressive immune compartments, such as myeloid-derived suppressor cells (MDSCs), tumor-associated macrophages (TAM), and regulatory T cells (**Figure 2**). One of the most studied immune-suppressive cell types associated with tumor progression is regulatory T cells (Treg), characterized by their expression of CD4, high CD25 (CD4<sup>+</sup>CD25<sup>+</sup>CD127low/ neg) as well as the transcription factor forkhead box P3 (FoxP3) (129). The expansion of Treg population is promoted in different cancers and their accumulation correlates with impaired immune cell function and poor prognosis (130–135). *In vitro*, NK cells are suppressed by Treg cells in a cell contact-dependent manner where membrane-bound TGF-β is utilized to attenuate their cytotoxicity (136). In line with this, inverse correlation between NK cell activity and Treg cell expansion has been observed in patients with gastrointestinal stromal tumor (GIST) (136) as well as in hepatocellular carcinoma patients (137). Treg cells express the high-affinity IL-2 receptor alpha (CD25, IL-2Rα) and need IL-2 for their full function. Recent studies have indicated that NK cell proliferation, accumulation, and activation can be limited by Treg cells through hampering the availability of IL-2 released by activated CD4<sup>+</sup> T cells (138, 139). Consequently, inadequate IL-2 levels in the tumor microenvironment limits the extent of NK cell-mediated tumor rejection.

Another group of immunosuppressive cells in the tumor is the MDSCs. MDSCs are heterogeneous precursors of the myeloid cells, granulocytes, macrophages, and immature dendritic cells with immunosuppressive activity (140). Recently, MDSCs have been proposed as a key immunoregulator in various solid and hematologic malignancies (141, 142). MDSCs are divided into two groups that can originate from granulocytic (grMDSCs) and monocytic precursors (moMDSCs) (143). In human beings, distinct phenotypes of MDSCs are associated with different types of cancers (144–148). Their suppressive function is mediated by a few different mechanisms such as production of suppressive cytokines including IL-10 and TGF-β, depletion of arginine in the tumor or production of reactive oxygen species (ROS) (144, 149–151). Additionally, recent studies investigated the induction mechanism of MDSCs and how they suppress T cells *in vitro* (152–154). Furthermore, several studies have characterized cytokines that can induce MDSCs from healthy human PBMCs. We found that prostaglandin E2 treated healthy monocytes resemble patient-derived moMDSCs and suppress NK cell responses through TGF-β-dependent mechanism (155). In patients with hepatocellular carcinoma, NK cells were shown to be suppressed by monocytic MDSC in a cell contact-dependent manner, but did not rely on the arginase activity of MDSCs, which is a hallmark function of these cells; however, MDSC-mediated inhibition of NK cell function was revealed to be mainly dependent on the NKp30 on NK cells (146). Moreover, a negative correlation between increased CD33<sup>+</sup>-MDSC accumulation and functional loss of NK cells has been demonstrated in patients with myelodysplastic syndromes (156).

Macrophages are the dominant myeloid-derived population that is found in the tumor microenvironment. TAM has been identified as regulators of solid tumor development based on their capacity to enhance angiogenic, invasive, and metastatic programing of neoplastic tissue (157–160). TAMs could be found in several types of human cancer correlating with poor clinical outcome (161, 162). The immune-suppressive mechanisms applied by TAMs on NK cells in the tumor microenvironment can be different, such as recruitment of Treg, prostaglandin E2-mediated inactivation, and production of IL-10 (163–165). Furthermore, tumors are able to escape NK cells by releasing indoleamine 2,3-dioxygenase and prostaglandin E2, which inhibit the expression of activating receptors of NCRs and NKG2D (166). These molecules are also released by mesenchymal stem cells to inhibit NK cell function in the tumor microenvironment (167). There is a direct association between the surface density of NCRs (NKp46) and the intensity of anti-tumor cytolytic activity of the NK cells (168).

As mentioned earlier, the tumor microenvironment plays a significant role in suppressing NK cell responses against cancer. Therefore, therapies aim to target immunosuppressive cell

populations are emerging (169–174). In the next section, some of the alternative ways aiming to enhance tumor-specific targeting and NK cell survival in order to overcome immunosuppressive effect of the tumor microenvironment on NK cells and to improve intra-tumoral NK cell responses will be discussed.

# FUTURE PERSPECTIVES

## Genetically Modified NK Cells

In the last decade, several NK cell based anti-cancer products have been taken to clinical trial stage with promising clinical outcomes. However, in order to manufacture more efficient NK cell therapy products, it is essential to develop novel potential strategies such as genetic modification of NK cells (**Figure 3**). Although NK cells are inherently resistant to retroviral infections (96, 175–177), our group has significantly enhanced retroviral and lentiviral gene delivery to NK cells through enhanced proliferation and targeting intracellular viral defense mechanism by small molecule inhibitors (96). Therefore, it is easier to design genetically modified NK cells expressing cytokine transgenes, silenced inhibitory receptors, overexpressing activating receptors, or retargeting NK cells by expression of CARs on the cell surface. By genetically modifying NK cells to produce cytokines such as IL-2 or IL-15, their survival capacity and proliferation increase and their activation and anti-tumor activity *in vivo* are enhanced (83, 87, 88, 178, 179). To enhance the specificity for the target cells, NK cells can be modified to recognize antigens specifically expressed on the tumor cells.

Furthermore, another approach aiming to enhance tumor specificity is to make use of ADCC. The constant region of the tumor-specific monoclonal antibodies (mAbs) targeting the tumor cells can engage to the FcγRIIIa receptor (CD16a) on the NK cell, activating the NK cell. However, NK-92 cell line cannot perform ADCC since they lack CD16a expression (80, 85). This defect on NK-92 cells can be reverted by the introduction of CD16a through genetic modification so that they are able to perform ADCC in antibody combination treatments (180). Finally, CAR-modified NK cell lines can also function as tumor-specific standardized and characterized NK cell-based therapy products. Most of the NK cell lines require further *in vivo* characterization with a potential to become standard NK cell-based products for certain tumors.

# Monoclonal Antibodies

When the antigen-binding fraction (Fab) of the antibody binds to the tumor target cell and the constant region (Fc) of the antibody binds to CD16 on the NK cells, NK cells get activated and ADCC is triggered. Several different mAbs have been developed for targeting specific tumor antigens, such as anti-CD20 (retuximab), anti-Her2 (trastuzumab), anti-CD52 (alemtuzumab), anti-EGFR (certuximab), and anti-CD38 (daratumumab) (181). Daratumumab treatment of patients with relapsed myeloma has mild infusion-related reactivity, complete or very good partial responses with reduced bone marrow plasma cell levels (182). mAbs bind to the target tumor cell plus engaging CD16 on NK cells and other cell types resulting in killing of tumor cell by ADCC both *in vivo* and *in vitro* [reviewed in Ref. (183)]. New generations of mAbs have been developed to increase ADCC and complement-dependent cytotoxicity. Second-generation anti-CD20 mAbs, such as veltuzumab (hA20) (184, 185) and ofatumumab (HuMax-CD20) (186–193), have the advantage of being humanized or of fully human origin. Both veltuzumab and ofatumumab had promising preliminary outcomes in various studies (184, 186, 187, 189, 190, 193). The benefit of third-generation anti-CD20 mAbs, ublituximab (TG-1101), ocaratuzumab (AME-133) (194, 195), and obinutuzumab (GA-101) (196–200), is that they are both humanized and that their Fc regions have been modified for increased binding affinity to CD16a. So far, the most studied third-generation anti-CD20 mAbs is obinutuzumab. The overall response rate for obinutuzumab is 44.6%, which is higher than the overall response rate for rituximab treatment which is 33.7% (200). In the same study, the progression-free survival did not promote obinutuzumab over rituximab. By increased affinity between CD16a and mAb better NK cell cytolysis can be induced by ADCC. Ublituximab, ocaratuzumab, or obinutuzumab-treated NK cells from CLL patients or healthy donors have more efficient ADCC compared to same cells treated with first- or second-generation anti-CD20 mAb *in vitro* (201–203).

Monoclonal antibody therapies in combination with already existing treatments can potentially enhance NK cell activity in anti-tumor therapy. The completely human IgG4 anti-KIR antibody, IPH-2102, has been tested in several clinical trials for hematological diseases both as single treatment and as combination (204, 205). Some clinical trials for combination treatment of

advanced solid tumors with anti-KIR antibodies are done as well, for example, in combination with anti-CTLA antibody or anti-PD1 antibody (NCT01750580 and NCT01714739, respectively). Thus, use of mAbs enhancing ADCC and stimulation of NK cells as well as blocking NK cell inhibition could potentially improve outcome of clinical anti-cancer NK cell products (**Figure 3**).

# Bi- and Trispecific Antibodies

Likewise designing CARs through tumor-specific mAbs can be used to engineer bi- and trispecific antibodies crosslinking CD16 with tumor-specific mAbs in order to enhance NK cell tumor reactivity (**Figure 3**). Briefly, the design of bi- and trispecific antibodies, fusing the Fab region of the antibody targeting the tumor cell antigen, such as CD19, CD20, and CD33, in combination with another Fab region recognizing CD16 on NK cell leads to stimulation of the NK cells followed by tumor cell killing. This technology makes it possible to select the amount of NK cells that should be activated as well as it is possible to add more Fab regions targeting other tumor-associated antigens. These Fab regions can be exchanged to other tumor-associated antigen-recognizing antibody parts, as long as the part crosslinking CD16 on the NK cell is present (206, 207).

## Chimeric Antigen Receptors (CARs)

Design of CARs using antigen-specific variable part of these tumor antigen antibodies fused with intracellular lymphocyte stimulatory molecules (CD3ξ, CD28, 4-1BB) enables highaffinity specific recognition of tumor antigens and tumors. CAR modifications of T cells have been studied extensively and have led to several phase I and phase II clinical trials (208–211). NK cells are less explored and so far only two clinical trials using CAR NK cells have been approved. The first study (NCT00995137) at St. Jude Children's Research Hospital is completed and was a phase I clinical trial with 14 relapsed or refractory B-lineage ALL patients below 18 years. Haploidentical NK cells were expanded by co-culture with irradiated K562 cell line expressing IL-15 and 4-1BB ligand on the surface to be transduced with a signaling receptor binding CD19 (anti-CD19 CAR). The second study (NCT01974479) is a phase II pilot study, which is still recruiting refractory B-lineage ALL patients in all ages. NK cells are expanded by co-culture with K562 cells as the previous trial, together with IL-2 before transduction with the same construct. The patients will also receive IL-2 after NK cell administration to support NK cell viability and expansion. Although CAR T cell studies have been extremely promising, CARs designed for T cell therapies are still suboptimal for NK cells. Thus, it is essential to further optimize the construct design, especially the intracellular stimulatory adapter molecules, in order to trigger most efficient NK cell responses.

## Immunomodulatory Drugs (IMiDs)

Immunomodulatory drugs (IMiDs) such as thalidomide, lenalidomide, and pomalidomide, can stimulate both NK cells and T cells, potentially resulting in better targeting cancer cells (212). Lenalidomide upregulates TRAIL molecules on NK cells and enhances anti-tumor activity (14, 15). So far, several different malignancies, both solid and hematological, have been treated with IMiDs. A large part of the nearly 100 clinical trials with IMiDs that has been reported with results to clinicaltrials. gov is treatment of myeloma, lymphoma, and leukemia. IMiDs can be used as combination treatment, such as lenalidomide in combination with IPH-2102, anti-inhibitory KIR antibody therapy (205). Lenalidomide expands and activates the NK cells, while anti-inhibitory KIR antibody (IPH2101) promotes NK cell recognition and lysis of tumor cells. This combination could give a better therapeutic outcome.

## Combination Treatments

It is possible that NK cell products cannot fully eliminate tumor cells due to several immunosuppressive effects of tumor microenvironment as well as reduced *in vivo* expansion and cytotoxicity. These obstacles could be overcome by combination treatments using NK cell therapy products together with other drugs either directly targeting tumor cells or modulating cytotoxic activity of NK cells. As mentioned earlier, use of mAbs and IMiDs together with appropriate NK cell products could enhance tumor targeting and elimination. Another way to enhance NK cell-mediated killing is to combine drug therapy with NK cell stimulating cytokines such as IL-2, IL-12, IL-15, and IL-21 (213).

Furthermore, chemotherapy in combination with NK cell infusions is an alternative way to overcome tumor-induced dysfunctions. NK cells from haploidentical donor require combination treatments with the intense chemotherapy drugs high-dose fludarabin and cyclophosphamide (Hi-Cy/Flu) plus daily infusion of IL-2 to be able to expand *in vivo* (38). Total body irradiation could help to create immunological space for expanding NK cells in addition to chemotherapy after short-term *ex vivo* activation of NK cells (214).

# CONCLUSION

In this review, we have summarized current NK cell-based therapy strategies as well as some of the challenges that need to be addressed. Even though NK cell-based therapies represent one of the most promising strategies to combat cancer, to our knowledge, no clinical trial has clearly demonstrated a significant benefit in patients with malignancies. This is in part due to the lack of prospective large-scale clinical trials and partly due to a lack of consensus in which NK cell product preparation would show the best effect. Further comparative clinical studies are definitely warranted; however, the design of such clinical trials is challenging due to the advanced therapy regulations in major countries such as European Union member states and the United States of America. Although cell therapy clinical trials are reaching a log-linear expansion, the number of NK cell-based therapies is not aligned with this increase. Nevertheless, there is a lot of promise in early clinical and pre-clinical data that cannot be omitted. In the near future, different NK cell-based products will reach multicenter clinical trial stage and we will start to see efficacy data.

Separately, NK cell-based therapies are in theory complementary to many different upfront, maintenance, and late-line therapies. Further studies clarifying the complementary efficacies and synergies have to be initiated to conclusively state if there is any place for these intriguing cells in search for an effective treatment of cancer.

# AUTHOR CONTRIBUTIONS

All the authors performed the review of the literature, wrote, and edited the manuscript.

# REFERENCES


# ACKNOWLEDGMENTS

This research was supported by generous grants from the Swedish Cancer Fund (Cancerfonden) (Reference number 141022), Swedish Cancer Society in Stockholm (Cancerföreningen/Radiumhemmets forskningsfonder), VINNOVA (Project number 2010-00501), and State of Florida Department of Health. The authors acknowledge the linguistic comments from Kim Kusser at Nova Southeastern University and Elizabeth Henry Alici.


renal cell carcinoma patients. *Int J Cancer* (2004) **109**:393–401. doi:10.1002/ ijc.11730


following T-cell-depleted stem cell transplantation. *Blood* (2015) **125**:784–92. doi:10.1182/blood-2014-07-592881


breast cancer induces immune activation and cytokine release: a phase I/II trial. *Bone Marrow Transplant* (2003) **32**:177–86. doi:10.1038/sj.bmt.1704086


hematopoietic transplants. *Science* (2002) **295**:2097–100. doi:10.1126/ science.1068440


previously treated follicular lymphoma. *Leuk Lymphoma* (2015) **56**:42–8. doi :10.3109/10428194.2014.911859


persisting after allogeneic hematopoietic stem cell transplantation. *Blood* (2013) **122**:4129–39. doi:10.1182/blood-2013-08-519413


**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 Dahlberg, Sarhan, Chrobok, Duru and Alici. 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.*

# **Natural killer cell immunotherapy: from bench to bedside**

*Anna Domogala1,2, J. Alejandro Madrigal 1,2 and Aurore Saudemont 1,2 \**

*<sup>1</sup> Anthony Nolan Research Institute, London, UK, <sup>2</sup> University College London, London, UK*

The potential of natural killer (NK) cells to target numerous malignancies *in vitro* has been well documented; however, only limited success has been seen in the clinic. Although NK cells prove non-toxic and safe regardless of the cell numbers injected, there is often little persistence and expansion observed in a patient, which is vital for mounting an effective cellular response. NK cells can be isolated directly from peripheral blood, umbilical cord blood, or bone marrow, expanded *in vitro* using cytokines or differentiated *in vitro* from hematopoietic stem cells. Drugs that support NK cell function such as lenalidomide and bortezomib have also been studied in the clinic, however, the optimum combination, which can vary among different malignancies, is yet to be identified. NK cell proliferation, persistence, and function can further be improved by various activation techniques such as priming and cytokine addition though whether stimulation pre- or post-injection is more favorable is another obstacle to be tackled. Here, we review the various methods of obtaining and activating NK cells for use in the clinic while considering the ideal product and drug complement for the most successful cellular therapy.

#### *Edited by:*

*Francisco Borrego, Cruces University Hospital, Spain*

#### *Reviewed by:*

*Ulrike Koehl, Medical School Hannover, Germany Olatz Zenarruzabeitia, BioCruces Health Research Institute, Spain*

#### *\*Correspondence:*

*Aurore Saudemont, Anthony Nolan Research Institute, Fleet Road, London NW3 2QU, UK aurore.saudemont@anthonynolan.org*

#### *Specialty section:*

*This article was submitted to NK Cell Biology, a section of the journal Frontiers in Immunology*

> *Received: 12 March 2015 Accepted: 13 May 2015 Published: 03 June 2015*

#### *Citation:*

*Domogala A, Madrigal JA and Saudemont A (2015) Natural killer cell immunotherapy: from bench to bedside. Front. Immunol. 6:264. doi: 10.3389/fimmu.2015.00264* **Keywords: natural killer cells, cancer, immunotherapy, activation, proliferation, persistence**

# **Introduction**

Natural killer (NK) cells are unique lymphocytes, distinct from B and T cells, which bridge the innate and adaptive immune systems. They have the unique capacity to exert immunoregulatory and cytotoxic functions against transformed and infected cells without prior sensitization. NK cells are characterized by the expression of CD56 and absence of CD3 and can be further subdivided into a CD56bright population, which is predominantly cytokine producing and a CD56dim population, which is cytolytic and provides antibody-dependent cell-mediated cytotoxicity via CD16 (1). NK cells operate by detecting information, which is missing on the target. This phenomenon is known as the "missing self hypothesis" and postulates that NK cell cytotoxicity inversely correlates with the target expression of major histocompatibility complex class I (MHC-I) (2, 3). In addition, NK cell activity is further regulated by a complex array of inhibitory and activating receptors such as killer cell immunoglobulin-like receptors (KIR), natural cytotoxicity receptors (NKp44, NKp30, and NKp46), and C-type lectins (CD94/NKG2A/NKG2C/NKG2D). These properties equip NK cells with the tools to actively eliminate susceptible targets (4). Taking into account, the cytotoxic potential of these cells numerous attempts have been made to transfer NK cell immunotherapy into the clinic. Here, we review which methods to consider for obtaining cells for therapy, drug complements, and pre-infusion activation techniques. We also summarize current clinical trials and outcomes and postulate where success in NK immunotherapy may lie.

# **NK Cells and Cancer**

Natural killer cells were first implicated as playing a role in cancer immunosurveillance when one large epidemiologic study found that low NK cell cytotoxicity forecasted an increased risk in developing cancer (5). There have since been numerous studies, which demonstrate that NK cells can target human tumors *in vivo* making them a desirable candidate for therapeutic use (6). Clinical trials using autologous NK cells have shown the therapy to be nontoxic, however, they fail to prove efficacy (7), which could be the result of inhibition by self-MHC-I. Allogeneic treatment therefore has potential to offer an alternative therapy with improved effect. The direct involvement of allo-reactive NK cells in inducing antitumor effect in hematopoietic transplants was first demonstrated in 2002 (8). NK cells showed to enhance engraftment; providing graft vs. leukemia (GvL) effect while suppressing graft vs. host disease (GvHD) particularly when a KIR ligand mismatch in the donor to host direction was observed. Reduced GvHD was hypothesized to be attributed to the lysis of the recipient's antigen presenting cells (APCs) reducing the incidence of GvHD while maintaining GvL effect. This was later successfully translated into an *in vivo* model using acute myeloid leukemia (AML)-engrafted NOD/SCID mice infused with allo-reactive NK cells. Tumor clearance was achieved implicating NK cells in preserving the GvL effect (9).

Miller and colleagues later translated NK cell therapy alone into the clinic where allogeneic NK cells were infused into patients with advanced cancer alongside IL-2 administration. This demonstrated that NK cell infusions were feasible and safe and led to complete remission in 5/19 patients with poor prognosis AML (10). Additionally, the efficacy of haploidentical NK cell therapy in the refractory disease was further improved by depleting host regulatory T cells with IL-2 diphtheria toxin preventing their immunosuppressive effect (11). NK cell allo-reactivity could also be utilized in other scenarios besides hematopoietic stem cell transplantation (HSCT) with studies in malignant glioma and neuroblastoma patients demonstrating that NK cell infusions are safe and partially effective (12, 13). Numerous types of cancer could therefore benefit from NK cell immunotherapy and current clinical trials include pancreas, lungs, head/neck, breast, and renal cell carcinomas.

# **Clinical Conditioning**

Not only chemotherapy and/or radiotherapy are required for the success of HSCT but also cellular immunotherapy. Such treatments are necessary to reduce tumor burden and suppress the immune system of the patient to prevent rejection of the cellular therapy. Defining the correct conditioning regimen is therefore critical. In a transplantation setting, common regimens are referred to as myeloablative, non-myeloablative, and reduced intensity and their use will depend on patient age and disease severity; however, any decrease of leukemia recurrence is often at the expense of an increase in toxicity (14).

The use of new conditioning agents termed as "novel agents" have become increasing popular in cancer immunotherapy as a result of their immunomodulatory and direct tumor targeting

mechanisms. In combination with cellular therapy, they offer the potential for a more personalized and less toxic treatment regimen as these specialized drugs have been shown to not only reduce tumor burden but also enhance the function of cellular therapies. Although chemotherapy has revolutionized the treatment of cancer, its side effects include the development of refractory disease and severe toxicity. Novel agents provide an alternative option of harnessing the immune system to tackle malignancies.

Thalidomide was one of the first novel agents to be well studied; it is a synthetic glutamic acid derivative that is capable of immunomodulatory, anti-inflammatory, and anti-angiogenic effects. Although proven successful in targeting multiple myeloma the exact mechanism of action of thalidomide is yet to be elucidated although anti-inflammatory effects have been attributed to inhibition of TNF-α production by monocytes and antiproliferative capabilities to disruption of the bone marrow (BM) microenvironment preventing multiple myeloma cellular development (15). Although extended anti-angiogenic characteristics make a desirable option in limiting tumor development its immunomodulatory properties have not been so well defined. Lenalidomide is an immunomodulatory compound with a dual mechanism of action. It is capable of targeting the tumor directly through stromal support disruption, induction of tumor suppressor genes, and activation of caspases (16). It is also able to stimulate the cytotoxic functions of NK cells and T lymphocytes while limiting the immunosuppressive impact of regulatory T cells (17). Additionally, bortezomib is a proteasome inhibitor proven popular by up-regulating expression of TRAIL death receptors and altering caspase-8 activity rendering tumors susceptible to NK cell lysis. However, intriguingly these tumors became resistance to T cell cytotoxicity (18). The specific mechanisms by which novel agent function offer a promising future for the treatment of a variety of malignancies as these agents target not only the tumor themselves but also offer potential to enhance the immune system. This provides the possibility of coupling cellular therapy with novel agents to provide personalized treatment regimens to target an individual's condition.

# **NK Cell Sources**

It is considered that the success of NK cell immunotherapy is dependent on obtaining high numbers of functional NK cells that have the potential to survive *in vivo*. Numerous attempts have therefore been made to obtain high levels of NK cells from a variety of sources. One option is to isolate cells directly from peripheral blood (PB) or cord blood (CB), however, as NK cells make up only 10% of circulating lymphocytes in PB and 20% in CB the number of cells obtained can be limited and could potentially prevent the option for multiple infusions. Doses of 1–2 *<sup>×</sup>* <sup>10</sup><sup>7</sup> cells/kg have been identified as safe (19); however, higher doses of 2 *<sup>×</sup>* <sup>10</sup><sup>8</sup> /kg have been shown to be well tolerated and non-toxic (20). Several techniques have therefore been explored to increase cell numbers. This includes expanding isolated cells *in vitro* using different combinations of cytokines with or without feeder layers, the use of NK cell lines and differentiating NK cells from hematopoietic stem cells (HSCs).

# **NK Cell Expansion**

Numerous methods of expanding mature NK cells *in vitro* have been explored and have been reviewed previously (21); however, these products seem to produce limited clinical success. This may be because of wide variations in expansion rate and distribution of NK cell subpopulations (22) or expanding mature cells produces effectors with a more finite lifespan unable to proliferate with lower cytotoxicity post infusion (23). NK cell expansion using aAPCs particularly the GMP-compatible genetically modified form of the K562 myeloid leukemia cell line engineered to express membrane-bound interleukin 15 and the ligand for the co-stimulatory molecule 4-1BBL has been rising in popularity due to the potential to rapidly expand an NK cell product with an up-regulation of activating receptors and improved killing capacity (24). However, a first-in-human trial carried out by Shah and colleagues in 2014 performing the adoptive transfer of donor-derived IL-15/4-1BBL activated NK cells showed interesting results. Surprisingly, 5/9 patients experienced acute GvHD and as the T cell content of the infusion was well below the specified threshold for GvHD development the group concluded that the aNK-DLI contributed to the effect by stimulating underlying T cell allo-reactivity (25). This is the first time in a clinical setting NK cells have been implicated in the role of induction or aggravation of GvHD, which could be a result of lack of immunosuppressive drugs post transplant or infusion of IL-2, which expands immunoregulatory populations. This coupled with the infusion of an expanded NK cell population with such a high up-regulation of activating receptors could be the reason for such unfavorable results.

# **NK Cell Lines**

The use of NK cell lines have been seen as an attractive option due to the availability of a clinical grade frozen stock and their homologous nature. The most prominent NK cell line currently in focus is NK-92, which was established from a patient with non-Hodgkin's lymphoma and has demonstrated the capability of lysing leukemia, lymphoma, and myeloma *in vitro* (26). Current clinical trials have proven non-toxic; however, they have shown limited success in demonstrating efficacy (27, 28). This could be a result of the necessity to irradiate a cell line prior to infusion for safety requirements, the cells could therefore be incapable of proliferation *in vivo* severely limiting their persistence and potential to target the tumor.

# **Differentiation of NK Cells**

Differentiating NK cells*in vitro* from HSCs or induced pluripotent stem (iPS) cells are alternative options for obtaining high numbers of functional cells. Different sources of cryopreserved HSCs have been used to differentiate NK cells *in vitro* including human embryonic stem cells (hESC), BM, mobilized peripheral blood stem cells (mPBSC), and cord blood stem cells (CBSC). hESC are a controversial source due to the ethical dilemma posed by obtaining cells from a 5- to 7-day-old embryo. However, the H9 hESC cell line has been used to produce NK cells that express activating and inhibitory receptors, including KIRs, and are able to produce

cytokines and mediate cytotoxicity *in vitro* and *in vivo* (29). The invasive collection procedure limits the use of BM and has therefore mainly been used to study NK cell development (30, 31). Differentiating NK cells from induced pluripotent cells offers potential due to the ready availability of a donor and the non-invasive cell harvesting methods. A recent study identified a method of differentiating mature and functional NK cells using a combination of embryoid body formation and membrane-bound interleukin 21-expressing aAPCs (32) and a thorough review of the potential uses of such cells in the clinic was published last year (33). The possibility of reprograming cells is a promising one; however, there is the possible limitation that the differentiated NK cells will be suppressed by self-MHC and therefore have little cytotoxic effect. The use of NK cells differentiated from CD34<sup>+</sup> progenitors was first shown to be feasible in the clinic by Yoon and colleagues in 2010 (34). This led to interest in the use of umbilical cord blood CD34<sup>+</sup> cells as a source of NK cells with the focus being on generating a readily available, non-invasive, off the shelf cellular product (35). Our group modified a published protocol (36) and compared the use of mPBSC, fresh CBSC, and frozen CBSC at differentiating NK cells *in vitro* (37). This work demonstrated frozen CB CD34<sup>+</sup> cells to be the best source of NK cells over fresh CB CD34<sup>+</sup> and frozen mPBSCs. This was due to higher fold expansion and therefore higher NK cell numbers generated without compromising on phenotype, cytokine production, or cytotoxicity. Additionally, the cells are capable of further proliferation *in vitro* and more importantly could persist for longer and in higher numbers*in vivo*. Considering that proliferation and persistence of NK cells *in vivo* is fundamental for the development of a clinically relevant cellular product this makes the differentiation of NK cells from CB HSCs *in vitro* an attractive candidate for NK cell immunotherapy.

# **NK Cell Activation**

As reviewed in **Table 1**, there have been many studies that well document the expansion of NK cells*in vitro*, however, we are yet to obtain a clinically successful product, which proliferates and persists *in vivo* inducing consistent efficacy. This could be because we are yet to identify the optimum activation method and status of the cells before infusion. As seen in **Figure 1**, whether the cells should be incubated with cytokines, genetically engineered, differentiated into a "memory-like" phenotype, or primed using NK nonsusceptible cell lines are all options that need to be considered.

Cytokine activation has always been a popular method of stimulating NK cells as it is a well-documented pathway of activation *in vivo* and different cytokines can give rise to the same signaling patterns while differing in their effects on development, activation, and proliferation. IL-2 stimulates cellular proliferation and enhances cytotoxicity, however, it has been noted that this only affects a small sub-population for an extended period (55). IL-15 significantly improves NK cell survival although it only stimulates minimal expansion (56). Furthermore, the toxic effects of the *in vivo* administration of cytokines cannot be ignored, IL-2 risks vascular leak syndrome caused by the stimulation of endothelial cells through the IL-2 receptor (57) and preferentially expands T regulatory cells, which mediate immune suppression (58). Studies with IL-15 in non-human primates have only shown transient

#### **TABLE 1| NK cells in the clinic: trials so far**.


NK cell immunotherapy

Domogala et al.

*(Continued)*

#### **TABLE 1| Continued**


*ALL, acute lymphoblastic lymphoma; AML, acute myeloid leukemia; ATG, anti-thymocyte globulin; B-CLL, B cell chronic lymphocyte leukemia; BOR, bortezomib; CLL, chronic lymphocyte leukemia; CR, complete response; CRC, colorectal carcinoma; CY, cyclophosphamide; DSRCT, desmoplastic small round cell tumor; EWS, Ewing sarcoma; FLU, fludarabine; G-CSF, granulocyte colony stimulating factor; HC, hydrocortisone; MEL, melphalan; MM, multiple myeloma; mPred, methylprednisolone; MRC, metastatic renal carcinoma; N/D, not determined; NSCLC, non-small cell lung carcinoma; PEN, pentostatin; PR, partial response; RCC, renal cell carcinoma; RMS, rhabdomyosarcoma; RTX, rituximab; TBI, total body irradiation.*

toxicity; however, its reduced half life may suggest the need for more frequent dosing in therapeutic applications (59).

A method to avoid such life threatening conditions through *in vivo* administration of cytokines is expanding or stimulating the cells *in vitro*. NK cells have always been considered a member of the innate immune system incapable of producing memory. However, in 2006, it was first observed that NK cells could mediate a long-lived antigen-specific adaptive response independently of other lymphocytes (60). Sun and colleagues (61) later identified an immunological memory in NK cells from MCMV infected mice and it was demonstrated that NK cells pre-activated with IL-12 and IL-18 infused into a naïve host and later re-stimulated showed enhanced IFN-y production (62). This model was later transferred to an *in vitro* model stimulating human cells showing the same results (63). This improvement in cytokine production offers the potential for enhanced GvL effect and a clinical trial is currently in progress targeting relapsed and refractory AML (NCT01898793).

It has been reported that resting NK cells require a two-stage activation process known as "priming" and "triggering" (64). This states that tumors resistant to NK cell killing evade lysis by failing to prime the cell; however, Mark Lowdells group were able to identify a cell line, which could prime the cell without triggering cytokine production or cytolytic activity. This led to the development of an NK cell priming technique that readied the cells for killing, which was still maintained post cryopreservation (65). Primed NK cells from patients with multiple myeloma have also been proven to kill NK cell resistant malignant plasma cells (66). Preliminary data from an ongoing transitional phase I/II clinical trial showed that without cytokine administration primed NK cells from HLA haploidentical-related donors can persist *in vivo* with no toxic effects (67).

## **Genetic Engineering**

Although currently restricted to pre-clinical models the use of chimeric antigen receptor (CAR)-expressing NK cells has the potential to offer enhanced effector cell function of increased specificity. Anti-CD19 CAR T cells have effectively demonstrated their ability to induce long-term remission in patients with B cell malignancies (68). However, concern associated with CAR T cell therapy extends to GvHD, on target/off tumor effects, and tumor lysis syndrome. By contrast, allogeneic CAR-engineered NK cells are expected to induce anti-tumor effects and dissipate after a few days (23). As previously reviewed by Glienke et al. in 2015, current work in the field has focused mainly on targeting CD19 and CD20, however, CARs, which target CS1 and CD138 for multiple myeloma, GD2 and CD244 for neuroblastoma, HER-2 for epithelial carcinomas, and GPA7 for melanoma, are also beginning to indicate promising results.

#### **Immune Escape Mechanisms**

Not all tumors are susceptible to NK cell mediated killing, as some cancer cells have developed the ability to escape detection by the immune system. Mechanisms that regulate the evasion of tumor cells by NK cells extends to the down-regulation of activating receptor ligands for NKG2D (69), the production of soluble stressinduced ligands, such as MICA, which degrades NKG2D leading to NK cell inhibition (70) and the release of suppressive cytokines such as IL-10 and TGF-β (71). Some success has been seen by NK cell immunotherapy targeting hematological malignancies; however, this has not been transferred to solid tumors. This could be a result of the increased concentration of immunosuppressive cytokines and ligands around a tumor mass, method to overcome such escape mechanisms could provide further potential for NK cells to not only target hematological malignancies but also solid tumors.

# **Concluding Remarks**

Natural killer cell immunotherapy has been a promising option for providing specialized and target specific treatment for a therapy

# **References**


in its own right and as a supportive one in infection or transplantation. Although some mechanisms of NK cell biology are yet to be elucidated as we make progress in the field an effective clinical NK cell immunotherapy will become more achievable. A standard clinical regimen is still to be elucidated and obstacles such as cell dose, activation status, method of expansion, drug complement, and source are still to be determined.

It has always been thought that high numbers of NK cells are necessary for a successful clinical product. However, numerous groups have managed to successfully generate high numbers of functional NK cells *in vitro* although the lack of clinical effect and significant cost implications cannot be ignored. The highcell number requirement is likely to be the result of a "success in numbers" approach with there being a significant loss of cells through *in vivo* targeting and just a small sub-population of effector cells that will target the tumor. Perhaps work should therefore be refocused on the infusion of a small population of cells with optimum pre-activation status that will traffic to the tumor site and would not be suppressed by tumor evasion mechanisms. This is a significant goal to achieve considering the variety of NK cell populations occurring naturally in the body. However, the absence of a labor-intensive long-term culture system would mean this method would pose significantly reduced cost implications. Once techniques have been optimized and streamlined there would therefore be a greater possibility of NK cell immunotherapy being routinely adopted as a clinical therapy in the future.

# **Acknowledgments**

This work was funded and supported by Anthony Nolan.


but requires CD15-mediated CD2 ligation and natural cytotoxicity receptors. *J Immunol* (2011) **187**:6227–34. doi:10.4049/jimmunol.1101640


**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 Domogala, Madrigal and Saudemont. 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.*

# **Revving up natural killer cells and cytokine-induced killer cells against hematological malignancies**

*Gianfranco Pittari <sup>1</sup> \*, Perla Filippini <sup>2</sup> , Giusy Gentilcore<sup>2</sup> , Jean-Charles Grivel <sup>2</sup> and Sergio Rutella<sup>3</sup> \**

*<sup>1</sup> Department of Medical Oncology, National Center for Cancer Care and Research, Hamad Medical Corporation, Doha, Qatar, <sup>2</sup> Deep Immunophenotyping Core, Division of Translational Medicine, Sidra Medical and Research Center, Doha, Qatar, <sup>3</sup> Clinical Research Center, Division of Translational Medicine, Sidra Medical and Research Center, Doha, Qatar*

#### *Edited by:*

*Raquel Tarazona, University of Extremadura, Spain*

#### *Reviewed by:*

*Roberto Biassoni, Istituto Giannina Gaslini, Italy Björn Önfelt, Karolinska Institutet, Sweden*

#### *\*Correspondence:*

*Gianfranco Pittari, Department of Medical Oncology, National Center for Cancer Care and Research, Hamad Medical Corporation, P.O. Box 3050, Doha, Qatar gpittari@hamad.qa; Sergio Rutella, Clinical Research Center, Division of Translational Medicine, Sidra Medical and Research Center, P. O. Box 26999, Doha, Qatar srutella@sidra.org*

#### *Specialty section:*

*This article was submitted to NK Cell Biology, a section of the journal Frontiers in Immunology*

> *Received: 27 March 2015 Accepted: 29 April 2015 Published: 13 May 2015*

#### *Citation:*

*Pittari G, Filippini P, Gentilcore G, Grivel J-C and Rutella S (2015) Revving up natural killer cells and cytokine-induced killer cells against hematological malignancies. Front. Immunol. 6:230. doi: 10.3389/fimmu.2015.00230* Natural killer (NK) cells belong to innate immunity and exhibit cytolytic activity against infectious pathogens and tumor cells. NK-cell function is finely tuned by receptors that transduce inhibitory or activating signals, such as killer immunoglobulin-like receptors, NK Group 2 member D (NKG2D), NKG2A/CD94, NKp46, and others, and recognize both foreign and self-antigens expressed by NK-susceptible targets. Recent insights into NK-cell developmental intermediates have translated into a more accurate definition of culture conditions for the *in vitro* generation and propagation of human NK cells. In this respect, interleukin (IL)-15 and IL-21 are instrumental in driving NK-cell differentiation and maturation, and hold great promise for the design of optimal NK-cell culture protocols. Cytokine-induced killer (CIK) cells possess phenotypic and functional hallmarks of both T cells and NK cells. Similar to T cells, they express CD3 and are expandable in culture, while not requiring functional priming for *in vivo* activity, like NK cells. CIK cells may offer some advantages over other cell therapy products, including ease of *in vitro* propagation and no need for exogenous administration of IL-2 for *in vivo* priming. NK cells and CIK cells can be expanded using a variety of clinical-grade approaches, before their infusion into patients with cancer. Herein, we discuss GMP-compliant strategies to isolate and expand human NK and CIK cells for immunotherapy purposes, focusing on clinical trials of adoptive transfer to patients with hematological malignancies.

**Keywords: natural killer cell, cytokine-induced killer cell, interleukin-2, interleukin-15, good manufacturing practice, leukemia, immunotherapy**

# **Biological Features of NK, LAK, and CIK Cells**

Natural killer (NK) cells comprise 5–25% of peripheral blood (PB) lymphocytes and were initially recognized for their ability to kill cancer cells without prior sensitization. The reader is referred to previously published papers for a thorough review of NK development and function (1–3). Briefly, NK cells originate from bone marrow (BM) CD34<sup>+</sup> hematopoietic stem cells and can also be differentiated *in vitro* from highly immature CD34*−* umbilical cord blood (UCB) cells (4). NK cells acquire function (killing or cytokine production) after encountering and recognizing self-human leukocyte antigen (HLA) molecules during a process termed "licensing" or NK-cell education. However, 10–20% of NK cells remain unlicensed, as they lack receptors for self-major histocompatibility complex (MHC) and are functionally hyporesponsive. Importantly, unlicensed NK cells can become alloreactive upon encounter with cytokines in a recipient environment, e.g., after adoptive transfer into hematopoietic stem cell transplantation (HSCT) recipients.

The function of NK cells is governed by a set of germlineencoded activating or inhibitory receptors referred to as killer immunoglobulin-like receptors (KIRs). The extracellular domain determines which HLA class I molecule NK cells recognize, whereas the intracytoplasmic domain transmits either an activating or an inhibitory signal. KIRs are monomeric receptors with either 2 (KIR2D) or 3 (KIR3D) immunoglobulin-like domains, and are further subdivided into those with long (L) cytoplasmic tails (KIR2DL and KIR3DL) and short (S) cytoplasmic tails (KIR2DS and KIR3DS) (5–7). Long-tail KIRs generate an inhibitory signal through the recruitment of the SH2-domaincontaining tyrosine phosphatase 1 protein (SHP1) (8–11). Shorttail KIRs possess truncated portions that transduce activating signals via tyrosine phosphorylation of DAP12 and other proteins (12–14).

Natural killer cells also express other activating receptors that recognize "stress ligands" on virally infected or malignant cells. For instance, NKG2D, a C-type lectin receptor that belongs to the NK group 2 (NKG2) of receptors as member D (15), is constitutively expressed on NK cells and recognizes MHC class I chain-related genes A and B (MICA and MICB) (16), as well as unique long 16 (UL16) binding protein family members (ULBPs) (17). Other activating molecules include natural cytotoxicity receptors (NCRs) NKp30, NKp44, and NKp46 (18, 19). It has been shown that killing of tumors of non-epithelial origin, including leukemia cell lines, involves synergism between NCRs and NKG2D (20). Activating KIRs, such as KIR2DS1, are likely involved in the anti-leukemia effect of NK cells (21, 22). In 2002, investigators from Perugia demonstrated superior diseasefree survival (DFS) in patients with acute myeloid leukemia (AML) receiving BM grafts from HLA-haploidentical donors who expressed KIR binding to MHC class I molecules absent in the host (i.e., KIR-ligand mismatch in the GVH direction) (23, 24). The most notable inhibitory receptors recognize HLA class I proteins (including groups of HLA-A, HLA-B, and HLA-C) and differ in both their transmembrane and intracytoplasmic domains (25–29).

Human leukocyte antigen-C is the predominant class I isotype involved in the inhibitory and activating regulation of human NK cells (1, 22). Individuals may have up to 15 KIR genes that reside in a single complex on chromosome 19p13.4. KIR genes can be divided into A or B haplotypes. The A haplotype consists of five inhibitory KIRs and a single activating KIR, KIR2DS4. By contrast, the B haplotype contains both inhibitory and several activating KIRs that are further subdivided into two separate regions, centromeric and telomeric. In the "missing self " model (30), donor NK cells express inhibitory KIRs for which HLA class I molecules are missing in the recipient. Donors with KIR B vs. KIR A haplotypes improve the clinical outcome for patients with AML by reducing the incidence of leukemia relapse and prolonging DFS (31). The centromeric KIR B genes were dominant over the telomeric ones, and included the genes encoding inhibitory KIRs that are specific for the C1 and C2 epitopes of HLA-C. When the authors examined a cohort of 1,532 T-cell-replete HSCT,

relapse protection associated with donor KIR B was enhanced in recipients with one or two C1-bearing HLA-C allotypes compared with homozygous recipients. This implies that a deeper understanding of the interaction between donor KIRs and recipient HLA class I will allow the selection of the "best donor" to improve outcomes of unrelated HSCT and adoptive NK infusion for AML. Intriguingly, KIR B haplotype donors were recently shown to confer a reduced risk for relapse after haploidentical HSCT in children with ALL (32), an effect that is not seen in adult ALL (33). In allogeneic HSCT, particularly from HLA-mismatched donors, NK cells reportedly influence clinical outcome by exerting anti-tumor effects without inducing graft-versus-host disease (GVHD) (34). However, NK cells reconstituting after allogeneic HSCT may be dysfunctional, likely as a result of low IL-2 levels (35). Some groups are attempting to improve NK-cell reconstitution following HSCT by depleting the graft of αβ<sup>+</sup> T cells and CD19<sup>+</sup> B cells, but leaving NK progenitors untouched (36). Using this approach, very high numbers of haploidentical NK cells and NK-like (CD56+) T cells can be infused into patients with malignant disorders (37).

Another family of human NK receptors is composed of a common subunit (CD94), covalently linked to a distinct chain encoded by a C-type lectin NKG2 family gene. Among the Ctype lectin NK receptors, CD94/NKG2A is inhibitory, whereas other heterodimers are activating receptors. CD94/NKG2A binds the non-classical class I molecule HLA-E (38). The binding of a unique peptide/HLA-E complex to the activating CD94/NKG2C receptor is of higher affinity than the binding to the inhibitory CD94/NKG2A ensuring the predominance of inhibitory signals when the same NK cells express both activating and inhibitory receptors recognizing HLA molecules (39).

In 1980, Rosenberg and co-workers demonstrated that incubation of heterogeneous lymphocyte populations with high-dose (800-1,000 U/ml) interleukin-2 (IL-2) generates lymphokineactivated killer (LAK) cells with prompt *in vitro* cytotoxicity to syngeneic and autologous fresh tumors (40–42). NK cells were identified as precursors of LAK cells, and LAK activity was found to be mainly, albeit not uniquely, mediated by activated NK cells (43, 44). LAK cells comprise CD3*−*CD56<sup>+</sup> NK cells, MHCunrestricted cytotoxic CD3+CD56<sup>+</sup> T cells, and CD3+CD56*<sup>−</sup>* T cells. However, LAK cells had limited expansion *in vitro* and low cytolytic activity *in vivo*. Furthermore, LAK therapy required high doses of IL-2 *in vivo* and was associated with relevant toxicity. Modifications in culture conditions, i.e., provision of agonistic αCD3 (OKT3) monoclonal antibodies (mAbs), IL-2 and interferon (IFN)-γ, translated into *>*1,000-fold expansion of peripheral blood mononuclear cells (PBMCs) with potent cytokine-induced killer (CIK) activity. CIK cells share phenotypic and functional properties of both T cells and NK cells, as they co-express CD3 and CD56, and are rapidly expandable in culture like T cells, while not necessitating functional priming for*in vivo* activity, analogous to NK cells. Interestingly, CIK cells do not recognize target cells through the T-cell receptor (TCR) and do not require the presence of MHC molecules on target cells, as suggested by the observation that cytotoxicity is not affected by antibody masking of the TCR or MHC class I or class II molecules. CIK cells also express activating NK receptors, including NKG2D, DNAX accessory molecule-1 (DNAM), and NKp30 (45, 46).

Evidence for an *in vivo* activity of CIK cells derives from studies in a murine severe combined immune deficiency (SCID)/human lymphoma model, where co-administration of CIK cells with B-lymphoma cells had favorable effects on mice survival, with a 1.5–2.0-log cell kill and only marginal toxicity against normal hematopoietic precursors (47). CIK cells reportedly protect against syngeneic and allogeneic tumors also in other experimental models, including nude mice xenografted with human cervical carcinoma cells (48–50). CIK cells are detected in the lungs 30 min after injection, followed by distribution to other sites, such as the liver and spleen and, by 72 h, the tumor site, where CIK cells may remain for more than 9 days (51).

# **Current NK-Cell Manufacturing Practices**

A direct comparison of NK manufacturing techniques is hampered by differences in starting materials, technologies, and manipulation strategies (52, 53). Classically, GMP-compliant NK-cell products have been generated from PBMCs collected by apheresis (**Table 1**). It has been shown that NK cells obtained from granulocyte colony-stimulating factor (G-CSF)-mobilized leukapheresis products have reduced functional capacity (54). Conceivably, non-mobilized blood may be preferable over G-CSF mobilized blood as a source of NK cells for immunotherapy trials. A variety of cellular media have been used to culture NK cells, including X-VIVO serum-free medium, AIM V, or stem cell growth medium (SCGM), typically supplemented with 5–10% human AB serum to enhance NK function. Because the limited number of NK cells in leukapheresis products restricts clinical applicability, *in vitro* methods to expand NK cells are intensely being developed. In this respect, IL-15 promotes NK-cell proliferation and survival, and has been variably used in GMP-grade laboratory protocols, as further detailed below. Alternative methods of expansion rely on human feeder cells, including artificial antigen presenting cells (APCs) that are modified with costimulatory molecules, such as CD137 ligand, and membrane-bound (mb) IL-15 or IL-21. However, expanded NK cells undergo exhaustion, as shown by telomere shortening and replicative senescence.

In 2001, Carlens and co-workers described a cytokine-based technique for *in vitro* enrichment of human NK cells from bulk PBMCs of healthy individuals (79). PBMCs were incubated in SCGM containing 5% human serum and varying concentrations of IL-2. In addition, stimulation with OKT3 at 10 ng/ml was provided during the first 5 days of the culture. Supplementation with 500 U/ml IL-2 yielded a median 193-fold cell expansion in 21 days. Fifty-five percent of the expanded cells had a CD3*−*CD56<sup>+</sup> phenotype, and prolongation of the culture beyond 3 weeks did not allow further NK-cell enrichment. Moreover, expansion of the NK-cell compartment was comparable in cultures containing IL-2 concentrations ranging from 100 to 1,000 U/ml. Expanded cells could efficiently kill the NK-susceptible K562 line. This protocol was subsequently applied to PBMCs from patients with multiple myeloma (MM), an incurable plasma cell malignancy with a unique ability to subvert anti-tumor immune responses (80). Following an initial non-proliferative phase of 5 days, patientderived NK cells expanded 1,625-fold on average after 20 days of culture (71). NK cells from MM patients displayed increased expression of multiple activating receptors, including 2B4, NKp46, NKp44, NKp30, NKG2D, and DNAM-1, and were efficiently cytotoxic to K562 cells and primary autologous MM cells, but not to autologous CD34<sup>+</sup> cells (71). Mobilized PBMCs from patients with MM have also been used to expand NK cells (81). After a 7-day culture with serum-free AIM V media, IL-2 and OKT3, polyclonal populations of cytotoxic lymphocytes were detected, including CD4<sup>+</sup> T cells, CD8<sup>+</sup> T cells, CD8+CD56<sup>+</sup> T cells, and CD56<sup>+</sup> NK cells. Culture bags provided a two- to threefold expansion of immune effectors that retained their cytotoxicity after cryopreservation and thawing.

Notably, *ex vivo* expansion of NK cells from PBMCs incubated with IL-2 was also pursued under GMP-compliant conditions. Using an automated bioreactor system, bulk PBMCs from healthy donors and MM patients could expand 77-fold on average, and acquired enhanced cytotoxicity that positively correlated with the up-regulation of the NKp44 activating receptor. However, the expanded culture contained a significant proportion of T cells, necessitating further T-cell depletion prior to clinical use (61). Furthermore, purified CD56<sup>+</sup> populations were positively selected from PBMCs of healthy individuals using CD56 magnetic microbeads, and cultured in X-VIVO 10 medium containing 10% human AB serum and 500 U/ml IL-2 *±* 10 ng/ml IL-15 for 2 weeks. Appreciable proliferation occurred 5–7 days from the start of the culture, although with remarkable donor-to-donor variability. Expansion of CD3+CD56<sup>+</sup> NK-like T cells was two to three times greater than that of CD3*−*CD56<sup>+</sup> NK cells and was not affected by IL-15. Compared with the NK-92 cell line, *ex vivo* expanded CD56<sup>+</sup> cells had lower lytic activity against both K562 and Raji target cells (66).

The natural nicotinamide adenine dinucleotide (NAD)<sup>+</sup> precursor and NAD+-dependent enzyme inhibitor nicotinamide (NAM) has been recently shown to induce a 60- to 80-fold NK-cell expansion when added to feeder-free cultures containing IL-2 and IL-15 (82). In this study, NAM also affected NK cell anti-tumor capabilities and trafficking properties by modulating expression of CD200R and PD1, two immune regulatory receptors that transmit inhibitory signals upon interaction with cognate ligands on cancer cells. In addition, NAM promoted surface expression of L-selectin, an adhesion molecule mediating interactions with vascular endothelium and lymph nodes.

#### **CD3**<sup>+</sup> **T-Cell Depletion with or without CD56 Enrichment**

The CE-approved, partially automated Clini-MACS® instrument from Miltenyi allows the enrichment of NK cells under GMPcompliant conditions (58). After a single step of magnetic CD3<sup>+</sup> T-cell depletion, PBMCs are stimulated and expanded with irradiated autologous cells in the presence of OKT3 and IL-2, resulting in a highly pure population of functional CD3*−*CD16+CD56<sup>+</sup> NK cells that lack cytotoxicity against allogeneic non-tumor cells (83) (**Table 1**). Immunomagnetic CD3<sup>+</sup> T-cell depletion with either the 2.1 or the 3.1 programs can be combined with CD56 cell enrichment (84).When CD56<sup>+</sup> cells are magnetically isolated, the expansion of CD3+CD56<sup>+</sup> cells in culture may outweigh that of CD3*−*CD56<sup>+</sup> cells, since CD3<sup>+</sup> cells are not depleted upfront (66). Furthermore, CD56 expansion in cultures supplemented with IL-2, either alone or in combination with IL-15, shows

#### **TABLE 1 | Current GMP-compliant NK-cell manufacturing methods are detailed**.


*(Continued)*

#### **TABLE 1 | Continued**


*PBMC, peripheral blood mononuclear cells; LK, leukapheresis; d, day; UCB, umbilical cord blood; SCGM, stem cell growth medium; MM, multiple myeloma.*

substantial inter-donor variability. Each of the above programs translates into differences in depletion efficiency and recovery of NK cells, with NK purification being improved after sequential processing with the Clini-MACS T-cell depletion programs D2.1 and D3.1. Not unexpectedly, absolute NK-cell numbers after manipulation may correlate with the pre-harvest NK-cell content of the PB (85), implying that donors with high NK-cell counts are likely to provide NK-cell products with the highest cell numbers. A clinical-scale procedure to isolate NK cells for infusion in pediatric patients was developed under clean-room conditions (70). One-hour leukapheresis collections from unstimulated healthy donors were used to positively select CD56<sup>+</sup> cells and negatively deplete T cells, ultimately leading to cell therapy products enriched in NK cells and containing only 0.09% remaining T cells. A similar procedure consisting of two rounds of CD3 depletion and one round of CD56 selection has been used to obtain clinically applicable numbers of NK cells for immunotherapy (86). In that study, NK cells were expanded with IL-2 for 10–14 days to achieve the desired cell dose for potential clinical application in three children with relapsed or refractory leukemia after haploidentical HSCT.

Natural killer cells can also be expanded with irradiated autologous feeder cells, IL-2, IL-15, and anti-CD3 antibodies. Using these systems, NK cells acquire a CD56intCD16int phenotype and increase an average of 117-fold in 3 weeks (65). IL-2 and IL-15 mediate better NK expansion and viability compared with cultures nurtured with IL-2 only. Importantly, the number of residual contaminating T cells may be significantly lower after NK-cell exposure to IL-2 and IL-15 compared with IL-2 alone. NK cells activated with IL-2 and IL-15 may display higher cytotoxicity against K562 cells when kept in culture at a low effector-to-target ratio (66). In order to selectively expand alloreactive NK cells, KIR<sup>+</sup> cells can be isolated from Clini-MACS-purified CD3*−*CD56<sup>+</sup> NK cells using cell sorting, and then stimulated with the same cytokine cocktail (65). GMP-sorted and expanded single KIR<sup>+</sup> cells were cytolytic against AML blasts, an effect that was more pronounced than that mediated by bulk NK cells in an HLA-mismatched setting.

Interleukin-21 can offer theoretical advantages for the expansion of NK cells. The temporal exposure of IL-2/IL-15-stimulated NK cells to IL-21 determines the extent to which NK-cell proliferation and function are promoted (87). Specifically, NK cells stimulated with IL-21 during the first week of culture were shown to have strong proliferative response and cytotoxic activity compared with control cultures. The short-term expanded NK cells had longer telomeres than NK cells maintained with IL-21 continuously. IL-21 has also been used in combination with IL-15 to activate HLA-mismatched NK cells derived from CD34<sup>+</sup> hematopoietic progenitors with SCF, Flt3-L, IL-15, and hydrocortisone (72).

#### **Use of Feeder Cells**

While the minimum necessary NK-cell number for therapeutic efficacy is still controversial, the consistent generation of large amounts of functional cells is crucial to develop clinical protocols of adoptively transferred NK cells. Different feeder cell types have been used to expand NK cells, including irradiated PBMCs, EBV-transformed lymphoblastoid cell lines (EBV-LCL), genemodified K562 cells expressing NK cell-stimulatory molecules such as 41BB-ligand and mbIL-15 (67). Compared with IL-2 mediated activation, NK-cell expansion in the presence of feeder cells may also result in increased anti-tumor cytotoxic functions, with comparable *in vivo* survival (69, 88).

K562 cells were transduced with constructs encoding mbIL-15 (IL-15 + CD8α) and human 41BB-ligand (both containing green fluorescent protein). NK-cell recovery was 21.6-fold after 7 days of culture and increased to 152-fold and 277-fold after 14 and 21 days of culture, respectively. Importantly, the median recovery of NK cells was comparable when mononuclear cells from patients with acute leukemia were used in the co-culture. The expanded NK cells were cytotoxic against both AML cell lines and primary AML blasts. When compared with IL-2-stimulated NK cells, the cytotoxicity of expanded NK cells was greater at all effector-totarget ratios (67). In a mouse model of AML, multiple injections of expanded NK cells vigorously suppressed leukemia growth, with some mice achieving long-term control of the disease in the absence of xenogeneic GVHD. Finally, a master cell bank of K562**–** mb15**–**41BBL cells was established following GMP guidelines. The transduced NK cells were used to expand NK cells from leukapheresis collections at a 1:10 NK cell-transduced K562 cell ratio. The expansion of NK cells ranged from 33- to 141-fold after 7 days in culture. The overall yield of NK cells was higher than that observed in small-scale experiments.

A GMP-compliant NK-expansion methodology was also applied to patients with metastatic melanoma or renal cell carcinoma. A 278- to 1,097-fold NK-cell expansion was obtained when OKT3-loaded, 30-Gy-irradiated autologous PBMCs were used as feeders in AIMV medium containing 10% human AB serum and 600 U/ml IL-2 for 21–26 days. Following adoptive transfer to patients treated with a lymphodepleting regimen, expanded NK cells persisted for multiple days, likely representing the majority of NK cells in the circulation 1 week after infusion (68). Autologous PBMCs have also been used as feeders for the expansion of NK cells from healthy donors. Feeder cells obtained from the NKdepleted fraction of donor leukapheresis collections were used at a 10:1 feeder/NK-cell ratio for a GMP-compliant expansion procedure in Baxter LifeCell culture bags containing SCGM Cell-Gro medium, 5% human AB serum, and 200 U/ml IL-2 with or without IL-15 supplementation. This protocol was successful in propagating cultured NK cells, which expanded 117 *±* 20-fold after 19 days in the presence of 10 ng/ml IL-15 (65). More recently, a similar NK-cell expansion efficiency was reported when NK cells from healthy donors or patients with ALL in CR were co-cultured with autologous PBMCs in CellGro SCGM medium containing IL-2 and IL-15 (respectively, 34.9- vs. 39.5-fold average expansion after 14 days) (89).

Allogeneic PBMCs have been used as feeder cells for large-scale expansion of clinical-grade NK cells (62, 69). Allogeneic PBMCs and NK cells were co-cultured in X-VIVO 20 medium containing 500 U/ml IL-2 (69), or 100 U/ml IL-2 and 10 ng/ml IL-15 (62). In these studies, a similar 80- to 100-fold NK-cell expansion was achieved in 14–15 days. In an interesting study from Kim and colleagues, autologous PBMC feeders from cancer patients or PBMCs from healthy donors were compared (90). Co-cultures containing PBMCs from healthy donors could more efficiently propagate NK cells than those containing PBMCs from cancer patients (respectively, 300- vs. 169.4-fold average expansion after 14 days).

Pittari and colleagues described a novel technique for selection, deposition, and high-efficiency cloning of individual NK cells displaying surface receptor repertoires of choice. Cells were selected by FACS, deposited into *U*-shaped polystyrene 96-well plates (one cell per well) containing CellGro SCGM medium supplemented with 10% human AB serum and without exogenous cytokines. Propagation of NK clones from single cells was driven by *trans*-presentation of IL-15 by BaF/3 pre-B-lymphocytes double transfected with human IL-15Rα and human IL-15 (BaF/3 IL-15Rα/IL-15). Additional feeder cells were EBV-BLCL (JY) and PBMCs from three allogeneic donors (91). In this pre-clinical design, the technique allowed for prompt propagation of NK clones from NK-cell populations potentially involved in the control of leukemia relapse, i.e., expressing the KIR2DS1 activating receptor (22), regardless of their frequency (**Figure 1**). After 3 weeks, propagated NK clones typically reached 0.25–4 *<sup>×</sup>* <sup>10</sup><sup>6</sup> cells, with an overall cloning efficiency as high as 35–40%.

The replicative potential of NK cells expanded with genetically modified K562 cells can be further enhanced by enforcing the expression of human telomerase reverse transcriptase (TERT) gene (92). After stimulation with K562 cells for 1 week, NK cells were transfected with a retroviral vector containing human *TERT*. At variance with the control cultures that underwent replicative senescence after 16 population doublings, TERT-NK cells continued to expand *in vitro* for more than 1,000 days, if periodically re-stimulated with K562 cells. However, NK cells accumulated genetic changes at late time-points, including gain in genes on chromosome 1 and losses in genes on chromosome 16, suggesting that genetic instability may be a limiting factor in immortalization of NK cells.

Gas-permeable cell culture devices (G-Rex) are being evaluated for the expansion of T cells and tumor cells. In these systems, gas exchange across the base of the culture allows increased volumes of medium per unit area, augments the rate of cell expansion, and decreases cell death, minimizing cell manipulation. Using this strategy, up to 19 *<sup>×</sup>* <sup>10</sup><sup>9</sup> functional NK cells were produced starting from leukapheresis products, within 8–10 days of culture (93). The contaminating T cells mostly comprised CD8<sup>+</sup> T cells and could be removed by magnetic depletion. When compared with conventional gas-permeable bags, the G-Rex yielded higher fold expansions of NK cells, requiring no *interim* manipulation or feeding during the culture period. The NK cells were viable and functional, even after 12 months of cryopreservation.

#### **Use of Cord Blood and Other Stem Cell Sources to Expand NK Cells**

Umbilical cord blood is an emerging source of NK cells for clinical applications and also provides an *in vitro* system to analyze NK development (4). Banked UCB units represent an ideal "off-theshelf " source of NK cells for adoptive immunotherapy. Importantly, NK cells from PB and UCB differentially express cytokine receptors, with IL-15Rα being preferentially detected on UCB NK cells and IL-12Rβ1 and IL-18α receptors being primarily found on PB NK cells (94, 95). The combination of IL-15 and IL-18 optimally stimulates the proliferation of UCB NK cells and potentiates the release of IFN-γ and TNF-α. The lower responsiveness of UCB NK cells to IL-2 observed in these studies may be the result of lower expression of IL-2 receptors and of decreased phosphorylation of STAT5 as compared with PB NK cells. This implies that, at variance with PB NK cells that are fully activated by IL-2 alone, UCB NK cells may require additional cytokine stimuli (96). For instance, the addition of tacrolimus and low-molecular-weight heparin significantly enhances NK-cell expansion induced by IL-2, IL-15, and anti-CD3 mAbs (56). Using this protocol, approximately 40 *<sup>×</sup>* <sup>10</sup><sup>6</sup> NK cells were obtained from 1 *<sup>×</sup>* <sup>10</sup><sup>6</sup> unmanipulated UCB cells, in the absence of feeder cells, corresponding to *>*1,000-fold expansion. Bioreactors have been used to expand UCB-derived NK cells as well. This approach resulted into the generation of a clinically relevant dose of NK cells with *>*2,000 fold expansion, purity of *>*90%, high expression of activating receptors and cytolytic activity against K562 leukemia cells (55).

It has been shown that UCB-derived NK cells actively migrate to the BM, spleen, and liver 24 h after infusion in NOD-SCID-IL-2Rγ-null mice (97). NK cells were differentiated in 3–4 weeks from CD34<sup>+</sup> hematopoietic progenitors exposed to multiple cytokines, and were found to express CXCR4, CXCR3, and CCR6, which likely accounted for their ability to home to BM and inflamed tissues. A single NK-cell infusion combined with *in vivo* low-dose IL-15 resulted in inhibition of leukemia growth and prolongation of mice survival.

Finally, human embryonic stem cells (ESCs) as well as induced pluripotent stem cells (iPSCs) are potential sources of phenotypically mature and functional NK cells (98). ESCs and iPSCs were first used to produce hematopoietic progenitors with the "spin embryonic body (EB)" method, in which defined numbers

of cells were spin-aggregated in serum-free medium. This strategy removed the need for murine stromal support, and led to hematopoietic cell development and proliferation. Spin EBderived cells were then tested in a feeder-free and serum-free system containing NK-cell promoting cytokines, i.e., IL-3, IL-7, IL-15, SCF, and Flt3-L. Within the first 2 weeks of culture, both non-adherent CD31<sup>+</sup> endothelial cells and CD73<sup>+</sup> mesenchymal stromal cells were detected. Importantly, NK cells developed in similar numbers, phenotype, and functional characteristics as those differentiated with the use of murine stromal cells (98). Artificial APCs engineered to express mbIL-21 additionally expanded NK cells. As the expected requirement for NK-cell adoptive transfer protocols is approximately 2 *<sup>×</sup>* <sup>10</sup><sup>7</sup> NK cells/kg (see below), genetically modified APCs allow the use of a starting population of *<*10<sup>6</sup> ESCs/iPSCs per patient, corresponding to a lower number of cells compared with that required for NK-cell expansion from the PB.

#### **Use of Genetically Engineered NK Cells**

Although lentiviral (LV) vectors have been successfully used to transduce both T cells and NK-cell lines, LV transduction of both freshly isolated and *ex vivo*-expanded NK cells may be challenging. Chimeric antigen receptors (CARs) are synthetic engineered receptors that target surface molecules in their native conformation, independent of MHC and of antigen processing by the target cells (99). The generations of CARs are typically classified based on the intracellular signaling domains, with first-generation CARs including only CD3ζ, second-generation CARs including one single costimulatory domain and third-generation CARs including two costimulatory domains, such as CD28 and 41BB.

Natural killer cells can be transduced with mRNA encoding for anti-CD19 CARs. The expression of a receptor containing CD3ζ and 41BB signaling molecules (anti-CD19-BB-ζ) was induced in human NK cells with a clinical-grade electroporator. The cytotoxicity of the transfected NK cells was evaluated both *in vitro* Pittari et al. NK cells for hematological malignancies

and in a mouse model of leukemia. Receptor expression was already detectable 6 h after electroporation, reaching maximum levels at 24–48 h. Toxicity against CD19-expressing targets was specifically observed at 96 h. Median anti-CD19-BB-ζ expression 24 h after electroporation was 40.3 and 61.3% in freshly purified and in expanded NK cells, respectively. NK cells expressing anti-CD19-BB-ζ secreted IFN-γ in response to CD19<sup>+</sup> target cells and had enhanced cytotoxicity against B-cell malignancies (100). Transduced NK cells were consistently more cytotoxic than nontransduced NK cells. A large-scale, GMP-compliant protocol was also developed and showed that median percentage of genetically modified NK cells with receptor expression was 82% after 24 h. NK cells transfected under these conditions exerted *in vivo* cytotoxicity in NSG mice with B-cell leukemia, and suppressed leukemia progression compared with mice inoculated with mocktransfected NK cells (100). Interestingly, NK cells can acquire anti-CD19 CARs from donor cells via trogocytosis (101). When co-cultured with live K562 cells transduced to express anti-CD19- BB-ζ, NK cells acquired anti-CD19 CARs, peaking at 1 h and declining thereafter. NK cells displayed enhanced degranulation in response to leukemia cell lines compared with NK cells cocultured with control cells.

Genetically modified NK-cell lines, such as NK-92 cells, have been tested for *in vitro* and *in vivo* efficacy against MM. IL-2-independent derivatives of NK-92 cells, i.e., NK-92MI cells, have been transduced with a first-generation CAR targeting CD138, an integral membrane protein expressed on differentiated plasma cells (102). Genetically modified NK-92MI cells harbored a CAR consisting of an anti-CD138 single-chain variable fragment (scFv) fused to CD3ζ chain. The retargeted NK cells (NK-92MI-scFv) released IFN-γ and granzyme-B, and lysed CD138-expressing MM cell lines. When assayed in a xenograft NOD-SCID mouse model, transduced NK cells exerted more potent anti-tumor activity toward CD138-expressing MM cells than NK-92MI-mock. Importantly, NK cells could be detected in the MM microenvironment more than 20 days after their adoptive transfer.

CS1 is another surface protein highly expressed on MM cells and is amenable to targeting with CS1-specific CARs (103). CS1 co-localizes with CD138 on polarized plasma cells where it promotes adhesion, clonogenic growth, and tumorigenicity. Compared with mock-transduced NK cells, CS1-CAR-transduced NK cells had increased cytotoxic activity against CS1-expressing MM cells and showed heightened IFN-γ production. In an orthotopic MM xenograft model, adoptively transferred CS1-CAR-NK-92 cells suppressed the growth of human IM9 MM cells and significantly prolonged mouse survival (103). Overall, studies with CAR-NK cells point to the efficacy of this approach. The safety profile of CAR-NK cells may be advantageous compared with that of CAR-T cells, because of lack of *in vivo* clonal expansion and cytokine storm. Also, CAR-NK cells should not induce GVHD, while potently mediating GVL effects.

Natural killer cells can also be transduced to express mbIL-15 (104). Compared with NK cells expressing wild-type IL-15, mbIL-15 NK cells secreted low amounts of IL-15 in culture supernatants. Membrane-bound IL-15 appeared to be mostly occupying autologous receptors, suggesting that mb-IL-15 preferentially stimulates cells in *cis*, i.e., by direct binding to receptors expressed in the same cell. Genetically modified NK cells were maintained and expanded in culture without exogenous IL-2. When tested *in vitro* and *in vivo*, mbIL-15 NK cells displayed enhanced survival and cytotoxicity, being capable of inhibiting the growth of AML and sarcoma cells in NOD-SCID IL-2Rγ-null mice.

A complementary approach to existing methods of genetic modification of NK cells is offered by a retroviral vector-based, gene transfer protocol (105). Using a SFα11GFP viral vector, transduced NK cells were visible as GFP-expressing cells by fluorescence microscopy. The median transduction efficiency after one or two rounds of transduction was 27 and 47%, respectively. On day 21 of culture, transduction efficiencies averaged 52 and 75%, respectively. The gene transfer procedure did not affect NKcell phenotype or function, suggesting that retroviral vectors can be successfully applied to immunotherapy trials.

Clinical experience with CAR-engineered NK cells is in its infancy (106). Two clinical trials are currently open with the aim at exploring the therapeutic benefit of haploidentical NK cells modified with anti-CD19 CARs in children with B-cell precursor ALL (ClinicalTrials.gov NCT00995137) and in children and adults with refractory ALL (ClinicalTrials.gov NCT01974479). In these studies, NK cells will be expanded by co-culture with irradiated K562 cells modified to express mbIL-15 and 41BB-ligand. The expanded NK cells will be then transduced with a signaling receptor that binds to CD19 (anti-CD19-BB-ζ).

#### **Use of Expanded NK-Cell Lines**

Several malignant NK-cell lines have been established and used for clinical trials in China, Japan, and Western Countries, as reviewed elsewhere (107). A potential drawback of this approach is that differences in HLA molecules across different ethnicities may translate into the production of HLA-specific antibodies by the recipient. NK-92 cells, the most extensively characterized NK-cell line, were established in 1994 from the PB of a male Caucasian patient with non-Hodgkin lymphoma, are IL-2-dependent and harbor a CD2+CD56+CD57<sup>+</sup> phenotype (108–112). The adoptive transfer of NK-cell lines has theoretical advantages related to lack of expression of inhibitory KIRs, lack of immunogenicity, and ease of expansion. The optimal conditions for large-scale *ex vivo* expansion of NK-92 cells were recently defined. The protocol uses X-VIVO 10 serum-free media, supplemented with 450 U/mL of pharmaceutical grade rhuIL-2, and 2.5% allogeneic or autologous human serum or plasma (64). Cells maintained in gas-permeable culture bag systems with regular addition of fresh supplemented media achieve *>*200-fold expansion in 15–17 days, from a starting population of 6.25 *<sup>×</sup>* <sup>10</sup><sup>6</sup> cells to approximately 1.5 *<sup>×</sup>* <sup>10</sup><sup>9</sup> total cells per 1 L-culture. Patients with solid tumors or leukemia/lymphoma (*n* = 2) were treated with two infusions of escalating doses of NK-92 cells given 48 h apart, with no infusion-related or long-term side effects being observed (63). NK-92 cell doses ranged from 1 *<sup>×</sup>* <sup>10</sup><sup>9</sup> to 1 *<sup>×</sup>* <sup>10</sup><sup>10</sup> cells/m<sup>2</sup> . The dose of 10<sup>10</sup> cells/m<sup>2</sup> was considered the maximum expandable cell dose. NK-92 cells persisted *in vivo* for at least 48 h, as shown by Y chromosome-specific PCR in two female patients. Some responses were observed in patients with lung cancer. Only one patient developed anti-HLA antibodies, despite the allogeneic nature of NK-92 cells. NK-92 cells (Neukoplast® ) will continue to be tested in patients with solid tumors, e.g., Merkel cell cancer and renal cell carcinoma, and with hematological malignancies<sup>1</sup> (**Table 1**).

Since several decades, EBV-immortalized B-lymphoblastoid cells (EBV-BLCL) are known to robustly support NK cell *in vitro* expansion and anti-tumor activity (113–115). Escudier and colleagues used 35-Gy-irradiated LAZ 388 EBV-BLCL for the *ex vivo* expansion of NK cells from patients with metastatic renal cell adenocarcinoma. NK cells were initially cultured in V-bottom microplates, at a 4:1 feeder cell to NK-cell ratio, in DMEM medium supplemented with 200 U/mL IL-2. Two to five days before clinical use, NK cells were transferred to Baxter bags, where they received an additional 250 U/ml IL-2 boost. On average, expansion of cultured NK cells was limited to 50-fold after 21 days. However, some clinical responses were observed when autologous NK cells were used as consolidation treatment for patients in partial remission (116). Berg and co-workers described a GMPcompliant protocol involving a 20:1 EBV-BLCL feeder to NKcell ratio and 500 U/ml IL-2. This system allowed for a 300- to 930-fold NK-cell expansion. EBV-BLCL feeders prevalently drove such an extensive phenomenon, as the use of PBMCs in similar conditions yielded inferior results (69). Based on this protocol, a phase I clinical trial is currently investigating technical feasibility and clinical efficacy of large-scale NK infusions (up to <sup>1</sup> *<sup>×</sup>* <sup>10</sup><sup>9</sup> /kg) in cancer patients receiving bortezomib administered with the scope of increasing susceptibility of tumor cells to NKmediated lysis (117, 118).

K562 engineered to express mbIL-15 and 41BB-ligand (K562**–** mb15**–**41BBL) may be used to efficiently propagate NK cells with enhanced anti-leukemia properties. NK cells typically reach a *>*20-fold expansion after 7 days of co-culture, and a *>*1,000 fold expansion after 3 weeks, with no concomitant T cell propagation (67, 119). NK cells from patients with MM may also efficiently grow when co-cultured with K562**–**mb15**–**41BBL (120). When grown in GMP-compliant gas-permeable static cell culture flasks (G-Rex), as many as 19 billion unmanipulated NK cells can be obtained in 8–10 days starting from 150 million NK cells (93). Importantly, K562**–**mb15**–**41BBL cells have been successfully used to expand NK cells transduced with an anti-CD19-BBζ CAR, which display enhanced reactivity to CD19<sup>+</sup> leukemia cells (119). Similar to K562**–**mb15**–**41BBL, K562 genetically modified to express mbIL-21, or to co-express the ligand for 41BB and the NKG2D ligand MICA (K562**–**4-1BBL**–**mMICA), have been shown to promote large-scale expansion of NK cells with enhanced anti-tumor *in vitro* reactivity (121–123).

#### **Impact of Expansion Methods on NK-Cell Function and Homing Potential**

There are theoretical concerns that extensive *in vitro* expansion may affect the replicative potential and long-term viability of *in vivo-*infused NK cells. For instance, both Fas expression and susceptibility to apoptosis are increased after culture of NK cells with IL-2 or with feeder cells (124). In addition, expanded NK cells down-regulate receptors required for homing into secondary lymphoid organs, such as CCR7, a member of the G protein-coupled receptor family, and CD62L. In line with this, NK cells expanded with genetically modified K562 cells were shown to predominantly express a CD16+CD56<sup>+</sup> phenotype, with no detectable CCR7 (125). To obviate this, NK cells have been cultured with genetically modified, IL-21/CCR7 expressing K562 cells. These culture conditions reportedly resulted into transfer of CCR7 to 80% of expanded NK cells by trogocytosis, a fast, contactdependent uptake of membrane fragments, and molecules from "donor" to "acceptor" cells (126). CCR7 conferred migratory properties to NK cells by enhancing lymph node homing upon adoptive transfer to athymic nude mice. NAM dose-dependently increases CD62L expression on IL-2/IL-15-stimulated NK cells (82). NK cells expanded with NAM displayed better *in vitro* cytotoxic activity against a variety of tumor cell lines, including leukemia cells, and enhanced homing, as well as*in vivo* persistence in NOD-SCID mice.

Recently, two GMP-grade NK cells products manufactured at different production assistance for cellular therapies (PACT) facilities were evaluated for homing characteristics, i.e., freshly activated (FA)-NK, used by the Minnesota group, and *ex vivo*expanded (Ex)-NK, developed by the Baylor College of Medicine group (93, 127). Although the two preparations had phenotypic differences, cytotoxicity against NK-sensitive targets was similar. *In vivo* recovery after the infusion of thawed products was lower compared with the infusion of fresh NK cells. Whereas the negative impact of cryopreservation on FA-NK was rescued by overnight culture with IL-2, this strategy was less effective on Ex-NK cells, suggesting the need for optimized cell processing methods (127). NK cells could be detected at day 7 but failed to further expand between day 7 and day 14. Interestingly, higher numbers of functional NK cells with enhanced expression of NKG2A were recovered in mice infused with Ex-NK cells and given IL-15. The homing pattern of the two products was different, with higher numbers of NK cells being detected in the BM of mice given Ex-NK cells and IL-15 compared with Ex-NK cells and IL-2. Conversely, mice receiving FA-NK cells had more NK cells in the spleen when given IL-15. This study emphasizes the importance of continued cytokine stimulation for *ex vivo*-expanded cells, and suggests that differences in the manufacturing process affect *in vivo* homing and clinical efficacy of the NK-cell product.

# **Clinical Trials with NK Cells in Hematological Malignancies**

#### **Autologous NK Cells**

Early clinical studies exploited LAK-based immunotherapy in the autologous setting. One hundred eight patients with refractory metastatic cancer received LAK cells generated from autologous PBMCs incubated with 1,000 U/ml IL-2 for 3–4 days. Systemic high-dose IL-2 was given to support LAK cells *in vivo* (128, 129). Objective tumor regression occurred in 22% of 106 evaluable patients. Median response duration was 10 months for eight patients achieving complete remission (CR). Further prospective studies assessing the therapeutic effects of high-dose IL-2 and LAK cells indicated a possible survival advantage for patients with melanoma treated with LAK cells (130).

<sup>1</sup> http://www.conkwest.com/nk-92

Immunotherapy with systemic IL-2 and autologous LAK cells was also given as consolidation treatment after autologous bone marrow transplantation (BMT). Sixteen patients with lymphoma received, 12–14 days post-transplantation, LAK cells generated from PBMCs incubated with IL-2 for 5 days (131). In a similar setting, NK cells obtained prior to transplant and activated with IL-2 for 6 days were infused into 12 patients with advanced cancer and post-BMT pancytopenia (132). Concomitant with NKcell transfer, sequential high to low-dose systemic IL-2 was also administered for over 90 days. This approach was well tolerated and resulted in the early enhancement of NK-cell activity in four recipients (132).

In general, trials with high-dose systemic IL-2 to support circulating LAK or NK cells were limited by severe and potentially lethal toxicities (e.g., vascular leak syndrome, oliguria, hypotension, myocardial infarction), counterbalancing the beneficial anticancer effects of LAK activity (129, 131, 132). On the other hand, chronic low-dose IL-2 treatment was relatively well tolerated (133–135), but unable to activate NK cells as robustly as high-dose *ex vivo* IL-2, or IL-2 at concentrations that engage the intermediate-affinity IL-2 receptor on NK cells (134, 136, 137). Subsequent studies sought to maximize NK-mediated anti-tumor effects. *Ex vivo* IL-2-activated NK-cell infusions were compared with supplemental intravenous IL-2 boluses on days 28 and 35 during daily subcutaneous IL-2 administration in patients with relapsed lymphoma or metastatic breast cancer. Both treatment conditions induced strong NK-cell anti-tumor reactivity, and boosted circulating cytokines, without any consistent impact on clinical outcome compared with matched patients from the Autologous Blood and Marrow Transplant Registry database (138).

The proliferation potential of NK cells isolated from cancer patients may be similar to that of NK cells from healthy donors, reassuring about the feasibility of manufacturing autologous NKcell products. Although autologous NK cells persist *in vivo* for at least 1 week after infusion, they express lower levels of NKG2D, a key activating receptor, and may necessitate *in vitro* re-activation with IL-2 to lyse tumor targets (68).

Collectively, the analysis of phase II immunotherapy studies with autologous NK cells failed to show efficacy (139). Several factors may have accounted for the disappointing results, including competition with the recipient's lymphocytes for cytokines and "space"; inhibition of autologous NK cells by self-MHC (30, 140, 141); chronic immunosuppression induced by the tumor on host immunity; and expansion of Treg cells by IL-2 (127). Autologous NK cells are currently being tested in patients with hematological malignancies and solid tumors (NCT00720785; **Table 2**) (142). In this trial, which is recruiting participants, patients will receive immune suppressive therapy with pentostatin, followed by bortezomib to sensitize tumor cells to NK cytotoxicity (143), escalating doses of autologous NK cells and IL-2.

#### **Allogeneic NK Cells**

Initial clinical trials on allogeneic T-cell depleted (TCD) haploidentical HSCT for patients with AML showed enhanced NKmediated cytotoxicity when KIR-HLA class I mismatch occurred (23). In 2005, Miller and co-workers administered haploidentical NK cells in a non-transplantation setting to 43 patients with advanced cancer. Three pharmacological regimens of different intensity were used to prevent immunological rejection (57). After a single leukapheresis, CD3<sup>+</sup> T cells were depleted under GMP conditions using CD3 microbeads. The TCD product was activated overnight with IL-2 before infusion. NK cells were enriched to 40% on average after processing. The final IL-2-activated product contained an NK-cell dose of 8.5 *<sup>×</sup>* <sup>10</sup><sup>6</sup> cells/kg of recipient's body weight and a final T-cell dose of 1.75 *<sup>×</sup>* <sup>10</sup><sup>5</sup> cells/kg. A lowintensity immune suppressive regimen was administered on an outpatient basis to the first 17 patients, followed by the infusion of escalating doses of NK cells. Importantly, no dose-limiting toxicity occurred in this patient cohort. Using RT-PCR primers for donorspecific MHC class I alleles, donor cells were shown to persist for 5 days and to comprise *<*1% of circulating PBMCs, likely due to immune rejection. Alternative low-intensity (fludarabine alone) and high-intensity immune suppressive regimens (highdose cyclophosphamide and fludarabine) were subsequently used in seven patients with renal cell carcinoma and 19 patients with poor-prognosis AML, respectively. None of the patients given fludarabine alone engrafted with donor NK cells. By contrast, 8 of 15 evaluable patients with AML showed at least 1% engraftment of donor cells after NK-cell infusion. Overall, five patients achieved a morphological CR. Among the four patients with a KIR-ligand mismatch in the graft-versus-host direction, three achieved CR. This was paralleled by the observation that CR was obtained only in 2 of 15 patients without alloreactivity. Finally, IL-15 serum levels were significantly higher in patients receiving higher-intensity, lymphodepleting immune suppression, suggesting that a rise in endogenous IL-15 may be required for the *in vivo* expansion and persistence of infused NK cells.

Subsequent studies suggested the possibility that IL-2 administered after NK infusions also expands Treg cells, potentially interfering with *in vivo* NK-cell proliferation (76). To address this issue, IL-2 diphtheria toxin (IL-2DT) was administered to 15 patients with relapsed/refractory AML, 1 day before the enriched NK product (145). IL-2DT is a recombinant cytotoxic fusion protein composed of the amino acid sequences for diphtheria toxin followed by truncated amino acid sequences for IL-2. IL-2DT reportedly depletes IL-2 receptor α-chain (CD25)-expressing cells, including Treg cells. In this study, three processing methods were used to prepare NK-cell products, including CD3 depletion alone, CD3 depletion followed by CD56 selection, and single-step CD3/CD19 depletion. Higher NK-cell doses were obtained after depletion of CD3<sup>+</sup> and CD19<sup>+</sup> cells from a 5-h donor apheresis collection. Among the 42 patients who did not receive host Treg depletion, 21% achieved remissions. By contrast, Treg depletion with IL-2DT resulted in remissions at day 28 for eight of 15 patients (53%). The magnitude of NK-cell expansion was also higher after Treg depletion. The ability of IL-2DT to deplete Treg cells was further supported by reductions in serum IL-35 concentrations 14 days after adoptive transfer. Finally, 7 out of 10 patients (70%) with detectable donor NK cells attained CR by day 28 compared with only 1 of 5 patients (20%) lacking detectable donor NK cells at day 7, suggesting that NK-cell persistence is required for clinical efficacy. Interestingly, this study showed no correlation between achievements of CR- and KIR-ligand mismatch.

#### **TABLE 2 | Completed and ongoing clinical trials with NK cells for patients with hematological malignancies are listed**.


*(Continued)*

#### **TABLE 2 | Continued**


*AML, acute myeloid leukemia; MRD, minimal residual disease; ALL, acute lymphoblastic leukemia; LY, lymphoma; MY, multiple myeloma; JMML, juvenile myelomonocytic leukemia; NHL, non-Hodgkin lymphoma; CY, cyclophosphamide; FLU, fludarabine; UCBT, umbilical cord blood transplantation; RIC, reduced intensity conditioning; MRD, matched related donor; MUD, matched unrelated donor; NCI, National Cancer Institute; ASH, American Society of Hematology. \*Clinical-grade CTV-1 lysate that primes NK cells ex vivo; see main text for further details. N.A. , not available. www.clinicaltrials.gov website, last accessed March 2015.*

Haploidentical, KIR-ligand-mismatched NK cells have been safely infused in elderly patients with high-risk AML, with some evidence of clinical benefit, especially for patients treated in CR or for those with molecular disease relapse (75). Approximately, 40% of the screened patients had a KIR-ligand-mismatched donor, suggesting that this strategy may be applicable to a significant proportion of patients with AML. Another study attempted to exploit KIR-ligand-mismatched NK cells from haploidentical family donors in patients with relapsed MM (74). The apheresis products were TCD and then cultured with IL-2, either overnight or during incubation with anti-CD3 beads. Patients received melphalan and fludarabine as conditioning regimen. After NKcell infusion, IL-15 levels increased. The response rate (RR) was 50%, with no patient developing GVHD. However, donor chimerism was eventually lost in conjunction with the appearance of host–anti-donor immune responses. The clinical application of allogeneic NK cells to patients with MM is further encouraged by the observation that most MM cell lines are susceptible to NK attack *in vitro*, showing no evidence for HLA class I loss (146).

Resting human NK cells can be primed to kill NK-resistant tumor cells by co-incubation with a clinical-grade lysate of the leukemia cell line CTV-1 (CNDO-109). CNDO-109-activated NK cells remain primed, with no requirement for IL-2 treatment, and can be cryopreserved. The safety, outcome and NK chimerism data from an ongoing phase I/II transitional clinical trial of CNDO-109-NK cells have been recently reported (77). This 3 *×* 3 dose-escalation phase 1 study was opened in 2013 for patients with high-risk AML in first CR and with no conventional treatment options available. Patients received preparative chemotherapy consisting of cyclophosphamide and fludarabine on study day-6 to -2, followed by a single dose of CNDO-109-activated NK cells on day 0. Patients were given different doses of NK cells (cohort <sup>1</sup> <sup>=</sup> <sup>3</sup> *<sup>×</sup>* <sup>10</sup><sup>5</sup> , cohort 2 <sup>=</sup> <sup>1</sup> *<sup>×</sup>* <sup>10</sup><sup>6</sup> , and cohort 3 <sup>=</sup> up to 3 *<sup>×</sup>* <sup>10</sup><sup>6</sup> cells/kg). CNDO-109-NK cells were manufactured from a single apheresis collection from HLA-haploidentical-related donors. NK cells were isolated with anti-CD56 microbeads and co-incubated overnight with CNDO-109 lysate under GMP conditions (Coronado Biosciences)<sup>2</sup> . Residual T-cell contamination (defined as *<*10<sup>4</sup> cells/kg patient body weight) was considered a safety criterion for lot release. Seven eligible patients were enrolled. No infusion-related toxicities or adverse events directly attributed to NK therapy were observed, including GVHD. Patients experienced transient myelosuppression lasting approximately 2 weeks. Three patients relapsed early post-treatment (average time to relapse from CR1 being 104 days). In five of seven evaluable patients, persistence of activated donor NK cells was observed from day 7 post-infusion to as late as day 56 in one patient. The comparison of donor and patient endogenous NK cells showed a mature activated phenotype of donor NK cells. Two of the three patients evaluated had persistence of low levels of activated autologous NK cells (~10–20% of circulating NK cells), exceeding the numbers circulating pre-treatment. This observation may indicate that NK-cell therapy induces endogenous NK activation and enhances innate immunity to AML in the absence of exogenous cytokine administration. When the study was published in abstract form, four of the seven patients enrolled were relapse-free.

Natural killer cells have also been administered pre-emptively to patients with high-risk cancer, after TCD haploidentical HSCT (73). Sixteen patients were treated in a prospective phase II study with purified NK cells on day 3, 40, and 100 after HSCT. The median dose of NK cells was 12.1 *<sup>×</sup>* <sup>10</sup><sup>6</sup> /kg. With a median follow-up of approximately 6 years, 4 out of 16 patients were alive. The four patients who developed acute GVHD had received *<sup>&</sup>gt;*0.5 *<sup>×</sup>* <sup>10</sup><sup>5</sup> /kg contaminating T cells. Compared with a historical cohort of patients treated with haploidentical HSCT without NK donor lymphocyte infusions (DLIs), NK cells apparently exerted no effect on disease relapse.

<sup>2</sup> http://www.coronadobiosciences.com/research-development/cndo-109.cfm

#### Trials in Children and Young Adolescents

A landmark study from St. Jude Children's Hospital (NKAML) has shown the safety, feasibility, and engraftment potential of haploidentical NK cells for children with favorable and intermediaterisk AML (59). Patients received a mild conditioning regimen, consisting of cyclophosphamide (60 mg/kg on day-7) and fludarabine (25 mg/m<sup>2</sup> /day from day-6 to -2), followed by KIR-HLA-mismatched NK cells (median number of cells infused <sup>29</sup> *<sup>×</sup>* <sup>10</sup><sup>6</sup> /kg) and six doses of IL-2 (1 *<sup>×</sup>* <sup>10</sup><sup>6</sup> U/m<sup>2</sup> ). Donor PBMCs obtained by leukapheresis were depleted of CD3<sup>+</sup> T cells and then enriched for CD56<sup>+</sup> cells. The manipulated product contained a very low number of contaminating B cells and T cells. The resulting NK population was infused fresh, without any incubation with IL-2. All patients were in CR, as shown by minimal residual disease status. NK infusions were well tolerated, with no GVHD observed. NK-cell engraftment was transient in all patients, with a median peak NK-cell chimerism of 7%. One patient had prolonged NK engraftment with 2% of donor NK cells being detected at day 189, in association with delayed neutrophil and platelet engraftment. With a median follow-up of approximately 32 months, all patients remained in remission. The 2-year EFS estimate was 100%.

The therapeutic potential of NK cells has also been tested in young adolescents. A two-step T-cell depletion strategy with a final CD56 enrichment step was pursued to isolate NK cells from steady-state leukapheresis collections (147). Purified NK cells were expanded for 2 weeks in X-VIVO 10 medium containing 10% human AB serum and 1,000 U/ml IL-2. This procedure resulted in a median 95% NK-cell purity, 99% viability, and enhanced cytotoxicity against the K562 line as well as primary leukemic blasts obtained from patients. T-cell contamination was negligible (*<*0.1%). Three children with multiply relapsed ALL or AML were treated with IL-2-stimulated NK cells after haploidentical HSCT. Directed KIR mismatches in the GVL direction were present in all three cases. Remission was achieved in all cases, although patients ultimately died of infectious complications or disease relapse. Another study in children and young adolescents with ultra-high-risk solid tumors has shown that donor-derived NK cells activated with IL-15 and 41BBL can be safely administered after HLA-matched TCD HSCT (148). NK cells displayed potent killing capacity. However, five of nine transplant recipients developed acute GVHD, with grade III GVHD being observed in three patients. The unexpected occurrence of GVHD in this report may be attributed to timing of NK-cell infusion, lack of posttransplantation immune suppression or activation of NK cells that were expanded on IL-15-secreting feeder cells (149). The observation that GVHD developed in all four patients given unrelated donor transplants compared with one of five patients given related donor transplants points to a T-cell-driven mechanism, mediated by minor antigens and accentuated by the infused NK cells.

#### *In vivo* **Targeting of NK Cells with Antibodies**

IPH2101 is a first-in-class, non-depleting human IgG<sup>4</sup> mAb directed against inhibitory KIRs, and functions by blocking inhibitory KIR–ligand interactions, leading to restored or augmented NK-cell function against tumor cells. A phase I trial of IPH2101 (#NCT00552396) was conducted in 32 patients with relapsed/refractory MM (150). IPH2101 was given intravenously every 28 days in sevn dose-escalated cohorts (0.0003–3 mg/kg) for up to four cycles. This study identified doses of IPH2101, which conferred KIR2D occupancy *in vivo*, with no concomitant dose-limiting toxicity or identification of a maximally tolerated dose (MTD). With one exception, adverse events were mild and transient and mainly consisted of self-limited infusion reactions. Although IPH2101 enhanced *ex vivo* patient-derived NK-cell cytotoxicity against MM, no objective responses were observed. Another phase I study of escalating doses of IPH2101 in 23 elderly patients with AML in first CR showed a correlation between IPH2101 exposure and KIR occupancy (151). Adverse events were mild and transient, consisting mainly of infusion syndrome and erythema. The study drug did not affect the numbers and distribution of lymphocyte subsets and NK cell receptor expression. At the highest dose levels, TNF-α and MIP-1β serum concentrations, as well as CD69 expression on NK cells, transiently increased. Overall and relapse-free survival compared favorably to reports in other patient populations with similar characteristics (151).

An immunization study of transgenic mice bearing human immunoglobulin loci with different combinations of KIR2DLs has recently led to the identification of the 1-7F9 mAb, based on binding to soluble, recombinant KIR2DL1, -2, and -3 by ELISA (152). The 1-7F9 antibodies bind to human NK cells, γδ<sup>+</sup> T cells, and CD8<sup>+</sup> T cells, consistent with KIR expression patterns. Using *in vitro* assays, 1-7F9 blocked the binding of HLA class I to inhibitory KIR2DLs, augmenting NK cell-mediated killing of HLA-C-expressing targets. 1-7F9 also enhanced the cytotoxicity of NK cells from an HLA-C-matched donor against AML blasts. Pre-treatment with 1-7F9 of patient-derived NK cells translated into a two- to threefold increase in cytotoxicity against autologous AML cells. Finally, studies in NOD-SCID mice showed that 1-7F9 potentiates NK-mediated killing and promotes mice survival compared with co-infusions of NK cells and AML blasts alone (152).

#### **Bi-Specific and Tri-Specific Antibodies**

A novel class of therapeutics uses either all or part of the antibody structure to deliver enhanced effector activity to the tumor site. The fusion of two (bi-specific) or three (tri-specific) portions of the fragment of antigen-binding (Fab) region of a traditional antibody yields reagents with high level of antigen specificity and cross-links tumor antigens with potent immune effectors. Bi-specific killer engagers (BiKEs) are constructed with a singlechain Fv against CD16 and a single-chain Fv against a tumorassociated antigen. The mechanisms by which BiKEs and TriKEs potentiate NK effector functions include intracellular calcium mobilization through direct CD16 signaling (153). Co-culture of reagent-treated resting NK cells with Raji targets also translates into increased NK-cell degranulation, target cell death, and NK production of IFN-γ, TNF-α, GM-CSF, IL-8, MIP-1α, and RANTES.

Fully humanized CD16 *×* 33 BiKEs have been shown to trigger NK-cell activation *in vitro* against CD33<sup>+</sup> AML cell lines and primary refractory CD33<sup>+</sup> AML targets (154). Combination treatment with BiKEs and ADAM17 inhibitor to prevent CD16 shedding further enhanced NK-cell function. BiKEs were also effective at activating NK cells from recipients of double UCB transplantation. BiKEs enhance degranulation and cytokine production by NK cells derived from patients with myelodysplastic syndromes and cultured with CD33<sup>+</sup> AML cell lines, irrespective of disease stage and age stratum (155). A potential drawback of this approach is the relatively short halflife of the antibody constructs, with limited trafficking to the tumor site.

#### **Other Approaches to Target NK Cells** *In vivo*

A first in-human trial of *Escherichia coli*-produced rhIL-15 has been recently published (156). The primary objectives of this single-institution, open-label, non-randomized 3 + 3 design, phase I dose-escalation study were to assess safety, dose-limiting toxicity, and MTD of rhIL-15 given as intravenous bolus at 3.0, 1.0, and 0.3 μg/kg/day for 12 consecutive days to patients with metastatic malignant melanoma or renal cancer. Reductions in the number of lymphocyte subsets were evident within 20 min of the infusion of IL-15, with the most dramatic decline being observed with NK cells, γδ<sup>+</sup> T cells and CD8<sup>+</sup> memory T cells. The acute efflux of CD8<sup>+</sup> cells from the circulation was most pronounced for transitional and effector memory subsets. An influx of NK cells to the blood was detected by 4 h, followed by a normalization of cell counts by 2–3 days. The initial influx of cells mainly resulted from redistribution, because no evidence of proliferation was observed until 48 h. CD8<sup>+</sup> cells showed evidence of activation, including proliferation and increased expression of CD38 and HLA-DR. Doselimiting toxicities in patients receiving 3.0 and 1.0 μg/kg/day included grade 3 hypotension, thrombocytopenia, and elevations of ALT and AST, resulting in 0.3 μg/kg per day being determined the MTD. Common cytokine-related adverse events, including fever, rigors, and hypotension occurred much less frequently in patients treated with the 0.3 μg/kg dose level. Cytokines such as IFN-γ, TNF-α, IL-6, and IL-8 increased up to 50-fold in patient serum after rhIL-15 administration. Overall, there were no clinical responses in this study, with stable disease being recorded as a best response. However, five patients manifested a decrease between 10 and 30% in their marker lesions, with two of these patients experiencing clearing of lung lesions (156). This study proves that rhIL-15 administration, as an intravenous bolus dose, is associated with clinical toxicities due to marked cytokine secretion. The authors initiated a dose-escalation trial of continuous intravenous infusion of rhIL-15 to patients with metastatic malignancies. Furthermore, the Cancer Immunotherapy Trials Network (CITN<sup>3</sup> ) is conducting a phase I doseescalation trial of subcutaneous rhIL-15 administered 5 days per week for 2 weeks.

# **Current Manufacturing Practices for CIK Cells**

Current protocols to differentiate CIK cells are based on the combination of 1,000 IU/ml IFN-γ on day 1 of culture followed 24 h later by 50 ng/ml OKT3 and 300 IU/ml IL-2 (157). After 21–28 days, CD3+CD56<sup>+</sup> cells, derived from CD3+CD56*<sup>−</sup>* cells, acquire cytotoxicity against various tumor cell targets, including AML, chronic myeloid leukemia, and B-cell and T-cell lymphoma. The expression of CD56 on CIK cells is thought to result primarily from IFN-γ priming with subsequent IL-12 production from monocytes. Recently, a GMP-grade protocol (ITG2) that incorporates thymoglobulin® (TG) was used to prepare CIK cells (**Figure 2**; **Table 3**) (45). TG is a purified, pasteurized preparation of polyclonal γ immunoglobulin raised in rabbits against human thymocytes (45). TG expanded CIK cells more efficiently than the

<sup>3</sup> http://citninfo.org/citn-science/clinical-studies.html


**TABLE 3 | Completed and ongoing clinical trials with CIK cells for hematological malignancies are listed**.

*PBMC, peripheral blood mononuclear cells; AL, acute leukemia; DLBCL, diffuse large B-cell lymphoma; LK, leukapheresis; CML, chronic myeloid leukemia; MDS, Myelodysplastic syndromes; MPD, myeloproliferative disorders; d, day; UCB, umbilical cord blood; SCGM, stem cell growth medium; MM, multiple myeloma. www.clinicaltrials.gov website last accessed March 2015.*

anti-CD3 mAb when provided to clinical-grade cultures in combination with IFN-γ and IL-2. Higher levels of NKG2D, NKp46 triggering receptor, and killer-like immunoglobulin receptors KIR2DL1 and KIR2DL2/DL3 were detected on CIK cells differentiated with TG compared with those obtained with αCD3 antibodies. CIK cells were capable of lysing tumor cell targets in an MHC-unrestricted manner and released high quantities of bioactive IL-12p40. The use of the ITG2 protocol was not associated with the emergence of Treg cells*in vitro*, thus reassuring against the infusion of excessive numbers of tumor-suppressive T-cell populations.

The administration of bulk CIK cells is not associated with GVHD in mice, even after sequential infusions (158). The same study also showed that bulk CIK cells are more effective than selected CD56<sup>+</sup> CIK cells or CIK cells depleted of potentially alloreactive αβ<sup>+</sup> CIK cells.

Several investigators have successfully obtained CIK cells from UCB units. The washouts of UCB units may yield approximately 500 *<sup>×</sup>* <sup>10</sup><sup>6</sup> CIK cells after a standardized 21-day expansion culture (159). Compared with PB-derived CIK cells, UCB CIK cells demonstrate lower immunogenicity and higher proliferative capacity and anti-tumor activity in pre-clinical models of cancer (160). Interestingly, UCB-derived CIK cells released higher amounts of IL-2 and IFN-γ and expressed higher levels of CCR6 and CCR7, pointing to a better ability to traffic to tumor sites and secondary lymphoid organs. CIK cells differentiated from PB and UCB also differ in receptor expression and in cytotoxic activity against ALL cells, suggesting that the source of CIK cells may impact on therapeutic efficacy (161).

# **Clinical Trials with CIK Cells**

An International Registry on CIK Cells (IRCC) has been established with the aim of reporting results from clinical trials centered on adoptively transferred CIK cells. In the first IRCC publication (167), eleven clinical trials with autologous or allogeneic CIK cells were identified, with 426 patients enrolled. Most trials included male patients with hepatocellular carcinoma, gastric cancer, and relapsed lymphoma. A clinical response was reported in 384 patients, who received up to 40 infusions of CIK cells. The total RR was 24% and a decrease of tumor volume was documented in three patients. DFS rates were significantly higher in patients treated with CIK cells than in a control group without CIK treatment. An update published in 2014 (168) enlists the results obtained in 2,729 patients from 45 phase I/II studies. A total of 1,520 patients with 22 different tumor entities were treated with CIK cells, either alone or in combination with chemotherapy. Allogeneic CIK cells were employed in the majority of the trials (41 out of 45). The number of CIK cells infused varied among the different trials, averaging 7.7 *<sup>×</sup>* <sup>10</sup><sup>9</sup> cells. Data on patient survival were available for 19 out of 45 trials; 15 of these 19 trials were paired, as they also included control patients receiving none or standard therapy alone. Overall, a beneficial effect of CIK cells emerged in patients with hepatocellular carcinoma, renal cell carcinoma, non-small cell lung cancer (NSCLC), colorectal carcinoma, and breast cancer. High numbers of CIK cells at time of infusion were associated with a better prognosis. Quality of life was also improved in four of the five trials for which data were disclosed. Ancillary biological data were provided in 23 studies. The absolute numbers of T cells, as well as serum IFN-γ, were increased after immunotherapy compared with baseline. Some studies also reported a decrease of blood Treg cells after CIK infusions. Immunotherapy was generally well tolerated, with fever occurring in 41% of cases and headache and fatigue reported in 30% of cases (168). Mild GVHD was observed in seven of 52 patients treated with allogeneic CIK cells and was responsive to steroid therapy. Based on these findings, the IRCC currently recommends the use of at least 10 *<sup>×</sup>* <sup>10</sup><sup>9</sup> CIK cells with at least 30% CD3+CD56<sup>+</sup> cells per infusion, every 2–4 weeks, for at least six times. The IRCC website<sup>4</sup> (last accessed March 2015) lists 1,787 patients treated with CIK cells, mostly for hepatocellular carcinoma, gastric cancer, ovarian cancer, renal cell cancer, MM, and NSCLC.

China has the largest population of patients with malignant tumors and is the country where most clinical trials with CIK cells were conducted (169). Using the VIP database of Chinese scientific and technological journals<sup>5</sup> , 24 articles were selected. Information on the total number of CIK cells used was available in 16 studies and ranged from 6 *<sup>×</sup>* <sup>10</sup><sup>6</sup> to 1.5 *<sup>×</sup>* <sup>10</sup><sup>10</sup> in one single treatment course. As far as hematological malignancies are concerned, 1 study specifically dealt with non-Hodgkin's disease (12 patients) and 4 studies with AML (51 patients). Only 14 out of 24 studies contained details about clinical outcome. Of the reported 563 patients, 40 had a CR, 126 had a partial response, 125 had a minimal response, 135 had stable disease, and 58 had progressive disease. The remaining 76 patients did not reach an objective response. The overall RR was 51.7% (291/563) (169). Information on patient outcome was provided in 10 studies. Four studies reported a 1-year OS rate of 72.5%, six studies reported 2-year OS rate of 66.3%, one study reported 3-year overall RR of 75.5%, and two studies reported 5-year OS rate of 38.2%.

A phase I study of allogeneic CIK cells has been conducted in 11 patients with hematological malignancies relapsing after allogeneic HSCT (157). Six patients had received other salvage treatments before CIK administration, including DLIs, without significant clinical response. The median number of CIK cells infused was 12.4 *<sup>×</sup>* <sup>10</sup><sup>6</sup> /kg, with no infusion-related toxicities recorded. Acute GVHD occurred in four patients 30 days after the last CIK infusion, and progressed into extensive chronic GVHD in two cases. Disease progression and death were observed in six patients. One patient had stable disease, one experienced hematologic improvement, and three obtained complete responses. The same authors differentiated CIK cells from UCB samples and administered them to five patients relapsed after UCB transplantation (162). In three patients, chemotherapy had been given before CIK administration to reduce disease burden. Infusions of a median of 1.5 *<sup>×</sup>* <sup>10</sup><sup>6</sup> /kg CIK cells were provided early after leukemia relapse. Some clinical response was observed in one patient who also developed acute intestinal GVHD. The remaining four patients experienced disease progression and died. GVHD was managed with steroids and mesenchymal stromal cells but the patient ultimately succumbed to leukemia relapse (162).

In another study (170), CIK cells were generated for 24 patients with advanced-stage hematological malignancies, mostly from allogeneic donors, either under steady-state conditions or after stem cell mobilization. Overall, 55 infusions were given to 16 patients at doses ranging from 10 to 200 *<sup>×</sup>* <sup>10</sup><sup>6</sup> CD3<sup>+</sup> cells/kg. Notably, the proportion of the CD3+CD56<sup>+</sup> subset was higher in CIK cultures derived from patients than in those differentiated from healthy donors. The median expansion of CD3<sup>+</sup> T cells and CD3+CD56<sup>+</sup> NK-like T-cells was 9.33-fold and 27.77-fold, respectively. Responses attributable to CIK infusions were documented in five patients, including two with ALL, two with Hodgkin disease, and one with AML. In five patients, the response to CIK cells could not be assessed as salvage chemotherapy or dasatinib was concomitantly administered. Acute GVHD occurred in three patients and was manageable. For three of the six patients failing to respond, leukemia cells were available for *in vitro* killing assays, which showed only modest cytotoxicity of donor CIK cells against tumor targets. Overall, this study in advanced-stage hematological malignancies suggests that CIK administration may translate into some clinical efficacy with a modest toxicity and low incidence and severity of GVHD.

Finally, autologous immune effector cells generated by TG, IFN-γ, and IL-2 (ITG2) can be safely administered to patients with advanced and/or refractory solid tumors (**Figure 2**) (171). After 2–3 weeks in culture, a median of 4.65 *<sup>×</sup>* <sup>10</sup><sup>6</sup> immune effectors/kg of recipient's body weight was infused intravenously without observing any toxicity. One patient with advanced melanoma died because of disease progression before the infusion of CIK cells. The target dose of at least 2.5 *<sup>×</sup>* <sup>10</sup><sup>6</sup> CIK cells/kg of recipient's body weight was reached in four out of five evaluable patients. The median survival was 4.5 months (range 1–13) from the first infusion of CIK cells (171).

# **Closing Remarks**

Although the field of NK cell and CIK cell-based immunotherapy is rapidly advancing, some pre-clinical and clinical issues need to be clarified before this immunotherapy approach is widely offered to patients with hematological malignancies and solid tumors (172).

Cell therapy products enriched for NK cells using CD3 depletion and CD56 selection contain variable percentages of monocytes. It is now established that both monocytes and monocyte-derived DCs can support NK-cell proliferation and function (173). This implies that different NK-cell manufacturing protocols may affect the cellular composition of the final products and impact on NK-cell function. Also, the optimal number of NK cells to be infused remains to be determined. Patients reportedly tolerate target doses of 2 *<sup>×</sup>* <sup>10</sup><sup>7</sup> NK cells/kg without any serious side effect (78). This number of NK cells can be routinely obtained from a 1-day large-volume leukapheresis. However, it is possible that higher doses and/or multiple infusions of NK cells may be required for optimal clinical efficacy. Whether NK cells may exert their beneficial effects pre-emptively in patients with disease remission, or rather in the context of HSCT warrants particular attention (73). In this respect, KIR ligand mismatch is expected to mediate a more powerful effect in T-cell-depleted HSCT, since alloreactive T cells reportedly blunt NK reactivity

<sup>4</sup> http://www.cik-info.org/index.php?kat=tableclinicaltrials

<sup>5</sup> http://oldweb.cqvip.com

(37, 174). Finally, recent discoveries on NK-mediated allorecognition will guide the choice of the optimal NK-cell donor in order to maximally exploit the anti-tumor effect (33). The isolation of single KIR<sup>+</sup> NK cells under GMP conditions is feasible and yields clinically applicable numbers of alloreactive NK cells (65).

# **References**


Current clinical evidence also points to CIK cells as a potentially useful immunotherapy approach for cancer patients. Similar to other forms of immunotherapy, the infusion of CIK cells may be more efficacious at disease stages where the tumor burden is relatively low or in an adjuvant setting, rather than for advanced disease (175).


expanded CD8+ NK-T cells in the treatment of lymphoma. *Biol Blood Marrow Transplant* (2001) **7**:532–42. doi:10.1016/S1083-8791(01)70014-6


but does not mediate tumor regression. *Clin Cancer Res* (2011) **17**:6287–97. doi:10.1158/1078-0432.CCR-11-1347


optimized manufacturing protocol. *Front Oncol* (2013) **3**:118. doi:10.3389/ fonc.2013.00118


patients relapsing after allogeneic stem cell transplantation: a phase I study. *Haematologica* (2007) **92**:952–9. doi:10.3324/haematol.11132


175. Rutella S, Locatelli F. Is there a role for cytokine-induced killer cells in cancer immunotherapy? *Immunotherapy* (2012) **4**:867–9. doi:10.2217/imt.12.89

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

# The Application of Natural Killer Cell Immunotherapy for the Treatment of Cancer

#### *Katayoun Rezvani1 \* and Rayne H. Rouce2,3*

*1Department of Stem Cell Transplantation, Division of Cancer Medicine, The University of Texas MD Anderson Cancer Center, Houston, TX, USA, 2Department of Pediatrics, Texas Children's Cancer and Hematology Centers, Baylor College of Medicine, Houston, TX, USA, 3Center for Cell and Gene Therapy, Baylor College of Medicine Houston Methodist Hospital and Texas Children's Hospital, Houston, TX, USA*

*Edited by:* 

*Rafael Solana, University of Cordoba, Spain*

#### *Reviewed by:*

*Kamalakannan Rajasekaran, Blood Research Institute, USA Alejandra Pera, Instituto Maimonides de Investigaciones Biomédicas de Córdoba, Spain*

> *\*Correspondence: Katayoun Rezvani krezvani@mdanderson.org*

#### *Specialty section:*

*This article was submitted to NK Cell Biology, a section of the journal Frontiers in Immunology*

> *Received: 24 September 2015 Accepted: 29 October 2015 Published: 17 November 2015*

#### *Citation:*

*Rezvani K and Rouce RH (2015) The Application of Natural Killer Cell Immunotherapy for the Treatment of Cancer. Front. Immunol. 6:578. doi: 10.3389/fimmu.2015.00578*

Natural killer (NK) cells are essential components of the innate immune system and play a critical role in host immunity against cancer. Recent progress in our understanding of NK cell immunobiology has paved the way for novel NK cell-based therapeutic strategies for the treatment of cancer. In this review, we will focus on recent advances in the field of NK cell immunotherapy, including augmentation of antibody-dependent cellular cytotoxicity, manipulation of receptor-mediated activation, and adoptive immunotherapy with *ex vivo*-expanded, chimeric antigen receptor (CAR)-engineered, or engager-modified NK cells. In contrast to T lymphocytes, donor NK cells do not attack non-hematopoietic tissues, suggesting that an NK-mediated antitumor effect can be achieved in the absence of graft-vs.-host disease. Despite reports of clinical efficacy, a number of factors limit the application of NK cell immunotherapy for the treatment of cancer, such as the failure of infused NK cells to expand and persist *in vivo*. Therefore, efforts to enhance the therapeutic benefit of NK cell-based immunotherapy by developing strategies to manipulate the NK cell product, host factors, and tumor targets are the subject of intense research. In the preclinical setting, genetic engineering of NK cells to express CARs to redirect their antitumor specificity has shown significant promise. Given the short lifespan and potent cytolytic function of mature NK cells, they are attractive candidate effector cells to express CARs for adoptive immunotherapies. Another innovative approach to redirect NK cytotoxicity towards tumor cells is to create either bispecific or trispecific antibodies, thus augmenting cytotoxicity against tumor-associated antigens. These are exciting times for the study of NK cells; with recent advances in the field of NK cell biology and translational research, it is likely that NK cell immunotherapy will move to the forefront of cancer immunotherapy over the next few years.

Keywords: natural killer cells, adoptive immunotherapy, CAR NK cells, ADCC, anti-KIR antibody, NK-92, transplantation

# INTRODUCTION

Natural killer (NK) cell-mediated cytotoxicity contributes to the innate immune response against various malignancies, including leukemia (1, 2). The antitumor effect of NK cells is a subject of intense investigation in the field of cancer immunotherapy. In this review, we will focus on recent advances in NK cell immunotherapy, including augmentation of antibody-dependent cytotoxicity, manipulation of receptor-mediated activation, and adoptive immunotherapy with *ex vivo*-expanded, chimeric antigen receptor (CAR)-engineered, or engager-modified NK cells.

# BIOLOGY OF NK CELLS RELEVANT TO ADOPTIVE IMMUNOTHERAPY

Natural killer cells are characterized by the lack of CD3/TCR molecules and by the expression of CD16 and CD56 surface antigens. Around 90% of circulating NK cells are CD56dim, characterized by their distinct ability to mediate cytotoxicity in response to target cell stimulation (3, 4). This subset includes the alloreactive NK cells that play a central role in targeting leukemia cells in the setting of allogeneic hematopoietic stem cell transplant (HSCT) (5). The remaining NK cells, predominantly housed in lymphoid organs, are CD56bright, and although less mature ("unlicensed") (3, 6, 7), they have a greater capability to secrete and respond to cytokines (8, 9). CD56bright and CD56dim NK cells are also distinguished by their differential expression of FcγRIII (CD16), an integral determinant of NK-mediated antibody-dependent cellular cytotoxicity (ADCC), with CD56dim NK cells expressing high levels of the receptor, while CD56bright NK cells are CD16 dim or negative (6). In contrast to T and B lymphocytes, NK cells do not express rearranged, antigenspecific receptors; rather, NK effector function is dictated by the integration of signals received through germ-line-encoded receptors that can recognize ligands on their cellular targets. Functionally, NK cell receptors are classified as activating or inhibitory. NK cell function, including cytotoxicity and cytokine release, is governed by a balance between signals received from inhibitory receptors, notably the killer Ig-like receptors (KIRs) and the heterodimeric C-type lectin receptor (NKG2A), and activating receptors, in particular the natural cytotoxicity receptors (NCRs) NKp46, NKp30, NKp44, and the C-type lectin-like activating immunoreceptor NKG2D (9).

The inhibitory KIRs (iKIRs) with known HLA ligands include KIR2DL2 and KIR2DL3, which recognize the HLA-C group 1-related alleles characterized by an asparagine residue at position 80 of the α-1 helix (HLA-CAsn80); KIR2DL1, which recognizes the HLA-C group 2-related alleles characterized by a lysine residue at position 80 (HLA-CLys80); and KIR3DL1, which recognizes the HLA-Bw4 alleles (9, 10). NK cells also express several activating receptors that are potentially specific for self-molecules. KIR2DS1 has been shown to interact with group 2 HLA-C molecules (HLA-C2), while KIR2DS2 was recently shown to recognize HLA-A\*11 (10, 11). Hence, these receptors require mechanisms to prevent inadvertent activation against normal tissues, processes referred to as "tolerance to self." Engagement of iKIR receptors by HLA class I leads to signals that block NK-cell triggering during effector responses. These receptors explain the "missing self " hypothesis, which postulates that NK cells survey tissues for normal levels of the ubiquitously expressed MHC class I molecules (12, 13). Upon cellular transformation or viral infection, surface MHC class I expression on the cell surface is often reduced or lost to evade recognition by antitumor T cells. When a mature NK cell encounters transformed cells lacking MHC class I, their inhibitory receptors are not engaged, and the unsuppressed activating signals, in turn, can trigger cytokine secretion and targeted attack of the virusinfected or transformed cell (13, 14). In parallel, cellular stress and DNA damage (occurring in cells during viral or malignant transformation) results in upregulation of "stress ligands" that can be recognized by activating NK receptors. Thus, human tumor cells that have lost self-MHC class I expression or bear "altered-self " stress-inducible proteins are ideal targets for NK recognition and killing (14–16). NK cells directly kill tumor cells through several mechanisms, including release of cytoplasmic granules containing perforin and granzyme (16–18), expression of tumor necrosis factor (TNF) family members, such as FasL or TNF-related apoptosis-inducing ligand (TRAIL), which induce tumor cell apoptosis by interacting with their respective receptors Fas and TRAIL receptor (TRAILR) (16–19) as well as ADCC (9).

# INTERACTION BETWEEN NATURAL KILLER CELLS AND OTHER IMMUNE SUBSETS

Increasing understanding of NK cell biology and their interaction with other cells of the immune system has led to several novel immunotherapeutic approaches as discussed in this review. NK cells produce cytokines that can exert regulatory control of downstream adaptive immune responses by influencing the magnitude of T cell responses, specifically T helper-1 (TH1) function (20). NK cell function, in turn, is regulated by cytokines, such as IL-2, IL-15, IL-12, and IL-18 (21), as well as by interactions with other cell types, such as dendritic cells, macrophages, and mesenchymal stromal cells (10, 22, 23). IL-15 has emerged as a pivotal cytokine required for NK cell development and maintenance. Whereas mice deficient in IL-2 (historically the cytokine of choice to expand and activate NK cells) have normal NK cells, IL-15-deficient mice lack NK cells (24).

Several cytokines are also known to inhibit NK cell activation and function, thus playing a crucial role in tumor escape from NK immune surveillance. Recently, considerable attention has been paid to the inhibitory effects of transforming growth factor-beta (TGF-β*)* and IL-10 on NK cell cytotoxicity (12, 25, 26). Several groups have shown that secretion of TGF-β by tumor cells results in downregulation of activating receptors, such as NKp30 and NKG2D, with resultant NK dysfunction (25, 26). Similarly, IL-10 production by acute myeloid leukemia (AML) blasts induces upregulation of NKG2A with significant impairment in NK function (3).

# MODULATION OF ANTIBODY-DEPENDENT CELLULAR CYTOTOXICITY

The CD56dim subset of NK cells expresses the Fcγ receptor CD16, through which NK cells mount ADCC, providing opportunities for its modulation to augment NK effector function (27, 28). In fact, a number of clinically approved therapeutic antibodies targeting tumor-associated antigens (such as rituximab or cetuximab) function at least partially through triggering NK cell-mediated ADCC. Several studies using mouse tumor models have established that efficient antibody–Fc receptor (FcR) interactions are essential for the efficacy of monoclonal antibody (mAb) therapy, a mainstay of cancer therapy (28, 29). Based on this premise, Romain et al. successfully engineered the Fc region of the IgG mAb, HuM195 targeting the AML leukemia antigen CD33, by introducing the triple mutation S293D/ A330L/I332E (DLE). Using timelapse imaging microscopy in nanowell grids (TIMING, a method of analyzing kinetics of thousands of NK cells and mAb-coated targets), they demonstrated that the DLE-HuM195 antibody increased both the quality and quantity of NK cell-mediated ADCC by recruiting NK cells to participate in cytotoxicity via CD16-mediated signaling. NK cells encountering DLE-HuM195-coated targets induced rapid target cell apoptosis by promoting conjugation to multiple target cells (leading to increased "serial killing" of targets), thus inducing apoptosis in twice the number of targets as the wild-type mAb (27).

Additional approaches under investigation to enhance NK cell-mediated ADCC include antibody engineering and therapeutic combination of antibodies predicted to have synergistic activity. For example, mogamulizumab (an anti-CCR4 mAb recently approved in Japan) is defucosylated to increase binding by FcγRIIIA and thereby enhances ADCC. Mogamulizumab successfully induced ADCC activity against CCR4-positive cell lines and inhibited the growth of EBV-positive NK-cell lymphomas in a murine xenograft model (30). These findings suggest that mogamulizumab may be a therapeutic option against EBV-associated T and NK-lymphoproliferative diseases (30). Obinutuzumab (GA101) is a novel type II glycoengineered mAb against CD20 with increased FcγRIII binding and ADCC activity. In contrast to rituximab, GA101 induces activation of NK cells irrespective of their inhibitory KIR expression, and its activity is not negatively affected by KIR/HLA interactions (31). These data show that modification of the Fc fragment to enhance NK-mediated ADCC can be an effective strategy to augment the efficacy of therapeutic mAbs (31).

Although enhanced NK-mediated ADCC occurs in the presence of certain mAbs, in the case of non-engineered mAbs (such as rituximab), this NK-mediated cytotoxicity is typically still under the jurisdiction of KIR-mediated inhibition. However, ADCC responses can be potentiated *in vitro* in the presence of antibodies that block NK cell inhibitory receptor interaction with MHC class I ligands (32). These include the use of anti-KIR Abs to block the interaction of iKIRs with their cognate HLA class I ligands. To exploit this pathway pharmacologically, a fully humanized anti-KIR mAb 1-7F9 (IPH2101) (33) with the ability to block KIR2DL1/L2/L3 and KIR2DS1/S2 was generated. *In vitro*, anti-KIR mAbs can augment NK cell-mediated lysis of HLA-C-expressing tumor cells, including autologous AML blasts and autologous CD138<sup>+</sup> multiple myeloma (MM) cells (34). Additionally, in a dose-escalation phase 1 clinical trial in elderly patients with AML, 1-7F9 mAb was reported to be safe and could block KIRs for prolonged periods (35). A recombinant version of this mAb with a stabilized hinge (lirilumab) was recently developed. Lirilumab is a fully humanized IgG4 anti-KIR2DL1, -L2, -L3, -S1, and -S2 mAb. The iKIRs targeted by lirilumab collectively recognize virtually all HLA-C alleles, and the blockade of the three KIR2DLs allows targeting of every patient without the need for prior HLA or KIR typing (33, 34). Furthermore, the combination of an anti-KIR mAb with the immunomodulatory drug lenalidomide was shown to potentiate ADCC and is being tested in a phase 1 clinical trial in patients with MM [NCT01217203 (35)]. A potential concern is related to how inhibitory KIR blockade may impact on the ability of NK cells to discriminate self, healthy cells from abnormal virally infected or cancerous cells. Preliminary *in vitro* data suggest that Ab blockade of iKIRs will preferentially augment the ADCC response, without increasing cytotoxicity against self healthy cells (32). It is reassuring that in the IPH2101 phase 1 studies, no alterations in the expression of major inhibitory or activating NK receptors or frequencies of circulating peripheral lymphocytes were reported, indicating that the Ab does not induce clinically significant targeting of normal cells by NK cells (35). Lin et al. recently reported on the application of an agonistic NK cell-targeted mAb to augment ADCC (36). Following FcR triggering during ADCC, expression of the activation marker CD137 is increased. Agonistic antibodies targeting CD137 have been reported to augment NK-cell function, including degranulation, secretion of IFN-γ, and antitumor cytotoxicity in *in vitro* and *in vivo* preclinical models of tumor (36–39). The combination of the agonistic anti-CD137 antibody with rituximab is currently being evaluated in a phase 1 trial in patients with lymphoma [NCT01307267 (35–37)].

Other factors, such as specific CD16 polymorphisms and NKG2D engagement, can also influence ADCC, with certain polymorphisms (such as FcγRIIIa-V158F polymorphism) resulting in a stronger IgG binding (40). These findings are clinically relevant, as supported by the observation that patients with non-Hodgkin lymphoma (NHL) with the FcγRIIIa-V158F polymorphism experienced improved clinical response to rituximab (41, 42). In summary, several antibody combinations designed to boost ADCC have shown promising results in preclinical and early clinical trials, thus warranting further study of this strategy to enhance NK cell activity against tumor cells.

# ADOPTIVE TRANSFER OF AUTOLOGOUS NK CELLS

The early studies of adoptive NK cell therapy focused on enhancing the antitumor activity of endogenous NK cells (43). Initial trials of adoptive NK therapy in the autologous setting involved using CD56 beads to select NK cells from a leukapheresis product and subsequently infusing the bead-selected autologous NK cells into patients (43, 44). Infusions were followed by administration of systemic cytokines (most commonly IL-2) to provide additional *in vivo* stimulation and support their expansion. This strategy met with limited success due to a combination of factors (44). Although cytokine stimulation promoted NK cell activation and resulted in greater cytotoxicity against malignant targets *in vitro*, only limited *in vivo* antitumor activity was observed (43–45). Similar findings were observed when autologous NK cells and systemic IL-2 were given as consolidation treatment to patients with lymphoma who underwent autologous BMT (46). The poor clinical outcomes observed with adoptive transfer of *ex vivo* activated autologous NK cells followed by systemic IL-2 were attributed to three factors: (1) development of severe life-threatening side effects, such as vascular leak syndrome as a result of IL-2 therapy; (2) IL-2-induced expansion of regulatory T cells known to directly inhibit NK cell function and induce activation-induced cell death (47–49); and (3) lack of antitumor effect related to the inhibition of autologous NK cells by self-HLA molecules. Strategies to overcome this autologous "checkpoint," thus redirecting autologous NK cells to target and kill leukemic blasts are the subject of intense investigation (33–35). These include the use of anti-KIR Abs (such as the aforementioned lirilumab) to block the interaction of inhibitory receptors on the surface of NK cells with their cognate HLA class I ligand.

# EXPLOITING THE ALLOREACTIVITY OF ALLOGENEIC NK CELLS – ADOPTIVE IMMUNOTHERAPY AND BEYOND

An alternative strategy is to use allogeneic instead of autologous NK cells, thus taking advantage of the inherent alloreactivity afforded by the "missing self " concept (13). Over the past decade, adoptive transfer of *ex vivo*-activated or -expanded allogeneic NK cells has emerged as a promising immunotherapeutic strategy for cancer (24, 50–52). Allogeneic NK cells are less likely to be subject to the inhibitory response resulting from NK cell recognition of self-MHC molecules as seen with autologous NK cells. A number of studies have shown that infusion of haploidentical NK cells to exploit KIR/HLA alloreactivity is safe and can mediate impressive clinical activity in some patients with AML (50–52). In fact, algorithms have been developed to ensure selection of stem cell donors with the greatest potential for NK cell alloreactivity for allogeneic HSCT (50).

Promising results in the HSCT setting suggest that the application of this strategy in the non-transplant setting may be a plausible option. Miller et al. were among the first to show that adoptive transfer of *ex vivo-*expanded haploidentical NK cells after lymphodepleting chemotherapy is safe, and can result in expansion of NK cells *in vivo* without inducing graft-vs.-host disease (GVHD) (50). In a phase I dose-escalation trial, 43 patients with either hematologic malignancies (poor prognosis AML or Hodgkin lymphoma) or solid tumor (metastatic melanoma or renal cell carcinoma) received up to 2 × 107 cells/kg of haploidentical NK cells following either low intensity [low-dose cyclophosphamide (Cy) and methylprednisolone or fludarabine (Flu)] or high intensity regimens (Hi-Cy/Flu). All patients received subcutaneous IL-2 after NK cell infusion. Whereas adoptively infused NK cells persisted only transiently following low intensity regimens, AML patients who received the more intense Hi-Cy/Flu regimen had a marked rise in endogenous IL-15 associated with expansion of donor NK cells and induction of complete remission (CR) in five of 19 very high-risk patients. The superior NK expansion observed after high-dose compared to low-dose chemotherapy was attributed to a combination of factors including prevention of host T cell-mediated rejection and higher levels of cytokines, such as IL-15. These findings provided the first evidence that haploidentical NK cells are safe and can persist and expand *in vivo*, supporting the proof of concept that NK cells may be applied for the treatment of selected malignancies either alone or as an adjunct to HSCT (50).

Another pivotal pilot study, the NKAML trial (Pilot Study of Haploidentical NK Transplantation for AML), reported that infusion of KIR-HLA-mismatched donor NK cells can reduce the risk of relapse in childhood AML (51). Ten pediatric patients with favorable or intermediate risk AML in first CR were enrolled following completion of 4–5 cycles of chemotherapy. All patients received a low-dose conditioning regimen consisting of Cy/Flu prior to infusion of NK cells (median, 29 × 106 /kg NK cells) from a haploidentical donor, followed by six doses of IL-2. NK infusions were well tolerated with limited non-hematologic toxicity. All patients had transient engraftment of NK cells for a median of 10 days (range 2–189 days) with significant expansion of KIRmismatched NK cells. With a median follow-up of 964 days, all patients remained in remission, suggesting that donor-recipient HLA-mismatched NK cells may reduce the risk of relapse in childhood AML (51).

Other strategies currently under investigation include the infusion of KIR-ligand-mismatched haploidentical NK cells as part of the pre-HSCT conditioning regimen (NCT00402558), and NK cell infusion to prevent relapse or as therapy for minimal residual disease in patients after haploidentical HSCT (NCT01386619).

# ADOPTIVE NK CELL THERAPY IN SOLID MALIGNANCIES

Natural killer cell-based immunotherapies are also a promising therapeutic option for solid tumors. A number of studies have shown that the presence of intratumoral NK cells correlates with delayed tumor progression and improved outcomes (53–55). However, the successful application of NK cell-based therapies in the solid tumor setting poses a special challenge. In addition to the immune evasion strategies common to hematologic malignancies, such as secretion of immunosuppressive cytokines and downregulation of activating ligands (55–57), additional challenges specific to solid tumors exist; NK cells must not only traffic to sites of disease, but also penetrate the tumor capsule in order to exert their effector function. Furthermore, tumor targets must be inherently susceptible to NK-mediated cytotoxicity (58). Several groups have focused on strategies to alter the tumor microenvironment by targeting myeloid-derived suppressor cells or regulatory T cells (Treg) rather than the tumors themselves (58, 59). In fact, the prospect of combining NK cell-based immunotherapy with approaches to target the immunosuppressive tumor microenvironment or immune checkpoints, such as KIR blockade, is especially relevant to the treatment of solid tumors (55, 58). Several early phase clinical trials have demonstrated the feasibility of adoptive therapy with autologous or allogeneic *ex vivo* activated/expanded NK cells in patients with refractory solid malignancies [NCT01875601 (60)]; however, outside of the post-HSCT setting (namely in neuroblastoma), limited data on the clinical efficacy of NK cells in eradicating solid tumors exist. Currently, several trials are actively recruiting patients with refractory solid tumors for adoptive NK therapy (including NCT01807468, NCT02130869, and NCT0210089).

# THE IDEAL MANUFACTURING STRATEGY FOR *EX VIVO* ACTIVATION OF NK CELLS

Recent approaches to adoptive NK therapy focused on infusion of NK cells that have undergone a process of *ex vivo* cytokine activation and expansion (61). A number of cytokines (IL-2, IL-12, IL-15, IL-18, IL-21, and type I IFNs) have been studied to activate and expand NK cells *ex vivo* (62–65). The most extensively studied cytokine is IL-2 (62, 63). This is not surprising, considering IL-2 was the only cytokine available in clinical grade until recently. Nevertheless, NK cells expanded in the presence of IL-12, IL-15, and IL-18, either alone or in combination, have shown remarkable activity against tumor targets in experimental models and offer an attractive strategy for clinical expansion of NK cells (64, 65). IL-15, in particular, is appealing as it does not stimulate Tregs (65). IL-15 has been tested in preclinical models with promising results; however, very high doses were necessary to observe any meaningful *in vivo* antitumor effects, and toxicity of systemic cytokine administration and cytokine-induced NK-cell apoptosis remained major issues (65). Recently, Miller et al. compared the persistence and *in vivo* efficacy of adoptively infused freshly activated NK cells (FA-NK) and *ex vivo*-expanded NK cells (Ex-NK) in a xenotransplantation model. They showed that *in vivo* NK cell persistence is cytokine dependent, with IL-15 being superior to IL-2. They also reported that cryopreservation of FA-NK or Ex-NK was detrimental to NK cell function, and that culture conditions influence homing, persistence, and expansion of NK cells *in vivo* (66).

Although the results from the abovementioned trials proved that transient persistence of adoptively transferred NK cells obtained via apheresis is feasible and safe, the requirement of a willing, available donor precludes the widespread applicability of this approach. Hence, more recent efforts have focused on optimizing methods for *ex vivo* expansion of NK cells from peripheral blood mononuclear cells (PBMCs) collected by a simple blood draw, with a goal of producing large quantities of purified, functionally active NK cells for clinical use. These expansion strategies include the use of "feeder cells," such as monocytes in the form of irradiated PBMCs, EBV-transformed lymphoblastoid cell lines (EBV-LCLs) or gene-modified, irradiated K562 cells expressing membrane-bound IL-15 or IL-21 and 41BB ligand for costimulation (61, 66–69) in gas-permeable large-scale expansion flasks. These techniques have dramatically increased the yield and activation status of NK cells, potentially overcoming the need for leukapheresis. Because the feeder cells used in these manufacture methods are lethally irradiated prior to use in culture (leaving the remaining feeder cells to be lysed by the expanding NK cells), the risk of infusing viable feeder cells is negligible. However, a number of safeguards have also been incorporated that include monitoring the growth rate of feeder cells and testing for the presence of viable feeder cells at the end of the culture period. Clinical products are, therefore, only released if no viable gene-modified K562 cells or transformed LCLs are present, with strict cutoff values for contaminating B cells and monocytes at the end of the culture period as well (67).

Although these expansion methods can produce large numbers of functionally active NK cells, concomitant expansion of competing cells with immunosuppressive properties, such as Tregs remains problematic. Early studies reported that NK cell infusions from haploidentical donors are able to induce remissions in some patients with AML, but not others (50–52). Several groups, therefore, set out to explore factors that may contribute to the failure of NK expansion *in vivo*. Bachanova explored the effect of competition between Tregs and NK cells in 57 patients with refractory AML who received lymphodepleting chemotherapy followed by NK cell infusion and IL-2 administration [NCT00274846 and NCT01106950 (70)]. Fifteen patients also received the IL-2-diphtheria toxin fusion protein (IL2DT) to deplete Tregs prior to NK cell infusion. IL2DT treatment was associated with increased donor NK cell persistence and improved CR and disease-free survival at 6 months (33 vs. 5% in patients not receiving IL2DT; *P* < 0.01). In the IL2DT cohort, NK cell expansion correlated with higher post-chemotherapy serum IL-15 levels (*P* = 0.002) and effective peripheral blood (PB) Treg depletion (<5%) at day 7 (*P* < 0.01). This study shed light on the importance of optimizing the cytokine milieu to facilitate the *in vivo* expansion of adoptively transferred NK cells and identifying ways to abrogate the immunosuppressive elements, such as regulatory T cells.

Although these data are encouraging, adoptive transfer of NK cells under good manufacturing practices (GMP) requires significant infrastructure and specialized processing equipment, thus limiting the availability and scalability of these NK cell therapies to a few specialized institutions (61). Nonetheless, the feasibility of centralized processing and safe delivery of *ex vivo*-manufactured NK cells for infusion at remote clinics have been demonstrated, suggesting that the practice might become more widespread as procedures are optimized (71). For example, in order to improve access to *ex vivo* activated NK cells and ease the burden associated with producing cellular products at individual treatment centers, the National Heart, Lung, and Blood Institute (NHLBI, Bethesda, MD, USA) sponsored the Production Assistance for Cellular Therapies (PACT) program. Using this approach, activated NK cells have been sent to other centers for infusion into patients (72, 73).

Since the initial reports of successful adoptive transfer of NK cells (50–52), many groups continue to perform extensive preclinical exploration of the ideal manufacturing strategy for *ex vivo* activation and expansion of NK cells. Several expansion methods optimized in the preclinical setting have been successfully scaled up for the clinic (61, 67–70). In addition to the six clinical trials of adoptive NK cell therapy for leukemia that have reported their data (48–52, 70), there are currently 12 active clinical trials enrolling patients with hematologic malignancies for NK cell adoptive therapy, a number that is steadily rising.

# ALTERNATIVE SOURCES OF NK CELLS FOR ADOPTIVE TRANSFER – IS CORD BLOOD THE ANSWER?

Although the majority of clinical studies of NK cell immunotherapy have used PB NK cells, several alternative sources of NK cells exist. These include bone marrow, human embryonic stem cells (hESCs), induced pluripotent stem cells [iPCSs (74, 75)], and umbilical cord blood (CB). While the generation of NK cells from hESCs or iPCS has been largely experimental to date, clinicalgrade generation and expansion of NK cells from CB-derived CD34<sup>+</sup> cells has been successfully achieved (76).

Umbilical CB as a source for NK cells lends additional clinical advantages. CB contains a high percentage of NK cells (77, 78) and serves as an immediate "off-the-shelf " source of NK cells, with less stringent requirements for HLA matching, and lower risk of causing GVHD following infusion due to the naivety of the cord T cell repertoire (77, 78). Although no direct comparison of PB- and CB-derived NKs has been performed in the clinical setting, *in vitro* studies have identified a number of differences between CB and PB NK cells. CB NK cells form weaker conjugates with target cells due to the lower membrane expression of adhesion molecules on their surface (79, 80). CB NK cells also express higher levels of lectin-like inhibitory receptors (CD94/ NKG2A) and lower levels of KIRs, indicating an immature phenotype (81). CB NK cells are similarly sensitive to cytokines for *in vivo* expansion and persistence (82). However, it appears that the requirements for *in vitro* expansion of CB NK cells may be different to those required for PB NKs. CB NKs are less responsive to IL-2 stimulation, which may be related to the lower expression of IL-2Rα and reduced activation of the STAT5 signaling pathway as compared with PB NK cells (83). The combination of IL-15 and IL-18, however, can induce significant proliferation and cytokine production by CB NK cells, while the killing capacity of CB NK cells is significantly enhanced after stimulation with IL-15 (83). As with PB-derived NK cells, T-cell contamination is a concern, but can be ameliorated by CD3 depletion. T-cell contamination should be limited to <1–5 × 105 /kg (61) to minimize the risk of GVHD. In addition, CD56+ selection reduces B-cell contamination to <1%, which minimizes passenger B lymphocyte-mediated complications, such as EBV-related post-transplant lymphoproliferative disorder (PTLD) and acute hemolytic anemia.

More recently, efforts have focused on optimizing the largescale expansion of purified CB-derived NK cells. Shah et al. were the first to describe a strategy for expanding NK cells from cryopreserved CB units in which they employed K562-based artificial antigen-presenting cells (aAPCs) expressing membranebound IL-21 (clone 9.mbIL21) (77, 84). The clone 9.mbIL21 cell line is GMP-grade and expresses membrane-bound IL-21, 4-1BB ligand, CD64 (FcγRI), and CD86. After only 14 days of culture in a gas-permeable culture system, mean-fold expansion of CB-NK cells was 1848-fold from fresh and 2389-fold from cryopreserved CB with >95% purity for CD56<sup>+</sup>CD3<sup>−</sup> NK cells. aAPC-expanded CB-NK cells displayed a phenotype similar to that of expanded PB-NK cells and maintained strong expression of the transcription factors eomesodermin and T-bet. Furthermore, CB-NK cells formed functional immune synapses and efficiently killed various MM targets *in vitro*. Finally, aAPC-expanded CB-NK cells showed significant *in vivo* activity against MM in a xenogenic mouse model. These findings highlight a clinically applicable strategy for the generation of highly functional CB-NK cells using an aAPC platform, which can be potentially extended to other hematologic malignancies and solid tumors (77). A number of phase I/II clinical trials are underway to test the feasibility and efficacy of CB-NK cell adoptive therapy in patients with hematologic malignancies (NCT01619761, NCT01729091 NCT02280525, NCT01914263, and NCT00412360) (summarized in **Table 1**).

# HUMAN NK CELL LINES AS A SOURCE OF NK IMMUNOTHERAPY

The adoptive transfer of NK cell lines has several theoretical advantages over the use of patient- or donor-derived NK cells. These are primarily related to the lack of expression of iKIRs, presumed lack of immunogenicity, ease of expansion and availability as an "off-the-shelf " product (85). Several human NK cell lines, such as NK-92 and KHYG-1, have been documented to exert antitumor activity in both preclinical and clinical settings (86–88). NK-92, the most extensively characterized NK-cell line, was established in 1994 from the PB of a male Caucasian patient with NHL. NK-92 cells are IL-2-dependent, harbor a CD2<sup>+</sup>CD56<sup>+</sup>CD57<sup>+</sup> phenotype and exert potent *in vitro* cytotoxicity (86). Infusion of up to 1010 cells/m2 NK-92 cells into patients with advanced lung cancer and other advanced malignancies was well tolerated and the cells persisted for a minimum of 48 h with encouraging clinical responses (86, 88–91). However, potential limitations of using NK cell lines, such as NK-92 cells, include the requirement for irradiation to reduce the risk of engrafting cells with potential *in vivo* tumorigenicity, and the need for pre-infusion conditioning to avoid host rejection. Furthermore, infusion of allogeneic NK cell lines may induce T and B cell alloimmune responses, limiting their *in vivo* persistence and precluding multiple infusions. A number of studies are testing NK-92 cells (Neukoplast®) in patients with solid tumors, such as Merkel cell cancer and renal cell carcinoma, as well as in hematological malignancies (85).

While results from clinical studies of NK cell adoptive therapy are encouraging (48–52, 70), significant gaps remain in our understanding of the optimal conditions for NK cell infusion. Based on the pioneering work from Rosenberg et al. demonstrating the importance of lymphodepletion to support the expansion of tumor-infiltrating T cells (92) and given its emergence as a key determinant of efficacy with CAR therapy, several groups are actively investigating the ideal preparative regimen to promote the expansion and persistence of adoptively infused NK cells (53, 69, 70, 75). Available data support the use of high-dose Cy/ Flu regimen as the frontrunner, considering it is reasonably well tolerated and shown to support the *in vivo* expansion of NK cells

#### TABLE 1 | Published results of NK adoptive immunotherapy trials in hematologic malignancies.


*HR, high risk; haplo, haploidentical; LR, low risk; IR, intermediate risk; Hi-Cy/Flu, high-dose cyclophosphamide and fludarabine; CR, complete remission; ALL, acute lymphoblastic leukemia; AML, acute myeloid leukemia; MDS, myelodysplastic syndromes; SCT, hematopoietic stem cell transplant; HL, Hodgkin lymphoma; NHL, non-Hodgkin lymphoma; pedi, pediatric; CRi, complete remission with incomplete platelet recovery; IL2DT, IL-2-diphtheria fusion protein.*

*aNK cells infused outside of the setting of hematopoietic cell transplantation.*

(51, 70). IL-15 is an ideal candidate cytokine for the expansion of

, especially since it does not promote expansion

of regulatory T cells (66), which have been shown to suppress NK cell effector function in IL-2-based trials (69, 70). In a recent

carcinoma, rhIL-15 was shown to activate NK cells, monocytes, phase 1 study in patients with metastatic melanoma or renal cell

, and CD8 T cells (93). However, as an intravenous bolus dose,

rhIL-15 proved too difficult to administer because of significant

93). Based on these promising data, alternative

NK cells

γδ

clinical toxicities (

CH

the specificity of T cells against leukemia (

an anti-CD19-BB-

with acute lymphoblastic leukemia (

the risk of GVHD. T cell-depleted allogeneic NK cells, by con activated T cells from an allogeneic source are likely to increase -

malignancies treated with haploidentical NK cells (

for patients with cancer.

and, provide an excellent source of off-the-shelf cellular therapy cells provide attractive candidate effector cells to express CARs their shorter lifespan and potent cytolytic function, mature NK

The feasibility of genetically engineering NK cells to express CARs has been shown in the preclinical setting (97, 98). Primary

engineered to express CARs against a number of targets includ human NK cells, as well as NK-92 cells, have been successfully ing CD19, CD20, CD244, and HER2 (97). CAR-transduced

NK cells mediate efficient

*in vitro* and

targets (97, 98) although to date, no clinical data of CAR NK cell

expression of a receptor containing CD3

molecules (anti-CD19-BB-

ζ) in human NK cells after mRNA

reported adequate transfection efficiency 24 h after electropora electroporation using a clinical-grade electroporator. The authors -

tion, with median anti-CD19-BB-

anti-CD19-BB-

purified and 61.3% in expanded NK cells. NK cells expressing ζ secreted IFN-γ in response to CD19-positive

transfect expanded NK cells, achieving excellent receptor expres retroviral transduction. A large-scale protocol was developed to after mRNA transfection were comparable to those achieved by target cells. Interestingly, the levels of CAR expression in NK cells -

sion and considerable cytotoxicity of CAR-transduced NK cells

in xenograft models of B-cell leukemia (

(

cytotoxic against a variety of hematologic and solid malignancies NKG2D ligands on the surface of tumor cells, rendering NK more strategy is the development of CAR-modified NK cells that target

100). NK cells have also been successfully engineered to target

99). Another interesting

ζ expression of 40.3% in freshly

therapy have been reported. Shimasaki et al. recently tested the ζ and 4-1BB signaling

*in vivo* killing of tumor

50

–52). Given

murine models, as well as in patients with leukemia and solid trast, should not cause GVHD, as predicted by observations in

95, 96); however, infusions of

94). Recently, use of

ζ receptor transduced into autologous or allo

geneic T cells produced dramatic clinical responses in patients


Chimeric antigen receptors have been used extensively to redirect

clinical trial for AML (NCT01385423).

dosing strategies are being investigated, including continuous IMERIC ANTIGEN RECEPTOR-MODIFIED NK CELLS

infusion of donor NK cells are currently being tested in a phase I intravenous infusions. To this effect, systemic IL-15 along with

*in vivo*

(51, 70). IL-15 is an ideal candidate cytokine for the expansion of NK cells *in vivo*, especially since it does not promote expansion of regulatory T cells (66), which have been shown to suppress NK cell effector function in IL-2-based trials (69, 70). In a recent phase 1 study in patients with metastatic melanoma or renal cell carcinoma, rhIL-15 was shown to activate NK cells, monocytes, γδ, and CD8 T cells (93). However, as an intravenous bolus dose, rhIL-15 proved too difficult to administer because of significant clinical toxicities (93). Based on these promising data, alternative dosing strategies are being investigated, including continuous intravenous infusions. To this effect, systemic IL-15 along with infusion of donor NK cells are currently being tested in a phase I clinical trial for AML (NCT01385423).

# CHIMERIC ANTIGEN RECEPTOR-MODIFIED NK CELLS

Chimeric antigen receptors have been used extensively to redirect the specificity of T cells against leukemia (94). Recently, use of an anti-CD19-BB-ζ receptor transduced into autologous or allogeneic T cells produced dramatic clinical responses in patients with acute lymphoblastic leukemia (95, 96); however, infusions of activated T cells from an allogeneic source are likely to increase the risk of GVHD. T cell-depleted allogeneic NK cells, by contrast, should not cause GVHD, as predicted by observations in murine models, as well as in patients with leukemia and solid malignancies treated with haploidentical NK cells (50–52). Given their shorter lifespan and potent cytolytic function, mature NK cells provide attractive candidate effector cells to express CARs and, provide an excellent source of off-the-shelf cellular therapy for patients with cancer.

The feasibility of genetically engineering NK cells to express CARs has been shown in the preclinical setting (97, 98). Primary human NK cells, as well as NK-92 cells, have been successfully engineered to express CARs against a number of targets including CD19, CD20, CD244, and HER2 (97). CAR-transduced NK cells mediate efficient *in vitro* and *in vivo* killing of tumor targets (97, 98) although to date, no clinical data of CAR NK cell therapy have been reported. Shimasaki et al. recently tested the expression of a receptor containing CD3ζ and 4-1BB signaling molecules (anti-CD19-BB-ζ) in human NK cells after mRNA electroporation using a clinical-grade electroporator. The authors reported adequate transfection efficiency 24 h after electroporation, with median anti-CD19-BB-ζ expression of 40.3% in freshly purified and 61.3% in expanded NK cells. NK cells expressing anti-CD19-BB-ζ secreted IFN-γ in response to CD19-positive target cells. Interestingly, the levels of CAR expression in NK cells after mRNA transfection were comparable to those achieved by retroviral transduction. A large-scale protocol was developed to transfect expanded NK cells, achieving excellent receptor expression and considerable cytotoxicity of CAR-transduced NK cells in xenograft models of B-cell leukemia (99). Another interesting strategy is the development of CAR-modified NK cells that target NKG2D ligands on the surface of tumor cells, rendering NK more cytotoxic against a variety of hematologic and solid malignancies (100). NK cells have also been successfully engineered to target antigens on a variety of solid tumors. For example, an NK-CAR targeting the ganglioside GD2 (present on neuroblastoma cells) has been tested in preclinical studies (101, 102). GD2 is also expressed on breast cancer stem cells, thus raising the potential for its widespread use as a target for immunotherapy (103). Additional antigens targeted by NK CARs include HER2 (overexpressed in a number of solid tumors), CD138, and CS1 (overexpressed in MM) (104, 105).

Although these data support the use of CAR engineering to redirect the specificity of NK cells to augment their cytotoxicity, a number of challenges remain. These include the relative difficulty in expressing exogenous genes in primary human NK cells and the need to expand NK cells in culture to achieve adequate numbers for clinical studies of immunotherapy. To counteract this difficulty, some groups have expressed CARs in the human NK-like cell line NK-92, in an attempt to engineer a uniformly cytolytic effector cell population (106). As previously mentioned, NK-92 cells can be easily expanded in culture and their safety has been shown in phase I clinical trials in human subjects. Thus, CAR-expressing NK-92 cells may offer a practical source of cells for NK cell-based immunotherapeutic trials. In order to prevent the risk of engrafting cells with potential *in vivo* tumorigenicity, however, NK-92 cells must be irradiated prior to infusion, which may in turn significantly impact their *in vivo* persistence and long-term antitumor efficacy. Although limited *in vivo* persistence could prove beneficial once the alloreactive NK cells have eradicated the tumors, a number of studies of adoptive therapy with NK cells and CAR-modified T cells have reported the importance of cell persistence in inducing long-term antitumor response (50, 95, 96).

As with CAR-modified T cell therapy, a number of variables can affect the activation, antitumor efficacy, and persistence of CAR-NK cells. Second and third generation CAR constructs incorporating additional costimulatory domains (e.g., CD28, OX-40, or 4-1BB) have been shown to enhance both *in vitro* and *in vivo* activation, and the persistence of CAR T cells. Further studies exploring the optimal vector, construct and transduction method are necessary to identify the "perfect NK CAR."

# SAFETY CONCERNS RELATED TO ADOPTIVE TRANSFER OF CAR-MODIFIED NK CELLS

When considering the use of CAR-modified effector cells, one must take into account their safety profile. Many of the same concerns raised with CAR-modified T cells may be relevant to CAR-NK cells. These include on-target/off-tumor effects, GVHD, cytokine release syndrome, tumor lysis syndrome, and toxicity to normal tissues due to limited selectivity of the target antigen (107–109). Thus, the necessity of equipping CAR-modified NK cells with a "safety switch" or suicide gene is an important question to explore. While mature allogeneic CAR-engineered NK cells are expected to be short lived, data on the persistence of more immature NK cells, such as those derived from CB, are lacking. Interestingly, a recent study reported that IL15/4-1BBLactivated NK cells infused early after T-depleted allogeneic stem

TABL

Reference Burns et al. (43)

aMiller

et al. (50)

aRubnitz and

Fresh-activated NK (FA-NK)

*N* = 10

LR/IR AML (pedi)

Haplo-related donors

Haplo-related HSCT donors

Pre-SCT conditioning

regimen (Bu/Flu/thymo)

(from CD34+ fraction)

Hi-Cy/Flu

Inaba (51)

Yoon et al. (49)

aCurti and

CD56+ selected NKs

AML-CR and

Haplo-related donors

Hi-Cy/Flu

5 × 106 cells/kg followed by

6/13 (46%) remain in CR

IL-2 × 6 doses

relapsed (adult)

*N* = 13

Ruggeri (52)

Stern et al. (48)

NKs

Klingemann and

Apheresis-mobilized CD56

HL, NHL, MM

Haplo donors

None

1 × 105–2 × 107 cells/kg

remission

1° endpoint safety/feasibility; 7/13 in

*N* = 13

Grodman (71)

aBachanova (70)

Choi et al. (116)

*protein.*

*aNK cells infused outside of the setting of hematopoietic cell transplantation.*

Apheresis-mobilized, *ex vivo*

*N* = 41

Haplo donors

Bu/Flu/ATG *HR, high risk; haplo, haploidentical; LR, low risk; IR, intermediate risk; Hi-Cy/Flu, high-dose cyclophosphamide and fludarabine; CR, complete remission; ALL, acute lymphoblastic leukemia; AML, acute myeloid leukemia; MDS,* 

*myelodysplastic syndromes; SCT, hematopoietic stem cell transplant; HL, Hodgkin lymphoma; NHL, non-Hodgkin lymphoma; pedi, pediatric; CRi, complete remission with incomplete platelet recovery; IL2DT, IL-2-diphtheria fusion* 

Median 1 × 108 cells/kg

Reduced leukemia progression 46 vs. 74%

IL-15/21 induced NK cells

Treg depletion

NK infusion w/IL-2 ± IL2DT

AML

Haplo donors

Hi-Cy/Flu

Mean 2.6 ± 1.5 × 107 cells/

IL-2 alone: 9/42 (21%) CR/CRi

IL2DT: 8/15 CR/CRi (53%)

kg

*N* *N*

= 15 (+IL2DT)

= 42 (IL-2 alone)

selection

1–3 doses positively selected

ALL, AML (adult and

Haplo donors

Pre-SCT conditioning

Median 1.2 × 107 cells/kg

4/16 (25%) alive

regimen

pedi)

*N* = 15

IL-7/15/21 *ex vivo* cultured NKs

 HR ALL/AML/MDS

(adults)

*N* = 14

*Ex vivo* IL-2 activated

Relapsed lymphoma

Autologous

None

autologous NKs or bolus IL-2

IL-2 activated NK cells

HR AML (adults)

Haplo-related donors

Hi-Cy/Flu

1 × 106–2 × 107 cells/kg

5/19 (26%) CR

followed by 14 days IL-2

Median 29 × 106 cells/kg

10/10 (100%) CR at 964 days

followed by IL-2 × 6 doses

Median 2.2 × 106 cells/kg

1° endpoint safety/feasibility; (no toxicity;

low-grade GVHD); 4/14 (28%) alive and well

*N* = 19

*N* = 29

Approach

E 1 | Published results of NK adoptive immunotherapy trials in hematologic malignancies.

Disease

NK source of cells

Conditioning regimen

Dose of cells 4 × 107–8 × 107 cells/kg

1° endpoint safety/feasibility; no change in

outcome compared to historical controls

Outcome

cell transplantation in patients not receiving immunosuppressive prophylaxis could contribute to acute GVHD (110). To this effect, the insertion of a suicide safety switch system, as employed with CAR-modified T cells (111, 112), can provide an efficient means for depletion of these cells if needed. Inducible suicide systems have safely and effectively eradicated GVHD in patients receiving adoptively transferred T cells without causing deleterious effects (112). However, these systems have not been extensively studied in NK cells, and in the absence of clinical data on the *in vivo* persistence of CAR-modified NK cells, the necessity of a suicide switch in this setting remains unknown.

Despite the growing wealth of preclinical experience with CAR-engineered NK cells, to date, only two clinical studies (both targeting CD19+ malignancies using a retroviral transduced anti-CD19-BB-ζ NK-CAR) have obtained regulatory approval: one is a recently completed pediatric study at St. Jude Children's Research Hospital, where haploidentical NK cells modified with anti-CD19- BB-ζ CAR were infused into patients with B-ALL (ClinicalTrials. gov.NCT00995137) and the other is an ongoing study at the National University Hospital in Singapore (ClinicalTrials.gov. NCT01974479) using IL-2-activated haploidentical CAR-modified NK cells in pediatric and adult patients with refractory B-ALL (99). The results of these studies have not been reported to date.

# BISPECIFIC AND TRISPECIFIC ENGAGERS

An innovative immunoglobulin-based strategy to redirect NK cytotoxicity towards tumor cells is to create either bispecific or trispecific antibodies (BiKE, TriKE) (113). BiKEs are constructed by joining a single-chain Fv against CD16 and a single-chain Fv against a tumor-associated antigen (BiKE), or two tumorassociated antigens (TriKE). Gleason et al. showed that bispecific (bscFv) CD16/CD19 and trispecific (tscFv) CD16/CD19/CD22 engagers directly trigger NK cell activation through CD16, significantly increasing NK cell cytolytic activity and cytokine production against various CD19-expressing B cell lines. The same group also developed and tested a CD16 × 33 BiKE in refractory

AML and demonstrated that the potent killing by NK cells could overcome the inhibitory effect of KIR signaling (113, 114).

Notably, activated NK cells lose CD16 (FcRγIII) and CD62L through a metalloprotease called ADAM17, which is expressed on NK cells, which may in turn impact on the efficacy of Fc-mediated cytotoxicity (115). Romee et al. recently showed that selective inhibition of ADAM17 enhances CD16-mediated NK cell function by preserving CD16 on the NK cell surface, thus enhancing ADCC (115). Additionally, Fc-induced production of cytokines by NK cells exposed to rituximab-coated B cell targets can be further enhanced by ADAM17 inhibition. These findings support a role for targeting ADAM17 to prevent CD16 shedding and to improve the efficacy of therapeutic mAbs. The same group subsequently discovered that ADAM17 inhibition enhances CD16 × 33 BiKE responses against primary AML targets (114).

## NK CELLS – WHAT DOES THE FUTURE HOLD?

Recent advances in the understanding of NK cell immunobiology have paved the way for novel and innovative anti-cancer therapies. Here, we have discussed a representation of these novel immunotherapeutic strategies to potentiate NK cell function and enhance antitumor activity including ADCC-inducing mAbs, *ex vivo* activated or genetically modified NK cells and bi- or trispecific engagers (**Figure 1**).

Although experience has shown that adoptive immunotherapy with allogeneic NK cells may be more efficacious than

## REFERENCES


with autologous NK cells, to date, their long-term antitumor benefits have been modest (3). Expansion and persistence of NK cells following infusion appear to be the main determinants of clinical response (50–52, 70), thus underscoring the importance of identifying ways to enhance their persistence and antitumor activity. It is likely that the combination of high-dose lymphodepleting chemotherapy with additional modifications (such as Treg depletion, *in vivo* administration of cytokines, such as IL-15 or enhancement of CD16-mediated antigen targeting) may maximize NK persistence and efficacy.

In addition, the possibility of third-party "off-the-shelf " products with partially HLA-matched NK cells from CB, thirdparty donors, or NK cell lines allow the advantage of unlimited sources of cells to improve the practicality of cell therapy. With increasing focus on genetically modifying NK cells to redirect their specificity or engager-modified NK cells, it is likely that NK cells will move to the forefront of cancer therapy over the next few years.

# AUTHOR CONTRIBUTIONS

RR and KR wrote and edited the manuscript.

### FUNDING

The authors' research efforts are funded in part by the Lymphoma Research Foundation (RHR), Lymphoma SPORE (RHR), MDACC Leukemia SPORE Grant CA (KR), and SINF (KR).


monocyte-derived dendritic cells. *Blood* (2008) **112**(5):1776–83. doi:10.1182/ blood-2008-02-135871


and responsiveness to IL-2 but not cytolytic activity. *J Immunol* (2002) **169**(8):4230–6. doi:10.4049/jimmunol.169.8.4230


antigen receptors compares favorably with antibody-dependent cellular cytotoxicity. *OncoImmunology* (2013) **2**(10):e26527. doi:10.4161/onci.26527


**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 Rezvani and Rouce. 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.*

# **Utilizing chimeric antigen receptors to direct natural killer cell activity**

*David L. Hermanson 1,2 and Dan S. Kaufman 1,2 \**

*<sup>1</sup> Department of Medicine, Division of Hematology, Oncology, and Transplantation, University of Minnesota, Minneapolis, MN, USA <sup>2</sup> Stem Cell Institute, University of Minnesota, Minneapolis, MN, USA*

Natural killer (NK) cells represent an attractive lymphocyte population for cancer immunotherapy due to their ability to lyse tumor targets without prior sensitization and without need for human leukocyte antigens-matching. Chimeric antigen receptors (CARs) are able to enhance lymphocyte targeting and activation toward diverse malignancies. CARs consist of an external recognition domain (typically a small chain variable fragment) directed at a specific tumor antigen that is linked with one or more intracellular signaling domains that mediate lymphocyte activation. Most CAR studies have focused on their expression in T cells. However, use of CARs in NK cells is starting to gain traction because they provide a method to redirect these cells more specifically to target refractory cancers. CAR-mediated anti-tumor activity has been demonstrated using NK cell lines, as well as NK cells isolated from peripheral blood, and NK cells produced from human pluripotent stem cells. This review will outline the CAR constructs that have been reported in NK cells with a focus on comparing the use of different signaling domains in combination with other co-activating domains.

#### *Edited by:*

*Francisco Borrego, Cruces University Hospital, Spain*

#### *Reviewed by:*

*Evelyn Ullrich, Goethe University Frankfurt, Germany Cristina Eguizabal, Basque Center for Transfusion and Human Tissues, Spain*

#### *\*Correspondence:*

*Dan S. Kaufman, Department of Medicine, Stem Cell Institute, University of Minnesota, 420 Delaware Street SE, MMC 480, Minneapolis, MN 55455, USA kaufm020@umn.edu*

#### *Specialty section:*

*This article was submitted to NK Cell Biology, a section of the journal Frontiers in Immunology*

> *Received: 23 March 2015 Accepted: 08 April 2015 Published: 28 April 2015*

#### *Citation:*

*Hermanson DL and Kaufman DS (2015) Utilizing chimeric antigen receptors to direct natural killer cell activity. Front. Immunol. 6:195. doi: 10.3389/fimmu.2015.00195* **Keywords: chimeric antigen receptors, natural killer cells, cancer immunotherapy, NK-92 cells, induced pluripotent stem cells**

# **Introduction**

Natural killer (NK) cells are an important component of the innate immune system due to their ability to lyse infected or malignant cells without prior sensitization and without human leukocyte antigens (HLA)-restriction (1). They play an important role in immune surveillance and early control of many malignancies. NK cells recognize infected or transformed cells through multiple cell surface receptors including NKG2D, CD16 [the receptor that mediates antibodydependent cellular cytotoxicity (ADCC)], and natural cytotoxicity receptors (NCRs) such as NKp44, NKp46, and NKp30 (2). These receptors activate signaling adapter proteins such as DAP10, DAP12, and CD3ζ, which contain immuno-tyrosine activation motifs (ITAMs) that initiate the release of cytolytic granules containing perforin and granzymes, as well as mediate production and release of cytokines and chemokines such as IFN-γ and TNF-α (3). Importantly, NK cellmediated cytotoxicity does not rely on the presentation of self HLA. Therefore, NK cells hold significant clinical interest as a cell-based therapy for cancer because of their ability to be used in an allogeneic setting and potentially provide an off-the-shelf cellular product. Clinical trials using NK cells obtained from haploidentical donors demonstrate long-term remissions in patients with refractory acute myelogenous leukemia (4). Trials against solid tumors such as breast cancer and ovarian cancer have also demonstrated efficacy (5). NK cell lines (NK-92 cells) (6) and NK cells derived from umbilical cord blood (7) have also been tested in clinical trials (NCT01729091, NCT02280525).

Chimeric antigen receptors (CARs) are engineered proteins designed to activate lymphocytes, particularly T cells, upon target recognition. CARs contain a single-chain variable fragment (scFv) fused to a variety of possible intracellular signaling domain(s). The scFv is designed to target antigens either overexpressed or unique to tumor cells. The signaling domain initially tested was the ζ chain of the T cell receptor complex CD3 in first generation CARs (8). Second generation CARs employ co-activating proteins such as CD28, CD137 (4-1BB), or CD134 (OX40) in combination with CD3ζ to increase T cell activation and proliferation (9, 10). Finally, CAR constructs incorporate multiple co-activation domains and CD3ζ in the third generation CARs (11). The clinical success of anti-CD19 CAR-expressing T cells for treatment of Bcell malignancies has fueled the design and evaluation of CARs for T cell therapy toward other antigens and malignancies (12). While development of T cell-CAR-based therapies seems to be revolutionizing tumor immunotherapy, one major obstacle with this approach is the need to collect and utilize autologous cells. A second concern with the use of T cells is their long-term persistence, resulting in chronic on-target-off-tumor effects such as B-cell aplasia with the anti-CD19 CARs being used currently in clinical trials.

Natural killer cells provide an alternative to the use of T cells for adoptive immunotherapy since they do not require HLA matching, so can be used as allogeneic effector cells (13). Clinical trials of adoptively transferred allogeneic NK cells demonstrate these cells can survive in patients for several weeks to months (4, 13). Additionally, expression of CARs in NK cells may allow these cells to more effectively kill solid tumors that are often resistant to NK cell-mediated activity compared to hematologic malignancies (especially acute myelogenous leukemia) that are typically more NK cell-sensitive (4, 5). As such, CAR-expressing NK cells have gained significant interest to provide a targeted, allogeneic, "universal" cell population for treatment of refractory malignancies. This review will focus on the CAR constructs and activating domains that have been reported to be used in NK cell lines (such as NK-92 cells) and peripheral blood (PB) NK cells. Additional work using NK cells produced from human pluripotent stem cells is also discussed.

## **Chimeric Antigen Receptors Used in NK Cell Lines**

Natural killer cell lines have been utilized to evaluate CARs targeting several different antigens. By far, the most commonly studied NK cell line has been the NK-92 cell line, which has been previously (6) and is currently being used in clinical trials (NCT00900809 and NCT00990717). Other NK cell lines include NKG, YT, NK-YS, HANK-1, YTS cells, and NKL cells (14). Working with NK cell lines has several advantages such as providing a more homogeneous cell population compared to NK cells isolated from PB. In addition, the NK-92 cells have been well-defined and there is no need to perform any isolation of the NK cells from donors. However, NK cell lines also have distinct disadvantages. NK-92 cells lack the expression of several typical NK cell activating receptors such as CD16, NKp44, and NKp46 (15, 16). Also, NK-92 cells are tumor cell lines with multiple cytogenetic

#### **TABLE 1 | CAR constructs utilized in NK cell lines (NK-92)**.


abnormalities (17) and are latently infected with Epstein–Barr virus (18). Therefore, for safety purposes, these cells must be irradiated prior to infusion.

The majority of studies to express CARs in NK-92 cells have used first generation CAR constructs that contain CD3ζ as their sole signaling domain. The scFvs of these CARs have targeted CD20 (19–21), CD19 (20–22), ErbB2 (HER2) (23–25), GD2 (26), and CD138 (27) (**Table 1**). In addition to directly targeting cell surface proteins, CARs can also recognize HLA-peptide complexes such as HLA-A2 expressing the melanoma-associated gp100 peptide (28). CARs directed toward CD19 and CD20 are designed to target B-cell malignancies and have also been studied extensively in T cells (29). The only other difference in the anti-CD19 or anti-CD20 CAR constructs used in NK-92 cells is the transmembrane region. One study used the CD3ζ transmembrane sequence (19) while another used the CD8 transmembrane sequence (22). However, without a direct comparison it is unknown if one construct is superior. Another study used an HLA-A2 transmembrane region coupled to a CD3ζ signaling domain (28), suggesting the transmembrane region may be easily altered without impacting CAR expression and functionality. Interestingly, comparison of CAR transfected NK-92 cells with ADCC function using NK-92 cells engineered to express CD16 found that the anti-CD20 CAR engineered cells lysed primary CLL cells more effectively than NK-92 cells acting through ADCC using rituximab (21). This study suggests that even first generation CARs may be an improvement over ADCC-mediated anti-tumor activity by NK-92 cells. It is important to note that NK-92 cells require transfection of CD16 in order to perform ADCC. This leaves open the possibility that PB-NK cells may still be better equipped to perform ADCC better than NK-92 cells.

Solid tumor antigens can also be targeted by first generation CAR constructs expressed in NK-92 cells. An anti-ErbB2 CAR construct against HER2-positive breast, ovarian, and squamous cell carcinoma cell lines mediated improved killing ability of NK-92 cells (23). Additionally, this study showed a reduction in tumor growth using ErbB2-expressing NIH 3.3 cells mixed with NK-92s in a subcutaneous mouse model (23). An anti-GD2 CAR using just the CD3ζ transmembrane and signaling domains was able to target primary glioblastoma cells as well as GD2-positive melanoma and breast carcinomas (26). NK-92 cells can also be targeted against multiple myeloma (MM) using an anti-CD138 CAR with only CD3ζ as a signaling domain (27). Notably, mice bearing a subcutaneous tumor treated with CAR-expressing NK-92 cells survived significantly longer than NK-92 cell alone in a CD138-positive tumor model; whereas, when a CD138-negative MM tumor was used no difference was detected (27). These data clearly demonstrate that first generation CARs are an effective means to induce target cell lysis in NK-92 cells both *in vitro* and in mouse models; however, many of the tumor models are subcutaneous, which may fail to properly recapitulate the complete tumor environment or NK cell trafficking issues.

Second generation CARs expressing a second signaling domain in conjunction with CD3ζ vastly improves the overall activity CAR-expressing T cells (9). This has generated interest in using second generation CARs in NK cells. Similar to first generation CARs, several different scFvs have been used with second generation CARs including EpCAM for multiple carcinomas including breast and ovarian cancer (30), an HLA-A2 EBNA3C complex for Epstein–Barr virus (31), CS1 for MM (32), and ErbB2 for HER2 positive cancers (24, 25). The most common second generation CAR utilized in NK-92 cells pairs the CD28 intracellular domain with CD3ζ (**Table 1**). Notably, NK cells do not naturally express CD28 (35); therefore, the effect that this domain has in NK cells is unclear. Other second generation CARs combine CD137 (4-1BB) intracellular domain with CD3ζ. Similar to first generation CARs, all of the constructs lead to antigen specific killing of target cells, displaying the diverse set of tumor antigens CARs can target. Comparison of an ErbB2 scFv fused with CD3ζ alone, CD28/CD3ζ, or CD137/CD3ζ tested head-tohead against breast cancer cells found that both of the second generation constructs improved killing compared to the first generation CARs (25). Specifically, the CD28/CD3ζ had 65% target lysis in ErbB2-positive MDA-MB453 while the CD137/CD3ζ lysed 62% and CD3ζ alone killed 51% (25). Another modification in their construct design was the modification of a cysteine to a serine in the CD8α signaling peptide used, which the authors suggest improves surface expression of the CAR in NK-92 cells. Finally, CD28/CD3ζ was compared to DAP12 alone using an anti-PSCA CAR in YTS NK cells for prostate cancer (34). In 293T cell lines engineered to express PSCA, a significant increase in cell killing was observed with the DAP12 containing CAR compared to the CD28/CD3ζ CAR, suggesting DAP12 may provide a better signaling domain than CD3ζ (34).

# **Chimeric Antigen Receptor use in Peripheral Blood NK Cells**

Chimeric antigen receptors have also been evaluated in PB-NK cells, which can be isolated from donors through simple blood draws or by apheresis if larger numbers of cells are needed. In contrast to NK-92 cells, activated PB-NK cells express a wider range of activating receptors, such as CD16, NKp44, and NKp46 as well as KIRs, which play an important role in NK cell licensing (36). In addition, PB-NK cells can be given without irradiating the cells so have the ability to expand *in vivo*, which has been correlated with effectiveness in trials involving AML (4). A greater variety of CAR constructs have been used and directly compared in PB-NK cells targeting CD19 (37–39), CD20 (33), or ErbB2 (40, 41) (**Table 2**). Imai et al. describe the use of two first generation anti-CD19 CARs, CD3ζ, or DAP10 signaling, and one second generation CAR with CD137 and CD3ζ. Compared to CD3ζ, DAP10 induced a much weaker response in PB-NK cells, and addition of the CD137 domain to the CAR resulted in augmented killing of RS4:11 and 380 (ALL) cell lines (37). Another study compared CD3ζ or 2B4 alone, 2B4 combined with CD3ζ, and a CD137/CD3ζ anti-CD19 CAR and tested them against the leukemic cell line REH. *In vitro* studies demonstrated the 2B4 alone CAR was slightly less active compared to CD3ζ alone. Comparing the second generation CARs, both were significantly better than CD3ζ alone while similar activity was observed in the 2B4/CD3ζ and CD137/CD3ζ CARs (38). When this work was extended to an anti-GD2 CAR for neuroblastoma with just the CD3ζ and 2B4/CD3ζ endodomains, again the 2B4/CD3ζ was significantly better than CD3ζ alone (38). Another study compared CD3ζ alone with a CD28/CD3ζ CAR using ErbB2 as a target.

#### **TABLE 2 | CAR constructs utilized in PB-NK cells**.


While no direct lysis experiment was performed, similar levels of INF-γ production were observed in PB-NK cells engineered with just CD3ζ or CD28/CD3ζ (41). While different measures were used, the finding that CD28/CD3ζ does not improve activity in PB-NK cells whereas the same construct was found to be more active in NK-92 suggests there may be differences in CAR activation of PB-NK and NK-92 cells.

One unique approach to CAR creation was to use the ectodomain of NKG2D, an NK cell activation receptor, and link it directly to CD3ζ (42). This approach utilizes natural NKG2D ligands commonly overexpressed on malignant cells to active the CAR. Further, NKG2D associates with DAP10 providing a secondary signaling molecule. Indeed, co-expression of DAP10 with the NKG2D/CD3ζ CAR increased surface expression. This CAR was tested against multiple cell lines derived from several malignancies with the best responses demonstrated against ALL, osteosarcoma, prostate carcinoma, and rhabdomyosarcoma (42).

A third source of NK cells suitable for CAR expression are NK cells derived from human pluripotent stem cells – both induced pluripotent stem cells (iPSCs) or human embryonic stem cells (hESCs) (43–47). These NK cells display a similar phenotype to PB-NK cells (43, 44, 48), and hESC/iPSC-NK cells can be grown on a clinical scale (48). iPSC-derived NK cells engineered with a CD4/CD3ζ CAR are able to inhibit HIV replication (49). In these studies, the CAR was expressed in the iPSC cells, which were then differentiated into CAR-expressing iPSC-NK cells. The CD4/CD3ζ iPSC-NK cells were shown to suppress the *in vitro* replication of HIV, providing a platform from which to work for the further development of CAR positive iPSC-NK cells. iPSCderived NK cells combine the best of PB-NK and NK-92 cells since the cells express NKp44, NKp46, and KIRs, are a homogeneous population with no evidence of undifferentiated iPSCs or T cells in the expanded NK cell population. Additionally, CARs can be easily expressed in hESC and/or iPSC-derived NK cells using nonviral gene transfer methods (49, 50). This is in contrast to PB-NK cells that are much more challenging to achieve high levels of stable CAR expression.

# **Outlook**

As the interest in using CARs in not only T cells (10) but also in NK cells continues to grow, there are still a number of questions that remain to be answered. Perhaps most important is what CAR constructs mediate optimal anti-tumor (or anti-viral) activity. Limited studies in NK-92 cells and in PB-NK cells directly compare first and second generation CARs. Second generation CARs in PB-NK cells are generally more active than first generation CARs. Additionally, the use of CD3ζ seems better than DAP10 as the signaling domain (37, 38). In NK-92 cells, DAP12 outperformed a CD28/CD3ζ CAR, but it remains unclear if NK-92 cells provide a good model for how CARs may function in PB-NK cells or hESC/iPSC-derived NK cells. Since NK cells do not naturally express CD28 (35, 51), it is not clear if CD28 is functioning in CAR-expressing NK cells. Different CAR constructs may be required to provide optimal NK cell activation depending on the tumor type or target antigen. More direct comparisons using various intracellular signaling domains and scFvs are needed to best resolve these questions.

Additional research is also needed to determine whether use of an NK cell line (such as NK-92 cells), PB-NK cells, or iPSC-NK cells will provide the best overall benefit. Both 4-1BBL/IL-15 (52) and mbIL-21 (53) artificial antigen presenting cells (aAPCs) can be used to expand PB-NK or iPSC-NK cells (48). Therefore, production of enough NK cells from these sources for clinical use is not a problem. However, it remains to be determined if one aAPC leads to an improved population for adoptive transfer, and the methods to engineer PB-NK cells still need to be further improved. iPSC-NK cells represent an attractive population of cells for NK-CAR therapy because once engineered the iPSC line can be maintained indefinitely and provide an almost limitless supply of NK cells. In addition, careful monitoring of the insertion site of the CAR can be achieved. Finally, NK cell lines provide another alternative but in general express fewer natural NK cell receptors and must be irradiated prior to infusion, which limits *in vivo* expansion and persistence of NK cells.

The method for CAR incorporation provides another important consideration. To get stable expression of CARs, retro- and lentivirus methods have dominated. However, following transduction of NK-92 cells a selection step is usually required to get a pure CAR-expressing population. In PB-NK cells, the efficiencies of gene transfer were at best 69% (37) and ranged as low as 13–24% (38) with most reporting around a 50% transduction efficiency. One way around this issue is the possibility of expressing the CAR in iPSCs and subsequent differentiation into mature NK cells (49), which is done via nucleofection with transposon and avoids the hazards of viral methods. Another consideration is whether the use of suicide systems, such as Cas9 or thymidine kinase (TK), will need to be put in place if unexpected toxicities arise despite the expectation that CAR-expressing NK cells will only circulate for a few weeks (14).

Despite the questions that remain, the ability to engineer NK cells with CARs holds great promise as a novel cellular immunotherapy against refractory malignancies and potentially chronic infectious diseases. The success of T cell-CARs in cases of ALL and CLL has revolutionized the prospects for cell-based immunotherapy. CAR-NK cells can build upon this success to provide important benefits as CAR-based therapy expands. Notably, NK cells can provide a homogenous, offthe-shelf, standardized product that can be used in as an allogeneic product to treat patients. Therefore, this process does not need to be done on a patient-specific basis, as with current T cell-CAR-based therapies. The ability to more potently direct NK cell-mediated cytotoxicity against refractory tumors through the expression of CARs can continue to revolutionize cancer treatment.

## **Author Contributions**

DH performed the review of the literature and wrote the manuscript. DK wrote and edited the manuscript.

# **Acknowledgments**

DH Ph.D., would like to acknowledge the Cancer Research Institute for support through the Irvington Postdoctoral Fellowship award.

# **References**


antigen receptors compares favorably with antibody-dependent cellular cytotoxicity. *Oncoimmunology* (2013) **2**:e26527. doi:10.4161/onci.26527


positive carcinomas. *Proc Natl Acad Sci U S A* (2008) **105**:17481–6. doi:10.1073/ pnas.0804788105


**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 Hermanson and Kaufman. 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.*

# Genetic manipulation of NK cells for cancer immunotherapy: techniques and clinical implications

#### *Mattias Carlsten\* and Richard W. Childs*

*Hematology Branch, National Heart, Lung and Blood Institute, National Institutes of Health, Bethesda, MD, USA*

Given their rapid and efficient capacity to recognize and kill tumor cells, natural killer (NK) cells represent a unique immune cell to genetically reprogram in an effort to improve the outcome of cell-based cancer immunotherapy. However, technical and biological challenges associated with gene delivery into NK cells have significantly tempered this approach. Recent advances in viral transduction and electroporation have now allowed detailed characterization of genetically modified NK cells and provided a better understanding for how these cells can be utilized in the clinic to optimize their capacity to induce tumor regression *in vivo*. Improving NK cell persistence *in vivo* via autocrine IL-2 and IL-15 stimulation, enhancing tumor targeting by silencing inhibitory NK cell receptors such as NKG2A, and redirecting tumor killing via chimeric antigen receptors, all represent approaches that hold promise in preclinical studies. This review focuses on available methods for genetic reprograming of NK cells and the advantages and challenges associated with each method. It also gives an overview of strategies for genetic reprograming of NK cells that have been evaluated to date and an outlook on how these strategies may be best utilized in clinical protocols. With the recent advances in our understanding of the complex biological networks that regulate the ability of NK cells to target and kill tumors *in vivo*, we foresee genetic engineering as an obligatory pathway required to exploit the full potential of NK-cell based immunotherapy in the clinic.

#### Keywords: NK cells, genetic manipulation, viral transduction, electroporation, cancer immunotherapy

#### Introduction

Natural killer (NK) cells are immune cells primarily found in the blood, liver, spleen, bone marrow, and to a lesser extent, in lymph nodes (1). They were initially identified based on their ability to lyse tumor cells without a need for priming (2–5). NK cells are now known to play an important role in host immunity against both cancers and certain viral infections (6–8).

NK cells can mediate cytotoxicity via multiple distinct mechanisms. Degranulation is the most studied cytotoxicity pathway, where NK cells release cytotoxic granules upon contact with the target. Cytotoxicity via this pathway is dictated by a balance of signals from an array of germline encoded activation and inhibitory cell surface receptors. Most activation receptors need simultaneous co-stimulation by other activation receptors to trigger NK cell cytotoxicity (9). One exception from this rule is the Fc receptor CD16, which alone can trigger NK cell degranulation against antibody-coated target cells via antibody-dependent cellular cytotoxicity (ADCC) (9). Other routes by which NK cells can kill targets are the death receptor pathways TRAIL/TRAIL-R and

#### *Edited by:*

*Francisco Borrego, Cruces University Hospital, Spain*

#### *Reviewed by:*

*Claudia Rossig, University Children's Hospital Münster, Germany Antonio Di Stasi, University of Alabama at Birmingham, USA*

#### *\*Correspondence:*

 *Mattias Carlsten, Hematology Branch, National Heart, Lung, and Blood Institute, National Institutes of Health, 10 Center Drive (MSC 1230), Building 10-CRC/3E-5332, Bethesda, MD 20892, USA mattias.carlsten@nih.gov*

#### *Specialty section:*

*This article was submitted to NK Cell Biology, a section of the journal Frontiers in Immunology*

> *Received: 19 March 2015 Accepted: 13 May 2015 Published: 10 June 2015*

#### *Citation:*

*Carlsten M and Childs RW (2015) Genetic manipulation of NK cells for cancer immunotherapy: techniques and clinical implications. Front. Immunol. 6:266. doi: 10.3389/fimmu.2015.00266*

Fas/FasL. Instead of triggering release of cytotoxic granules, death receptor pathways prompt apoptosis via caspase activation in target cells.

More than a decade has passed since initial reports established the anti-tumor potential of NK cells in patients with cancer. These studies showed that haplo-identical donor NK cells could prevent relapse in acute myeloid leukemia (AML) following hematopoietic stem cell transplantation (HSCT) and that adoptively infused mature donor NK cells could induce remission in AML patients (6, 10). Despite this revelation, doubts remain about the true therapeutic potential of NK cells in cancer immunotherapy. In contrast to therapy utilizing T cells, enthusiasm for NK cell-based immunotherapy has been tempered by uncertainties about their *in vivo* persistence, and doubts regarding their ability to migrate to tumor tissues following adoptive infusions. Although recent data have shown CMV reactivation reduces the risk for AML relapse following HSCT (11) potentially caused by CMV-induced NK cells cross-reacting with AML cells, NK cells, unlike T-cells, lack antigen specificity, further tempering enthusiasm for their use as immune effectors in cellular therapy.

Genetic manipulation of NK cells to improve their persistence, cytotoxicity, tumor targeting capacity, and ability to home to disease sites *in vivo* holds potential to advance the efficacy of NK cell-based cancer immunotherapy. However, until relatively recently, the genetic manipulation of NK cells has proven to be challenging. Viral transduction, successfully used for T cells, has been associated with low levels of transgene expression and unfavorable effects on cell viability when used with NK cells. Recent optimization of viral transduction and the establishment of electroporation technologies for efficient gene transfection have revived the enthusiasm for studies evaluating genetic modification of NK cells. Investigators around the world are now exploring the potential of multiple different NK cell modalities to genetically reprogram with the overall aim of further improving upon their capacity to kill tumors in cancer patients. One example of how this technique can be utilized is to introduce genes into NK cells coding for gamma-cytokines (IL-2 and IL-15) to induce independence from the obligate need of exogenous cytokines for proper *in vivo* persistence and expansion post infusion. This and similar strategies may further improve the efficacy of NK cell-based immunotherapy, as tumor regression following adoptive NK cell infusions in AML patients has been reported to be dependent on their ability to expand *in vivo* (6), while being limited by regulatory T cells also mobilized following exogenous cytokine administration (12, 13). The introduction of chimeric antigen receptors (CARs) and the down-regulation of inhibitory NK cell receptors such as NKG2A are additional examples of specific genetic manipulations that can be utilized to improve the outcome of adoptive NK cell immunotherapy.

Given their rapid and efficient method of recognizing tumor cells, NK cells represent a unique immune cell to genetically reprogram in an effort to improve the outcome of cell-based cancer immunotherapy. This review focuses on methods for introducing transgenes into NK cells and the advantages and limitations of such strategies. It also gives an overview of strategies for genetic reprograming of NK cells that have been evaluated to date and an outlook on how these specific strategies may be best utilized in clinic to maximize the anti-tumor potential of NK-cell based immunotherapy.

# Methods and Challenges with Genetic Manipulation of NK Cells: Viral Transduction Versus Transfection

Genetic manipulation of T cells has successfully been used in both preclinical and clinical research (14). In contrast, studies on genetically engineered NK cells have historically been limited by poor efficacy of transgene delivery and substantial procedure-associated NK cell apoptosis. In this section, we discuss available approaches for gene delivery into NK cells, characterizing how each approach developed over time while highlighting the positive and negative aspects of each method (**Box 1**).

#### Viral Transduction

The reduced efficacy of viral transduction of NK cells compared to T cells may in part be related to the innate properties that characterize NK cells. Innate immune receptors, such as pattern recognition receptors that recognize foreign genomic material, are likely involved in triggering apoptosis of NK cells following viral transduction (15). Best results from studies of viral transduction of NK cells have been achieved using either NK cell lines or primary NK cells that have undergone expansion *ex vivo* (**Table 1**). In contrast, viral transduction of primary resting human NK cells typically




*a Only those studies that have reported transgene expression following genetic manipulation of NK cell are reported in this table. n.r., not reported.*

results in substantially lower transduction efficiencies. Most studies on viral transduction of NK cells have utilized retro- and lentiviral vectors. Although adenoviral- and vaccinia virus vectors have been utilized for transduction of NK cells, their use has been limited and they will not be discussed further in this review.

*Retroviral vectors* were the first viral vectors used to genetically modify NK cells. The first report on retroviral transduction of NK cells was published in the late 1990s and focused on genetic manipulation of the NK cell line NK-92 (16). This study reported a transduction efficacy of only 2–3%. Optimization of retroviral transduction approaches over the past decade has resulted in higher transduction efficiencies, especially when used with human NK cells that have undergone *ex vivo* expansion (**Table 1**). A recent report showed that retroviral transduction of *ex vivo* expanded NK cells with genes coding for either IL-15 or membrane bound IL-15 (mbIL-15) resulted in an average 69 and 71% transduction efficiency, respectively (25). Although retroviral transduction of NK cells has been reported to not alter the function, phenotype, and proliferative capacity of NK cells (20, 23), their viability following retroviral transduction has rarely been reported. A significant deleterious impact on the viability of primary NK cells undergoing retroviral transduction may preclude utilizing this approach in a clinical setting. Further, retroviral transduction also requires active cell division, impeding the use of this method with primary non-activated NK cells. This limitation is less important when retroviral transduction is utilized with NK cell lines such as NK-92 that have continuous and unlimited proliferation capacity. However, as discussed later in this review, it is important to note that this NK cell line does have phenotypic and function differences from primary human NK cells, which may have therapeutic implications for clinical therapy.

#### Lentiviral Vectors

More recently, studies evaluating transduction of NK cells using lentiviral vectors have been pursued. In contrast to transduction with retro- and adenoviral vectors, lentiviral vectors can incorporate transgenes into the genome of non-dividing cells. Further, lentiviral vectors allow for gene modification of NK cells without altering their phenotypic and functional properties as occurs following stimulation with i.e., cytokines. The first report on the successful use of lentiviral vectors for genetic modification of NK cells was performed in primary murine NK cells (42), with subsequent studies establishing that lentiviral transduction of human NK cells could also be achieved (**Table 1**). Although most studies have reported lentiviral transduction of NK cell lines with efficiencies of 15–40% (27, 28), the efficiency highly varies from only a few percent to nearly 100%, and in some cases, multiple rounds of transduction are required (26, 29). Recent data indicate that transduction efficiencies of primary human NK cells can be increased by drug-induced inhibition of intracellular innate immune receptors in NK cells (15). Unfortunately, and similar to studies utilizing retroviral transduction, the viability of NK cells after lentiviral transduction has rarely been reported. Using an optimized protocol, our lab has achieved a maximum transgene expression in up to 60% of *ex vivo* expanded NK cells 3 days after lentiviral transduction with GFP without incurring any deleterious effects on NK cell viability, phenotype, or function (Personal communication, R. Childs).

In summary, viral transduction of NK cells results in variable transduction efficacies and may require multiple rounds of transduction and/or post transduction cell enrichment to achieve acceptable transgene expression. Further, viral associated cell death and the need for post-transduction enrichment may compromise the clinical utility of this approach. Finally, although the risk may be low, the possibility of viral-induced insertional mutagenesis and immunogenicity (43, 44) occurring post transduction must be considered when utilizing this methodology in the clinic. Nevertheless, viral transduction of NK cells does achieve stable transgene expression which, depending on how the NK cell is being genetically modified, might be required to induce a durable and long-term clinical response.

#### Transfection

Compared to viral transduction, transfection of NK cells appears to be associated with lower degrees of apoptosis, less inter-individual and inter-experimental variability, with transgene delivery efficiency being completely independent of cellular division. In most cases, this approach results in a more rapid albeit transient expression of the transgene as compared to viral transduction where genes must first be incorporated into the cellular genome before expression can occur. Gene transfer using transfection can be achieved by either *electroporation* (including *nucleofection*) or *lipofection*. Since the latter has been used only in a few studies (45), this review will focus on strategies utilizing the electroporation approach.

Electroporation is a method where genetic material is delivered into cells following a short electric pulse that temporarily induces small pores in the cell membrane, allowing charged molecules such as DNA and RNA to move into the cell. This technology was first used with NK cell lines in the late 1990s (32, 36–38, 46, 47) and more recently has been used to genetically manipulate primary NK cells to express CARs (35, 39, 48) or cytokines for autocrine growth stimulation (49). With technological advances and the use of mRNA instead of cDNA, transfection efficiencies have increased dramatically, reaching up to 90% or more while having only a minimal deleterious effect on cell viability (**Table 1**). Remarkably, using mRNA electroporation, transfection efficiencies of 80–90% can be achieved in not only *ex vivo* expanded cells but also in primary resting (non-cytokine activated) human NK cells (40). Despite this remarkable advance, a detailed characterization on the effects of electroporation on the phenotype, function, and proliferative capacity of NK cells following electroporation has yet to be published.

As electroporation does not involve viral vectors, its use in the preclinical and clinical setting is associated with less regulatory issues. Also, as indicated above, electroporation most often leads to transient transgene expression, which may be viewed favorably from a safety perspective when new transgenes with unknown potential toxicities are being explored in early clinical trials. Regimens that use DNA electroporation technology have been employed to generate stable transgene expressing cells. Although the efficacy of this approach is typically lower than that achieved with viral transduction, it may be improved if combined with targeted integration techniques that avoid random integration in inactive heterochromatin regions. Such strategies also reduce the risk for off-target effects, including gene silencing due to random integration in active genes and integration in hot-spots that may trigger malignant transformation. With advantages in design of guiding RNAs and by having better on-target specificity compared to other gene editing technologies such as Zink-Finger nuclease (ZFN) and the transcription activator-like effector nucleases (TALEN) technologies, the recently developed clustered regularly interspaced short palindromic repeats (CRISPR) technique has rapidly become a popular tool for targeted gene integration (50). The CRISPR/ Cas9 system induces permanent modifications at specific sites of the genome via double-strand breaks (DSBs), and can be used to integrate new genes at specific sites via homology-directed recombination (50). Although only moderate degrees of genome integration are currently being achieved with this technique today, the CRISPR/ Cas9 system could be used to produce stably transduced NK cells by gene editing of primary NK cells prior to their *ex vivo* expansion.

# Gene Modification Strategies Aimed at Improving the Efficacy of NK Cell-Based Cancer Immunotherapy

With new advances in the field, genetic manipulation of NK cells has opened up possibilities to study many different pathways involved in NK cell tumor targeting and the ability to genetically modify NK cells to improve their tumor cytotoxicity. Here, we will discuss reported gene modification strategies that can improve *in vivo* persistence and expansion, tumor tissue migration, and the tumor targeting capacity of adoptively infused NK cells (**Figure 1**, **Table 2** and **Box 2**).

#### Strategies to Improve Persistence and Expansion of Infused NK Cells

*In vivo* persistence and expansion of infused NK cells have been shown critical for inducing tumor regression following adoptive NK cell infusion (6). Using retroviral transduction of the IL-2 gene into NK-92 cells, Nagashima et al. were able to propagate this NK cell line for up to 5 months *in vitro* without the addition of exogenous cytokines (16). Further, IL-2-expressing NK-92 cells where shown to also have enhanced tumor cytotoxicity compared to non-transduced parental NK-92 cells that were stimulated with exogenous IL-2. In line with these *in vitro* findings, these genetically modified cells showed improved *in vivo* persistence and anti-tumor responses when infused into tumor-bearing mice. Similar data with IL-2 gene delivery in expanded NK cells were reported by Konstantinidis et al. (51). As observed with IL-2 transduced NK-92 cells, retroviral transduction of *ex vivo* expanded NK cells with the mbIL15 gene also dramatically increased their survival *in vitro*; median cell recovery was 85% for mbIL-15 NK cells after 7 days in culture without IL-2, whereas mock-transduced NK cells were

FIGURE 1 | Schematic overview of how genetic manipulation can be can be used to improve the efficacy of NK cell-based cancer immunotherapy in the clinic. Genetic engineering of NK cells to promote persistence and expansion by autocrine cytokine stimulation, migration to the tumor tissue via introduction of receptors involved in cellular homing (i.e., chemokine receptors and adhesion molecules), as well as bolstering their anti-tumor cytotoxicity via introduction of CARs or activating NK cell receptors (aNKRs) or via silencing of inhibitory NK cell receptors (iNKRs), protection from suppressive cytokines in the tumor environment, and boosted function via autocrine cytokine stimulation.

TABLE 2 | Overview of strategies evaluated for improving the anti-tumor efficacy of primary human NK cells and NK cell lines *in vitro* and in preclinical animal models.


*RV, retroviral transduction; LV, lentiviral transduction; EP, electroporation; ADCC, antibody-dependent cellular cytotoxicity; HA-CD16, high-affinity CD16; DNT*β*RII, double negative TGF-*β *RII.*

# Box 2 | Examples of NK Cell Modalities to Gene Manipulate for Improved Clinical Efficacy.


essentially undetectable (25). Hence, the strategy of introducing genes coding for gamma-cytokines to improve *in vivo* NK cell persistence and expansion following infusion independent of exogenous cytokine administration appears promising.

#### Strategies to Enhance Migration of Infused NK Cells

Proper tumor tissue homing of infused NK cells is a prerequisite for their ability to induce tumor regression. However, studies characterizing the *in vivo* migration capacity of adoptively infused NK cells have been largely overlooked (60). Recent evidence suggests non-expanded and expanded NK cells have different migration patterns when infused into animal models (61). Moreover, using trogocytosis to transfer premade cell surface molecules from a feeder cell line to NK cells, Somanshi et al. have shown that migration of infused NK cells can be redirected by equipping them with the lymph node homing receptor CCR7 (62). Despite these data, no study has so far used gene modification techniques to actively direct infused NK cells to selected organs. Based on data from Somanshi et al., we have been able to use mRNA transfection to genetically engineer NK cells with the CCR7 receptor to improve their migration toward one of its ligands CCL19 (Carlsten M., Manuscript in preparation, April 2015). Other strategies may involve utilizing chemokine receptors, such as CXCR3 to improve NK cell migration to inflamed tissues, such as those infiltrated with metastatic tumors (63).

#### Strategies to Increase Tumor Cytotoxicity by Infused NK Cells

The majority of reports on expression of transgenes in NK cells have characterized the effects of CARs in NK cell lines, expanded NK cells, and primary non-expanded NK cells (**Table 2**). CARs are engineered receptors that have the extracellular specificity of an antibody combined with potent intracellular signaling adaptors such as CD3ζ, CD28 and/or 4-1BB. Importantly, these receptors do not require stimulation through co-receptors to trigger robust anti-tumor cytotoxicity. The recent breakthrough success of anti-CD19 CAR T cell therapy in patients with B cell malignancies has stimulated the research community to develop and investigate a wide array of CARs against multiple different epitopes expressed on numerous tumor types (64). Several of these CARs have been explored in NK cells (**Table 2**). CD19 and CD20 specific CARs against B cell malignancies (39–41, 53, 54), and CARs targeting CD33 on leukemia cells (38), CS1 and CD138 on myeloma cells (24, 48, 55), GD2 on neuroblastoma cells (23, 56), Her2/Neu and erbB2 on breast cancer cells (22, 35), carcinoembryonic antigen (CEA) on colon cancers (36), EpCAM on epithelial tumors (29), GPA7 on melanoma (59), NKG2D ligand on leukemia and solid tumors, and TRAIL-R1 on various tumor targets (58) have all been shown to have the capacity to redirect NK cell cytotoxicity against their target antigens. The majority of these studies have used viral vectors to transduce CARs into the NK cell, albeit electroporation has also been used in a few studies (**Table 2**).

Based on clinical data showing superior response rates in rituximab-treated lymphoma patients homozygous for the highaffinity CD16-158V polymorphism (HA-CD16) compared to those who carry the low-affinity CD16-158F (LA-CD16) polymorphism (65, 66), several groups have recently addressed whether introduction of the HA-CD16 gene into NK cells lacking this polymorphism can be used as a strategy to augment ADCC against tumors. This approach has appeal as only a minority of patients is homozygous for HA-CD16 (67). Moreover, in contrast to CAR NK cells, infusions of NK cells genetically modified to express HA-CD16 may be used to improve the outcome of virtually any malignancy for which there is an FDA approved IgG1 antibody, without the expectation for any severe off target side-effects. *In vitro* experiments conducted by Binyamin and colleagues showed significantly improved cytotoxicity against a rituximab-coated B lymphoma cell line following stable transduction of the CD16 negative NK-92 cell line with HA-CD16 compared to NK-92 cells were equipped with LA-CD16 (19). Recently, our group explored a similar approach, where *ex vivo* expanded NK cells from CD16-158F/F (LA-CD16) donors were found to have substantially augmented ADCC following electroporation with mRNA coding for the HA-CD16 (68). These data suggest the addition of the HA-CD16 gene to patient NK cells that already express endogenous CD16 can be used to augment their ability to induce ADCC, and that this approach could be used as a strategy to improve the efficacy of antibody-based therapies for cancer patients.

Introduction of genes that render NK cells insensitive to suppressive cytokines such as TGF-β, thereby preserving their cytotoxicity, has also been studied. Yang et al. generated an NK-92 cell line resistant to the suppressive effects of TGF-β by genetically modifying them to express the dominant negative mutant form of TGF-β type II receptor (DNTβRII) on their surface (34). Adoptive transfer of these TGF-β insensitive NK-92 cells in lung cancer-bearing mice was associated with increased levels of IFN-γ released from the infused cells and resulted in increased survival rates compared to mice treated with wild-type NK-92.

Genetic reprograming of NK cells may also be directed to achieve specific protein silencing with the aim of improving tumor targeting by circumventing NK cell inhibitory signals induced upon interaction with tumor cells. Initial studies have focused on the use of shRNA technology for this purpose. In this context, shRNAs expressed inside cells are processed by the Dicer endonuclease complex to generate double-stranded small interfering RNAs that prevent translation of their target mRNAs (69), shRNAs have been used successfully to knock-down expression of the HLA-E-binding inhibitory NK cell receptor NKG2A (31). Using an inducible vector in IL-2 activated NK cells, Figueiredo et al. observed a 40% increased killing capacity against the HLA-E expressing cell line K562 HLA-E. Using a similar approach with the NK cell line NKL, our group observed increased killing capacity of HLA-E expressing 721.221 cells *in vitro* and in a preclinical mouse model (70). Further details on protocols for shRNA-mediated protein silencing in NK cells can be found in Purdy et al. (71). To date, studies utilizing CRISPR, ZFN, or TALEN to genetically modify NK cells to silence their inhibitory receptors for the same purpose of enhancing the anti-tumor capacity of NK cells have not yet been reported.

In conclusion, an array of gene modification strategies for NK cells has now been reported. Several of them hold promise for improving clinical responses of NK cell-based cancer immunotherapy. However, to date, few have been translated into clinical studies. The following section will discuss how these strategies can be incorporated in clinical NK cell cancer immunotherapy.

# Considerations for the Development of Clinical Protocols using Genetically Engineered NK Cells

Challenges associated with genetic manipulation of NK cells have significantly delayed the debut of this strategy in clinical cancer therapy. While recently initiated trials (NCT00995137 and NCT01974479) exploiting the role for CAR19-expressing *ex vivo* expanded NK cells in patients with B cell malignancies will give us a first insight into the potential of this approach; further optimization of clinical compliant methods for genetic modifications of NK cells is needed to exploit the full clinical potential of this approach. Moreover, additional research on the multiple aspects of NK cell tumor targeting that could be modified with this technique is needed. Although clinical responses following infusion of NK cells may be further improved by simply augmenting their tumor targeting capacity, studies evaluating the potential of this technology to improve the persistence of infused cells as well as avenues to promote proper NK cell migration and homing to the tumor tissue are also warranted (**Figure 1**).

Genetic engineering of NK cells to make them cytokine independent and thereby improve persistence, while boosting their cytotoxic capacity, may be one avenue to further explore. The advantage with this approach would be that exogenous cytokines would be unnecessary following NK cell infusion, which may reduce the risk of mobilizing regulatory T cells that directly suppress NK cell cytotoxicity (13). Challenges with taking this approach to a clinical context include the risk of inducing a cytokine release syndrome due to massive and unregulated NK cell proliferation. This approach also comes with the potential risk of inducing malignant transformation of the NK cells due to permanent autocrine growth stimulation, as have been observed for IL-2 engineered T cells (72). However, such scenarios may be avoided if genes coding for IL-2 or IL-15 are only temporarily introduced via mRNA electroporation of NK cells. Should stable transgene expression be required to induce proper tumor regression, an alternative strategy to prevent runaway NK cell proliferation would be to introduce an inducible suicide gene in the modified cells (73).

Migration to the tumor tissue is another aspect governing proper tumor targeting. This aspect has been largely overlooked and could potentially improve clinical outcome if infused NK cells are redirected to the tumor site instead of circulating nonspecifically into mostly non-tumor-bearing tissues. No studies aimed at improving the *in vivo* homing of infused gene engineered NK cells have yet been published.

As discussed above, numerous strategies for redirecting or boosting NK cell tumor killing *in vitro* have been explored. Introduction of CARs represent the most studied and developed approach that has recently reached clinical evaluation (**Table 2**). Expression of the high-affinity CD16 may soon also be tested in a clinical setting as this approach can be combined with already clinically available monoclonal antibodies that target an array of antigens expressed on a variety of different tumor types. Bolstering NK cell cytotoxicity via autocrine cytokine stimulation or via silencing of inhibitory NK cell receptors will likely require additional evaluation in preclinical animal models before they can be incorporated in clinical protocols. Once all these strategies are fully characterized pre-clinically, they may be combined to further improve the full anti-tumor potential of adoptively transferred NK cells. For instance, introduction of a CAR while simultaneously silencing the NKG2A inhibitory receptor may represent one such future approach. One can also consider adding autocrine cytokine stimulation to further improve cytotoxicity while simultaneously supporting their *in vivo* persistence. As NK cell degranulation is regulated by a balance of activating and inhibitory signals from well-defined cell surface receptors, it may also be possible to add CARs or other activation receptors together with selected receptors that mediate inhibition via ligands that are expressed on normal tissues (and not tumor cells), thereby giving genetically reprogramed NK cells an additional layer of target specificity. However, many additional preclinical studies will be needed before these approaches can reach clinic.


The choice of method for genetic reprograming of NK cells is another important factor that needs to be considered when taking genetically engineered NK cells to clinical evaluation. Viral transduction has the advantage of stable expression; however, as mentioned above, viral transduction of NK cells, especially primary cells, does not always lead to a satisfactory level of transgene expression and may require multiple rounds of transduction followed by selection of transgene positive cells. Moreover, proper expression of transgenes induced by viral transduction can take days, which may be of disadvantage since the lifespan of an NK cell may be relatively short following adoptive transfer (i.e., weeks). Future studies are warranted to better understand if multiple infusions of transfected NK cells can compensate transient transgene expression or if stable transgene expression is a prerequisite for inducing clinical responses following adoptive transfer of genetically engineered NK cells. Studies are also needed to fully understand the lifespan of NK cells, particularly those that have undergone *ex vivo* manipulation.

The optimal method for genetic manipulation of NK cells to be used in a clinical trial may also depend on what NK cell preparation is used (**Box 3**). The advantage with NK cell lines is that they can be utilized as an off-the-shelf product stably transduced to express the gene or genes of interest. They may also be long-lived if given the proper cytokine support. However, the downside with using NK cell lines, like NK-92, is the requirement for irradiation (10 Gy) prior to infusion to avoid the risk of engrafting cells that are potentially tumorigenic *in vivo* (74). Moreover, patients treated with infusions of NK cell lines would also need moderate to high level of preconditioning to suppress host immunity to avoid rejection of these allogeneic cells. Moreover, infusion of allogeneic cells can raise humoral immunity and lead to adaptive T-cell immune responses specifically against alloantigens, precluding repeated infusions even with the use of preconditioning. Similar allo-reactivity can be induced with the use of primary allogeneic NK cell infusions. The use of autologous NK cells circumvents these risks and precludes the need for preconditioning. The potential draw back with using autologous NK cells is that efficient tumor targeting can be prevented by inhibitory KIR interactions with self-HLA. A potential advantage with using an NK cell line versus primary NK cells is that large numbers of NK cells from the NK cell line can be infused, whereas the number of primary cells available for infusion are typically much more limited. However, this limitation has recently been circumvented by a number of highly efficient methods to expand primary NK cells *ex vivo* for clinical infusion (60). Ideally, infusion of autologous gene-modified NK cells can be used to avoid the rejection risk and the prerequisite for preconditioning. One approach to overcome limitations of autologous NK cell inactivation via self-HLA is to genetically modify these effectors to silence inhibitory self-HLA binding receptors, such as NKG2A and KIRs, which alone or in combination with for instance CARs, can improve the tumor targeting capacity of NK cells in the autologous setting.

# Concluding Remarks

Anti-tumor antibodies and CAR T cells have established immunotherapy as a viable treatment option for patients with cancer. Given their rapid and efficient method of recognizing tumor cells, NK cells represent a unique immune cell to genetically reprogram in an effort to improve the outcome of cell-based cancer immunotherapy. However, technical and biological challenges associated with gene delivery into NK cells have significantly tempered this approach. Viral transduction of NK cells initially resulted in low transgene delivery efficiencies that often required multiple rounds of transduction and/or cellular enrichment to achieve acceptable numbers of transgene expressing cells. Nevertheless, recent improvements in retro- and lentiviral transduction of NK cells have led to a flurry of preclinical studies on gene engineered NK cells. A number of studies have also shown that NK cells can be genetically reprogramed using mRNA electroporation. In contrast to viral transduction, this approach offers high transfection efficiencies without compromising their viability and does not require high-level biosafety laboratories. Although promising preclinical data on mRNA electroporated NK cells have emerged recently, concerns have been raised regarding the clinical utility of this approach as it only results in transient transgene expression.

Recently initiated clinical trials will soon give insight into the potential effectiveness of cell-based cancer immunotherapy strategies that utilize genetically modified NK cells. Nevertheless, further optimization of both viral transduction and electroporation of NK cells is still needed before this approach can be fully exploited in the clinic. With the recent advances in our understanding of the complex biological networks that regulate the capacity of NK cells to target and kill tumors *in vivo*, and with rapid developments in clinically compliant techniques to genetically manipulate NK cells, we foresee genetic engineering as an obligatory pathway to exploit the full potential of adoptive NK cell immunotherapy in patients with cancer.

# Acknowledgments

The authors wish to acknowledge all the members of the NHLBI Laboratory of Transplantation Immunotherapy, the NHLBI

### References


Division of Intramural Research (DIR), the Swedish Research Council, the Dean R. O'Neill and Edward Rancic Memorial Cancer Research Fellowship for their many contributions and support for original research described in this manuscript.


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

*Copyright © 2015 Carlsten and Childs. 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.*

# Natural Killer Cells for Immunotherapy – Advantages of the NK-92 Cell Line over Blood NK Cells

*Hans Klingemann\*, Laurent Boissel and Frances Toneguzzo*

*NantKwest, Inc., Culver City, CA, USA*

Natural killer (NK) cells are potent cytotoxic effector cells for cancer therapy and potentially for severe viral infections. However, there are technical challenges to obtain sufficient numbers of functionally active NK cells from a patient's blood since they represent only 10% of the lymphocytes and are often dysfunctional. The alternative is to obtain cells from a healthy donor, which requires depletion of the allogeneic T cells to prevent graft-versus-host reactions. Cytotoxic cell lines have been established from patients with clonal NK-cell lymphoma. Those cells can be expanded in culture in the presence of IL-2. Except for the NK-92 cell line, though, none of the other six known NK cell lines has consistently and reproducibly shown high antitumor cytotoxicity. Only NK-92 cells can easily be genetically manipulated to recognize specific tumor antigens or to augment monoclonal antibody activity through antibody-dependent cellular cytotoxicity. NK-92 is also the only cell line product that has been infused into patients with advanced cancer with clinical benefit and minimal side effects.

#### *Edited by:*

*Francisco Borrego, BioCruces Health Research Institute – Cruces University Hospital, Spain*

#### *Reviewed by:*

*Mar Vales-Gomez, Consejo Superior de Investigaciones Cientificas, Spain Jacki Kornbluth, Saint Louis University School of Medicine, USA*

*\*Correspondence:*

*Hans Klingemann hans.klingemann@nantkwest.com*

#### *Specialty section:*

*This article was submitted to NK Cell Biology, a section of the journal Frontiers in Immunology*

> *Received: 29 September 2015 Accepted: 23 February 2016 Published: 14 March 2016*

#### *Citation:*

*Klingemann H, Boissel L and Toneguzzo F (2016) Natural Killer Cells for Immunotherapy – Advantages of the NK-92 Cell Line over Blood NK Cells. Front. Immunol. 7:91. doi: 10.3389/fimmu.2016.00091*

Keywords: NK-92 cells, immunotherapy, cancer therapy, ADCC, cellular cytotoxicity

The remarkable responses recently achieved with T cells expressing chimeric antigen receptors (CARs) to target tumor antigens, especially in patients with lymphoid malignancies (1–3), highlight the ability of immune cells to become powerful therapeutic agents. However, in a significant number of patients, CAR-T-cell treatment was associated with adverse events including a potentially fatal "cytokine release syndrome" requiring ICU admission. In addition, the logistics and costs of this treatment pose a significant challenge for making it available for a larger number of patients. An increasing number of investigators believe that cellular therapy with natural killer (NK) cells obtained from the peripheral blood of either the patient (autologous) or a healthy donor (allogeneic) may represent safer effector cells for targeted cancer cell therapy than T cells.

However, there are biological, logistical, and financial challenges for the application of blood NK cells as a treatment modality for cancer patients (**Figure 1**). Autologous NK cells are typically not very effective as they are functionally silenced when they encounter self-MHC antigens, and they are also frequently compromised by the underlying disease and its treatment. On the other hand, allogeneic NK-cell infusions carry the risk of graft-versus-host (GvH) reactions even after the CD3 lymphocytes have been depleted (4). "Supply" is also limited, in part, because only about 10% of circulating blood lymphocytes are NK cells: to collect sufficient numbers of NK cells, patients or donors often have to undergo repeated leukaphereses that at times requires placing a central venous line, which is a major inconvenience for patients. This also usually limits the number of collections of NK-cell products for treatment to one or two. Moreover, to reach therapeutically meaningful numbers, NK cells have to be expanded *ex vivo*. This is most effectively done by culturing the cells (for allogeneic cells, this is after T-cell depletion) on a genetically engineered feeder layer of K562

cells that has been modified to express stimulatory molecules, such as IL-15 or IL-21 and 4-1BB (5–7). While expansion of NK cell can be achieved, some of these protocols result in NK-cell telomere shortening and reduction in cytotoxicity. Additionally, and in contrast to T-cell therapies, the ability to target blood NK cells through a CAR type mechanism is challenging due to the low transfection efficiency of blood NK cells even when viralbased methods are used.

Recognizing the significant challenges being faced in the use of blood-derived NK cells for therapeutic purposes, investigators have been trying to generate stable cell lines from blood NK cells. These efforts have generally been unsuccessful as those (frequently EBV-transformed) NK cells undergo only limited number of divisions before they experience apoptosis. The derivation of functional NK cells from embryonic stem cells and/or iPSC cells may be another avenue to generate sufficient numbers of NK cells for infusion. However, these studies are still at a relatively early stage and require additional characterization of the final product, as well as standardization of protocols, before this approach can be considered clinically relevant (8–10).

Another way of generating larger numbers of cytotoxic NK cells for treatment is *via* a clonal cell line immortalized from a patient who has developed a NK-cell lymphoma. However, NK-cell lymphoma is a relatively rare disease, and importantly, the clonal outgrowth of a cell line is a rare event. Over the past 20 years, only a handful of clonal NK-cell lines have been established (11–17) (**Table 1**). Those cell lines generally consist of "pure" NK cells, which proliferate and expand easily in culture, with a doubling time of 2–4 days and hence can be given to patients repeatedly on a flexible schedule. Most of those NK-cell lines do not display a robust and more universal cytotoxicity that would warrant their further development with the exception of NK-92, which is the only cell line that is consistently and highly cytotoxic to cancer targets (13). NK-92 cells have undergone extensive preclinical development (18–21) and have completed phase I trials in cancer patients [(22, 23), clinical trials NCT00900809 and NCT00990717]. Importantly, NK-92 cells – in stark contrast to blood NK cells – can be easily engineered by non-viral transfection methods to express specific receptors or ligands that can retarget them toward malignant cells.

Infusing cells of malignant origin may be counterintuitive, but a large body of evidence suggests that it is indeed safe as the cells are irradiated before infusion. Irradiation prevents *in vivo* proliferation while maintaining their ability to kill target cells and produce immune active cytokines. For NK-92, functional cytotoxicity is maintained after irradiation with 1000 cGy, a dose that completely abrogates proliferation (24). A large dataset in immunocompromised SCID mice has demonstrated that NK-92 cells are not tumorigenic (20, 21, 25, 26). This is supplemented with data from close to 50 patients who have now been treated with repeated infusions of irradiated NK-92 cells without any short- or long-term complications, especially tumor formation. Those phase I studies also confirmed that even with cell numbers as high as 10 billion cells/m2 , infusions are safe with no severe unexpected side effects (22, 23). At higher doses, responses were observed even for unselected end-stage patients.

Relatively few cell lines comply with the commonly accepted definition of NK cells as summarized in **Table 2**. YT cells, for example, do not express CD56 but are generally considered "NK-like" because they kill the MHC negative cell line K562. On the other hand, NKL and NKG cell lines are more closely related to NK-92. In fact, the NKG cell line was established by using identical culture conditions, as described for NK-92, i.e., the combination of fetal calf and horse serum, β-mercaptoethanol, and hydrocortisone as base constituents for the medium. Both the NKG and NKL cell lines have demonstrated *in vitro* cytotoxicity against a variety of malignant target cells, but these cells have never been administered to patients (12, 15).

The remainder of the NK-cell lines listed in **Table 1** has variable cytotoxicity toward cancer cell lines or primary malignant cells. One explanation may be that these cell lines express inhibitory KIR receptors, which are missing on NK-92 (less well characterized for NKL and NKG). For NK cells to engage and release their cytotoxic granules, adhesion molecules and the expression of activating receptors (such as NKp30, NKp44, and NKp46) are also essential. The combination of expression of activating


TABLE 2 | Operational definition of NK-cell lines.


receptors and adhesion molecules, together with the lack of most of the currently known KIRs, accounts for the broad cytotoxicity of NK-92 (27).

## PRECLINICAL STUDIES IN SCID MICE WITH NK-92

A large number of SCID mice studies with infusing either irradiated NK-92 (1000 cGy to mirror the clinical protocols) or non-irradiated NK-92 cells have been reported for a spectrum of human cancer xenotransplanted malignancies. In addition to AML (21), myeloma (28), and melanoma (20) using the parental NK-92 cells, CAR-modified NK-92 have been shown to eliminate AML [CD33.CAR (29)], lymphoma [CD19.CAR (18)], myeloma [CS1.CAR (25)], prostate cancer [EpCAM.CAR (30)], breast cancer [Her-2.CAR (31)], neuroblastoma [GD2.CAR (32)], and glioblastoma [EGFR.CAR (33)]. In those studies, CAR-modified NK-92 cells (now called taNK = targeted NK cells) eliminated the human tumor and significantly improved survival without any side effects in the recipient mice (34).

An advantage of the NK-92 platform is the ability to transfect the cells with a gene of interest without using retrovirus or lentivirus, as is necessary for T cells and peripheral blood NK cells. NK-92 can be genetically engineered by simple electroporation. Since the cells are highly IL-2 dependent, this can be used as a selection marker: the gene/construct of interest is cloned into a bicistronic vector with an IL-2 variant that is restricted to the endoplasmic reticulum and thus avoids any safety issues associated with secreted IL-2. Only those cells that are successfully transfected will grow out in a medium without IL-2, a huge advantage of a cell line over blood cells that makes the NK-92 cell platform an "off-the-shelf " engineered cellular product (35).

# CLINICAL TRIALS WITH NK-92

Four phase I trials in three different countries (US, Canada, and Germany) for different malignancies have been conducted with NK-92. All patients had treatment-resistant advanced cancer. The initial trials in Chicago and Frankfurt enrolled patients with renal cell and lung cancer and other solid tumors (22, 23). Two to three infusions of escalating dose levels of NK-92 were given 48 h apart. The MTD in these trials was largely based on the number of NK-92 cells that could be expanded over 2–3 weeks, and 1010 cells/m2 was considered the highest dose level. Except for some mild fever reactions in the occasional patient, the infusions were well tolerated. Despite the advanced disease, clinically significant responses were seen in patients with melanoma, lung cancer, and kidney cancer.

The study at Princess Margaret in Toronto (Keating, unpublished) enrolled patients with hematological malignancies, some of whom had relapsed after an autologous stem cell transplant. Again, those infusions were well tolerated and some clinically significant responses were noted. A phase I study at Pittsburgh Cancer Center is currently enrolling the last cohort of patients with relapsed/treatment-resistant AML. Those patients had a high leukemic blast infiltration in the bone marrow, posing a potential risk for tumor lysis syndrome, which, however, was not observed. Some patients showed a decrease or stabilization of their bone marrow blast count.

Despite the allogeneic nature of NK-92 cells and repeated infusions, the formation of HLA antibodies only occurred in less than half of all patients. This is likely related to the fact that cancer patients are immunocompromised, but it also mirrors earlier *in vitro* data suggesting that NK-92 cells are only mild stimulators in a mixed lymphocyte reaction (NantKwest, unpublished).

The costs of preparation and administration of NK-92 are significantly less compared to autologous or allogeneic NK cells and, particularly, compared to CAR.T cells, a treatment that has garnered significant attention recently. In contrast to CAR.T cell protocols, which involve highly selected patients and are believed to cost on the order of \$250,000 or more, infusion cycles with engineered NK-92 cells are generally less than \$20,000, with the option of repeated treatment cycles (**Table 3**).

## THE NEXT GENERATION OF ENGINEERED NK-92: haNK AND taNK

The parental NK-92 cells do not express the FcγRIIIa receptor (CD16) (**Figure 2**). Therefore, NK-92 cells cannot mediate antibody-dependent cellular cytotoxicity (ADCC). A NK-92 variant that expresses the high-affinity Fc receptor FcγRIIIa (158V) (haNK) is in clinical development to be combined with IgG1 monoclonal antibodies (mAbs). *In vitro* and *in vivo* studies have confirmed that combination with FcγRIIIa (V/V) augments mAb efficacy (36–39). The rationale for a treatment that combines mAb treatment with haNK infusions is based on the number of retrospective studies demonstrating an improved overall survival benefit in patients expressing the high-affinity FcγRIIIa receptor upon treatment with mAbs, such as Rituxan® (lymphoma), Herceptin® (breast cancer), and Erbitux® (colon cancer) (36, 37, 39, 40). Since only 10% of the population is homozygous for the high-affinity FcγRIIIa receptor (V/V), there is a clear rationale


for infusing haNK to those patients who carry the low- or intermediate-affinity FcγRIIIa receptor (90% of the population) (41) to maximize mAb efficacy.

The term "taNK" refers to targeted NK-92 cells (42). Those cells have been transfected with a gene that expresses a CAR for a given tumor antigen. A large body of preclinical murine data supports this approach as one with superior efficacy to the parental aNK cells [reviewed in Ref. (34)]. Further, the efficient transfection of NK-92 with mRNA (>80%) provides a route for quickly assessing the effectiveness of any given CAR construct for a particular indication (43). This approach may also ultimately provide a timely approach for personalized treatment based on a patient's particular tumor antigen/mutation.

# THE PATH TO PERSONALIZED CANCER THERAPY

Currently, only CARs that recognize common known tumor antigens are used to transfect T cells. What is needed are CARs that recognize patient-specific tumor antigens. This "missing link" can be achieved by using proteomic analysis of patient tumors to identify patient-specific neoantigens, followed by the screening of an antibody library for that particular antigen. Gene sequencing alone is not sufficient as many somatic DNA changes in tumors do not translate into expression of tumor antigens. Based on the nucleotide sequence of the antibody's antigen binding site, a

single chain Fv (scFv) for the CAR specific for the patient's cancer can be engineered and transfected into NK-92 cells *via* mRNA or other approaches.

The non-viral, mRNA-based off-the-shelf CAR technology allows to generate large numbers of taNK that are highly specific to the patients' tumor ("precision medicine," **Figure 3**). These cells can be frozen and shipped to the treatment site, on short notice. By identifying multiple patient-specific cancer antigens, the technology also enables the engineering of alternative and overlapping CARs in the event of a change in the tumor antigen profile ("escape").

# HOMING AND TARGET RECOGNITION

For NK cells to get to the site of tumor, they have to express certain "homing" molecules, such as CXCR4 for bone marrow and CCR7 for lymphoid tissue (44, 45). There is also some suggestion that CXCR2 is responsible for targeting cytotoxic cells to solid tumors (46). Although the expression of a CAR probably can account for some homing, the migration of cells from the blood stream into the bone marrow, lymph nodes, or solid tumors requires appropriate trafficking and homing receptors. Once the cells are at their "destination," the CAR will help targeting the malignant cells among the healthy ones.

# COMBINATION THERAPY

An off-the-shelf cell line, such as NK-92, with all its modifications lends itself to combination therapy. A recent review summarized

# REFERENCES

1. Brentjens RJ, Davila ML, Riviere I, Park J, Wang X, Cowell LG, et al. CD19 targeted T cells rapidly induce molecular remissions in adults with chemotherapy-refractory acute lymphoblastic leukemia. *Sci Transl Med* (2013) **5**:177ra38. doi:10.1126/scitranslmed.3005930

the additive and synergistic effect of certain drugs (bortezomib, IMiDs, and HDAC inhibitors) on the function of blood NK cells (47) and NK-92 cells (48, 49).

The checkpoint inhibitors (Keytruda®, Opdivo®, and Yervoy®) have recently shown some remarkable responses in several types of cancers. This beneficial effect is believed to be largely due to blocking of inhibitory molecules on T cells, such as CTLA-4 and PD1. Studies on the expression of checkpoint molecules on activated NK cells are somewhat inconclusive, but blood NK cells seem to express PD1 (50). By using checkpoint inhibitors in combination with NK-cell therapeutics, it could be expected that both the innate and the T-cell immune response can be further augmented.

The NK-92 platform clearly provides a base for targeting tumors through a multiplicity of approaches. The platform has been proven to be safe and effective even in its unmodified (parental) form. Additional improvements through genetic modifications will provide a combination therapy approach with mAb therapy (haNK) and a direct targeting approach through CAR modification (taNK). As an off-the-shelf therapy that can be administered universally to patients, this platform can provide a cell therapy modality that is not only versatile but that can be tailored to specific patient needs.

# AUTHOR CONTRIBUTIONS

HK contributed the main effort for writing the manuscript; LB and FT provided support for writing and review of the references, as well as for editing, formatting, and revisions.


cytokine-associated toxicity in a clinical trial of anti-CD19 chimeric-antigen-receptor-transduced T cells. *Blood* (2012) **119**:2709–20. doi:10.1182/ blood-2011-10-384388


patients treated with single-agent cetuximab. *J Clin Oncol* (2007) **25**:3712–8. doi:10.1200/jco.2006.08.8021


**Conflict of Interest Statement:** HK, LB, and FT are currently employed by Nantkwest, Inc.

*Copyright © 2016 Klingemann, Boissel and Toneguzzo. 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.*

# Correlation of Hsp70 serum levels with gross tumor volume and composition of lymphocyte subpopulations in patients with squamous cell and adeno non-small cell lung cancer

*Sophie Gunther1† , Christian Ostheimer2† , Stefan Stangl1 , Hanno M. Specht1 , Petra Mozes1 , Moritz Jesinghaus3 , Dirk Vordermark2 , Stephanie E. Combs1,4,5 , Friedhelm Peltz6 , Max P. Jung1 and Gabriele Multhoff1,4,5\**

#### *Edited by:*

*Francisco Borrego, Cruces University Hospital, Spain*

#### *Reviewed by:*

*Catharina C. Gross, University Hospital Münster, Germany Franz Rödel, Johann Wolfgang Goethe-University Frankfurt am Main, Germany*

#### *\*Correspondence:*

*Gabriele Multhoff gabriele.multhoff@tum.de*

*† Sophie Gunther and Christian Ostheimer have contributed equally to this work.*

#### *Specialty section:*

*This article was submitted to NK Cell Biology, a section of the journal Frontiers in Immunology*

> *Received: 15 September 2015 Accepted: 17 October 2015 Published: 02 November 2015*

#### *Citation:*

*Gunther S, Ostheimer C, Stangl S, Specht HM, Mozes P, Jesinghaus M, Vordermark D, Combs SE, Peltz F, Jung MP and Multhoff G (2015) Correlation of Hsp70 serum levels with gross tumor volume and composition of lymphocyte subpopulations in patients with squamous cell and adeno non-small cell lung cancer. Front. Immunol. 6:556. doi: 10.3389/fimmu.2015.00556*

*1Department of Radiation Oncology, Klinikum rechts der Isar, Technische Universität München (TUM), München, Germany, 2Department of Radiation Oncology, Martin Luther University Halle-Wittenberg, Halle (Saale), Germany, 3Department of Pathology, Klinikum rechts der Isar, Technische Universität München (TUM), München, Germany, 4Department of Innovative Radiotherapy (iRT), Helmholtz Zentrum München, Oberschleißheim, Germany, 5Department of Radiation Sciences (DRS), Helmholtz Zentrum München, Oberschleißheim, Germany, 6Pulmonary Division, 1 Medizinische Klinik, Klinikum rechts der Isar, Technische Universität München (TUM), München, Germany*

Heat-shock protein 70 (Hsp70) is frequently found on the plasma membrane of a large number of malignant tumors including non-small cell lung cancer (NSCLC) and gets released into the blood circulation in lipid vesicles. On the one hand, a membrane (m) Hsp70-positive phenotype correlates with a high aggressiveness of the tumor; on the other hand, mHsp70 serves as a target for natural killer (NK) cells that had been pre-stimulated with Hsp70-peptide TKD plus low-dose interleukin-2 (TKD/IL-2). Following activation, NK cells show an up-regulated expression of activatory C-type lectin receptors, such as CD94/NKG2C, NKG2D, and natural cytotoxicity receptors (NCRs; NKp44, NKp46, and NKp30) and thereby gain the capacity to kill mHsp70-positive tumor cells. With respect to these results, the efficacy of *ex vivo* TKD/IL-2 stimulated, autologous NK cells is currently tested in a proof-of-concept phase II clinical trial in patients with squamous cell NSCLC after radiochemotherapy (RCT) at the TUM. Inclusion criteria are histological proven, non-resectable NSCLC in stage IIIA/IIIB, clinical responses to RCT and a mHsp70-positive tumor phenotype. The mHsp70 status is determined in the serum of patients using the lipHsp70 ELISA test, which enables the quantification of liposomal and free Hsp70. Squamous cell and adeno NSCLC patients had significantly higher serum Hsp70 levels than healthy controls. A significant correlation of serum Hsp70 levels with the gross tumor volume was shown for adeno and squamous cell NSCLC. However, significantly elevated ratios of activated CD69+/CD94+ NK cells that are associated with low serum Hsp70 levels were observed only in patients with squamous cell lung cancer. These data might provide a first hint that squamous cell NSCLC is more immunogenic than adeno NSCLC.

Keywords: biomarker, tumor markers, biological, heat-shock protein 70, NSCLC, gross tumor volume, lymphocytes, immune responses

# INTRODUCTION

According to recent statistics, lung cancer is still among the most frequent causes of cancer-related deaths and the second most common cancer in both men and women in Western societies (1). The numbers of new cases are further increasing especially in Asia and Africa (2). According to the GLOBOCAN report 2000 (3), the incidence of lung cancer worldwide is 1,238,900 with a mortality of 1,103,100 and a 5-year prevalence of 1,394,400. One reason for this high mortality is that patients with lung cancer are frequently diagnosed in advanced tumor stages since the symptoms, such as dyspnea, coughing, or chest pain, are quite unspecific for a long period of time (4). Even after radical surgery, chemo-, and/or radiotherapy using up-to-date therapeutic approaches could not improve the outcome of locally advanced tumor stages. The progression-free and overall survival of nonsmall cell lung cancer (NSCLC) patients in stage IIIA and IIB is often <16 months (5). Therefore, there is a high medical need to explore new treatment modalities to increase life expectancy and to develop minimal invasive methods for an earlier detection of NSCLC.

In 2013, immunotherapy was elected as the "breakthrough of the year" for the treatment of cancer by the journal "Science" (6). The basis for this was the increase in knowledge in the detection of tumor-specific traits that have the potential to serve as tumor-specific targets for immunotherapeutic approaches. Along this line, our laboratory investigated the potential of the major stress-inducible heat-shock protein 70 (Hsp70, HSPA1A) as a tumor-specific target. Hsp70 is frequently overexpressed in many different tumor types like hematological malignancies, breast, prostate, colon, brain, and lung cancer (7, 8). Hsp70 assists protein folding, prevents protein aggregation and apoptotic cell death under physiological conditions and following stress (9, 10). Tumor cells compared to normal cells not only express significantly higher levels of Hsp70 in the cytosol (7, 8), but also exhibit an unusual plasma membrane localization of Hsp70 (11). Therefore, mHsp70 has the potential as a tumor-specific target for immunological approaches. Additionally, we have shown recently that mHsp70 positive tumor cells actively secrete Hsp70 in lipid vesicles, most likely exosomes, that mirror the membrane orientation of the cell from which they are derived (12). Based on these findings, liposomal Hsp70 which is found in the peripheral blood circulation can reflect the mHsp70 status of the tumor. We have established the lipHsp70 ELISA (13), which enables the detection of Hsp70 in the serum and plasma of patients. The use of the monoclonal antibody (mAb) cmHsp70.1 (14, 15) in this ELISA allows a quantitative determination of free and liposomal Hsp70 in the blood, whereas other commercially available Hsp70 ELISA tests only detect free Hsp70.

A mHsp70-positive tumor phenotype exerts dual functions, on the one hand, a high mHsp70 density is associated with a high aggressiveness of the tumor (16) and the potential of metastatic spread; on the other hand, mHsp70 on tumor cells serves as a target for activated natural killer (NK) cells, which have been incubated either with Hsp70 protein or TKD a 14mer peptide derived from Hsp70 in combination with low-dose IL-2 (TKD/ IL-2) (12, 17, 18). Following activation, these NK cells regain the capacity to kill mHsp70-positive tumor cells *in vitro* (19) and *in vivo* (15, 20) via granzyme B-mediated apoptosis (21).

For a better understanding of this duality of mHsp70, we addressed the question whether serum Hsp70 levels are associated with clinical parameters, such as gross tumor volume (GTV) at diagnosis and after radiochemotherapy (RCT), and whether serum Hsp70 levels can have impact on the immune phenotype of squamous cell and adeno NSCLC (18).

# MATERIALS AND METHODS

#### Patient Material

Blood samples of healthy human donors and NSCLC patients of the Klinikum rechts der Isar, TUM (patient collective #1; **Table 1**) and the Martin Luther University Hospital Halle-Wittenberg (patient collective #2, **Table 2**) were collected between 2008 and 2015. In patient collective #1, blood was taken from patients with squamous cell (*n* = 25) and adenocarcinoma (*n* = 18) of the lung at diagnosis and directly after RCT (*n* = 6), and from age- and gender-matched healthy human volunteers (*n* = 126) as a control group. Tumor biopsies were obtained from nine NSCLC patients, six in stage IV, and three in stage 3 (patient collective #1). The

#### TABLE 1 | Patient collective #1.


*Clinico-pathological characteristics of 43 NSCLC patients treated at the Klinikum rechts der Isar, TU München, Munich, Germany.*

#### TABLE 2 | Patient collective #2.


*Clinico-pathological characteristics of 55 NSCLC patients treated at the Martin-Luther University Hospital, Halle-Wittenberg.*

median age of all patients of patient collective #1 was 64 years and ranged from 23 to 95 years. In a screening study, NSCLC patients are stratified for their tumor stage and Hsp70 phenotype to enter a phase II clinical trial at the TUM, which is entitled "Targeted NK cell based adoptive immunotherapy for the treatment of patients with NSCLC after radiochemotherapy (RCT)" (18). In patient collective #2, blood was taken from 55 patients (median age 63 years, range 47–86 years) with advanced stage, inoperable NSCLC with an indication for primary RCT. These patients were recruited into a pilot study entitled "Potential plasma hypoxia markers in the radiotherapy of non-small cell lung cancer" (22). Characteristics of both patient collectives are summarized in **Tables 1** and **2**. Briefly, blood was collected in two EDTA KE/9 ml tubes and one Serum Z/9 ml separator tube (S-Monovette, Sarstedt, Nümbrecht, Germany). For the serum, blood was allowed to clot for 15 min at room temperature. After collecting 1.4 ml of EDTA blood for flow cytometry, plasma and serum were obtained by centrifugation at 750 *g* for 10 min. Aliquots of 100–300 μl were stored at −80°C for further analysis. The studies were approved by the local Ethics Committee of the Medical Faculties of both Universities (TUM, Halle-Wittenberg) and written informed consent was obtained from all patients before entering the trial. All procedures were performed in accordance to the Declaration of Helsinki, 1975, as revised in 2008. Clinical stage was determined according to the UICC TNM classification, seventh edition.

# Radiochemotherapy and Volumetric Parameters

Three-dimensional conformal RT (3D-RT) was given normofractionated (5 fractions/week) with curative intent (66 Gy total dose, 2 Gy single dose; Siemens Primus, Germany). Chemotherapy consisted of cisplatin (20 mg/m2 body surface on days 1–5) and vinorelbine (25 mg/m2 body surface on day 1) in treatment week 1 and 5 (2 courses). RT was CT based (Siemens Lightspeed RT, Germany) and all patients received a PET-scan (Philips Accel, USA) before RT. CT and PET images were merged and GTV was defined as the primary tumor and involved nodes (pathologic confirmed, highly suspicious by CT and PET). GTV was delineated by an experienced radiation oncologist at planning CT before RT and all image data were registered in the Oncentra Masterplan external beam planning software (Nucletron, USA) used for RT plan calculation.

# Detection of Hsp70 in Serum/Plasma Using the lipHsp70 ELISA

The Hsp70 content in the blood of NSCLC patients and healthy donors was determined using the lipHsp70 ELISA, which is equally suitable for serum and plasma samples (13). Using the monoclonal cmHsp70.1 antibody as a detection reagent (15), it is possible to detect both, soluble-free and lipid-bound Hsp70 in the serum/plasma of patients and healthy human individuals. This ELISA allows a quantitative analysis of the total amount of Hsp70 in the circulation blood (13). Briefly, 96-well MaxiSorp Nunc-Immuno plates (Thermo, Rochester, NY, USA) were coated overnight with 2 μg/ml rabbit polyclonal antibody (Davids, Biotechnologie, Regensburg, Germany), directed against human Hsp70 in sodium carbonate buffer (0.1 M sodium carbonate, 0.1 M sodium hydrogen carbonate, pH 9.6). After washing three times with phosphate-buffered saline (PBS, Life Technologies, Carlsbad, CA, USA) with 0.05% Tween-20 (Calbiochem, Merck, Darmstadt, Germany), wells were blocked with 2% milk powder (Carl Roth, Karlsruhe, Germany) in PBS for 1.5 h at 27°C. Following another washing step, serum samples diluted 1:5 in CrossDown Buffer (AppliChem, Chicago, IL, USA) were added to the wells for 2 h at 27°C. Then, the wells were washed again and incubated with 4 μg/ml of the biotinylated mouse mAb cmHsp70.1 (multimmune, Munich, Germany) in 2% milk powder in PBS for 2 h at 27°C. Finally, after another washing step, 0.2 μg/ml horseradish peroxidase-conjugated streptavidin (Pierce, Thermo, Rockford, IL, USA) in 1% bovine serum albumin (Sigma-Aldrich, St. Louis, MO, USA) was added for 1 h at 27°C. Binding was quantified by adding substrate reagent (R&D Systems, Minneapolis, MN, USA) for 30 min at 27°C and absorbance was read at 450 nm, corrected by absorbance at 570 nm, in a Microplate Reader (BioTek, Winooski, VT, USA). An eightpoint standard curve was determined for each ELISA test using 0–50 ng/ml recombinant Hsp70 diluted in CrossDown Buffer. Each sample was measured in triplicates.

# Immunohistochemical Staining

Immunohistochemical staining was performed on formalinfixed and paraffin-embedded specimen of lung tumors (*n* = 9). Sections were cut, dewaxed and hydrated, heated for 30 min in a microwave oven in 600 ml DAKO retrieval buffer, then washed for 5 min in H2O. After washing twice with T-PBS buffer, specimens were blocked for 1 h in 10% rabbit serum in PBS containing 1% BSA. Immunohistochemistry was done with streptavidine–biotin complex (StreptABC) using mouse mAb cmHsp70.1 (multimmune, Munich Germany) at a dilution of 1:200 overnight 2 h at 4°C.

# Analysis of the Lymphocyte Subpopulations with Flow Cytometry

In order to determine the proportion of different lymphocyte subpopulations, flow cytometric (FACS) analysis was performed using freshly collected EDTA blood (1.4 ml). Therefore, blood (100 μl) was transferred into 14 test tubes and then fluorescently labeled antibodies were added. The antibody combinations that were used for the FACS analysis are summarized in **Table 3**. After an incubation time of 15 min in the dark, the tubes were centrifuged for 5 min at 500 *g* at room temperature after adding 2 ml of PBS/10% FCS washing buffer. In order to eliminate erythrocytes, cells were incubated with lysing buffer (1:9 dilution of BD Lysing Solution Cat. 3490202 with millipore H2O) for 10 min at the room temperature in the dark. The respective percentages of B, T, and NK cell subpopulations are defined as the proportion of cells within the lymphocyte gate R1 (see **Figure 3**). For the determination of regulatory T cells, buffer A (1:10 dilution of component A with H2O) was added to the respective tubes. After two washing steps, cells were permeabilized with buffer C (1:50 dilution of buffer A with component B) for 30 min in the dark. Following another two washing steps, a PE-conjugated antibody

#### TABLE 3 | Panel of antibodies and 14 antibody combinations used in the study.


*APC, allophycocyanin; B, B lymphocyte; BD, Becton Dickinson Biosciences; BC, Beckman Coulter; CD, cluster of differentiation; COPD, chronic obstructive pulmonary disease; Ctrl, control; FITC, fluorescein isothiocyanate; NK, natural killer cell; PE, phycoerythrin; T, T lymphocyte; Treg, regulatory T cells.*

directed against the intracellular transcription factor forkhead box P3 (FoxP3) was added for another 30 min. After another two washing steps, 5 × 104 cells were analyzed on a FACScalibur instrument (Becton Dickinson, Heidelberg, Germany).

#### Statistical Analysis

Statistical analysis was performed using the IBM SPSS 20.0 software package for windows (SPSS Inc., USA). Statistically significant differences between Hsp70 levels of patients with high and low GTV and high and low CD94 expression, lymphocyte subpopulations of healthy donors, patients with squamous cell and adenocarcinoma as well as between the percentage of all lymphocytes of patients with high and low Hsp70 expression were determined with Mann–Whitney's *U*-test. Correlation between serum Hsp70 levels and GTV was evaluated using Spearman's Rank Correlation Coefficient. Potential differences in Hsp70 serum levels in NSCLC patients before and after RCT were determined with the Wilcoxon Rank-Sum Test. Comparison of Hsp70 levels in the serum of two patient groups (squamous cell and adeno NSCLC) and a group of healthy donors was also performed using the Kruskal–Wallis test with a Dunn multiple comparison test. A value of *p* < 0.05 was considered as statistically significant.

## RESULTS

#### Comparison of Hsp70 Levels in Patients with Squamous Cell, Adeno NSCLC, and Healthy Human Individuals

Serum samples derived from 25 squamous cell and 18 adeno NSCLC patients (patient collective #1), and 126 age- and gendermatched healthy donors were analyzed to determine the Hsp70 levels in the peripheral blood. As shown in **Figure 1A**, patients with squamous cell and adeno NSCLC (NSCLC; *n* = 43) have significantly higher serum Hsp70 levels (*p* ≤ 0.001) compared to healthy control controls (healthy; *n* = 126), as determined with the lipHsp70 ELISA, at diagnosis. The mean serum Hsp70 levels of patients with squamous cell and adeno NSCLC, 16.69 ± 2.7 ng/ ml and 14.51 ± 2.49 (median 12.15 vs. 12.20 ng/ml), respectively, did not differ significantly from each other (*p* = 0.825), but both tumor types differed significantly (*p* ≤ 0.001) from that of healthy controls (7.0 ng/ml) as shown in **Figure 1B**. A representative image of an Hsp70 positive tumor section of a squamous cell (upper graph) and an adeno (lower graph) NSCLC in stage IV is illustrated in **Figure 1C**. All nine tumor sections of NSCLC patients (six in stage IV and three in stage III) had elevated Hsp70 serum levels and exhibited a strong Hsp70 staining in the tumor cells but not in the connective tissue. Studies are ongoing that aim to analyze a potential correlation between cytosolic and serum Hsp70 levels in a larger panel of patients.

To investigate whether RCT impacts on serum Hsp70 levels in a small subgroup of patients (*n* = 6), blood was collected before start of therapy and directly after completion of therapy. Although Hsp70 levels after RCT remained significantly higher compared to those of healthy human individuals, a slight drop, which did not reach statistical significance (*p* = 0.463; Wilcoxon Rank-Sum Test), was detectable in the serum Hsp70 levels after completion of RCT (**Figure 1D**).

# Correlation of Hsp70 Levels at Diagnosis with Tumor Volume in NSCLC Patients

A comparison of free and lipid-bound Hsp70 in the circulation of tumor patients revealed that a major part of Hsp70 is bound to lipid vesicles, most likely exosomes, which are actively secreted by viable tumor cells carrying Hsp70 on their cell surface (12, 23). Therefore, we studied a potential correlation of the detected serum Hsp70 levels with the GTV of 55 NSCLC patients (patient collective #2; **Table 2**) that was determined by PET-imaging before start of RCT. The average tumor size of these patients was 219.9 ± 32.3 ml and the mean Hsp70 level was 11.2 ± 1.7 ng/ml. The Spearman's Rank Correlation Coefficient revealed a significant correlation (*p* = 0.03) between these two metric parameters. Regarding the median GTV of 143.6 ml, these patients were subdivided into a group with low (≤143.6 ml) and high (>143.6 ml) median GTV. As shown in **Figure 2**, patients in the high GTV group had significantly higher serum Hsp70 levels than patients with a low tumor volume (*p* < 0.05).

# Differences in the Immune Phenotype of Healthy Human Donors, Patients with Squamous Cell and Adenocarcinoma of the Lung

Nowadays, it is well accepted that an intact immune system plays a key role in long-term tumor control and in prevention of distant metastasis (24). Therefore, we comparatively investigated differences in the relative amount of lymphocytes and lymphocyte subpopulations, such as B, T, and NK cells, in the EDTA blood of healthy human donors (*n* = 10) and NSCLC patients (patient collective #1; *n* = 43; Table 1) at diagnosis, using multi-color FACS analysis. The panel of fluorescence-labeled antibodies and antibody combinations that were used in the study, is summarized in **Table 3**. Compared to blood of healthy human donors, the relative number of lymphocytes was significantly lower in patients with squamous cell (*n*= 25; *p*= 0.001) and adeno (*n*= 18; *p* = 0.008) NSCLC, although the percentage of lymphocytes in patients with different histology was very similar (22.4 ± 1.6% for squamous cell and 21.8 ± 2.4% for adenocarcinoma patients vs. 34.5 ± 1.77% in healthy controls) (**Figure 3**). The gating strategy of the lymphocytes is exemplified in the lower part of **Figure 3**. Gate R1 refers to the gated population of lymphocytes, whereas gate R2 represents granulocytes. The population of CD14+ monocytes is localized between the gates R1 and R2.

With respect to CD19<sup>+</sup> B cells, patients with squamous cell carcinoma had significantly lower percentages of B cells than healthy donors (*p* = 0.001) and adenocarcinoma patients (*p* = 0.02) (**Figure 4A**). A comparison of different CD3<sup>+</sup> T cell subpopulations, such as CD4<sup>+</sup>, CD8<sup>+</sup>, CD4<sup>+</sup>/CD25<sup>+</sup> regulatory, NKG2D<sup>+</sup>, and CD94<sup>+</sup> T cells revealed no major differences, apart from a significant increase in the subpopulation of CD3<sup>+</sup>/CD56<sup>+</sup> NKT cells in patients with adeno lung carcinomas compared to healthy controls (*p* = 0.038) (**Figure 4B**). The activation marker

that patients with squamous cell NSCLC had elevated ratios of CD3<sup>−</sup>/CD56<sup>+</sup> NK cells and CD3<sup>−</sup>/NKG2D<sup>+</sup> NK cells in general, compared to healthy controls (*p* = 0.053) and adenocarcinoma

Correlation of Hsp70 Serum Levels with Lymphocyte Subpopulations in Squamous

CD3<sup>−</sup>/CD94<sup>+</sup> NK cells were found to be significantly elevated in squamous cell carcinoma of the lung (**Figure 4C**). Furthermore, we have shown earlier that the C-type lectin receptor CD94 plays a key role in the recognition of Hsp70 by NK cells (17). To investigate the influence of serum Hsp70 on the ratio of CD3<sup>−</sup>/ CD94<sup>+</sup> NK cells, patients with squamous cell (*n* = 25) and adenocarcinoma (*n* = 18) were divided into groups with high (≥8.5% in squamous cell and ≥6.5% in adenocarcinoma) and low (<8.5% in squamous cell and <6.5% in adenocarcinoma) median percentages of CD3<sup>−</sup>/CD94<sup>+</sup> NK cells. It appeared that patients with squamous cell NSCLC with a high CD3<sup>−</sup>/CD94<sup>+</sup> NK cell ratio had significantly lower serum Hsp70 levels than the corresponding group with a low ratio of CD3<sup>−</sup>/CD94<sup>+</sup> NK cell ratio (*p* = 0.048) (**Figure 5A**). The CD94<sup>+</sup> NK cell population was also found to be positive for CD69, which indicates that these NK cells are active. Patients with high percentages of CD94<sup>+</sup>/CD69<sup>+</sup> NK cells have low Hsp70 serum levels and also a lower GTV, which indicates that these NK cells might be able to control the growth of mHsp70-positive tumor cells. In contrast, patients with adenocarcinoma showed no significant differences with respect to the ratio of CD3<sup>−</sup>/CD94<sup>+</sup> NK cells and serum

Cell and Adenocarcinoma Patients

patients (**Figure 4C**).

FIGURE 1 | Hsp70 serum levels (nanogram per milliliter) in healthy human individuals and patients with squamous cell and adeno NSCLC at diagnosis and directly after RCT therapy. (A) Serum Hsp70 levels of healthy donors (healthy; *n* = 126; median Hsp70 level: 7.03 ng/ml, 95th percentile: 13.84 ng/ml) and NSCLC patients (NSCLC; *n* = 43; median Hsp70 level: 12.15 ng/ml, 95th percentile = 40.20 ng/ml) (patient collective #1) at diagnosis measured with the lipHsp70 ELISA; \*\*\*p < 0.001 (Mann–Whitney U-Test). (B) Serum Hsp70 levels of healthy donors (healthy; *n* = 126) and patients with squamous cell (squamous; *n* = 25; median Hsp70 level: 12.10 ng/ml; 95th percentile: 40.50 ng/ml) and adeno (adeno; *n* = 18; median Hsp70 level: 12.38 ng/ml; 95th percentile: 40.20 ng/ml) NSCLC (patient collective #1) at diagnosis measured with the lipHsp70 ELISA; \*\*\**p* < 0.001 (Kruskal–Wallis test with Dunn's multiple comparison test). (C) Representative immunohistochemical images of a squamous cell and adeno NSCLC section stained with cmHsp70.1 antibody; 20× magnification (patient collective #1). The upper graph shows a squamous cell NSCLC and the lower graph an adeno NSCLC section. Only the tumor tissue but not the surrounding tissue shows an Hsp70 staining. (D) Serum Hsp70 levels of NSCLC patients (*n* = 6; patient collective #1) at diagnosis (before) and directly after RCT (after).

CD69 appeared to be slightly, but not significantly, elevated on CD3<sup>+</sup> T cells of patients with squamous cell (*p* = 0.81) and adenocarcinoma (*p* = 0.197) patients compared to healthy controls (**Figure 4B**). A representative picture of the strategy to analyze CD3<sup>+</sup>/CD4<sup>+</sup> T cells is illustrated in the inset of **Figure 4B**. In contrast to the T cell subpopulations, significant differences were

observed with respect to CD3− NK cell subpopulations regarding the activation marker CD69 and the C-type lectin receptor CD94. Patients with squamous cell NSCLC had significantly higher percentages of CD3<sup>−</sup>/CD56<sup>+</sup>/CD69<sup>+</sup> (*p* = 0.016) and CD3<sup>−</sup>/CD56<sup>+</sup>/CD94<sup>+</sup> (*p* = 0.028) NK cells than healthy controls (**Figure 4C**). Although not significantly different, it also appeared

that patients with squamous cell NSCLC had elevated ratios of CD3<sup>−</sup>/CD56<sup>+</sup> NK cells and CD3<sup>−</sup>/NKG2D<sup>+</sup> NK cells in general, compared to healthy controls (*p* = 0.053) and adenocarcinoma patients (**Figure 4C**).

## Correlation of Hsp70 Serum Levels with Lymphocyte Subpopulations in Squamous Cell and Adenocarcinoma Patients

CD3<sup>−</sup>/CD94<sup>+</sup> NK cells were found to be significantly elevated in squamous cell carcinoma of the lung (**Figure 4C**). Furthermore, we have shown earlier that the C-type lectin receptor CD94 plays a key role in the recognition of Hsp70 by NK cells (17). To investigate the influence of serum Hsp70 on the ratio of CD3<sup>−</sup>/ CD94<sup>+</sup> NK cells, patients with squamous cell (*n* = 25) and adenocarcinoma (*n* = 18) were divided into groups with high (≥8.5% in squamous cell and ≥6.5% in adenocarcinoma) and low (<8.5% in squamous cell and <6.5% in adenocarcinoma) median percentages of CD3<sup>−</sup>/CD94<sup>+</sup> NK cells. It appeared that patients with squamous cell NSCLC with a high CD3<sup>−</sup>/CD94<sup>+</sup> NK cell ratio had significantly lower serum Hsp70 levels than the corresponding group with a low ratio of CD3<sup>−</sup>/CD94<sup>+</sup> NK cell ratio (*p* = 0.048) (**Figure 5A**). The CD94<sup>+</sup> NK cell population was also found to be positive for CD69, which indicates that these NK cells are active. Patients with high percentages of CD94<sup>+</sup>/CD69<sup>+</sup> NK cells have low Hsp70 serum levels and also a lower GTV, which indicates that these NK cells might be able to control the growth of mHsp70-positive tumor cells. In contrast, patients with adenocarcinoma showed no significant differences with respect to the ratio of CD3<sup>−</sup>/CD94<sup>+</sup> NK cells and serum

Hsp70 levels (*p* = 0.908) (**Figure 5B**). These findings might indicate that squamous cell NSCLC is more immunogenic than adeno NSCLC.

# DISCUSSION

granulocytes.

Many lung tumors are diagnosed at advanced stages, which often restrict curative-intent treatment. In bronchial carcinoma, first diagnosis can be delayed by unspecific symptoms like coughing or dyspnea, which is also seen in inflammatory diseases of the lung, such as chronic obstructive pulmonary disease (COPD) or pneumonia. Apart from that, the majority of patients with the diagnosis COPD are smokers who additionally have an increased risk of developing lung cancer. Consequently, there is an urgent need for novel tumor biomarkers that can distinguish malignant from benign diseases. In contrast to normal cells, tumor cells frequently present Hsp70 on their surface. Membrane Hsp70 positive tumor cells have the capacity to actively secrete Hsp70 in lipid vesicles with molecular characteristics of exosomes (8, 9, 12). In a large variety of different malignant tumor entities,

FIGURE 1 | Hsp70 serum levels (nanogram per milliliter) in healthy human individuals and patients with squamous cell and adeno NSCLC at diagnosis and directly after RCT therapy. (A) Serum Hsp70 levels of healthy donors (healthy; *n* = 126; median Hsp70 level: 7.03 ng/ml, 95th percentile: 13.84 ng/ml) and NSCLC patients (NSCLC; *n* = 43; median Hsp70 level: 12.15 ng/ml, 95th percentile = 40.20 ng/ml) (patient collective #1) at diagnosis measured with the lipHsp70 ELISA; \*\*\*p < 0.001 (Mann–Whitney U-Test). (B) Serum Hsp70 levels of healthy donors (healthy; *n* = 126) and patients with squamous cell (squamous; *n* = 25; median Hsp70 level: 12.10 ng/ml; 95th percentile: 40.50 ng/ml) and adeno (adeno; *n* = 18; median Hsp70 level: 12.38 ng/ml; 95th percentile: 40.20 ng/ml) NSCLC (patient collective #1) at diagnosis measured with the lipHsp70 ELISA; \*\*\**p* < 0.001 (Kruskal–Wallis test with Dunn's multiple comparison test). (C) Representative immunohistochemical images of a squamous cell and adeno NSCLC section stained with cmHsp70.1 antibody; 20× magnification (patient collective #1). The upper graph shows a squamous cell NSCLC and the lower graph an adeno NSCLC section. Only the tumor tissue but not the surrounding tissue

shows an Hsp70 staining. (D) Serum Hsp70 levels of NSCLC patients (*n* = 6; patient collective #1) at diagnosis (before) and directly after RCT (after).

elevated Hsp70 levels in the serum could be detected (4, 5) which reflect a mHsp70-positive tumor phenotype. Herein, we could show significantly elevated levels of Hsp70 in the peripheral blood circulation of patients with squamous cell and adeno NSCLC when compared to healthy individuals. Previous work of our group has demonstrated that differences exist in Hsp70 serum levels in patients with inflammatory diseases, such as chronic hepatitis or liver cirrhosis and tumors, such as hepatocellular carcinoma (HCC) (25). All patients exhibited elevated Hsp70 levels in the serum compared to healthy controls, but the highest Hsp70 levels were detected in the group of tumor patients. In line with these findings, it has been demonstrated that NSCLC patients have higher Hsp70 levels in the blood than patients with COPD (26). These findings might provide a first hint that Hsp70 could have the potential as a tumor-specific biomarker, which is able to distinguish inflammatory and tumor diseases.

Since liposomal Hsp70, which can be quantified in the serum and plasma of patients, is derived from viable tumor cells (13, 27), we were interested to study the impact of RCT on serum Hsp70 levels in a small cohort of six NSCLC patients diagnosed with NSCLC stage IIIA and IIIB from whom blood was taken at diagnosis and after RCT. Despite a slight drop directly after completion of RCT, serum Hsp70 levels remained significantly higher than those in healthy individuals. This means that it might be possible to determine the Hsp70 tumor phenotype in the serum of patients not only at diagnosis but also during RCT.

Tumor staging in the follow-up period (2–3 months after RCT) revealed clinical responses, such as partial response or stable disease in these patients. Future studies on a larger patient cohort will elucidate whether clinical responses can be determined by a drop in the serum Hsp70 levels since the major part of circulating Hsp70 is actively released in a lipid-bound form by viable tumor cells (13). In order to further test this hypothesis, the GTV was compared to the serum Hsp70 levels. Herein, we could show that a small GTV was associated with low Hsp70 and a large GTV with high serum Hsp70 levels in NSCLC patients (*n* = 55; collective #2). Furthermore, a significant correlation between serum Hsp70 levels and PET-based GTV was shown using Spearman's Rank Order Correlation. A potential correlation of the Hsp70 levels with the UICC stage has to be performed in a patient cohort with a more balanced distribution of different UICC stages. Studies of Zimmermann et al. (26) have shown that the Hsp27 and Hsp70 serum levels could discriminate clinical stages in NSCLC and the group of Bauer et al. (28) has shown that the tumoral expression of both HSPs might provide useful biomarkers for risk stratification of UICC stage I/II colon cancer. Considering the potential prognostic and predictive quality of tumor volume and its changes during RT of cancer (29, 30), serial GTV registrations at different time points before, during and after RT by CT, MRI, or PET will be determined together with serum Hsp70 levels in ongoing studies.

Nevertheless, further research is necessary to assess in more detail how homogeneously membrane Hsp70 is expressed in tumor cells within one tumor or in tumors of different patients in order to validate a direct correlation between serum Hsp70 levels and the viable tumor mass. Immunohistochemistry data reveal that tumor cells, but not the surrounding normal tissue, are Hsp70 positive. Equally important is to determine which factors can influence the active secretion of Hsp70-containing vesicles by tumor cells. In the tissue of patients with squamous cell carcinoma of the head and neck, a high membrane Hsp70 expression on viable tumor cells was found to be associated with high serum Hsp70 levels (31). Salamuta S. Mambula observed a re-binding of extracellular Hsp70 to the cell surface of prostate carcinoma cells after its release (32). This phenomenon might also have an impact on circulating levels of Hsp70.

Apart from the fact that high membrane Hsp70 expression levels are associated with and aggressive tumor phenotype, radioresistance (16), and tumor progression (33), Hsp70 can also provide a target for the innate immune system (11, 34, 35). In general, the immune system of each individual human blood donor is highly individual, depending on the genetic constitution combined with the exposition to various antigens during life. Patients with solid tumors often show an immunosuppressed immune phenotype due to a variety of tumor immune escape mechanisms (36). In the present study, we intended to detect differences in the immune phenotype of patients with NSCLC of different histology and healthy human individuals. Considering that patients with squamous cell and adenocarcinoma had significantly lower percentages of lymphocytes in the peripheral blood than healthy controls (**Figure 4A**), our findings confirm that the immune system is essential for tumor control. Since membrane Hsp70 acts as a recognition structure for Hsp70-peptide pre-activated NK cells (35), we asked the question whether lipid-bound, circulating Hsp70 has an impact on the immune phenotype of peripheral blood lymphocytes (PBL). Flow cytometric analysis of the blood of patients with squamous cell carcinoma showed decreased percentages of B cells but elevated percentages of activated NK cell subpopulations in patients with squamous cell, but not adeno NSCLC. Significantly increased percentages of CD69<sup>+</sup>/CD94<sup>+</sup> NK cells were found in these patients compared to the healthy donors and adenocarcinoma patients. We could show that high serum Hsp70 levels are associated with a larger GTV in squamous cell but not adeno NSCLC. Regarding **Figures 5A,B**, patients with a lower percentage of CD94<sup>+</sup>/CD69<sup>+</sup> activated NK cells have higher Hsp70 serum levels in squamous cell NSCLC. Since high Hsp70 serum levels are associated with a larger GTV we speculate that CD94<sup>+</sup>/CD69<sup>+</sup> activated NK might be able to control growth of membrane Hsp70-positive tumor cells. Depending on its subcellular localization Hsp70 exerts dual functions. On the one hand, high intracellular and membrane-bound Hsp70 levels protect tumor cells from apoptotic cell death and thus mediate therapy resistance; on the other hand, membrane Hsp70 acts as a recognition structure for activated NK cells. Highly malignant tumor cells that secrete large amounts of Hsp70 might escape protective antitumor immunity by inducing tolerance, and therefore high Hsp70 levels that are associated with a larger GTV might be associated with a suppression of C-type lectinpositive NK cells. Vice versa, high percentages of CD94<sup>+</sup>/CD69<sup>+</sup> NK cells can control growth of mHsp70-positive squamous cell carcinomas and thus serum levels of Hsp70 are lower. In case of adeno NSCLC, no correlation of the percentage of CD94<sup>+</sup>/ CD69<sup>+</sup> NK cells and serum Hsp70 levels were observed. This finding might be attributed to the fact that adenocarcinomas are less immunogenic.

Previously, it has been demonstrated that the cell surface density of the C-type lectin receptor CD94 was up-regulated on NK cells after stimulation with Hsp70-peptide TKD (aa450–463) and low-dose IL-2 (17). Apart from Hsp70, it is known that CD94 interacts with non-classical HLA-E molecules (37), and serves either as an activating or inhibitory receptor depending on the NKG2C or NKG2A co-receptor (38). In Hsp70 membranepositive SCCHN patients, even 2 years after surgery and radiation therapy, the expression density of CD94 and NKG2D on NK cells was found to be significantly up-regulated (31). An increased expression of CD69 on NK cells is associated with an increased cytotoxic activity, proliferation, TNF-α production and the induction of further activation markers, such as CD25 and ICAM-1 (39). Our present data indicate that an increased percentage of CD69 and CD94 positive NK cells is only present in the blood of patients with squamous cell but not of adeno NSCLC patients and a significant association of the CD94 expression with serum Hsp70 could be also only detected in the group of squamous cell NSCLC patients (**Figures 5A,B**).

According to the diversity of their gene expression patterns, adenocarcinoma can be divided into subgroups with different outcome in overall survival (40). In squamous cell carcinoma of the lung, a reinforcement of the innate immune response by danger signals, such as circulating Hsp70, might be favorable. NK cells are not only able to detect "missing self " on malignant cells (41), but also can recognize membrane Hsp70 if expressed in a tumor-specific manner (42). Experimental mouse models indicate that the development of tumor-specific CD8<sup>+</sup> cytotoxic T cell responses is highly dependent on the NK cell-mediated elimination of tumor cells (43, 44) through the secretion of IFN-γ. Also macrophages and dendritic cells are activated by IFN-γ and TNF. An 11-year follow-up epidemiologic survey has shown that the paucity of activated NK cells was associated with an increased risk to develop cancer (45). Taken together our data indicate that NK cells as the first line of defense might play a major role in the control of squamous cell NSCLC. The danger molecule Hsp70 in the presence of pro-inflammatory cytokines, such as IL-2, might support the immune system to reinforce immunity against cancer.

# CONCLUSION

We could show that Hsp70 detected by the lipHsp70 ELISA can serve as a tumor biomarker in liquid biopsies of patients with squamous cell and adeno NSCLC. Due to the fact that vesicular, lipid-bound Hsp70 predominantly originates from viable tumor cells, a correlation of serum Hsp70 levels with GTV was found. This finding is in accordance to the result that changes in tumor volume during radiotherapy in NSCLC patients have potential prognostic and predictive value (46, 47).

Compared to healthy individuals, NSCLC patients have decreased lymphocyte counts in general. However, a comparison of lymphocyte subpopulations in NSCLC patients with different histology revealed elevated percentages of CD69<sup>+</sup>/ CD94<sup>+</sup> NK cells in squamous cell but not adeno NSCLC patients. This might provide a hint that squamous cell NSCLC is more immunogenic than adeno NSCLC. High serum Hsp70 levels are associated with a larger GTV and lower percentage of CD69<sup>+</sup>/CD94<sup>+</sup> NK cells. This might indicate that activated NK cells might be able to control growth of mHsp70-positive tumors in squamous cell NSCLC patients. If tumor escape mechanism suppress NK cell activation mHsp70-positive tumors cannot be killed.

# AUTHOR CONTRIBUTIONS

SG and CO contributed equally to the study; SG, CO, DV, and GM conceived and designed the experiments and wrote the paper; SG, SS, HS, PM, MJ, and FP performed the experiments, and or analyzed the data; DV and SC did proof-reading of the paper and gave clinical advice.

# REFERENCES


# ACKNOWLEDGMENTS

The work of the team at the TUM was supported by the EU-CELLEUROPE (315963); BMBF (Strahlenkompetenz, 02NUK007E; 02NUK031B; Innovative Therapies, 01GU0823; NSCLC, 16GW0030), DFG – Cluster of Excellence: Munich Centre for Advanced Photonics; DFG –SFB824/2; DFG – INST 95/980-1 FUGG; DFG – INST 411/37-1 FUGG. The work of the team at the Martin Luther University Halle-Wittenberg was supported by the Wilhelm-Sander Foundation (FKZ: 2007.123.2).


**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 Gunther, Ostheimer, Stangl, Specht, Mozes, Jesinghaus, Vordermark, Combs, Peltz, Jung and Multhoff. 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.*

# Heat shock protein 70 (Hsp70) peptide activated Natural Killer (NK) cells for the treatment of patients with non-small cell lung cancer (NSCLC) after radiochemotherapy (RCTx) – from preclinical studies to a clinical phase II trial

**Hanno M. Specht <sup>1</sup> , Norbert Ahrens <sup>2</sup> , Christiane Blankenstein<sup>3</sup> ,Thomas Duell <sup>4</sup> , Rainer Fietkau<sup>5</sup> , Udo S. Gaipl <sup>5</sup> , Christine Günther <sup>6</sup> , Sophie Gunther <sup>1</sup> , Gregor Habl <sup>1</sup> , Hubert Hautmann<sup>7</sup> , Matthias Hautmann<sup>8</sup> , Rudolf Maria Huber <sup>9</sup> , Michael Molls <sup>1</sup> , Robert Offner <sup>2</sup> , Claus Rödel <sup>10</sup>, Franz Rödel <sup>10</sup> , Martin Schütz <sup>11</sup>, Stephanie E. Combs <sup>1</sup> and Gabriele Multhoff 1,12\***


#### **Edited by:**

Francisco Borrego, BioCruces Health Research Institute, Spain

#### **Reviewed by:**

Roland Jacobs, Hannover Medical University, Germany Olatz Zenarruzabeitia, BioCruces Health Research Institute, Spain

#### **\*Correspondence:**

Gabriele Multhoff , Experimental Radiation Oncology, Klinikum rechts der Isar, Technische Universität München (TU München), Ismaninger Str. 22, Munich 81675, Germany e-mail: gabriele.multhoff@lrz. tu-muenchen.de

Heat shock protein 70 (Hsp70) is frequently overexpressed in tumor cells. An unusual cell surface localization could be demonstrated on a large variety of solid tumors including lung, colorectal, breast, squamous cell carcinomas of the head and neck, prostate and pancreatic carcinomas, glioblastomas, sarcomas and hematological malignancies, but not on corresponding normal tissues. A membrane (m)Hsp70-positive phenotype can be determined either directly on single cell suspensions of tumor biopsies by flow cytometry using cmHsp70.1 monoclonal antibody or indirectly in the serum of patients using a novel lipHsp70 ELISA. A mHsp70-positive tumor phenotype has been associated with highly aggressive tumors, causing invasion and metastases and resistance to cell death. However, natural killer (NK), but not T cells were found to kill mHsp70-positive tumor cells after activation with a naturally occurring Hsp70 peptide (TKD) plus low dose IL-2 (TKD/IL-2). Safety and tolerability of ex vivo TKD/IL-2 stimulated, autologous NK cells has been demonstrated in patients with metastasized colorectal and non-small cell lung cancer (NSCLC) in a phase I clinical trial. Based on promising clinical results of the previous study, a phase II randomized clinical study was initiated in 2014. The primary objective of this multicenter proof-of-concept trial is to examine whether an adjuvant treatment of NSCLC patients after platinum-based radiochemotherapy (RCTx) with TKD/IL-2 activated, autologous NK cells is clinically effective. As a mHsp70-positive tumor phenotype is associated with poor clinical outcome only mHsp70-positive tumor patients will be recruited into the trial. The primary endpoint of this study will be the comparison of the progression-free survival of patients treated with ex vivo activated NK cells compared to patients who were treated with RCTx alone. As secondary endpoints overall survival, toxicity, quality-of-life, and biological responses will be determined in both study groups.

**Keywords: Hsp70-based immunotherapy, NSCLC patients, radiochemotherapy clinical trial, clinical phase II, NK cells**

#### **INTRODUCTION**

The major stress-inducible heat shock protein 70 (Hsp70) is known as a cytoprotective molecular chaperone, which is frequently overexpressed in a large variety of tumor cells. As a molecular chaperone Hsp70 supports the correct folding of nascent and misfolded proteins, prevents protein aggregation following stress and assists protein transport across membranes (1). High cytosolic Hsp70 levels in tumor cells are associated with poor prognosis, metastatic spread (2) and resistance to standard therapies, such as radiochemotherapy (RCTx) (3–7). Inside tumor cells Hsp70 contributes to tumor cell survival by interfering with apoptosis pathways (8, 9).

Apart from these intracellular chaperoning functions Hsp70 has been found to be expressed on the cell surface of highly aggressive primary and metastatic tumor cells. This finding was not expected since Hsp70 lacks a transmembrane domain. Initially, the membrane expression of Hsp70 was proven by selective cell surface iodination (10) and by biotinylation followed by proteomic profiling of cell surface bound proteins (11). Later a mouse monoclonal antibody was established [cmHsp70.1 mAb; (12)], which is able to detect mHsp70 highly selectively on the plasma membrane of viable tumor cells by flow cytometry. Screening of viable single cell suspensions of more than 1,000 freshly isolated tumor biopsies and their corresponding normal tissues revealed that more than 50% of all tumors, but none of the healthy normal tissues exhibited a mHsp70-positive phenotype (13, 14). A prognostic value of mHsp70 on tumor cells has been demonstrated in xenograft tumor mouse models (13), since metastases of orthotopically implanted primary tumor cells showed a significantly higher surface Hsp70 density than the primary tumor cells (15). Moreover, survival of patients with mHsp70-positive squamous cell carcinoma of the lung and lower rectal carcinomas revealed a significantly decreased overall survival (OS) (16) compared to their mHsp70-negative counterparts.

Evidence is accumulating that cell surface translocation of Hsp70 in tumor cells is mediated via non-classical vesicular pathways (17) since inhibitors of the classical ER-Golgi transport route (Brefeldin A, Monensin) did not affect the cell surface expression of Hsp70. Anchorage of Hsp70 in the plasma membrane of tumor cells is most likely enabled by tumor-specific lipid components, such as globotriaosylceramide Gb3 (18), which directly interact with Hsp70 and are compounds of lipid rafts. Whether mHsp70 in lipid rafts mediates chaperone activity for cell surface signaling receptors is still a matter of debate. Following stress, such as a nonlethal radio- or chemotherapy not only the intracellular Hsp70 levels but also the membrane density of Hsp70 was found to be increased in tumor cells (19–21). Following irradiation Hsp70 is predominantly co-located with phosphatidylserine in the plasma membrane of tumor cells and thus has to leave the cholesterol-rich microdomain signaling platforms (22). Although elevated Hsp70 membrane and cytosolic levels confer resistance of tumor cells to standard therapies such as radio- and chemotherapy (2, 5, 6, 8), it also has been shown that mHsp70 serves as a target structure for activated natural killer (NK) cells. It appears that mHsp70 can fulfill dual functions: on the one hand, it can mediate protection against lethal damage, which is induced by radio- and chemotherapy; on the other hand, it might provide a target structure for the cytolytic attack by the innate immune system (13, 14, 16, 19).

#### **PRECLINICAL FINDINGS**

#### **ACTIVATION OF NK CELLS WITH Hsp70 PEPTIDE TKD PLUS IL-2 AND IDENTIFICATION OF CD94 AS A SURROGATE MARKER FOR CYTOLYTIC ACTIVE NK CELLS**

Previously, we demonstrated that incubation of NK, but not T cells, with peptide-free recombinant Hsp70 protein in combination with pro-inflammatory cytokines, such as IL-2 or Il-15 can stimulate the cytotoxic, proliferative, and migratory capacity of NK cells against highly aggressive, mHsp70-positive tumor cells, *in vitro* (14, 23). Similar to full-length Hsp70 protein, a 14-mer

peptide (TKDNNLLGRFELSG, aa 450–463) also could activate the cytolytic and proliferative capacity of NK cells at equimolar concentrations (24). The stimulatory 14-mer peptide is an Nterminal extension of the 8-mer binding epitope of the antibody cmHsp70.1, which detects mHsp70 on the cell surface of tumor cells. Since the induction of the cytolytic activity of NK cells with the peptide is dose-dependent and saturable it is assumed that the interaction of NK cells with the peptide might be receptormediated. By antibody and protein/peptide blocking assays the C-type lectin receptor CD94 could be identified as a potential receptor, which mediates the interaction with the stimulatory Hsp70 peptide. CD94 forms a heterodimer either with the coreceptor NKG2A or NKG2C and thus acts as an inhibitory or activation receptor complex. Following incubation of NK cells with Hsp70 protein or Hsp70 peptide plus IL-2, the density of CD94 was found to be significantly up-regulated concomitant with an increased cytolytic activity against mHsp70-positive tumor cells (25, 26). Therefore, the density of CD94 on NK cells was considered as a surrogate marker for the cytolytic activity of NK cells against mHsp70-positive tumor cells.

#### **MODE OF TUMOR CELL KILLING OF mHsp70-POSITIVE TUMOR CELLS BY PEPTIDE PLUS IL-2 ACTIVATED NK CELLS**

It has been shown that cell membrane-bound Hsp70 renders tumor cells more susceptible to the lysis of NK cells that had been stimulated with Hsp70 protein/peptide plus low dose IL-2 (13, 14). In order to uncover the mechanism of lysis affinity chromatography, experiments were performed using lysates of activated NK cells on columns that were bound to either Hsp70 protein or Hsp70 peptide. Interestingly, the apoptosisinducing serine protease granzyme B has been found to show an interaction with Hsp70 protein and peptide as determined by matrix-laser desorption ionization time of flight mass peptide finger printing (MALDI-TOF) (27). The interaction of granzyme B with Hsp70 was previously confirmed by Western blot and flow cytometry (27).

Natural killer cells that have been stimulated with Hsp70 plus IL-2 show a significantly up-regulated production of granzyme B in their intracellular vesicles. In contrast, the levels of perforin were found to be up-regulated only moderately (25, 26). Therefore, it is assumed that mHsp70-positive tumor cells are predominantly killed by granzyme B. Incubation of isogenic tumor cell systems that differ in their mHsp70 expression levels indicate that granzyme B in the absence of perforin effectively lysed mHsp70 positive tumor cells, but not their mHsp70-negative counterparts. Regarding these results, we concluded that Hsp70-positive tumor cells are killed by Hsp70 plus IL-2 activated, CD94-positive NK cells via granzyme B-mediated apoptosis (27).

#### **PRECLINICAL MODELS SHOWING THE EFFICACY OF Hsp70 PLUS IL-2 ACTIVATED NK CELLS**

An incubation of purified human NK cells with Hsp70 peptide plus low dose IL-2 resulted in a specific tumor cell killing of mHsp70-positive, but not their mHsp70-negative counterparts, *in vitro*. Furthermore, activated NK cells compared to resting NK cells showed a significantly increased migratory capacity

toward mHsp70-positive tumor cells as demonstrated in a transwell migration system. In order to proof the Hsp70-based antitumor activity of NK cells, *in vivo* immunodeficient SCID/beige mice bearing Hsp70-positive colon carcinoma cells were injected intravenous (i.v.) into the tail vein with either peripheral blood lymphocytes (PBLs), CD3-positively sorted T lymphocytes or CD3-negative CD94-positive NK cells that had been stimulated *ex vivo* with Hsp70 peptide plus low dose IL-2. A single injection of activated NK, but not of T cells, was found to result in a significant reduction in the tumor weight of the mice (28). After stimulation of PBL with Hsp70 peptide plus IL-2 activation markers such as CD25 or CD69 were found to be increased predominantly on the NK cell fraction. This result indicated that it is possible to selectively stimulate NK cells within unsorted PBL. Furthermore, the amount of tumor cell killing appeared to correlate with the actual number of activated NK cells that show a high expression density of CD94 (25, 26).

*Ex vivo* activated human NK cells could also control metastasized mHsp70-positive pancreatic carcinomas in immunodeficient mice, as shown previously (15). These data indicate that mHsp70 acts as a universal tumor-specific target structure, which is not restricted to only one specific tumor entity. In contrast, unstimulated NK cells did not induce tumor cell killing. Interestingly, NK cells that had been stimulated with IL-2 only were significantly less efficient in the control of the tumor growth in mice. A single injection of mice with Hsp70 peptide plus IL-2 activated NK cells, but not with identically stimulated T cells or IL-2 activated NK cells was also able to significantly enhance the OS of tumorbearing mice (15). Following four repeated i.v. injections with Hsp70 peptide plus IL-2 pre-activated NK cells, the primary pancreatic tumor was found to be eliminated completely and hepatic metastases could be prevented (15).

#### **CLINICAL RESULTS**

#### **SUMMARY OF THE FINDINGS OF A PHASE I CLINICAL TRIAL USING Hsp70 PEPTIDE PLUS IL-2 ACTIVATED NK CELLS**

Based on promising preclinical results using *ex vivo* activated NK cells in different tumor mouse models, a phase I clinical trial in tumor patients was performed in 2002 (29). Mouse models also have indicated that pre-activated NK cells are well tolerated even at high numbers (28). A major goal of the previously performed clinical phase I study (29) was to test whether a treatment of tumor patients with autologous, *ex vivo* Hsp70 peptide plus IL-2 activated NK cells is safe and well tolerated. After i.v. injection of escalating numbers of *ex vivo* pre-activated NK cells and escalating numbers of treatment cycles (up to six cycles) using complete leukapheresis products none of the patients showed any severe toxicities [no toxicities ≥grade 2 according to common toxicity criteria (CTC)]. Biological and clinical responses were evaluated in patients with confirmed metastatic colorectal cancer (*N* = 11) and non-small cell lung cancer (NSCLC) (*N* = 1) who failed clinical responses to standard therapies such as chemotherapy, radiotherapy, or laser-induced thermotherapy. At the beginning of the NK cell-based therapy, patients suffered from multiple pulmonary, hepatic, soft tissue and bone metastases, or local relapses. Because of the advanced tumor stages of the patients, it was ethically not accepted to obtain fresh biopsies to determine the Hsp70 status

of the tumor during the study. Patients were enrolled in the study at least 4 weeks after the last standard therapy. Leukocyte concentrates were obtained from all patients by a 3–4 h leukapheresis at the Institute for Clinical Chemistry and Laboratory Medicine, University Hospital Regensburg. After sterile density gradient centrifugation lymphocytes were counted and re-suspended at cell densities of 5–10 × 10<sup>6</sup> /ml in fetal-calf serum (FCS)-free, GMP grade culture medium. For the dose escalation part, PBLs of the patients were frozen in aliquots. After simultaneous addition of Hsp70 peptide TKD (2µg/ml, Bachem) and recombinant IL-2 (100 IU/ml Aldesleukine, Chiron) the cell suspensions were transferred into 250 ml Teflon culture bags (VueLife-118) and incubated in an incubator at 37°C for 3–5 days under gentle rotation. After washing in physiological saline (0.9% NaCl) and harvesting, the cells were suspended in physiological saline (500 ml) conditioned with 100 IU/ml recombinant IL-2. Sterility testing of the cell products was performed before, on day 3 and directly before re-infusion of the cells (**Figure 1**). Four of the 12 patients who were treated within an intra-individual and inter-individual dose escalation schedule showed no adverse effects. Therefore, from patient 4 onward all patients received the complete leukapheresis product containing between 0.7 × 10<sup>9</sup> and 8.5 × 10<sup>9</sup> PBLs in a single injection. The number of activated NK cells, which were re-infused ranged from 0.1 × 10<sup>9</sup> to 1.5 × 10<sup>9</sup> cells in all patients. These treatments were repeated up to six times in individual patients without observing any toxic side effects. The laboratory parameters, which were taken before and after each re-infusion cycle showed no NK cell-based treatment associated changes. A gradual deterioration of bilirubin, lactate dehydrogenase, and liver enzymes in some patients could be related to the disease progression. Irrespectively

blood lymphocytes (PBLs) are stimulated ex vivo in a GMP laboratory with TKD/IL-2 for 3–5 days. After measuring of NK activation markers and sterility testing, the activated cells are washed and re-infused (i.v.) in the patient.

of the treatment cycle, the number of leukocytes, thrombocytes, and hemoglobin remained unaltered at each leukapheresis. After each re-infusion, vital parameters of all patients were monitored for at least 1 h. Only after the first re-infusion of the cells into the first patient, the patient was hospitalized overnight. A graphic overview of the NK cell activation process is shown in **Figure 1**. In step 1, patient received a leukapheresis, after gradient density centrifugation, patient-derived PBLs were tested for their activity by flow cytometry and functional assays. Then PBLs were incubated with TKD/IL-2 for 3–5 days in a GMP laboratory. After testing viability, sterility, and functional characteristics, cells were washed twice in 0.9% NaCl solution. Then activated cells were re-infused in 500 ml 0.9% NaCl conditioned with IL-2 (100 IU/ml) into the patient by i.v. injection within 30–60 min.

In addition to routine laboratory exams also specific laboratory parameters were determined from the blood product and from the patient before treatment, before leukapheresis, before cell reinfusion, and after cell re-infusion. The immune phenotype of the lymphocytes and the release of cytokines such as IFNγ, TNFα, and the apoptosis-inducing enzyme granzyme B were determined in the *ex vivo* cell culture before and after stimulation and in the blood of the patient before and after re-infusion of the activated cells. From day 1 to 4 of stimulation, the expression density of CD94 on activated NK cells continuously increased as it has been shown for healthy individuals previously. A comparison of the mean fluorescence intensity of the Hsp70 receptor CD94 after the first and the fourth treatment cycle also revealed a significant increase. These data indicate that similar to healthy controls the CD94 expression could also be increased on NK cells of tumor patients that had been treated with radio- and/or chemotherapy before. With respect to the cytolytic activity of *ex vivo* stimulated NK cells, 10 out of the 12 patients showed a significant increase after stimulation with Hsp70 peptide plus IL-2. In contrast, a stimulation of the patient's cells with IL-2 alone showed no significant increase in the cytolytic activity. Studies using either Hsp70 specific or CD94 specific antibodies to block the target structure or the NK receptor demonstrated that Hsp70 is recognized on tumor cells by CD94-positive NK cells of the patients. Most interestingly, we could show that the cytolytic activity of *ex vivo* stimulated NK cells could be confirmed in the blood of the patients even 24 h after re-infusion of the cells. A comparison of the activity of NK cells in the blood before and after the fourth re-infusion cycle revealed significantly increased cytolytic responses of the blood lymphocytes in three out of five patients.

Clinical tumor responses (one stable disease, one mixed response) could be observed in one patient with colorectal and one patient with NSCLC who received at least four treatment cycles (29). Despite the low numbers, these findings were not expected due to the fact that all patients suffered from advanced tumor stages and had progressive disease during their last standard therapy.

In summary, the phase I clinical trial showed that re-infusion of Hsp70 peptide TKD plus IL-2 activated autologous NK cells is feasible, safe, and well tolerated. The immunological and clinical responses warrant additional studies in patients with a lower tumor burden and a confirmed mHsp70-positive tumor phenotype (29).

#### **DESCRIPTION OF AN ONGOING PROOF-OF-CONCEPT CLINICAL PHASE II TRIAL USING Hsp70 PEPTIDE PLUS IL-2 ACTIVATED NK CELLS IN PATIENTS WITH NSCLC FOLLOWING RADIOCHEMOTHERAPY**

There is still a strong need to further improve the therapy of patients with non-resectable locally advanced NSCLC, since despite multimodal therapies the prognosis of those patients remains bad with a median OS of approximately 16 months. In this tumor stage, <20% of the patients survive more than 5 years (30). Several approaches to improve outcome have been evaluated, including systemic treatments or novel radiation techniques. The addition of chemotherapy did not enhance OS in these patients significantly (31). Although a distinct dose– response relationship is known for radiation therapy in lung cancer, escalated regimes have not improved outcome in NSCLC as shown in a prospective clinical trial (32). Immunotherapy seems to be a promising concept to improve the therapy of those patients. Since the membrane expression density of Hsp70 could be selectively enhanced on tumor cells following standard therapies such as ionizing irradiation and chemotherapy *in vitro*, we aimed to treat tumor patients with Hsp70 peptide TKD plus IL-2 activated (TKD/IL-2) autologous NK cells that had been treated with a RCTx. On the one hand, RCT should reduce the actual mass of viable tumor cells, and on the other hand, RCT should enhance the membrane density of Hsp70, which is recognized by pre-activated NK cells. Patients (*n* = 90) with NSCLC in non-metastasized but locally advanced stages IIIA and IIIB after RCTx (platinum based chemotherapy, 60– 70 Gy) will be enrolled into the randomized multicenter clinical phase II trial (EUDRA-CT:2008-002130-30). Previous findings have indicated that a mHsp70-positive tumor phenotype is associated with a significantly decreased OS (16). Therefore, this interventional phase II trial incorporates a 1:1 randomized control group of patients that receive no adjuvant NK cell-based immunotherapy in addition to the current standard of treatment (simultaneous RCTx), and also exhibit a mHsp70-positive tumor phenotype.

The major in- and exclusion criteria of the trial are summarized in **Table 1**. The scheme of the ongoing clinical phase II trial is shown in **Figure 2**. In a pre-study part, the Hsp70 phenotype of the tumor will be assessed in the blood of the patient by an Hsp70-specific ELISA and if available on tumor biopsies by flow cytometry using an Hsp70-specific mouse monoclonal antibody and the tumor stage will be determined. After successful RCTx (partial response, or at least stable disease) NSCLC patients in stage IIIA and IIIB will be randomized into the study. Patients in the interventional study arm receive four cycles of *ex vivo* TKD/IL-2 stimulated NK cells on a monthly schedule. Tumor assessment will be performed in both arms of the trial every 3 months for the first year and every 6-month thereafter until progression of disease.

The mHsp70 status on the tumor will be assessed in a pre-study screening part carried out in the Department of Experimental Radiation Oncology at the Klinikum rechts der Isar, Technische Universität München (TUM). Tumor biopsies obtained during the bronchoscopy at primary staging will be sent to the sponsor's laboratory. Freshly isolated single cell suspensions of biopsies are used for analysis of the Hsp70 tumor phenotype.


#### **Table 1 | In- and exclusion criteria of the ongoing phase II clinical trial: targeted NK cell-based adoptive immunotherapy for the treatment of patients with non-small cell lung cancer (NSCLC) after radiochemotherapy**.

If bronchoscopy has been performed *alio loco* prior to patient admittance, the mHsp70 expression will be determined by quantifying the amount of exosomal Hsp70 in the blood of the patients. Previously, we have shown that mHsp70-positive tumor cells actively release Hsp70 in exosomes, which present Hsp70 on their surfaces. Since most commercial Hsp70 ELISAs are unable to

detect lipid-bound, exosomal Hsp70 in the serum we established a novel lipHsp70 ELISA test. This ELISA detects and quantifies exosomal Hsp70 in the serum of patients with a high accuracy (patent filed). Using this Hsp70 ELISA, we could show that patients with tumors of different entities (including NSCLC patients) exhibit significantly higher Hsp70 serum levels than a group of age- and gender-matched healthy volunteers (33). Therefore, in the screening part of the study, Hsp70 serum levels will be measured in NSCLC patients before and after RCTx. In the therapeutic clinical phase II trial, Hsp70 serum levels will be assessed before and after NK cell-based immunotherapy for the treatment group and at identical fixed time points for the control group. The mHsp70 phenotype will be correlated with the Hsp70 serum levels and the tumor volume will be correlated with the Hsp70 serum level. Furthermore, the results of the Hsp70 serum values before and after therapy will help to elucidate, whether serum Hsp70 levels can improve the monitoring of the clinical outcome after therapeutic intervention. The results of the Hsp70 serum levels will also help to further validate the role of exosomal Hsp70 as a prognostic/diagnostic tumor biomarker, as it was suggested for head and neck, lung, colorectal, pancreatic, brain cancer, and leukemic patients before (33). Elevated levels of Hsp70 in the serum were also found in patients with squamous cell carcinoma of the head and neck (34) and glioblastomas (35).

Identical to the phase I clinical trial the leukapheresis products, which are used as source material for stimulated NK cells, are produced centralized at the Institute for Clinical Chemistry and Laboratory Medicine, Transfusion Medicine, University Hospital Regensburg, under GMP conditions in order to obtain comparable cell products. Afterwards cell processing is performed in a GMP cleanroom laboratory. For all manufacturing steps, the permission of the competent authorities was obtained. NSCLC patients in the treatment arm will be treated four times every 2–6 weeks with *ex vivo* TKD/IL-2 stimulated NK cells after RCTx. In case of significant toxicities, the treatment will be interrupted, the dose of re-infused cells will either be reduced or stopped. Patients in the control arm also have received standard RCTx prior to enrollment. Tumor assessment will be done for both study groups at enrollment, every 3 months during the first year and every 6-month thereafter until progression of disease. Response to treatment will be assessed centrally at TU München in order to prevent bias.

#### **REGULATORY ASPECTS AND PRODUCTION OF THE INVESTIGATIONAL MEDICINAL PRODUCT OF THE PHASE II CLINICAL TRIAL (APCETH GmbH UND Co. KG, MUNICH)**

The Investigational Medicinal Product (IMP) is classified as an advanced therapy medicinal product [ATMP, somatic cell product, Regulation (EC) No 1394/2007]. The process established in an academic institution for the phase I had to be transferred and adapted

to the actual requirements for ATMP's. The process required approximately 1.5 years and was completed with the manufacturer license granted to apceth by the local and national authority as a prerequisite for the clinical trial application by the sponsor. The manufacturing of TKD-activated autologous NK cells follows the principles of good-manufacturing practice (EU-GMP guidelines) in order to provide a robust and reproducible pharmaceutical product.

The implementation involved the development phase, the transfer to GMP, and the implementation of a GMP-compliant process. The development phase defined the process steps (closed systems wherever possible), the equipment and quality control (QC) methods, the definition of in-process controls, and the materials (should be available in adequate quality). The comparability to the process applied for phase I and the results obtained should be confirmed. The transfer phase involved a stepwise implementation to full GMP mainly relating to up-scaling from small-scale (microtiter-plates), medium-scale (bags with small volume of cell suspension), and large-scale in addition to the qualification of the materials and analytical methods. The GMP-process was established with the successful validation of the QC methods and the validation of the process by media-fills and qualification runs. The modifications to the phase I process relate to the exact specification of the starting material "apheresis product" including transport to the GMP-facility, the introduction of closed systems (density centrifugation of the starting material is now performed in a Sepax® device and harvesting/washing procedures in closed systems; in a class B room), release criteria, and test methods. Special attention was paid to the reconstitution buffer [Ringerlactate and 0.1% human serum albumin (HAS)] and the shelf-life (24 h) as the end product is delivered to several study centers. Most of the development work was attributed to the definition and validation of analytical QC methods (flow cytometry) and the qualification of ancillary and raw materials (for example, TKD and cytokines). Flow cytometry of the activated NK cell population was challenging and different approaches had to be evaluated resulting in a change of specification (CD94) and the introduction of a parallel culture for QC-testing to test for the mean fluorescence intensity. The test results according to the specifications for potency, purity, identity, and absence of microbial contamination/endotoxins/mycoplasma are the basis for the release of the final product for further application by the qualified person. This IMP is not cryo-preserved and has a limited shelf-life; at least the sterility results are not available at the time of application and requiring a well-defined aseptic manufacturing process.

The seven participating study sites have been initiated in 2014 and the study will last for approximately 2 years after the inclusion of the first patient (in 2015) into the interventional study part. Since the phase I trial (29) and pilot studies (36) have shown that four repeated infusion cycles of *ex vivo* stimulated, autologous leukapheresis products lead to elevated basal NK cell activity in the peripheral blood of the patients, four treatment cycles will also be administered in the clinical phase II trial. The NK cell activity in the peripheral blood of the patients will be determined prior to study entry and start of adjuvant immunotherapy, every 3 months after enrollment for the first year and every 6-month thereafter to determine the biological activity. Multiparameter **Table 2 | Panel of CE-certified, fluorescence-labeled antibodies (Beckman Coulter), which are used in the clinical trial**.


7-AAD (A07704) was used as a viability dye.

flow cytometric analysis will be performed with the peripheral blood of the patients to determine the activation status of the NK and T cells using a pre-fixed panel of antibodies as indicated in **Table 2**. The cytotoxic response of patient derived NK cells before and after RCTx and after immunotherapy will be assessed by Europium cytotoxicity assays using K562 cells as a classical NK cell target and by measuring the density of NK/T cell activation markers such as CD94 and CD69 in the laboratory of Gabriele Multhoff at the TUM. Serum Hsp70 levels will be measured as mentioned above. In parallel to the blood analysis, tumor response assessment will be performed and centrally reviewed according to the immune related Response Criteria (irRC). Patients will be excluded when they show progressive disease according to irRC [increase of tumor burden more than 25% relative to nadir (minimum recorded tumor burden, confirmation by a repeat, consecutive assessment no <4 weeks from the date first documented)].

#### **CONCLUDING REMARKS**

This review aims to summarize results from bench to bedside experiments that resulted in the initiation of a phase II clinical trial using *ex vivo* activated NK cells in a targeted immunotherapy of NSCLC patients after RCTx bearing Hsp70 membranepositive tumors. Hsp70 has been found to serve as a biomarker for highly aggressive tumors and metastases (5, 6, 19). Preclinical data have shown that stimulation of NK cells with Hsp70 peptide TKD plus IL-2 results in an increased migratory capacity of NK cells and an enhanced killing of Hsp70 membrane-positive tumor cells *in vitro* (24), and in relevant tumor mouse models (28). An increased expression density of the C-type lectin receptor CD94 has been identified as a useful surrogate for the cytolytic activity of NK cells (25, 26), and granzyme B-mediated apoptosis was found to be responsible for the killing of tumor cells presenting Hsp70 on their cell surface (27). A phase I clinical trial has shown that re-infusion of *ex vivo* TKD/IL-2 stimulated autologous NK cells in patients with late-stage colorectal cancers and NSCLC is feasible, safe, and well tolerated (29). As stress, such as RCTx has been found to increase the cell surface density of Hsp70 selectively on tumor cells (21), a proof-of-concept phase II clinical trial was initiated in NSCLC patients stage IIIA/B after RCTx.

#### **ACKNOWLEDGMENTS**

The authors want to thank Anett Lange for excellent editorial assistance. The study was supported by DFG – Cluster of Excellence: Munich-Centre for Advanced Photonics (MAP), SFB824/2 (DFG), DFG – INST95/980-1FUGG, DFG – INST411/37-1FUGG, German Federal Ministry of Education and Research (BMBF, 0313909), m<sup>4</sup> – Cluster of Excellence: Personalized Medicine (BMBF, 16EX1021C, 16GW0030), Kompetenzverbund Strahlenforschung (02NUK038A), Innovative Therapies (01GU0823), EU-CELLEUROPE (315963), Wilhelm Sander-Stiftung (2012.078.1).

#### **REFERENCES**


tumor specific recognition structure for the cytolytic activity of autologous NK cells. *Haematologica* (2003) **88**:474–6.


mature results of the Spanish lung cancer group 0008 study. *Lung Cancer* (2013) **81**:84–90. doi:10.1016/j.lungcan.2013.03.009


36. Milani V, Stangl S, Issels R, Gehrmann M, Wagner B, Hube K, et al. Anti-tumor activity of patient-derived NK cells after cell-based immunotherapy – a case report. *J Transl Med* (2009) **7**:50. doi:10.1186/1479-5876-7-50

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

*Received: 18 February 2015; accepted: 25 March 2015; published online: 15 April 2015. Citation: Specht HM, Ahrens N, Blankenstein C, Duell T, Fietkau R, Gaipl US, Günther C, Gunther S, Habl G, Hautmann H, Hautmann M, Huber RM, Molls M, Offner R, Rödel C, Rödel F, Schütz M, Combs SE and Multhoff G (2015) Heat shock protein 70 (Hsp70) peptide activated Natural Killer (NK) cells for the treatment of patients with non-small cell lung cancer (NSCLC) after radiochemotherapy (RCTx) – from preclinical studies to a clinical phase II trial. Front. Immunol. 6:162. doi: 10.3389/fimmu.2015.00162*

*This article was submitted to NK Cell Biology, a section of the journal Frontiers in Immunology.*

*Copyright © 2015 Specht , Ahrens, Blankenstein, Duell, Fietkau, Gaipl, Günther, Gunther, Habl, Hautmann, Hautmann, Huber, Molls, Offner, Rödel, Rödel, Schütz, Combs and Multhoff. 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.*

# NK cell-mediated antibodydependent cellular cytotoxicity in cancer immunotherapy

*Wei Wang1 , Amy K. Erbe1 , Jacquelyn A. Hank1 , Zachary S. Morris1 and Paul M. Sondel1,2\**

*1Department of Human Oncology, University of Wisconsin-Madison, Madison, WI, USA, 2Department of Pediatrics, University of Wisconsin-Madison, Madison, WI, USA*

#### *Edited by:*

*Susana Larrucea, BioCruces Health Research Institute, Spain*

#### *Reviewed by:*

*John T. Vaage, Oslo University Hospital and University of Oslo, Norway Todd A. Fehniger, Washington University School of Medicine, USA*

#### *\*Correspondence:*

 *Paul M. Sondel, University of Wisconsin-Madison, Department of Human Oncology & Department of Pediatrics, 1111 Highland Avenue, 4159 WIMR Building, Madison, WI 53705, USA pmsondel@humonc.wisc.edu*

#### *Specialty section:*

*This article was submitted to NK Cell Biology, a section of the journal Frontiers in Immunology*

> *Received: 03 June 2015 Accepted: 06 July 2015 Published: 27 July 2015*

#### *Citation:*

*Wang W, Erbe AK, Hank JA, Morris ZS and Sondel PM (2015) NK cell-mediated antibody-dependent cellular cytotoxicity in cancer immunotherapy. Front. Immunol. 6:368. doi: 10.3389/fimmu.2015.00368*

Natural killer (NK) cells play a major role in cancer immunotherapies that involve tumor-antigen targeting by monoclonal antibodies (mAbs). NK cells express a variety of activating and inhibitory receptors that serve to regulate the function and activity of the cells. In the context of targeting cells, NK cells can be "specifically activated" through certain Fc receptors that are expressed on their cell surface. NK cells can express FcγRIIIA and/ or FcγRIIC, which can bind to the Fc portion of immunoglobulins, transmitting activating signals within NK cells. Once activated through Fc receptors by antibodies bound to target cells, NK cells are able to lyse target cells without priming, and secrete cytokines like interferon gamma to recruit adaptive immune cells. This antibody-dependent cell-mediated cytotoxicity (ADCC) of tumor cells is utilized in the treatment of various cancers overexpressing unique antigens, such as neuroblastoma, breast cancer, B cell lymphoma, and others. NK cells also express a family of receptors called killer immunoglobulin-like receptors (KIRs), which regulate the function and response of NK cells toward target cells through their interaction with their cognate ligands that are expressed on tumor cells. Genetic polymorphisms in KIR and KIR-ligands, as well as FcγRs may influence NK cell responsiveness in conjunction with mAb immunotherapies. This review focuses on current therapeutic mAbs, different strategies to augment the anti-tumor efficacy of ADCC, and genotypic factors that may influence patient responses to antibody-dependent immunotherapies.

Keywords: natural killer cell, therapeutic monoclonal antibody, antibody-dependent cellular cytotoxicity, cancer, immunotherapy

#### Introduction

Natural killer (NK) cells have been described throughout the literature for their ability to kill virally infected and malignant cells without priming. Unlike B and T cells, NK cells do not require somatic gene rearrangements to produce highly specific receptors that recognize target cells (1, 2). Instead, mature NK cells reserve large amounts of cytotoxic granules containing perforin and granzymes, as well as the mRNA of IFNγ that is ready for translation if stimulated. As soon as the balance between inhibitory and activating signals within NK cells are skewed toward activation, NK cells are capable of forming synapses with target cells, allowing the release of the perforin and granzyme to lyse the target cells, as well as for IFNγ production (3). In addition, NK cells can initiate the transduction of death signals within target cells through death receptor/ligand ligation (4). Their capabilities of tumor cytotoxicity and inflammatory cytokine production enable NK cells to play an important role in different settings of cancer immunotherapy.

#### NK Cell Recognition and "Missing-Self" Hypothesis

Mature NK cells express a series of transmembrane receptors. The activating receptors allow them to recognize stress-induced ligands, while their inhibitory receptors prevent them from attacking normal cells. Activating receptors are often associated with adaptor proteins that have activation motifs in their cytoplasmic domains. Upon ligand binding, followed by phosphorylation, these activating receptors can activate down-stream kinases, leading to NK cell degranulation and cytokine secretion (5). Inhibitory receptors, on the other hand, have one or more inhibitory motifs in their cytoplasmic tails. Once phosphorylated, they recruit phosphatases and deactivate signaling kinases, resulting in NK cell inhibition (6). NK cell activity is tightly regulated through the balance between inhibitory and activating signals transduced by these receptors.

One family of receptors on human NK cells is that of the killer immunoglobulin-like receptors (KIR), which recognize HLA as their ligands (7). Some inhibitory and activating KIRs share the same ligand (8); however, most inhibitory KIRs have stronger binding affinity to their shared ligand (9, 10). One way to shift the activating-inhibitory balance toward NK cell activation is by decreasing the inhibitory KIR signaling. When cells are under stress or virally infected, the HLA expression on their surface is often downregulated in order to escape from T cell recognition. When an NK cell encounters these "target" cells, an immune mediated synapse can occur. If the target cell is missing the expression of the HLA ligand for the inhibitory receptors on that NK cell, and expresses ligands for the activating receptors, the interaction could lead to NK cell activation due to lack of inhibitory signals ("missing-self " hypothesis) (11, 12).

However, within an individual, not all NK cells have the same receptor expression profile, and each individual has a different KIR expression profile. As a result, not all NK cells express inhibitory KIR receptors (13, 14). The NK cells without any inhibitory KIR receptors have been shown to be hyporesponsive to HLA-null targets compared to NK cells expressing inhibitory KIR receptors that can recognize self-HLA molecules (15). NK cells that express no inhibitory KIR receptors, or KIR receptors that only recognize allogeneic HLA (i.e., which do not recognize or bind self-HLA) do not go through a "licensing" process during NK cell differentiation. Licensing plays a role in enabling the licensed NK cells to be more capable of killing targets that do not express inhibitory HLA ligands (such as HLA-null targets). Unlicensed NK cells are less potent in their activation by, or killing of, HLA-null targets, than are licensed NK cells (16). Interestingly, in mouse models, such unlicensed (hyporesponsive) NK cells, which express Ly49 receptors (the inhibitory receptors found on mouse NK cells that are the functional counterparts of human inhibitory KIRs) but have not seen "self-ligand," can be made to be functional by transferring these NK cells to an environment where cognate ligand is expressed (17, 18).

Besides the expression of HLA ligands for the inhibitory receptors on NK cells, there are other mechanisms by which self-normal cells are protected from being attacked by NK cells. Studies in both mice (19–23) and humans (24, 25) have shown that after continuous exposure to the ligand for NK-activating KIR receptors, NK cells expressing these activating receptors can become hyporesponsive to target cells that express the ligand. This is evidence that the NK cells can be desensitized through their continual receptor contact with activating ligands.

#### ADCC Mechanism

In the setting of tumor-targeting monoclonal antibody (mAb) therapies, the anti-tumor efficacy of many mAb are shown to be NK cell-dependent (26). Human NK cells can express both FcγRIIC/CD32c (27) and FcγRIIIA/CD16a (28), which bind to the Fc portion of human immunoglobulins. FcγRIIIA often associates with FcϵRI-γ chains or CD3-ζ chains within the cell membrane, or with a heterodimer of these two chains (5). Both FcϵRI-γ and CD3-ζ chains have immune tyrosine-based activating motifs (ITAM) in their cytoplasmic tails. Unlike most activating receptors on NK cells, FcγRIIC has an ITAM in its own cytoplasmic tail. Upon FcγR binding, these ITAMs are phosphorylated, and through signal transduction mechanisms (binding to tyrosine kinases ZAP-70 and Syk and activation of PI3K, NF-κb and ERK pathways) NK cell degranulation, cytokine secretion, and finally tumor cell lysis occur (29).

Antibody-dependent NK-mediated tumor killing occurs through several different pathways, including: (1) exocytosis of cytotoxic granules; (2) TNF family death receptors signaling; (3) pro-inflammatory cytokine release, such as IFNγ. Both the uptake of perforin and granzymes by target cells and TNF family death receptor signaling cause target cell apoptosis (29), while IFNγ released by NK cells activate nearby immune cells to promote antigen presentation and adaptive immune responses (30). IFNγ production and cytotoxicity have been considered two distinctive functions of different NK subsets (31, 32), but growing evidence shows that the main cytotoxic NK subset, CD56dimCD16+ NK cells, that are responsible for mAb-mediated tumor killing, are also able to produce IFNγ following activation (33, 34). In addition to inhibiting cell proliferation, angiogenesis, and increasing MHC surface expression (35), IFNγ was also shown to contribute to upregulation of TRAIL expression on NK cells (36), which suggests that one mechanism may interact with another to synergistically enhance tumor killing. A recent study indicates that NK-insensitive targets can become NK-sensitive via treatment with IFNγ, which induces lysis through ICAM-1 upregulation and increasing conjugate formation with NK cells (37). These mechanisms might work together to eliminate tumor targets through engagement of both innate and adaptive immunity; whether one is predominant over the others in tumor killing is still unknown (29–37).

#### NK Cell Fc**γ**Rs in ADCC

Genotypic variations (polymorphisms) exist in humans in both FcγRIIIA and FcγRIIC that influence FcR function. Thus, FcγRIIIA and FcγRIIC genotype can influence the interaction of these receptors with immunoglobulin, resulting in differential effectiveness of mAb therapy depending on an individual's genotype. In addition, immunoglobulin isotypes (IgG1, IgG2, IgG3, and IgG4), as well as fucosylation and glucosylation patterns, have varying influence on the affinities of these IgG molecules for both FcγRIIIA and FcγRIIC (38–41). Such factors (patient FcγR genotype and antibody Fc backbone) create the opportunity for considering therapeutic treatment options that may optimize the degree to which a patient will respond when administering mAb therapy. Selection of an optimized regimen may lead to a more effective trigger of the proper immune response.

A number of studies have shown that anti-tumor activity of certain tumor-specific mAbs is associated with higher affinity FcRs, based on the FcR genotype. These results suggest that these mAbs are acting through antibody-dependent cell-mediated cytotoxicity (ADCC) by the cells that express those FcRs. In particular, FcγRIIIA expressed on NK cells has a single nucleotide polymorphism (SNP) that results in FcγRIIIA polymorphic variants [FcγRIIIA with phenylalanine (F) at amino acid position 158, or FcγRIIIA with valine (V) at amino acid 158], which vary in their strength of binding to immunoglobulins. Several studies have shown that those individuals that are homozygous for V (FcγRIIIA-158-V/V) have improved clinical outcome (after treatment with ADCC-inducing tumor-reactive mAb) over those that are either heterozygous (FcγRIIIA-158-V/F) or homozygous (FcγRIIIA-158-F/F) for the lower affinity FcγRIIIA isoform (42–46). While not all such studies confirm these findings, several clinical analyses are consistent with this result. These are presented in the section below regarding a few of the mAbs that appear to be acting, at least in part, via ADCC.

### Therapeutic Monoclonal Antibodies for Cancer Treatment

Tumor-specific mAbs that recognize tumor-selective antigens on the surface of tumor cells are being used as cancer therapy. These therapeutic mAbs target and attack tumor cells through various mechanisms, including directing toxic molecules to target cells, inhibiting target cell proliferation, blocking inhibitory signals for immune cells, and directing immune cells to kill targets through ADCC (47). Some newer antibodies, such as bi-specific antibodies (bsAbs), work through promoting conjugate formation between target cells and NK or T cells (48). A more comprehensive list of anti-cancer mAbs has been summarized in other reviews (47, 49, 50). This review focuses on NK cell-mediated anti-tumor activity, via ADCC, and thus includes discussion of representative tumorspecific mAbs that have been shown in pre-clinical or clinical models to function, at least in part, via ADCC (**Table 1**).

#### Anti-GD2 mAb for Melanoma and Neuroblastoma Treatment

GD2 is a disialoganglioside expressed on human melanoma and neuroblastoma cells with restricted expression on normal tissues, which makes it a suitable target for mAb immunotherapy (**Table 1**). The first anti-GD2 antibody, 3F8, is a murine IgG3 mAb that was produced in 1985 from a mouse hybridoma (62). 3F8 is able to elicit complement activation and ADCC against human neuroblastoma cells. However, most patients in early clinical trials that received 3F8 developed human anti-mouse antibody (HAMA) response (63). HAMA may compete for the binding site of the therapeutic antibodies resulting in decreased binding to GD2, therefore dampening the anti-tumor efficacy and leading to acceleration of clearance of the therapeutic antibody from circulation (64).

Another murine anti-GD2 mAb, 14.18, which is also an IgG3, was generated separately by Mujoo et al. in 1987 (65). This antibody also displayed the capability to induce efficient *in vitro* ADCC and *in vivo* anti-tumor effects. An isotype variant of this murine anti-human GD2 antibody, 14.G2a (66), was tested clinically and showed some anti-tumor activity (67, 68), but HAMA response was still present in a significant portion of patients. While effective in targeting tumor and reducing tumor size in occasional patients, it became evident that it was necessary to improve the backbone of these initial mAb to increase efficacy and decrease the immunogenicity of this immunotherapeutic option.

In order to reduce the HAMA response and lengthen the antibody half-life in patients, efforts were made to create chimeric anti-GD2 antibodies, containing human constant regions with murine variable regions. Since a chimeric antibody has a majority of human epitopes, these epitopes should not be recognized by the immune system as foreign, and thus be less immunogenic than the fully murine antibodies. Dinituximab (formerly known as ch14.18) is a chimeric mAb comprising a fusion protein of the human constant portion of IgG1 and the GD2-reactive variable portion of the murine 14.18 mAb (69). Dinituximab has been shown to induce stronger ADCC than 14.G2a *in vitro* against GD2-positive neuroblastoma cells (70), and have anti-tumor activity against GD2-positive melanoma cells *in vivo* (71). In the initial published phase I clinical study of dinituximab treatment for pediatric neuroblastoma (72), no human anti-chimeric antibody (HACA) response was detected. Four out of nine children had anti-tumor response and one had a minor response. Thus, by modifying the backbone of the antibody, improved clinical outcome was observed.

To further improve antibodies, a fully human antibody was "grafted" with murine complementarity determining regions (CDRs), which confer antigen specificity. These humanized antibodies are considered less immunogenic than chimeric antibodies (73). However, even with humanized antibodies specific for GD2, pain and capillary leak were seen as significant toxicities. These toxicities limit the dose that can be administered, which restrains the possible anti-tumor effect that one would expect if a higher dose could be given. The toxicities are mainly attributed to complement activation (74), which is elicited by the CH2 domain on antibodies (75). Therefore, by reducing complement activation via a point mutation at amino acid position 322 in the CH2 domain of humanized antibody, complement activation is greatly reduced. Such reduction in complement activation, and thus reduced toxicities (76), allowed for higher treatment-dose to be administered to patients, while at the same time maintaining the anti-tumor ADCC effect (77). Both humanized 14.18K322A and humanized 3F8 are under clinical investigation (**Table 1**) (73, 78).



*a ALT-803 is a fusion protein consisting of mutated IL15 and IL15R*α*/Fc complex.*

*bRLI (IL15R*α*-linker-IL15) is a fusion protein linking the NH2-terminal domains of IL15R*α *to IL15 through a 20-amino acid linker.*

*c c.60C3 is a chimeric anti-GD2 mAb.*

*This table is a selected (not complete) list of therapeutic mAbs that are capable of inducing antibody-dependent cellular cytotoxicity.*

#### Herceptin/Trastuzumab

Trastuzumab is a humanized anti-HER2 mAb used to treat HER2-positive breast carcinoma (**Table 1**), as well as many other types of cancers that overexpress HER2, a member of the human epidermal growth factor receptor (EGFR) family. HER2 is a transmembrane tyrosine kinase with no known ligand. Dimerization of HER2 with certain EGFR family members leads to activation of signaling pathways that promote cell proliferation and survival (79). HER2 is overexpressed on a variety of tumors with limited expression on normal tissues, thus it is an ideal target for treatment of HER2-positive cancers.

Trastuzumab was first approved by the FDA in 1998 to treat HER2-positive metastatic breast cancer. Besides preventing HER2 from dimerization, trastuzumab was also shown to mediate ADCC against HER2-positive tumor cells *in vitro*, and the major effector cells were NK cells expressing FcγRIIIA (80, 81). A mutant trastuzumab that lost the ability to bind to FcγR lost anti-tumor activity *in vivo* in a xenograft breast tumor model, suggesting that ADCC is involved in the anti-tumor effect of anti-HER2 mAb therapy *in vivo* (54). In addition, Clynes et al. showed that the anti-tumor response to trastuzumab in a breast carcinoma xenograft mouse model was decreased in mice lacking the activating receptor FcγRIIIA, but enhanced in mice lacking the inhibitory receptor FcγRIIB (82). These experimental data demonstrate that Fc receptor recognition is responsible for at least part of anti-tumor efficacy of trastuzumab.

#### Cetuximab

Cetuximab is an FDA-approved chimeric mAb for treatment of EGFR-expressing metastatic colorectal cancer (mCRC) (55), metastatic non-small cell lung cancer, and head and neck cancer. It reacts against the human EGFR, and can interfere with tumor growth via receptor blockade from growth factor activation. *In vitro* studies indicate that some of the anti-tumor activity of cetuximab is mediated via ADCC (83, 84), and cetuximab-mediated *in vitro* ADCC is correlated with NK cell FcγR polymorphisms of the effector donors (46, 85). In addition, in mCRC patients treated with cetuximab and irinotecan, those who have higher affinity FcγR polymorphisms had longer progression-free survival (45, 86). In other studies of mCRC patients, the opposite association has been found, namely low affinity FcγR polymorphism was associated with better clinical outcomes (87–89). However, the patients in these studies either had different percentage of KRAS mutation or received other antibody concurrently with cetuximab, which indicates patient mutation profile as well as treatment regimen could also influence the impact of FcγR on cetuximab response (90). Nevertheless, both *in vitro* and some clinical data suggest that cetuximab-mediated ADCC through NK cells may contribute to its anti-tumor activity.

#### Rituximab and Obinutuzumab

Rituximab is a chimeric IgG1 mAb targeting CD20, a B cell differentiation antigen (**Table 1**). There are various mechanisms that may account for the anti-tumor effect of rituximab: complement-dependent cytotoxicity (CDC), direct target cell apoptosis, antibody-dependent phagocytosis and ADCC (91). In xenograft mouse models of B-cell lymphoma, the anti-tumor effect of rituximab was greatly reduced in FcRγ−/<sup>−</sup> nude mice (82), or in mice treated with FcγR block (92). Interestingly, clinical evidence gathered by Cartron et al. suggested that FcγRIIIA on NK cells plays an important role in anti-tumor effect of rituximab. Cartron et al. evaluated FcγRIIIA polymorphisms in follicular lymphoma patients treated with rituximab, and this was the first time that better response to rituximab was associated with higher affinity FcγRIIIA genotype (42). However, there have been mixed results as to the influence of FcR genotype on clinical response to rituximab. Some groups have found in B cell lymphoma patients with no association between FcγR polymorphism and outcome (93–95). There are various factors that could contribute to the differences in these findings, including patient population, genotyping methodology, rituximab treatment strategy, and whether or not patients received concurrent chemotherapy (96–99).

Modifications have been made to improve the binding affinity of therapeutic anti-CD20 mAb to activating Fc receptors, in hopes of maximizing the ADCC function of this mAb therapy. One such effort consists of using glycoengineered antibodies, which are produced in CHO cells that overexpress β-1,4-Nacetyl-glucosaminyltransferase III and Golgi α-mannosidase II, and will increase the binding affinity to both the higher affinity and to the lower affinity isoforms of FcγRIIIA (39, 41, 100). The first Fc-glycoengineered anti-CD20 humanized mAb, obinutuzumab, was shown to induce stronger ADCC and direct target cell death *in vitro*, and it also elicited better anti-tumor activity in a lymphoma xenograft mouse model compared to rituximab (101, 102). In a phase III clinical trial comparing obinutuzumab vs. rituximab, combined with chemotherapy, for treating chronic lymphocytic leukemia (CLL) patients (52), obinutuzumab plus chlorambucil significantly prolonged progression-free survival compared to rituximab plus chlorambucil. Obinutuzumab alone, or in combination with chemotherapy, is currently under clinical investigations for other B-cell malignancies as well (103–106).

#### How to Augment Anti-Tumor Effects of ADCC

Since ADCC is an important contributor to the anti-tumor activity of many mAb therapies, enhanced immune activation of the effector cells may be an ideal adjuvant therapy to augment ADCC activity of mAb. In addition to the ADCC capabilities of NK cells, they can also stimulate the activity of other immune processes through their release of cytokines (such as IFNγ), and thus can provide a link to initiate additional immune responses to attack target tumors.

#### mAb **+** Radiation Therapy

Approximately 60% of oncology patients receive radiation therapy as part of their cancer treatment. Radiation elicits an anti-tumor effect through the induction of DNA damage but may also increase tumor susceptibility to immune response (107). Consequently, the effect of radiation may be modulated by immune response (108, 109) and radiation may augment the efficacy of immunotherapies (110). The mechanisms by which radiation may interact with the immune system include radiation-induced production of inflammatory cytokines, release of tumor-specific antigens, phenotypic changes in tumor cell expression of immune susceptibility markers, and effects on vascular architecture that enhance immune surveillance (107, 110). Included among the phenotypic changes induced by radiation are the upregulation of MHC class I, NKG2D ligand, and the Fas death receptor; all of which may potentiate the ADCC response (111–114). In addition, radiation may impact the expression of antigens targeted by tumor-specific antibodies and this has been show to enhanced ADCC response *in vitro* (115, 116). The potential interaction of radiation and ADCC has not yet been clarified *in vivo*. Interestingly, however, a number of tumor-specific antibodies that are known to elicit ADCC (including cetuximab, trastuzumab, and dinituximab) are commonly administered to patients that also receive radiation therapy. The role of ADCC and NK cells in the clinical response to such combined modality treatment has not been defined. Further pre-clinical investigation is needed to evaluate whether the effects of radiation on tumor immune susceptibility may be leveraged to enhance ADCC response in the clinical setting.

#### mAb **+** Matrix Metalloproteases Inhibitor

Upon activation, NK cells have been shown to shed FcγRIIIA (also known as CD16) from their cell membrane, a process that is mediated by matrix metalloproteases (MMPs) activation (117, 118). *In vitro* treatment of NK cells with the MMP inhibitor, GM6001, rescues FcγRIIIA loss stimulated by K562 tumor cells, but does not interfere with NK cell degranulation (118), which indicates that FcγRIIIA shedding and degranulation of NK cells are dependent upon separate pathways. It is plausible, therefore, that maintaining FcγRIIIA expression on NK cell surface via an MMP inhibitor could enhance the ADCC function of NK cells without interrupting NK cell degranulation (119). Recently, one specific MMP, ADAM17, was identified by Romee et al. as the key MMP responsible for FcγRIIIA expression loss after NK cell activation (51). According to their report, an ADAM17-specific inhibitor not only rescues FcγRIIIA shedding stimulated by tumor targets but also improves NK cell degranulation as well as IFNγ production in the presence of tumor-specific mAb (51). These *in vitro* studies suggest that limiting the loss of FcγRIIIA on the cell surface is important for enhanced NK cell-mediated ADCC, and opens new possibilities for combination of both MMP and mAb therapies to further promote ADCC.

#### Anti-GD2 mAb **+** IL2 **+** GM-CSF

While anti-GD2 mAb alone produced some anti-tumor activity in neuroblastoma patients, combining the mAb with GM-CSF and interleukin 2 (IL2) to further activate immune effector cells enabled potent clinical anti-tumor efficacy of anti-GD2 mAb. Our lab previously showed that peripheral blood mononuclear cells (PBMCs) obtained from cancer patients pre-IL2 treatment had low levels of ADCC against a neuroblastoma cell line, even in the presence of both anti-GD2 mAb and IL2. However, PBMCs obtained from the same patient following 4 weeks of IL2 infusions mediated much higher ADCC of neuroblastoma cells, and the addition of IL2 *in vitro* dramatically boosted the anti-GD2 mAbmediated ADCC (120, 121). Moreover, depletion of FcγRIIIApositive cells eliminated ADCC completely (120). These data suggest that *in vivo* infusion of IL2 could overcome the immune suppression seen in some cancer patients; when using NK cells obtained from patients following therapy with IL2, the *in vitro* combination of IL2 and anti-GD2 mAb greatly boosted ADCC of neuroblastoma cells.

Besides IL2, GM-CSF also acts as an immune stimulator, especially following immune suppressive chemotherapy, to rescue bone marrow myeloid function (122–125). The combination of GM-CSF and the murine anti-GD2 mAb, 3F8, resulted in complete responses of GD2-positive cancer patients, particularly in patients with minimal residual disease (MRD) (126). In a phase III clinical trial for high-risk neuroblastoma patients following induction therapy and autologous transplant, patients were randomized into standard therapy (isotretinoin) or into a group that received the combined immunotherapy regimen of dinituximab + IL2 + GM-CSF in addition to isotretinoin. Patients in the immunotherapy treatment group had significantly improved event-free survival and overall survival as compared to the patients that received the standard of care alone (53). Other mAb therapies, such as rituximab and trastuzumab (81), are also being combined with IL2 in clinical trials to evaluate the antitumor efficacy (127–129).

#### mAb **+** IL15

IL15 is another cytokine that can activate immune cells, such as NK and T cells. IL15 receptors (IL15R) share the same β chain and common γ chain as IL2 receptors, but IL15R have a unique α chain that is specific for IL15, which is used for IL15 presentation to β/γ receptors on immune cells. Soluble IL15 has been shown to activate NK cells, enhance NK-mediated cytotoxicity, and cytokine production *in vitro* (130), and administration of recombinant human IL15 (rhIL15) to cancer patients resulted in *in vivo* NK cell proliferation and activation (131). While IL2 has the potential for T regulatory cell (Treg) maintenance, which could dampen activating immune response, IL15 does not stimulate Tregs (132). Moreover, IL15 does not trigger T cell death after activation, or lead to vascular leak syndrome (VLS), both of which are significant toxicities seen in pre-clinical models using IL2 (133–135). Therefore, in addition to IL2, IL15 is a potential candidate that can be used clinically combined with mAb, to improve NK cell-mediated ADCC without severe toxicity.

Indeed, the Caligiuri group reported enhanced NK-mediated ADCC by IL15 in 1994 (130). More recently, Moga et al. showed that *in vitro* activation of PBMCs by IL15 significantly improved rituximab-mediated ADCC against B lymphoma cells, and the major effector cells were NK cells (136). They also showed this enhancement of rituximab-mediated ADCC using PMBCs from CLL patients against a CLL cell line (137). In addition, they showed that IL15 treatment resulted in similar ADCC levels between individuals with the lower affinity FcγRIIIA and individuals that have the higher affinity FcγRIIIA (137). These findings suggested that the difference between the capabilities of the high and low affinity FcγRIIIA to mediate ADCC might be overcome by IL15 treatment. Such findings suggest that IL15 administration may compensate for mAb-affinity differences that are dependent upon FcγRIIIA polymorphisms.

Different from IL2, the activation of NK and T cells by IL15 is through trans-presentation by the IL15Rα chain, which is expressed independent of the β/γ chains. With efforts to increase the stimulation activity of soluble IL15, several different IL15 agonists have been generated and tested in pre-clinical models in order to boost NK-mediated ADCC efficacy. A fusion protein linking IL15Rα chain sushi domain and human IL15 (RLI) has been generated and shown superior stimulation potential *in vitro* and better anti-tumor effects *in vivo* than soluble IL15 (138, 139). In addition, an IL15 mutant has been identified as a better surrogate for soluble IL15 due to its enhanced activity to stimulate proliferation of cells expressing IL15R (140). Later on, the same group generated a fusion protein consisting of the IL15 mutant and an IL15Rα/Fc complex (141). This IL15 superagonist, ALT-803, could improve rituximab-mediated ADCC against B cell lymphoma both *in vitro* and *in vivo* [Maximillian (142)]. Thus, IL15 combined with tumor-specific mAb appears to merit clinical testing as a potential immunotherapy, which is currently underway (NCT02384954) (**Table 1**).

# Immunocytokines

#### Anti-GD2 Immunocytokines

Immunocytokines (IC) are fusion proteins made by linking immune-activating cytokine to a tumor-specific mAb. The initially described IC (ch14.18-IL2) linked IL2 to the C-terminus of the ch14.18 chimeric anti-GD2 mAb (143). The IL2 component on these IC has the same ability as soluble IL2 in terms of stimulating cell proliferation via IL2 receptors (IL2R) (144), and have been shown to mediate the conjugation between IL2R-positive cells and tumor target cells (145). IC also preserve FcR binding ability, and thus are capable of mediating *in vitro* ADCC (144). By targeting IL2 to the tumor microenvironment, IC might be superior at activating tumor infiltrating immune cells resulting in improved ADCC while causing less toxicity than soluble IL2. Lode et al. showed that ch14.18-IL2 had improved anti-tumor efficacy than the combination of ch14.18 and soluble IL2 in a spontaneous neuroblastoma metastases mouse model (146). They also showed that the mechanism behind this anti-tumor effect strongly depended on NK cells, since NK cell depletion completely abrogated the anti-tumor effect in this model (147).

A separate humanized IC was created by linking human IL2 to the humanized 14.18 mAb (hu14.18-IL2; **Table 1**). This IC was assessed for anti-tumor activity in a phase II clinical trial of refractory neuroblastoma patients divided into two strata of patients depending on disease burden. Stratum 1 included patients with disease measurable by computed tomography and/ or magnetic resonance imaging using standard radiographic criteria, while stratum 2 included patients with disease evaluable only by 123I-MIBG scintigraphy and/or BM histology. Of 36 Wang et al. NK cells in cancer immunotherapy

patients evaluable for response, no responses were seen in the 13 patients in stratum 1 while 5 complete responses were noted in the 23 patients in stratum II (59). In another phase II clinical trial of hu14.18-IL2 in 14 patients with measurable metastatic melanoma, 1 patient had a partial response to the immunotherapy (148). Both of these trials suggest that hu14.18-IL2 works better in clinical cases where MRD is present than in individuals that have bulky disease. Furthermore, in the phase II neuroblastoma study of hu14.18-IL2 noted above where 21.7% of stratum II patients with MRD responded, all of the responders had a favorable KIR/ KIR-ligand genotype (i.e., KIR-ligand missing) (58). This study indicates that NK cells play an important role in response to the hu14.18-IL2 treatment in neuroblastoma patients.

#### Anti-PSMA Immunocytokine

Prostate-specific membrane antigen (PSMA) is a surface antigen highly expressed by poorly differentiated prostate cancers, which makes it a suitable target for mAb therapies. Several anti-PSMA mAb and antibody-drug conjugates have been tested clinically, and a small fraction of patients showed anti-tumor response, which in part was due to ADCC (149). To improve ADCC function of anti-PSMA mAb, an anti-PSMA IC was generated by fusing IL2 to a mouse/human chimeric anti-PSMA mAb (60). This IC showed enhanced *in vitro* ADCC, and superior *in vivo* anti-tumor activity compared to a naked anti-PMSA mAb (60). Continued research is underway in both pre-clinical models and clinical testing to determine if superior anti-tumor efficacy is noted with this novel IC.

#### IL15-Linked Immunocytokine

One limitation of IL2-linked ICs is IL2-mediated toxicity that restricts the maximum-tolerated dose (MTD) that can be administered. In contrast to IL2, IL15 does not appear to elicit as severe adverse side effects. In addition, IL15 may have improved anti-tumor efficacy (133). Since IL15 is most active in a transpresentation form, a fusion protein of human IL15Rα and human IL15, called RLI, was generated. This fusion protein showed superior activation function than soluble IL15 both *in vitro* and *in vivo* (138, 150). Vincent et al. linked RLI to the end of chimeric anti-GD2 antibody (c.60C3) and generated an RLI-linked anti-GD2 IC (57). This IC had similar IL15-induced proliferative activity as RLI, and similar ADCC-inducing ability as anti-GD2 antibody *in vitro*. Moreover, anti-GD2-RLI exhibited better anti-tumor activity than either RLI or anti-GD2, alone or in combination, in an NXS2 mouse neuroblastoma model (57). The same group also showed that RLI-linked rituximab significantly prolonged survival as compared to administration of RLI and rituximab at the same time in a residual lymphoma mouse model (56). These pre-clinical data suggest that IL15-linked IC merit clinical investigation to evaluate both efficacy and toxicity relative to IL2-linked IC.

#### Novel Bi-Specific Antibodies or Single Chain Variable Fragment Targeting NK Cells

With evolving genetic engineering technologies, non-conventional antibodies that have dual or tri specificity have been constructed. They either target two different antigens on tumor cells, or facilitate conjugate formation between immune cells and tumor targets (151). There are two classes of bsAbs: Fc-containing bsAbs that are of similar size as conventional mAbs and those bsAbs without an Fc domain, which are of much smaller size. They are further divided according to their structures, specificities, or how they are constructed (152). In efforts to bring NK cells and tumor cells in proximity, several bsAbs or single chain variable fragment (scFv) have been constructed to bind both tumor antigen and FcγRIIIA (CD16). AFM13 is a bi-specific tetravalent antibody construct that targets both CD30 and CD16 (153), which is currently in a phase II clinical trial for relapsed Hodgkin lymphoma patients (NCT02321592) (**Table 1**). It has been shown to exhibit superior cytotoxicity than other CD30-targeting antibodies *in vitro* and its ADCC activity is independent of the FcγRIIIA allotypes (154). A CD20/CD16 bsAb also showed improved clearance of B cell malignancy compared to rituximab both *in vitro* and *in vivo*. Again, the efficacy of this antibody construct is not influenced by FcγRIIIA polymorphism (61). Two different CD33/CD16 bsAbs have been shown to efficiently kill acute myeloid leukemia (AML) cells or stem cells from myelodysplastic syndrome, a precursor of AML (155–157). A few CD19/CD16 bsAb or derivatives have also been generated to target leukemia and lymphoma cells and showed promising ADCC activity *in vitro* (158–161).

For solid tumors, a bsAb-targeting HER2/CD16 has been tested in mouse model against HER2-positive tumor cells and showed enhanced anti-tumor efficacy than trastuzumab against HER2 low expressing tumor. Interestingly, the efficacy of this bsAb is also FcγRIIIA polymorphism independent (162). In addition, an EpCAM/CD16 bi-specific scFv antibody fragment showed enhanced *in vitro* killing of human carcinomas with a broad range of origins, as well as efficient killing of NK-resistant carcinoma targets (163). The fact that a lot of these bsAbs discussed above have proficient anti-tumor effect regardless of FcγRIIIA affinity on NK cells enlarges the patient population that will potentially benefit from such type of immunotherapy.

#### Adoptive Transfer of *Ex Vivo*-Activated NK Cells

A separate approach to augment NK cell function and ADCC is to infuse NK cells that have been activated and expanded in cell number *ex vivo.* During the *ex vivo* expansion of NK cells, these activated effectors are primed to kill tumors more effectively, and they can then be infused into cancer patients. Autologous NK cells may be relatively tolerant to self-tumors and, in certain settings, may have less anti-tumor potential than allogeneic NK cells (164). There are several approaches being pursued in order to expand and activate NK cells *ex vivo*. In the presence of IL2 and irradiated feeder cells, such as EBV-LCL cells or genetically modified K562 cells, NK cells are preferentially expanded to 200- to 400-fold within a 21-day period (165, 166). NK cells can be potently expanded and activated by culturing them with K562 cells that have been modified to be far more stimulatory by expressing membrane-bound IL15 and 41BBL. NK cells that have been *ex vivo* expanded using these modified K562 cells have better anti-tumor activity *in vitro* and *in vivo* in mouse models (166). In addition, *ex vivo*-expanded NK cells have also been shown to be able to mediate ADCC, and in combination with tumor antigen-specific antibodies, they exerted better anti-tumor efficacy (167, 168). Different protocols using expanded NK cells are currently under clinical testing, either alone or in combination with tumor-specific mAb (169–171). These ongoing clinical trials may answer the questions of how long these NK cells may survive *in vivo*, their homing capacity, and tumor-targeting specificity.

# Genotypic Factors that Affect Antibody/ NK Cell Based Immunotherapy

Despite the success noted in some immunotherapies involving NK cells, some patients do not benefit from these immunotherapies. Therefore, it would be advantageous to be able to identify those patients who would likely respond, and those who would be less likely to benefit from immunotherapies that require NK cell response. There are many different genotypic factors that could potentially affect the efficacy of different immunotherapies in various cancer types. Here, we focus mainly on those genotypes that could affect NK cell function and ADCC efficacy.

#### Killer-Immunoglobulin Receptors on NK Cells and Their Ligands

Killer-immunoglobulin receptors are expressed on a subset of NK cells and on some T cell subsets. KIRs are a family of receptors with high polymorphisms inherited on chromosome 19 (172). There are 15 functional KIR genes and 2 KIR pseudogenes in the human genome. Of the functional KIR genes, some function as inhibitory receptors and others are activating receptors (173). For some KIR genes, the ligands are well defined (7). KIRs with defined ligands generally recognize distinct HLA class I molecules, but do so with less specificity than T cell receptors. Furthermore, and again in distinction to T cell receptors, KIRs appear to recognize class I HLA molecules without regard to peptides that may be presented in the cleft (pocket) of the HLA structure. Some inhibitory and activating KIRs with high homology in extra cellular domains share the same HLA ligand, with the inhibitory KIR/ KIR-ligand interaction being a much stronger binding pair. The interaction between KIRs and KIR-ligands influence NK cell education and function (15, 174). Since KIR haplotype, as well as KIR expression, is highly diverse among the population, and since KIR-ligands are inherited independently from KIRs, an individual's KIR/KIR-ligand genotype can affect NK cell function and ADCC in different settings. These differences in genotypes for KIR/KIR-L between different individuals have been associated with the individual's outcome in response to certain forms of immunotherapy, particularly those immunotherapies that involve NK cells.

The KIR haplotype of NK cell donors (175–177) as well as the relationship between the donor's inhibitory KIR and the recipient's KIR-ligand genotypes can predict clinical outcomes in allogeneic hematopoietic transplantation for AML patients; the impact of KIR and KIR-L genotypes may also play a role in some, but not all, settings of hematopoietic transplantation for acute lymphoid leukemia (ALL) patients (178–180). Recipients that are missing the KIR-ligands for the KIR genes that are present on the donor NK cells are predicted to have longer progressionfree survival post-transplantation than those who have the KIR-ligand genes present for the donor KIR genes (178–180). In an autologous hematopoietic transplantation setting for high-risk neuroblastoma patients, Venstrom et al. showed that patients who were missing any self-inhibitory KIR-ligand had improved progression-free survival and improved overall survival (181). Even though direct ADCC was not involved in these transplants for cancer patients, these data involving KIR/KIR-L relationships indicate that NK cells likely contribute to the anti-tumor response post-transplantation.

In some immunotherapies where ADCC is involved, patients missing their self-inhibitory KIR-ligand also had better clinical outcome. In neuroblastoma patients treated with anti-GD2 mAb or IC, and in lymphoma patients treated with rituximab, inhibitory KIR-ligand missing was associated with improved clinical outcome (58, 181–183). These findings suggest that self KIR/ KIR-ligand genotypes not only affect NK cell function but also affect NK-mediated ADCC effects in the clinical setting.

The effects of inhibitory KIR/KIR-ligand interactions on NK cell function have been investigated for over a decade, but the role of activating KIR/KIR-ligand interactions in immunotherapies involving NK cells has only recently been reported. In 2011, Scquizzato et al. evaluated the impact of recipient KIR-ligand missing for both inhibitory and activating donor KIR genes, in a group of patients with various hematopoietic malignancies who received allogeneic hematopoietic stem cell transplantation (184). They found that the cohort of recipient patients missing the KIR-ligand for donor KIR genes had better disease-free survival, but only when patients who had KIR-ligands missing for their activating KIR were excluded from the analysis (184). This finding suggests that activating KIR/KIR-ligand interactions may have differential impact on patient outcome than inhibitory KIR/KIR-ligand interactions, thus the KIR-ligand missing model to predict patient outcome appears to be applicable only to inhibitory KIRs, with exclusion of activating KIRs.

In 2012, Venstrom et al. reported that in a large group of AML patients, those who received allografts from donors that had the activating KIR gene, KIR2DS1, and two copies of the ligand for KIR2DS1, HLA-C2, exhibited a higher relapse rate than patients who received grafts from KIR2DS1-positive donors and either one or no copies of HLA-C2 (24). Consistent with this clinical observation, Pittari et al. showed that KIR2DS1-positive NK cells from healthy donors that have two copies of HLA-C2 were hyporesponsive to tumor targets as compared to KIR2DS1 positive NK cells from donors with one or no copy of HLA-C2 (25). Both clinical and *in vitro* data suggest that NK cells that express activating KIR2DS1 are subjected to hyporesponsiveness via long-term contact of the activating KIR with its cognate ligand when two copies of the ligand are present. Whether or not NK cell hyporesponsiveness affects patient response in the context of ADCC, where mAb is administered to trigger NK cell activation through FcγRs, is still under investigation. As the interactions of activating KIRs and their ligands have not been studied as extensively as the interactions of the inhibitory KIRs and their ligands, much more characterization is required. Inhibitory KIR/ KIR-ligand genotypes and activating KIR/KIR-ligand interactions may also serve as potential predictors of clinical outcome in the future.

#### NK Cell Fc Receptor Genetic Variations

FcγRs function as receptors for the Fc portion of IgG immunoglobulins, and in doing so serve as a link of the innate immune system to the humoral system. In humans, there are three classes of FcγR, including variations of FcγRI, FcγRII, and FcγRIII. These three receptor classes are characteristically expressed on various immune cells. NK cells express both FcγRIIC and FcγRIIIA, which have low to intermediate affinity for IgGs (depending on FcR polymorphisms and IgG subclasses). As noted above, individuals with intact immune systems all express FcγRIIIA on most of their NK cells, less than half the population expresses FcγRIIC on their NK cells. Genetic variability exists within both FcγRIIC and FcγRIIIA, and this genotypic variation can vary the expression and avidity of these FcγRs for IgG molecules (185, 186).

As was mentioned above, several groups have found associations between SNP genotype of FcγRIII3A-158-V/F and patient response; those with the higher affinity V/V genotype respond better to some antibody therapies than those with V/F or F/F (42–46). In fact, FcγRIIIA-158-V has improved binding affinity for IgG1 subclasses as compared to FcγRIIIA-158-F, which is the most common IgG subclass used in mAb cancer immunotherapeutics. Interestingly, FcγRIIIA-158-V and FcγRIIIA-158-F have increased binding strength for IgG3 over IgG1 (187). This suggests that antibody IgG subclass variations may improve their interaction with NK cells, via improved engagement of FcγRIIIA, and thus improve tumor destruction. However, other FcRs (such as Fcwever) are expressed on other immune cells, such as neutrophils and monocytes/macrophages, that also play an important role in tumor killing, and have improved binding to IgG1 isotypes over IgG3 (188).

Another FcγR that is expressed on NK cells, that also has improved binding to IgG3 over IgG1, is FcγRIIC. FcγRIIC is the result of the unequal crossover of FcγRIIA (an activating FcR expressed on myeloid immune cell lineages) and FcγRIIB (an inhibitory FcR expressed on B cells, monocytes, and macrophages) (185). FcγRIIC also has SNPs within its nucleotide sequence that governs its expression (or non-expression) on the cell surface. SNP sites within exon 3 of FcγRIIC result in either an open reading frame, hence protein expression on the cell surface (FcγRIIC-ORF), or a stop codon, thus non-expression of FcγRIIC on the cell surface (FcγRIIC-STOP). As such, FcγRIIC is expressed in ~20–40% of the population, and co-expression of FcγRIIC with FcγRIIIA may result in enhanced ADCC capabilities of the NK cells (186).

Some of the FcγRs have been shown to be subject to copy number variability (CNV), and this CNV can result in variable expression of these FcRs on the cell surface. Both FcγRIIC and FcγRIIIA can be CNV, and CNV in these receptors correlates with differences in protein expression levels (189, 190), as well as increased ADCC function through enhanced NK cell activation (186, 190).

Given that the isotypes of FcγRIIIA have been shown to have differential binding affinity to IgG subclasses, that FcγRIIC expression on NK cells is variable within the population, and that these FcγRs can be CNV, using genotypic measures to pre-select patients based on their FcγR genotype for therapeutics that require NK cell may be of critical importance for future clinical investigations. Detailed simultaneous testing of polymorphisms and CNV in all three of these genes (FcγRIIA, FcγRIIC, and FcγRIIIA) has yet to be evaluated for associations with clinical outcome in clinical trials of ADCC-inducing tumor-reactive mAbs. However, it is likely that these factors, which should influence *in vivo* ADCC function, will be found to play a role in the clinical activity of ADCC-inducing mAb therapies.

# Concluding Remarks

Monoclonal antibodies utilize different mechanisms to destroy cancer cells, one of which is ADCC. As these treatments have continued to evolve from original mouse antibodies to chimeric antibodies to humanized and fully human antibodies, therapeutic mAbs for cancer treatment are still being engineered to achieve improved anti-tumor efficacy. Besides optimizing the characteristics of the mAb itself, there are other ways to improve anti-tumor effect, one of which is to improve ADCC by combining immune stimulatory therapies with the mAb. Since NK cells are considered a major player in mAb-mediated ADCC against tumor cells, reagents that can enhance NK cell activation, such as IL2, may be combined with mAb to improve the ADCC effect. However, due to IL2-associated toxicity in patients (191), novel IC therapeutics are being generated in an attempt to reduce toxicity and to allow for an increased MTD. Another cytokine that helps NK cell activation is IL15, and it is currently being tested clinically in various types of cancer. Besides cytokines or IC, newly discovered mechanisms that could potentially improve ADCC include MMP inhibitors to prevent FcR shedding, and ionizing radiation to make tumors more immunogenic and vulnerable to immune-mediated destruction. These approaches combined with tumor-specific mAb are still being explored in pre-clinical models to determine efficacy and optimize dosing regimens. Finally, genotypic profiles of NK cells also may contribute to our understanding of the magnitude of ADCC responses of NK cells in any given patient. Specifically, KIRs and FcR genotypes may help to predict clinical outcome to ADCC-inducing mAb therapy, allowing for more personalized treatment. More detailed analyses of associations of clinical outcome with NK cell genotype profiles are needed in order to determine the predictive value of this form of genotyping for distinct types of cancer and for different immunotherapies that involve NK cells.

## Acknowledgments

This work was supported, in part, by National Institutes of Health Grants CA032685, CA87025, CA166105, CA14520, a Stand Up To Cancer – St. Baldrick's Pediatric Dream Team Translational Research Grant (SU2C-AACR-DT1113), the University of Wisconsin-Madison Institute for Clinical and Translational Research Grant 1TL1RR025013-01, RSNA Research Resident Seed Grant, ASTRO Resident Seed Grant, and grants from the Midwest Athletes for Childhood Cancer Fund, The Crawdaddy Foundation, and Hyundai Hope on Wheels.

# References


antibody plus chemotherapy but not chemotherapy alone. *Haematologica* (2012) **97**:937–42. doi:10.3324/haematol.2011.050419


to upregulate antibody targets. *Br J Cancer* (2014) **110**:1472–80. doi:10.1038/ bjc.2014.79


transgenic mice. *Proc Natl Acad Sci U S A* (2000) **97**:11445–50. doi:10.1073/ pnas.200363097


in patients with relapsed or refractory Hodgkin lymphoma. *Blood* (2015) **125**(26):4024–31. doi:10.1182/blood-2014-12-614636


cancer patients: a phase I clinical study. *Immunotherapy* (2009) **1**:753–64. doi:10.2217/imt.09.47


monoclonal antibodies in cancer. *J Hematol Oncol* (2013) **6**:1. doi:10.1186/ 1756-8722-6-1


**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 Wang, Erbe, Hank, Morris and Sondel. 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.*

# Cetuximab reconstitutes pro-inflammatory cytokine secretions and tumor-infiltrating capabilities of sMICA-inhibited NK cells in HNSCC tumor spheroids

*Stephan Klöss1 \*, Nicole Chambron2 , Tanja Gardlowski1 , Sandra Weil3 , Joachim Koch3 , Ruth Esser1 , Elke Pogge von Strandmann4 , Michael A. Morgan5 , Lubomir Arseniev1 , Oliver Seitz2 and Ulrike Köhl1*

*<sup>1</sup> Institute of Cellular Therapeutics, Integrated Research and Treatment Center Transplantation (IFB-Tx), Hannover Medical School, Hannover, Germany, 2Department of Oral, Cranio-Maxillofacial and Facial Plastic Surgery, Klinikum Hanau GmbH, Hanau, Germany, 3Georg-Speyer-Haus Institute for Tumor Biology and Experimental Therapy, Frankfurt, Germany, 4Klinik I für Innere Medizin, Uniklinik Köln, Cologne, Germany, 5 Institute of Experimental Haematology, Hannover Medical School, Hannover, Germany*

Immunosuppressive factors, such as soluble major histocompatibility complex class I chain-related peptide A (sMICA) and transforming growth factor beta 1 (TGF-β1), are involved in tumor immune escape mechanisms (TIEMs) exhibited by head and neck squamous cell carcinomas (HNSCCs) and may represent opportunities for therapeutic intervention. In order to overcome TIEMs, we investigated the antibody-dependent cellular cytotoxicity (ADCC), cytokine release and retargeted tumor infiltration of sMICA-inhibited patient NK cells expressing Fcγ receptor IIIa (FcγRIIIa, CD16a) in the presence of cetuximab, an anti-epidermal growth factor receptor (HER1) monoclonal antibody (mAb). Compared to healthy controls, relapsed HNSCC patients (*n* = 5), not currently in treatment revealed decreased levels of circulating regulatory NK cell subsets in relation to increased cytotoxic NK cell subpopulations. Elevated sMICA and TGF-β1 plasma levels correlated with diminished TNFα and IFN-γ release and decreased NKG2D (natural killer group 2 member D)-dependent killing of HNSCC cells by NK cells. Incubation of IL-2 activated patient NK cells with patient plasma containing elevated sMICA or sMICA analogs (shed MICA and recombinant MICA) significantly impaired NKG2D-mediated killing by down-regulation of NKG2D surface expression. Of note, CD16 surface expression levels, pro-apoptotic and activation markers, and viability of patient and healthy donor NK cell subpopulations were not affected by this treatment. Accordingly, cetuximab restored killing activity of sMICA-inhibited patient NK cells against cetuximab-coated primary HNSCC cells via ADCC in a dose-dependent manner. Rapid reconstitution of anti-tumor recognition and enhanced tumor infiltration of treated NK cells was monitored by 24 h co-incubation of HNSCC tumor spheroids with cetuximab (1 μg/ml) and was

#### *Edited by:*

*Susana Larrucea, BioCruces Health Research Institute, Spain*

#### *Reviewed by:*

*Francisco Borrego, BioCruces Health Research Institute and Cruces University Hospital, Spain Frank Momburg, Deutsches Krebsforschungszentrum, Germany*

> *\*Correspondence: Stephan Klöss kloess.stephan@mh-hannover.de*

#### *Specialty section:*

*This article was submitted to NK Cell Biology, a section of the journal Frontiers in Immunology*

> *Received: 15 July 2015 Accepted: 09 October 2015 Published: 02 November 2015*

#### *Citation:*

*Klöss S, Chambron N, Gardlowski T, Weil S, Koch J, Esser R, Pogge von Strandmann E, Morgan MA, Arseniev L, Seitz O and Köhl U (2015) Cetuximab reconstitutes pro-inflammatory cytokine secretions and tumorinfiltrating capabilities of sMICAinhibited NK cells in HNSCC tumor spheroids. Front. Immunol. 6:543. doi: 10.3389/fimmu.2015.00543*

characterized by increased IFN-γ and TNFα secretion. This data show that the impaired NK cell-dependent tumor surveillance in relapsed HNSCC patients could be reversed by the re-establishment of ADCC-mediated effector cell activity, thus supporting NK cell-based immunotherapy in combination with antineoplastic monoclonal mAbs.

Keywords: ADCC, cetuximab-activated NK cells, HNSCC tumor spheroids, soluble MICA, TGF-β1

# INTRODUCTION

Natural killer cells are lymphoid effector cells important for the innate immune response against virally infected and malignant cells (1, 2). NK cells eliminate transformed target cells during cytotoxic interactions by releasing pro-inflammatory cytokines, especially IFN-γ and TNF-α (3). Similarly, tumor-infiltrating NK cells can trigger stimulating interactions via "cell cross-talk" with dendritic cells (DCs), possibly facilitating tumor antigen presentation and induction of tumor antigen-directed T-cell responses. This demonstrates the constitutive role of NK cells as mediators between the innate and acquired immune systems (4–6). In this respect, NK cell killing activity is regulated by both stimulatory and inhibitory receptors. Interaction between NK cell inhibitory receptors in the presence or absence of MHC class I molecules on normal and possible target cells was described as the "missing self " hypothesis (7, 8). Activating receptors include the natural cytotoxicity receptors (NCRs) NKp30, NKp44, and NKp46 with poorly characterized ligands, and the NKG2D receptor, which recognizes a variety of well-defined ligands expressed by transformed cells (9–12). Several studies confirmed the predominant relevance of NKG2D in efficient recognition and elimination of tumor- and "stressed" cells by targeted binding of MICA and MICB (9, 10, 12). Interestingly, elevated levels of soluble forms of these NKG2D ligands (sMICA and sMICB), generated by matrix metalloproteinase (MMP)-dependent proteolytic cleavage ("shedding") were detected in plasma/serum of cancer patients (13). These soluble NKG2D ligands are responsible for systemic reduction of NKG2D expression on the surface of various circulating blood lymphocytes, especially cytotoxic NK cells, NK-like T (NKT) cells, and CD8<sup>+</sup> αβ+- and γδ+-T cells. Thus, these immune modulating effects resulted in decreased tumor surveillance by attenuated recognition and elimination of malignant cancer cells (14–16).

Some reports described NK cell dysfunction in patients with head and neck squamous cell carcinoma (HNSCC). These highly aggressive solid tumors originate from the epithelial lining of the upper aero-digestive tract and are able to escape NK cell-mediated immunosurveillance. Tumor progression is accomplished by significantly reduced expression levels of NKG2D on effector cells (17–20). Detection of increased sMICA plasma levels monitored in HNSCC patients at advanced disease stages (stage IV) and poor clinical prognosis further supports the importance of diminished tumor surveillance in HNSCC progression (21, 22). Indeed, high sMICA levels coincide with increased frequencies of lymph node (LN) metastasis. Additionally, decreased survival rates in high-risk cancer patients are potentiated by high sMICB levels (19, 20, 23). Interestingly, the multiple dysfunctions of NK cells can be largely reversed by cancer antigen-targeted antibodies, which stimulate the antibody-dependent cellular cytotoxicity (ADCC)-mediated cytotoxicity of activated NK cells to selectively eliminate malignant cells. Therefore, multiple monoclonal antibodies (mAbs) have been designed to target diverse tumor surface molecules. Based on the assumption that mAbs interact specifically with tumor target molecules, mAbs could affect tumor cells by directly inhibiting essential signaling pathways initiated by target molecules and/or by stimulating effector cell cytotoxicity, resulting in tumor elimination. One highly investigated cancer antigen for treatment of solid tumors is the epidermal growth factor receptor (EGFR). EGFR is a member of the ErbB protein family, which consists of four transmembrane receptor proteins, including HER1 [EGFR, ErbB1: avian erythroblastic leukemia viral (v-erb-b) oncogene homolog, receptor for EGF], HER2 (ErbB2), HER3 (ErbB3), and HER4 (ErbB4) (24–26). EGFR overexpression is associated with poor clinical prognosis in multiple solid tumors and EGFR signaling plays an important role in malignant cell migration, evasion, and proliferation (27). Therefore, mAbs designed to target EGFR were developed. In the last decade, cetuximab, an anti-EGFR human-mouse chimeric IgG1 monoclonal antibody (mAb) was approved by the Food and Drug Administration (FDA) for treatment of metastatic colorectal cancer, metastatic non-small cell lung cancer, and HNSCC patients (28, 29). However, only a weak to moderate (10–20%) benefit was observed in clinical trials with high-risk cancer patients (30–35). In this context, HNSCC cells may have evaded NK cell immunosurveillance due to polymorphisms in the FcγRIIIa (CD16a) on effector cells that impact interaction with the IgG1 Fc, heavy-chain, portion of cetuximab. This partially elucidates patient-specific responses to cetuximab and underscores the essential importance of modulating immunological synapses (36–38). In addition, the tumor microenvironment can impact lymphocyte-dependent immunosurveillance, which correlated strongly with tumor infiltration as well as NK cell-mediated killing activity. The ability to control these factors could contribute to improved prognosis in some malignant diseases (39–42). Thus, NK cells as key players in ADCC-related cetuximab activity were able to infiltrate primary colorectal adenocarcinomas and NK cell infiltration was an independent predictor for response and progression-free survival in patients receiving cetuximab treatment (43).

Recently, we described a decreased anti-tumor recognition, cytokine release and a reduced NKG2D expression on NK cells from untreated HNSCC patients. *In vitro* blocking experiments revealed a synergistic negative effect of sMICA potentiated by TGF-β1 on the killing activity of patient NK cells (22). In the current study, cetuximab treatment reconstituted the tumor surveillance capacity of sMICA-inhibited NK cells from HNSCC patients (*n* = 5), thus demonstrating the potential usefulness of cetuximab in retargeted ADCC. In order to investigate specific NK cell-dependent tumor infiltrations and ADCC-related cetuximab response, we developed HNSCC tumor-like cell clusters and a tumor spheroid model derived from primary, singularized tumor cells from these HNSCC patients. Our results indicate a crucial relevance of enhanced sMICA levels in tumor surveillance and infiltrations of inhibited patient NK cells. Finally, we demonstrate that these immunosuppressive effects on NK cell-mediated killing activity could be bypassed using cetuximab-coated HNSCC cells.

## PATIENTS AND METHODS

#### HNSCC Patients

We analyzed five HNSCC patients (three male and two female, age range: 24–76 years) and five age-matched healthy individuals (three male and two female, age range: 26–58 years) served as controls. Histopathology confirmed that patients had stage II–IV HNSCC. Patients were included in this study after tumor recidivism but before initiation of any clinical treatment (**Table 1**). Corresponding patient blood samples (80–100 ml) were received shortly before the tumor surgery and associated tumor fragments were collected during tumor extractions from all patients. Informed consent was obtained from patients, caretakers, and healthy controls (HCs). Patient characteristics are summarized in **Table 1**. Blood samples were collected from HNSCC patients and healthy individuals in Heparin- and EDTA-coated tubes. Total leukocytes and the resultant subpopulations were counted by five-color flow cytometry (FCM) analysis as described previously (44). Immunocompetent cell subpopulation distributions were compared among patients and HCs (**Figure 1**).

# Target Cell Line

The human HNSCC cell line SCC-4 (ATCC: CRL-1624) (45, 46) was used to compare the cytotoxic activities of freshly purified patient and healthy NK cells and served as an internal control for scored intensities of comparable fluorescence staining's from different patient-derived primary HNSCC cells. Therefore, the SCC-4 was cultured in DMEM and GlutaMAX™ medium (GIBCO, Invitrogen, Germany) supplemented with 10% (v/v) heat-inactivated fetal bovine serum (FBS) and 2 mM l-glutamine (PAA Laboratories GmbH, Austria).

# Preparation of Single Cell Suspension from Primary Tumor Samples

Tumor samples from untreated HNSCC patients (*n* = 5) were collected post-surgery and washed twice in serum-free DMEM/ F12HAM/Glutamax supplemented with 100 U/ml penicillin, 100 μg/ml streptomycin, 0.25 μg/ml amphotericin B (antibiotic– antimycotic 100×, all purchased from Life Technologies, Gibco®, Darmstadt, Germany). After dissociation with 0.05% trypsin/ EDTA (Life Technologies, Gibco®, Darmstadt, Germany), tumor pieces were minced with scissors and scalpels in a sterile dish. The digestion was stopped with DMEM/F12HAM/Glutamax containing 10% AB-Serum (former: PAA, Linz, Austria), antibiotic–antimycotic and the sample was passed through a 70 or 100-μm nylon mesh cell strainer (BD Biosciences, Heidelberg, Germany) to achieve single cell suspensions. Cells were collected in a 50-ml conical tube and subsequently centrifuged. Suspended cells were counted with trypan blue, characterized with FCM and cultivated in DMEM/F12HAM/Glutamax/10% AB serum/ antibiotic–antimycotic in an incubator (37°C, 5% CO2, 90% humidity). Cultured tumor cells formed small tumor cell clusters after a few days, which resulted in tightly arranged HNSCC tumor spheroids (diameter: 1–3 mm) after cell cultivation of 1–2 weeks. Tumor cluster and spheroids derived from our HNSCC patients were used for NK cell-based cytotoxicity and tumor-infiltration assays monitored by fluorescence microscopy and time-lapsed transmitted imaging.

## Cytokine Analysis

The BD CBA Kit was utilized for scavenging soluble cytokines, especially IFN-γ and TNFα with beads of known size and fluorescence, allowing identification of soluble molecules in blood or supernatants from cell culture medium by FCM as described previously (44).

TABLE 1 | The clinical parameters and immune status of HNSCC (*n* **=** 5) patients summarized after tumor (TU) and lymph node (LN) surgery but before any acute clinical therapeutic regimens.


NK cells (**Figures 2** and **3**). Singularized HNSCC cells were also tested for MICA, HER2, and HER1 surface expression, the latter as a target for restored ADCC before initiation of described cytotoxicity assays. To determine the effect of cetuximab on NK cell-dependent killing activity, various cetuximab doses (Cetmab: 1 pg/ml–1 μg/ ml) were used to coat the corresponding HNSCC target cells (E:T ratio: 10:1). To inhibit putative effects of cetuximab toward NK cell-mediated cytotoxicity, the NK cells were pre-incubated for 20 min with anti-CD16 mAb (20 μg/ml). The effector-based cytotoxicity of these treated NK cell samples were analyzed against corresponding patient HNSCC cells in the indicated effectorto-target ratios (**Figures 3** and **4**). To avoid effector and target cell sedimentations or insufficient stirring of our co-incubated approaches during the cytotoxic reactions, the co-cultured cell suspensions were shaken in an CO2-incubator (CO2cell, 170-400 Plus, RS Biotech, Scotland) for 4 h (37°C, 5% CO2, 250 rpm). An optimized gating panel (Figure S1 in Supplementary Material) based on a no-wash single platform FCM procedure (FC500, Beckman Coulter, Germany) was applied. Treated NK cells were stained with several monoclonal antibodies (mAbs): CD45 FITC (fluorescein isothiocyanate), CD56 PE (phycoerythrin) or NKG2D PE, CD16 PC-7 (phycoerythrin-cyanin-7) in order to exclude the effector cells from primary HNSCC target cells stained with CD9 FITC, CD9 PE, HER1 PE, MICA PE, or CD81 PE. Effector and target cells were stained with mAbs as described previously (47, 48). Target cell elimination by effector cells was calculated as the total

loss of viable HNSCC target cells as follows (48, 49):

Fluorescence Microscopy

Time-Lapse Microscopy

Killing activity concentrationco-cultured HNSCC cells/ = − (1 <sup>µ</sup><sup>L</sup>

IL-2 activated patient NK cells were cultured (24 h) with PP [(high sMICA); 1:2 diluted with X-VIVO™10] or healthy plasma [HP (low sMICA); 1:2 diluted with X-VIVO™10]. Afterward, these treated NK cells were co-incubated (1 h, 37°C, 5% CO2, approximate E:T ratios of 5:1) on 4-well chamber slides (1.7 cm2 growth area/well, 0.5–1.0 ml working volume, Nunc™, USA) with corresponding adherent HNSCC cell clusters derived from HNSCC patients (*n* = 5, time of cultivation: 3–12 days) in presence or absence of 1 μg/ml cetuximab for indicated periods of time (**Figure 5A**). The capability of tumor infiltration from those treated patients NK cells (see above) in corresponding patient HNSCC tumor spheroids (*n* = 5) were assessed with and without 1 μg/ml cetuximab and analyzed after 24 h co-cultivation (37°C, 5% CO2) (**Figure 6**). 2D/3D confocal fluorescence microscopy (CFM) using FITC- and PE-conjugated mAbs was used to separate stained patient NK cells (CD45 FITC or NKG2D PE) and corresponding HNSCC cell clusters or tumor spheroids (CD9 FITC, HER1 PE, HER2 PE, or MICA PE), as described before (47).

Transmission microscopy was used to monitor infiltration of sMICA-inhibited NK cells into corresponding HNSCC tumor spheroids. Low numbers of tumor spheroids were grown scattered on chamber slides for 16 h (37°C, 5% CO2). Subsequently,

HNSCC control cells/ L concentration

/

<sup>µ</sup> ) % ×100

# Quantification of sMICA and TGF-**β**1 in HNSCC Patients

The *BAMOMAB* MICA-Sandwich ELISA kit for sMICA (AXXORA GmbH, Germany) was designed for quantification of soluble MICA (sMICA). The kit was utilized for detection and monitoring of immunosuppressive molecules in HNSCC patient blood plasma (*n* = 5), HCs (*n* = 5), and supernatants of cell culture medium during adherent growth phase of tumor cell clusters and generation of tumor spheroids as described previously (47). TGF-β1 levels in human blood plasma samples and cell culture supernatants were quantified by MTPL ELISA (Milenia Biotec, Version 3.0, Germany).

# Immunomagnetic Separation of CD56+CD3<sup>−</sup> NK Cells

Up to 100 ml heparinized blood from HNSCC patients (*n* = 5) was used to isolate viable mononuclear cells (MNC) in high yield and purity by Ficoll-Paque density gradient. Primary NK cells were separated from purified MNCs via "non-touched" depletion using the EasySep® Human NK Cell Enrichment Kit (STEMCELL Technologies SARL, Germany). Other leukocyte subsets were labeled with tetrameric antibody complexes against CD3, CD4, CD14, CD19, CD20, CD36, CD66b, CD123, HLA-DR, glycophorin A, and dextran-coated magnetic particles. The non-labeled cells were scavenged by an EasySep® hand magnet according to the manufacturer's recommendations. Freshly purified NK cells (purity: 95.1 ± 2.8%) were expanded and activated with 1000 IU/ ml IL-2 for 9–12 days as described previously (47).

# Production of Shed MICA

To generate shMICA, the DNA sequence encoding for full length MICA engineered with an N-terminal histidine-tag was cloned into a tet-on vector system and transfected into the UKF-NB3 tumor cell line using the Neon transfection system (Life Technologies, USA). Cells were selected with 0.25 mg/ml G418 and 1 μg/ml puromycin (Life Technologies, USA). MICA expression was induced by addition of 2 μg/ml doxycycline (Sigma, Germany). Further on cells were stressed by serum starvation for 72 h to induce MICA shedding. Finally shMICA was purified with Protino Ni-NTA agarose (Macherey-Nagel, Germany) from cell culture supernatants and concentrated using Amicon centrifugal filter units (Merckmillipore, Germany).

# Cytotoxicity Assays

FIGURE 1 | Phenotypical and functional characterizations of patient NK cell subsets and primary tumor cells from non-treated HNSCC patients (*n* **=** 5) after tumor relapse. (A) *In vitro* rearrangement of the NK cell phenotype was quantified in the PB before separation of NK cells and after IL-2 expansion (1000 IU/ml IL-2; 9–12 days). Shown are the absolute numbers of patient (HNSCCNK cells) and healthy donor (HDNK cells) CD56+/CD3− NK cells [cells/μl] [left graph area (NK)], the mean fluorescence intensity [MFI (%)] of distribution of resultant CD56bright/CD16dim&neg and CD56dim/CD16+ NK subpopulations [middle graph area (subsets)] and co-expressed NCRs [MFI (%) right graph area (NCRs)] among total NK cells. (B) SMICA and TGF-β1 levels were analyzed in blood plasma from corresponding HNSCC patients (PP) and compared to age-matched healthy donor plasma controls (HP). (C) Assessment of the basic killing activities between effector cells isolated from patient and healthy donor NK cells against SCC-4 target cells. Freshly isolated, non-stimulated NK cells from patients (HNSCC), and healthy controls (HC) were treated with corresponding HNSCC patient plasma (high sMICA/TGF-β1) or associated healthy control plasma (low sMICA/TGF-β1) and co-incubated for 4 h (37°C, 5% CO2, 250 rpm) with SCC-4 cells at the indicated E:T ratios and cytotoxicity (%) was measured by FCM. (D,E) Immunofluorescence staining and FCM-based characterization of relevant tumor antigen expression from primary tumor samples derived from corresponding HNSCC patients (*n* = 5). After preparation of single cell suspensions from primary tumor samples, tumor cells were cultured (1–2 days, 37°C, 5% CO2) on chamber slides and characterized phenotypically for CD9, MICA, HER1, and HER2 expression profiles by immunofluorescence microscopy. Depicted is the staining for one representative HNSCC sample. (D) The HNSCC cells were also analyzed for CD9 (FITC), MICA (PE), HER1 (PE), and HER2 (PE) surface expression levels by FCM (E). Nuclei were stained with DAPI (4′,6-diamidino-2-phenylindole, blue fluorescence). (F) Secretion of soluble immunosuppressive factors derived from five tumor samples. SMICA and TGF-β1 levels in expanded primary HNSCC cell cultures were analyzed by ELISAs in the collected cell medium supernatants at the indicated time frames. (A–F) Data

are shown as mean ± SD from two to four experiments for the five patients. Range of statistically significant differences: from \**p* ≤ 0.01 to \*\**p* ≤ 0.001.

To analyze the NK cell-mediated killing activity in presence and absence of cetuximab, we developed a matched effector-target cell system based on a FCM-based cytotoxicity assay. Therefore, we utilized only concordant patient NK cells, patient plasma (PP, high sMICA) and primary patient HNSCC tumor cells (*n*= 5). To demonstrate the sMICA-mediated inhibition, IL-2 expanded (9–12 days) primary patient NK cells were co-incubated overnight (24 h, 37°C, 5% CO2, 250 rpm) with 500 pg/ml shMICA, PP containing high MICA levels (PP, range: 220.9–870.7 pg/ml) and, as a comparative control, with HC plasma (HP, range: 2.8–22.0 pg/ml) diluted 1:2 in X-VIVO™10 medium (Biowhittacker™Cambex Bioscience, Belgium). Phenotypical cell characterizations were accomplished to detect altered expression patterns of CD16 and NKG2D on treated NK cells (**Figures 2** and **3**). Singularized HNSCC cells were also tested for MICA, HER2, and HER1 surface expression, the latter as a target for restored ADCC before initiation of described cytotoxicity assays. To determine the effect of cetuximab on NK cell-dependent killing activity, various cetuximab doses (Cetmab: 1 pg/ml–1 μg/ ml) were used to coat the corresponding HNSCC target cells (E:T ratio: 10:1). To inhibit putative effects of cetuximab toward NK cell-mediated cytotoxicity, the NK cells were pre-incubated for 20 min with anti-CD16 mAb (20 μg/ml). The effector-based cytotoxicity of these treated NK cell samples were analyzed against corresponding patient HNSCC cells in the indicated effectorto-target ratios (**Figures 3** and **4**). To avoid effector and target cell sedimentations or insufficient stirring of our co-incubated approaches during the cytotoxic reactions, the co-cultured cell suspensions were shaken in an CO2-incubator (CO2cell, 170-400 Plus, RS Biotech, Scotland) for 4 h (37°C, 5% CO2, 250 rpm). An optimized gating panel (Figure S1 in Supplementary Material) based on a no-wash single platform FCM procedure (FC500, Beckman Coulter, Germany) was applied. Treated NK cells were stained with several monoclonal antibodies (mAbs): CD45 FITC (fluorescein isothiocyanate), CD56 PE (phycoerythrin) or NKG2D PE, CD16 PC-7 (phycoerythrin-cyanin-7) in order to exclude the effector cells from primary HNSCC target cells stained with CD9 FITC, CD9 PE, HER1 PE, MICA PE, or CD81 PE. Effector and target cells were stained with mAbs as described previously (47, 48). Target cell elimination by effector cells was calculated as the total loss of viable HNSCC target cells as follows (48, 49):

Killing activity concentrationco-cultured HNSCC cells/ = − (1 <sup>µ</sup><sup>L</sup> HNSCC control cells/ L concentration / <sup>µ</sup> ) % ×100

# Fluorescence Microscopy

IL-2 activated patient NK cells were cultured (24 h) with PP [(high sMICA); 1:2 diluted with X-VIVO™10] or healthy plasma [HP (low sMICA); 1:2 diluted with X-VIVO™10]. Afterward, these treated NK cells were co-incubated (1 h, 37°C, 5% CO2, approximate E:T ratios of 5:1) on 4-well chamber slides (1.7 cm2 growth area/well, 0.5–1.0 ml working volume, Nunc™, USA) with corresponding adherent HNSCC cell clusters derived from HNSCC patients (*n* = 5, time of cultivation: 3–12 days) in presence or absence of 1 μg/ml cetuximab for indicated periods of time (**Figure 5A**). The capability of tumor infiltration from those treated patients NK cells (see above) in corresponding patient HNSCC tumor spheroids (*n* = 5) were assessed with and without 1 μg/ml cetuximab and analyzed after 24 h co-cultivation (37°C, 5% CO2) (**Figure 6**). 2D/3D confocal fluorescence microscopy (CFM) using FITC- and PE-conjugated mAbs was used to separate stained patient NK cells (CD45 FITC or NKG2D PE) and corresponding HNSCC cell clusters or tumor spheroids (CD9 FITC, HER1 PE, HER2 PE, or MICA PE), as described before (47).

## Time-Lapse Microscopy

Transmission microscopy was used to monitor infiltration of sMICA-inhibited NK cells into corresponding HNSCC tumor spheroids. Low numbers of tumor spheroids were grown scattered on chamber slides for 16 h (37°C, 5% CO2). Subsequently,

cells from untreated HNSCC patients (*n* = 5) were stimulated for 9–12 days with 1000 IU/ml IL-2. (A) Effect of high sMICA levels on NKG2D surface expression. IL-2 activated patient NK cells were incubated (37°C, 5% CO2) at the indicated points of time with shMICA (500 pg/ml), rMICA (500 pg/ml) and patient plasma containing high MICA levels (PP, range: 220.9–870.7 pg/ml). The NKG2D expression levels on total NK cells and both CD56dim/CD16+ and CD56bright/CD16dim&neg NK subsets were compared to non-treated control NK cells. (B) Phenotypical analyses on time-dependent impact of shMICA on NKG2D surface expression were determined by FCM. Exemplarily depicted here are representative dot plots for one time-dependent experiment.

FIGURE 3 | Determination of the sMICA impact on ADCC-stimulating and pro-apoptotic NK cell receptors. (A) Surface expression levels of CD16, TRAIL, FasL, and CD57 on the same effector cell samples were measured at the indicated time points and compared to non-treated control cells. As an additional control, effector cell viability was monitored over a 24-h period (B–D). The impact of sMICA on the NKG2D-mediated NK cell cytotoxicity and NK cell stability during effector-target interactions against primary HNSCC cells (*n* = 5) was analyzed by FCM. Therefore, the same overnight-treated NK cell samples (see above, section Figure 2A) were subsequently co-incubated for 4 h (37°C, 5% CO2) with corresponding patient HNSCC target cells at the indicated E:T ratios. (E) Inhibition of sMICA effects on NKG2D-mediated NK cell cytotoxicity. SMICA molecules in all treatment samples (shMICA, rMICA, and PP) were blocked by pre-incubation (20 min) with MICA-specific mABs (20 μg/ml, MAB13001). NKG2D-dependent killing rates of those incubated NK cells against primary HNSCC cells were measured after 4 h ("Blocking assays," ratio: 10:1, 37°C, 5% CO2) as described above. Data are shown as mean ± SD from three to four experiments for each patient. Statistically significant difference: \**p* ≤ 0.01 and \*\**p* ≤ 0.001.

#### FIGURE 4 | Continued

IL-2 activated patient NK cells (*n* = 5) were incubated for 24 h (37°C, 5% CO2) with (A) shed MICA (shMICA, 500 pg/ml), (B) with patient plasma containing high MICA levels (PP, *n* = 5, range: 220.9–870.7 pg/ml) and healthy plasma (HP, *n* = 5, range: 1.9–28.5 pg/ml) served as a control. (A) CD16-mediated ADCC of those treated patient NK cells was assessed by FCM-based cytotoxicity assay using cetuximab (Cetmab, range of dose titration: 1 μg/ml–1 pg/ml)-coated primary HNSCC cells and compared to killing activity of shMICA-treated NK cells against non-coated HNSCC cells. (B) To overcome sMICA effects against NK cell cytotoxicity, 1 μg/ml cetuximab was chosen based on the earlier titration experiments (see above). Inhibition of cetuximab-dependent ADCC was achieved by blocking CD16 epitopes on treated patient NK cells with anti-CD16 mAb [20 μg/ml, 20 min pre-incubation, graphs (A,B)]. (C) The effector cell stability of PP- and HP-treated viable NK cells during tumor cell lysis was analyzed in presence and absence of 1 μg/ml cetuximab and compared to non-treated effector cell controls quantified by single platform functionality assays. (D) FCM-based characterization was utilized to monitor CD16 and NKG2D surface expression patterns on treated NK cells (CD16/NKG2D, see overlay histograms) before (−PP) and after (+PP) overnight incubation with patient plasma (PP) containing high sMICA and TGF-β<sup>1</sup> levels. Analogously, HNSCC cell clusters were singularized for FCM-based characterization of target cell antigens (HER1/HER2/MICA) to assess alterations in the killing activity from different treated patient NK cells. (E) Effect of IL-2 on the killing activity of patient NK cells (*n* = 5) in presence (+IL-2/+Cetmab) and absence (+IL-2) of 1 μg/ml cetuximab (Cetmab). IL-2 expanded NK cells were incubated with patient plasma containing high MICA levels (PP, *n* = 5, range: 220.9–870.7 pg/ ml), healthy plasma (HP, *n* = 5, range: 1.9–28.5 pg/ml) and control medium. Cytotoxicity of treated NK cells was analyzed at the indicated ratios with and without Cetmab (1 μg/ml)-coated primary patient HNSCC cells and compared to killing activity of unstimulated NK cells against non-coated HNSCC cells (−IL-2). (A–D) Data represent the mean ± SD of three experiments for each patient. Statistically significant difference: \**p* ≤ 0.01 and \*\**p* ≤ 0.001.

HNSCC spheroids were co-incubated with freshly isolated, non-stimulated patient NK cells (1 × 106 effector cells/ml). Putative effector cell migration and tumor chemoattraction were monitored by time-lapse microscopy and imaging (Video S1 in Supplementary Material) as described previously (47).

#### Blocking Assays

To inhibit the restored NK cell-based cytotoxicity by cetuximab, patient NK cells were pre-incubated for 20 min with anti-CD16 mAb (clone 3G8, 20 μg/ml, A07766, Beckman Coulter, Germany). Afterward, killing activity of treated NK cells was analyzed with cetuximab-coated primary patient HNSCC cells over a time period of 4 h (37°C, 5% CO2, 250 rpm) at an E:T ratio of 10:1 and compared to corresponding controls (**Figure 4**) as described previously (47). To analyze the direct role of sMICA on NKG2Dmediated cytotoxicity, we incubated IL-2-activated patient NK cells with shMICA, rMICA, and PP containing high sMICA levels overnight in the presence or absence of specific MICA antibodies, respectively (20 μg/ml, MAB13001, R&D systems, Germany). Additionally, plasma from five healthy donors (low sMICA, see above) was utilized to incubate patient NK cells as a control.

#### Statistical Analyses

The Mann–Whitney non-parametric *U*-test was utilized to compare clinical and pathological parameters of plasma sMICA and TGF-β1 levels from HNSCC patients (*n* = 5) with healthy individuals (*n* = 5). The Student's *t* test was used to assess the significance of the killing activity of patient NK cells incubated under various conditions. A *p* level ≥0.01 was considered statistically as non-significant. Unless otherwise declared, results of statistical evaluations from functional assays are indicated as mean ± SD and represent three to four experiments for each patient.

## RESULTS

#### Characterization of Altered NK Cell Subsets and Expression of NCRs in HNSCC Patients

Compared to age-matched healthy individuals (50), HNSCC patients showed a broad range of leukocyte subpopulations and absolute numbers of lymphocytes and leukocytes (**Table 1**). Although median NK cell amounts (12.8%; range: 2.7–33.2%) did not differ from HCs (**Table 1**), the absolute NK cell numbers (cells/μl) differed widely in the peripheral blood (PB) of patients and healthy donors (left graph sector, **Figure 1A**). Moreover, the proportion of immunoregulatory NK cells (CD56bright/ CD16dim&neg) was markedly reduced in all patients [median: 2.4% (HNSCCNK cells) versus 11.8% in healthy donors (HDNK cells), middle graph sectors, **Figure 1A**]. In contrast, the cytotoxic NK cell subpopulation (CD56dim/CD16<sup>+</sup>) was strongly increased for all investigated HNSCC patients [median (HNSCCNK cells): 96.2% versus 86.8% (HDNK cells), middle graph sector, **Figure 1A**]. Moreover, freshly isolated patient NK cells revealed low to moderate expression levels of the NCRs, NKp30, NKp44, NKp46, and NKG2D compared to higher frequencies of IL-2 stimulated NK cells from HCs (right graph sector, **Figure 1A**). Nevertheless, the percentage of NK cells expressing NCRs increased (~4.7-fold, 3.8-fold, and 2-fold for NKp30, NKp44, and NKp46, respectively) during IL-2 activation over 9–12 days and was accompanied by ~60.7-fold higher expression levels of NKG2D (**Figure 1A**, right graphs) for all stimulated patient NK cells. The distribution of NK cell subpopulations shifted to higher CD56bright/CD16dim&neg NK cell subsets (median before IL-2: 2.4% versus median after IL-2: 12.5%) and consequently lower percentages of CD56dim/ CD16<sup>+</sup> NK cells (median before IL-2: 96.2% versus median after IL-2: 88.9%) (**Figure 1A**).

## Reduced NK Cell-Dependent Cytotoxicity and Increased Immunosuppressive Factors in HNSCC Patients

The well-defined immunosuppressive factors sMICA and TGF-β1, which are responsible for impaired immunosurveillance, were quantified in PB from our HNSCC patients. Higher levels of both soluble factors were detected in HNSCC patients compared to HCs (sMICA, median: 532.8 versus 5.9 pg/ml; TGF-β1, median: 48.9 × 104 versus 10.9 × 104 pg/ml, respectively) (**Figure 1B**). Healthy plasma samples showed sMICA and TGF-β1 levels close to the detection limits of this assay [sMICA (mean ± SD): 10.8 ± 11.2 pg/ml; TGF-β1: 9.5 ± 5.2 × 104 pg/ml; **Figure 1B**, HP], whereas sMICA (TGF-β1) in PP ranged between 220.9 and


*Fluorescence labeling of these tumor cells were compared to the SCC-4 cell line and intensities were scored as follows: weak (*+*/*−*), moderate (*+*), good (*++*), and strong (*+++*).*

870.7 pg/ml (25–64.8 × 104 pg/ml) (**Figure 1B**, PP). To compare the basic killing activity between patient NK cells and NK cells from healthy individuals, freshly isolated, non-stimulated NK cells from both, patients and HCs were co-incubated overnight (37°C, 5% CO2) with corresponding HNSCC PP (high sMICA/ TGF-β1) or associated HC plasma (low sMICA/TGF-β1). In both cases, the NK cell-mediated cytotoxicity was analyzed against the target cell line SCC-4 at the indicated ratios (4 h, 37°C, 5% CO2, 250 rpm) by FCM (**Figure 1C**). HNSCC plasma-treated patient NK cells (HNSCC, **Figure 1C**) exhibited significantly reduced cytotoxicity in all prepared E:T ratios when compared to NK cells from healthy donors pre-incubated with the corresponding HC plasma (**Figure 1C**).

# Phenotypical Analysis of Tumor-Relevant Expression Markers on Primary Patient HNSCC Cells

Expression levels of relevant surface antigens, especially HER1 and HER2, were examined on single tumor cell suspensions prepared from primary tumors surgically removed from HNSCC patients (*n* = 5, see above). Pre-characterized SCC-4 cells were utilized as a positive control for HER1 and HER2 expression because it was previously shown that these HNSCC cells displayed high amplification rates of HER1 genes in combination with enhanced HER2 and MICA expression levels (51). In addition to the HER1/2 levels, we analyzed expression levels of surface markers CD9 and MICA on the mono-dispersed patient tumor cells (**Figure 1E**). Representative overlay plots (FCM) were shown exemplarily in **Figure 1E** for prepared tumor cells derived from one HNSCC patient and were compared with qualitative evaluations determined by CFM. In **Figure 1D**, representative results of overlapping CFM exemplarily illustrated for cultured (1–2 days) adherent patient tumor cells after fluorescence labeling with similar stimulation energy and duration are shown. For counterstaining, we utilized DAPI as a fluorescent stain that binds strongly to A–T rich DNA regions of nuclei and chromosomes and emits blue fluorescence (**Figure 1D**). All tumor samples showed higher HER1 and CD9 levels in contrast to low-to-moderate expression of membrane-associated MICA and HER2 molecules. In summary, the antigens exhibited variable expression levels on all tumor samples of HNSCC patients and were slightly lower than on the SCC-4 control line as presented in **Table 2**.

Primary patient HNSCC cells (*n* = 5) were cultured over 2 weeks (37°C, 5% CO2) to investigate time-dependent accumulation of immunosuppressive factors (sMICA and TGF-β1), and cell culture supernatants were collected at the indicated time periods (**Figure 1F**). A time-dependent increase of both immunosuppressive molecules was identified in cell culture supernatants in which the mean of sMICA levels increased from 18.7 ± 9.5 to 218.7 ± 131.9 pg/ml, and the mean of TGF-β1 levels increased from 6.8 ± 4.8 × 104 to 40.9 ± 11.6 × 104 pg/ml between the first and last time points monitored (**Figure 1F**).

# SMICA Affects NKG2D Expression and Cytotoxicity of IL-2-Activated Patient NK Cells

Tumor-derived TGF-β1 potentiated the sMICA-mediated downregulation of NKG2D surface expression on various effector cells, especially NK cells, and resulted in decreased NKG2D-dependent immunity, thus reflecting the predominant role of the sMICA-NKG2D system (20, 52). Therefore, we co-cultured IL-2-activated patient NK cells overnight (24 h) with sMICA analogs (shMICA and rMICA, each in 500 pg/ml) and PP containing high sMICA levels (range: 220.9–870.7 pg/ml). All incubated NK cell samples from the five HNSCC patients exhibited a time-dependent downregulation of NKG2D expression on total NK cells and both NK cell subsets (**Figures 2A,B**), whereas surface expression of CD16, pro-apoptotic FasL, and TRAIL receptors as well as activation marker CD57 were largely unaffected and stable over the indicated time period on all NK cell fractions by the described sMICA analogs (**Figure 3A**). We detected a distinct effect of shMICA on NKG2D expression on total NK cells and resultant NK subpopulations analyzed during the indicated time frame of 24 h (37°C, 5% CO2) by FCM. The time-dependent NKG2D down-regulation is displayed in plots exemplarily shown for one experiment (**Figure 2B**). Although surface phenotype, cell proportions, and viability of NK cells were not affected by co-incubation with PP (24 h, high MICA, **Figure 3A**), the observed decrease of NKG2D surface expression was accompanied by reduced IFN-γ and TNFα secretion compared to IL-2 activated patient NK cells [IL-2 (9–12 days), **Figure 5D**].

Additionally, the impact of sMICA on the NKG2D-mediated NK cell cytotoxicity of overnight co-incubated NK cells was analyzed against corresponding primary tumor cells derived from HNSCC patients (*n* = 5) to assess the degree of killing activity. Effector and target cells were co-incubated for 4 h (37°C, 5% CO2, 250 rpm) at indicated E:T ratios and killing activity was subsequently compared to X-VIVO™10-incubated controls without sMICA analogs. The cytotoxicity of shMICA- and PP-treated NK cells was strongly suppressed compared to low inhibition frequencies of rMICA-incubated effector cells at all E:T ratios (**Figures 3B–E**). During the cytotoxicity assays of co-cultured effector and target cells, the NK cell viability was decreased in shMICA-, rMICA-, and PP-treated NK cells as compared to

cultured (24 h) with patient plasma (PP; 1:2 diluted with X-VIVO™10) containing high sMICA levels and co-incubated (1 h, approximate E:T ratios of 5:1) with corresponding HNSCC cell clusters (time of cultivation: 3–12 days) in (i) presence and (ii) absence of 1 μg/ml cetuximab. (iii) Activated patient NK cells treated for 24 h with healthy plasma (HP [low sMICA]; 1:2 diluted with X-VIVO™10) were used to monitor early effector-target cell interactions, served as a positive control. CD45 (green) for PP- and HP-treated patient NK cells were analyzed by staining with FITC-conjugated mAb, whereas co-incubated adherent HNSCC clusters were unlabeled. Before initiating these functional assays, effector and target cells were tested flow cytometrically for relevant surface markers, especially MICA, HER2 and HER1 on HNSCC cells (B), and CD16 and NKG2D levels on PP (high sMICA/TGF-β1)-treated NK cells (C). Variable IFN-γ and TNFα secretions levels during co-incubations of treated NK cells with HNSCC cell clusters or tumor spheroids in presence or absence of cetuximab were quantified and compared with untreated controls (D). Statistically significant difference: \**p* ≤ 0.01 and \*\**p* ≤ 0.001.

more stable effector cells in X-VIVO™10-incubated controls (**Figures 3B–E**). To confirm the direct sMICA impact on the NKG2D-mediated NK cell cytotoxicity, MICA-specific mAbs (20 μg/ml, MAB13001) were used to deplete sMICA molecules in different treatment mixtures (shMICA, rMICA, and PP) (**Figure 3E**). NK cell cytotoxicity of these treated effector cells co-incubated with corresponding primary HNSCC cells were compared with sMICA analog-treated NK cells co-cultured without sMICA-specific mAbs as a negative control. The diminished killing activity of sMICA-affected NK cells was partially restored when compared to untreated (X-VIVO™10) control effector cells and to NK cells pre-treated with healthy plasma (HP, low sMICA) (**Figure 3E**).

# Cetuximab Restores NK Cell-Dependent Cytotoxicity Against Primary HNSCC Cells via ADCC

Cetuximab is a therapeutic mAb directed against the HER1 epitopes on several types of high-malignant tumors (53). Consequently, cetuximab is a powerful stimulus of NK cell-mediated ADCC via activation of FcγRIIIa against cetuximab-coated tumor cells and for induction of cytokine release, especially IFN-γ and TNFα secretion. **Figure 4A** demonstrates significantly restored NK cellbased cytotoxicity against cetuximab-coated HNSCC cells with mAb concentrations of 1 μg/ml (mean: 43.8 ± 10.8%) and 1 ng/ ml (mean: 38.4 ± 6.8%) as compared to reduced killing activity of shMICA-incubated NK cells (mean: 13.4 ± 11.2%) against non-coated HNSCC cells and anti-CD16 mAb-blocked NK cells (mean: 28.8 ± 8.9%). Based on these titration experiments, only 1 μg/ml cetuximab were applied in subsequent experiments to restore the decreased NK cell cytotoxicity (mean: 41.5 ± 6.3%) after overnight incubation with PP (high sMICA, mean of cytotoxicity: 26.1 ± 6.3%) and blocking experiments with anti-CD16 mAb-treated NK cells (mean of cytotoxicity: 24.8 ± 5.7%) (**Figure 4B**). Untreated (X-VIVO™10) and healthy plasma (HP, low sMICA)-incubated patient NK cells were defined as unaffected control cells. NK cell viability was affected exclusively by PP-treatment and HP-incubated NK cells exhibited similar viability levels as untreated effector cells (**Figure 4C**). As internal effector cell controls, we assessed the activation marker NKG2D and IgG Fc receptor (CD16) of PP-treated (high sMICA) NK cells after 24 h overnight incubation by FCM. As shown in **Figure 4D** (upper row), the expression of NKG2D was significantly decreased but the CD16 expression levels were not changed on these effector cells. Otherwise, the phenotypical characterizations of target cell parameters showed higher levels of HER1 and HER2 antigens and only low-to-moderate MICA expressions on corresponding patient HNSCC cells (**Figure 4D**, lower row).

To determine whether IL-2 potentiates the cetuximab-mediated outcome on restored effector cell cytotoxicity against HNSCC cells, we analyzed the degree of ADCC from X-VIVO™10-, PP-, and HP-treated patient NK cells against cetuximab-coated and non-labeled HNSCC cells in the presence or absence of IL-2 (**Figure 4E**). Reconstituted NK cell cytotoxicity via ADCC was detected against cetuximab-coated HNSCC cells (IL-2 and Cetmab) compared to decreased cytotoxicity of PP-incubated NK cells against non-labeled HNSCC cells independently from the presence (+IL-2) or absence of IL-2 (**Figure 4E**, middle graph). Combination of IL-2 with cetuximab-coated or non-coated HNSCC cells (IL-2 with Cetmab or IL-2 alone) revealed no improvement of ADCC from X-VIVO™10- and HP-treated NK cells, but inclusion of IL-2 significantly enhanced NK cell cytotoxicity compared to untreated NK cells (**-**IL-2) (**Figure 4E**, left/ right graphs). The cetuximab-mediated reconstitution of NK cell cytotoxicity correlated with higher IFN-γ and TNFα secretion levels (24 h PP + 1 μg/ml Cetmab + 4 h HNSCC cells) compared to PP-treated NK cells [24 h PP (high MICA) **Figure 5D**].

# Cetuximab Reconstitutes NK Cell Infiltration into HNSCC Clusters and Tumor Spheroids

To assess the capability of tumor infiltrations from sMICAaffected patient NK cells in absence or presence of 1 μg/ml cetuximab, we developed different *in vitro* models by establishing co-cultures (37°C, 5% CO2). Therefore, activated patient NK cells were pre-incubated overnight with corresponding PP (high sMICA) or healthy plasma (HP, low sMICA) served as a positive control. Afterward, the early recognition and tumor-infiltration capabilities of these treated effector cells were monitored after 1 h co-incubation with primary HNSCC cell clusters (**Figure 5**) or after 24 h in HNSCC tumor spheroids derived from five HNSCC patients to identify specific tumor-infiltrated NK cells (**Figure 6**).

Early "effector-to-target" affinities of treated NK cells against adherent HNSCC cell clusters coated or not coated with 1 μg/ ml cetuximab were monitored (10 min–1 h) by CFM. The experiments showed clearly impaired and disordered "effectorto-target" interactions and decreased HNSCC cell cluster infiltrations from PP-treated NK cells in absence of cetuximab (**Figure 5A**, one row) compared to normal tumor infiltration capabilities of HP-treated (low sMICA) NK cells, served as a positive control [**Figure 5A**, lower row, right picture (control)]. In contrast, early effector cell-dependent infiltration within 20–30 min and restored anti-tumor reaction was observed by PP-treated NK cells against cetuximab-coated HNSCC clusters after 1 h (**Figure 5A**, upper row) accompanied by raised IFNγ and TNFα medium levels (24 h PP + 1 μg/ml Cetmab + 4 h HNSCC cells, **Figure 5D**). Otherwise, the cytokine secretions of PP-inhibited NK cells against non-coated HNSCC cell clusters (without cetuximab) revealed significantly degraded IFN-γ and TNFα concentrations analyzed in medium supernatants [24 h PP (high sMICA), **Figure 5D**].

In addition, phenotypical analysis of relevant effector and target cell parameters were analyzed before co-cultivations. HNSCC cells revealed high HER1 and CD9 expression levels (not shown) and low-to-moderate MICA levels on the target cell clusters (**Figure 5B**). Otherwise, treated NK cells with corresponding PP (high sMICA, 24 h) showed a decreased NKG2D surface expression as detected by FCM analysis, while CD16 surface expression was unaffected (**Figure 5C**). In this context, sMICA and TGF-β<sup>1</sup> accumulated time-dependently in the cell culture medium as summarized in **Figure 1F** analyzed for different time periods (0–2 up to 11–14 days).

Analysis of tumor-infiltration assays were also performed with PP-treated (high sMICA) NK cells co-cultured (24 h, 37°C, 5% CO2) against non-coated HNSCC tumor spheroids. As monitored for one spheroid by time-lapse imaging, effector-to-target cell interactions and tumor infiltrations of NK cells were strongly abolished and resulted in a lack of recognition of the HNSCC spheroid by non-functional NK cells, even though they were in close proximity to the spheroids (24 h, Video S1 in Supplementary Material). However, the PP-mediated inhibition of NK cell functions was also clearly restored by cetuximab-coated HNSCC spheroids (1 μg/ml cetuximab) via multiple effector-target interactions exemplarily shown for one representative overlay photograph (**Figure 6B**, right overlay plot containing yellow regions) generated by 2D CFM. Similarly, 3D fluorescence microscopy allowed visualization of the reconstituted specific tumor infiltrations by PP-treated (high sMICA) NK cells via putative ADCC in HNSCC tumor spheroids over 24 h in the presence of 1 μg/ ml cetuximab (green-yellow areas, **Figure 6C**). Moreover, the medium supernatants of those incubated samples (PP-cultured NK cells and cetuximab-coated HNSCC tumor spheroids) revealed significantly increased IFN-γ and TNFα secretion levels (24 h PP + 1 μg/ml Cetmab + 24 h HNSCC tumor spheroids, **Figure 5D**). However, the tumor infiltration [CD45<sup>+</sup> NK cells (green signals), **Figure 6D**] and cytokine release [24 h PP (high sMICA), **Figure 5D**] of same PP-inhibited NK cells were strongly reduced in 24 h-incubated samples in absence of cetuximab and consequently non-coated HNSCC tumor spheroids.

## DISCUSSION

Dysfunctional tumor surveillance in HNSCC patients is further hampered by tumor immune escape mechanisms, which may induce dysregulation of immunocompetent cell profiles. In contrast to literature reports describing decreased NK cell numbers in HNSCC MNCs (54), we did not observe marked alterations in total NK cells among the total lymphocytes in our cohort of five relapsed HNSCC patients not currently in treatment. Previously, we demonstrated disbalances in NK cell subpopulations in 55 patients with initial and relapsed HNSCC (22). However, the immunoregulatory NK cell population responsible for stimulation of immature DCs by TNFα and IFN-γ secretion (5, 55) was significantly decreased in comparison with age-matched healthy individuals. Accordingly, Wulff et al. reported reduced regulatory NK cells in many HNSCC patients at different tumor stages (56). We detected higher levels of cytotoxic NK cell subpopulations compared to HCs, which may reflect suppressed cytotoxic interactions against HNSCC cells and limited tumor-infiltration capacities of these inhibited NK cell subsets. Additionally, we demonstrated that alterations in NK cell subtypes in our HNSCC patients were accompanied by disrupted TNFα and IFN-γ secretion. Enhanced sMICA and TGF-β1 plasma levels in HNSCC patients also correlated strongly with NKG2D-dependent dysfunction of patient NK cells, resulting in suppressed killing activity against HNSCC cells and decreased NK cell viability during cytotoxic effector-target interactions. In accordance, Bose et al. described marked alterations in the Th1/Th2 cytokine ratios and significantly increased suppressor regulatory T cells in cultured MNCs from HNSCC patients resulted in decreased cytotoxicity of HNSCC effector cells (54). According to our cytotoxicity experiments, high sMICA/TGF-β1 levels derived from HNSCC PP also contribute to diminished effector cell stability. This was also confirmed by Rossi et al. via correlation between impaired NK cell viability, effector cell cytotoxicity and decreased NKG2D and NKp46 surface expression (57). Increased NK cell susceptibility also reflects the ability of the tumor to release apoptosispromoting factors (programmed death receptor ligand, PDL-1), which can abolish several effector cell functions within the tumor microenvironment as detected on PD-1<sup>+</sup> NK cells in cancer patients (58, 59).

We translated an activation protocol from a previous phase I/II trial (Clin-Gov-No-NCT01386619), which described stimulation of allogeneic NK cells (1000 IU/ml IL-2) with resultant increased distribution of NK cell subsets and high NCR expression levels (11), to NK cells isolated from our HNSCC patients. After IL-2 stimulation, the activated patient NK cells revealed improved distribution of increased immunoregulatory NK cells (CD56bright/ CD16dim&neg), enhanced TNFα and IFN-γ secretion and upregulated NKG2D expression levels that resulted in enhanced cytotoxicity against associated HNSCC cells. Increased expression of CD56 was coincident with higher levels of NCRs as was also detected in other studies, suggesting that tumor-infiltrating NK cells were activated effector cells, but characterized by poor functionality (39, 60, 61). Indeed, our incubations (24 h) of activated NK cells with sMICA analogs or PP containing increased sMICA resulted in decreased NKG2D expression followed by impaired NKG2D-dependent killing activity against associated HNSCC cells. However, NK cell viability and expression levels of CD16, pro-apoptotic FasL and TRAIL receptors as well as for the activation marker CD57 were not affected by sMICA. In contrast, the significantly decreased levels of FasL on cultured HNSCC MNC NK cells shown by Bose et al. might be due to marked upregulation of suppressor regulatory T cells followed by enhanced TGF-β1 secretion levels (54).

Our observations regarding inhibited cytotoxic functions of sMICA-treated NK cells due to suppressed HNSCC tumor infiltration indicate suppressed migratory capacity of NK cells toward HNSCC tumors as detected by transmitted time-lapse imaging (Video S1 in Supplementary Material). This is one of the numerous effector cell functionalities, impairment of which can result in TIEM allowing unhindered tumor growth. NKG2D is an activating receptor for NK, NKT, CD8<sup>+</sup>, and γδ+ T effector cells, and down-regulation of NKG2D in HNSCC patients seems to be a crucial mechanism of immune evasion. Soluble NKG2D ligands (NKG2DL) in association with growth factors, such as TGF-β1, released from mesothelioma cell-generated exosomes were described to potentiate the down-modulation of NKG2D surface expression on activated NK cells (62). Indeed, we observed significant correlations between time-dependent high sMICA and TGF-β1 secretion levels and increased HNSCC cell growth of our tumor clusters or spheroids, which also resulted in disrupted NKG2D-mediated immunosurveillance of blocked patient NK cells as shown previously in our HNSCC patient study (22). However, the negative impact of increased TGF-β<sup>1</sup> levels detected in these HNSCC patients could be inhibited by

FIGURE 6 | Reconstitution of NK cell-dependent tumor infiltration by cetuximab. Tumor recognition and infiltration of sMICA-affected NK cells in absence or presence of cetuximab (1 μg/ml) were analyzed by co-culturing (24 h, 37°C, 5% CO2) of corresponding patient NK cells, plasma (high sMICA) and HNSCC tumor spheroids derived from HNSCC patient tumors (*n* = 5). Activated NK cells were pre-incubated (24 h) with corresponding patient plasma containing high sMICA (PP: > 500 pg/ml; diluted 1:2 with X-VIVO™10) and co-incubated (24 h) with primary HNSCC spheroids (time of cultivation: 11–14 days) in the presence of 1 μg/ml cetuximab. CD45 (green) for PP-treated patient NK cells were identified by staining with FITC-conjugated mAb, whereas tumor spheroids were labeled with anti-HER2 PE-conjugated mABs (A). Tumor-infiltrating NK cells were illustrated by overlay plots (CFM). Exactly the same orientation of CD45+ NK cells (green) and HER2+ tumor spheroids (red) showed a positive, overlapping signal in yellow [see black arrows (B)] for ADCC-related effector-target cell interactions, whereas single positive, HNSCC cells (CD45−/HER2+) are shown as red signals and single positive NK cells are depicted by green signals in the corresponding photographs (A). 3D fluorescence microscopy visualized tumor-infiltrating patient NK cells (green-yellow areas) in corresponding HNSCC tumor spheroids in the presence (C) and absence (D) of 1 μg/ml cetuximab. DAPI [blue signals (A,C,D)] was used to stain DNA from methanol-fixed effector and target cells for analyses of the nuclear morphology.

neutralization antibodies against this immunosuppressive factor, indicating that TGF-β1 seemed to potentiate the sMICA-induced decrease of patient NK cell cytotoxicity and diminished NKG2D expression. According to our results, others have demonstrated restored NKG2D expression levels on NK and CD8<sup>+</sup> T cells after tumor resection in glioma patients, which was accompanied by enhanced killing activities of those effector cells against NKG2DL-positive tumor targets (63). Interestingly, NKG2D was markedly down-regulated on activated CD8<sup>+</sup> T cells but only if CD4<sup>+</sup> T cells and NKG2DLs, such as soluble MICB, were present (64). This observation supports the hypothesis that soluble NKG2DLs played a secondary role and the down-regulation of NKG2D was primarily caused by tumor-derived TGF-β1 (63). The immunosuppressive tumor environment was responsible for the high diversity of antigen presentation patterns in stromainfiltrating NK cells and tumor-infiltrating NK cells. The effector cells showed lower CD56 levels and higher CD16 expression during cytotoxic interactions with breast cancer cells and resulted in altered NK cell phenotypes with decreased functional capacities (65). In accordance with this report, we demonstrate stable CD16 expression levels, which were not affected by sMICA analogs or HNSCC plasma containing high amounts of sMICA and TGF-β1.

NK cell subsets have a stimulating Fc receptor for binding IgG (Fcγ RIIIa), which induces ADCC and may trigger TNFα and IFN-γ secretion to finally recognize and kill antibody-coated targets (66). Since several ErbB family members, including HER1 (EGFR), HER2 and HER3, seem to be strong predictors for the outcome of HNSCC (67), we stained different tumor samples post-surgery from corresponding relapsed patients for HER1/2 and collected PB for NK cell separation as well as blood plasma for quantification of sMICA and TGF-β1. The chimeric (humanmurine) IgG1 cetuximab directly affects HER1-positive tumor cells by activation immunocompetent cells (66). Therefore, we assessed our sMICA-affected NK cells regarding to cytotoxicity and tumor infiltrations. We found improved killing activity via ADCC for sMICA-inhibited NK cells against cetuximab-coated HNSCC cells. Additionally, enhanced cytokine (TNFα and IFNγ) release was observed compared to low cell lysis of the same effector cells co-cultured with untreated tumor cells. Negative effects of sMICA and TGF-β1 on NK cell cytotoxicity were overcome by cetuximab and correlated well with high IFN-γ and TNFα secretion levels described previously in HNSCC and other cancers (53, 68). In this context, it was reported that cetuximab-induced NK cells are able to activate DC maturation markers and antigen presentation machinery via IFN-γ secretion, thus allowing initiation of adaptive immune responses by NK cell stimulated DC maturation (69). We demonstrated that combining cetuximab with IL-2 (1000 IU/ml) re-established NK cell cytotoxicity and reconstituted TNFα and IFN-γ secretion. This is in agreement with other reports that described increased ADCC activity and cytokine release in the presence of cetuximab and several cytokines (IL-2, IL-12, IL-15, and IL-21) compared to cetuximab-mediated ADCC in absence of these cytokines (53, 65, 68). Prospective studies should clarify how to improve the cetuximab effect combined with pro-inflammatory cytokines, especially IL-2. Interestingly, it was demonstrated that ADCC- and IL-2 activated NK cells were less susceptible to immunosuppression by chemotherapy or other immunosuppressive drugs, especially mycophenolate mofetil (MMF), than non-stimulated NK cells in cancer patients (68, 70).

To investigate the potency of cetuximab on cytotoxic functionalities of sMICA-inhibited NK cells, we generated HER1-positive tumor spheroids to simulate the *in vivo* tumor microenvironment. This enabled us to analyze the restoration of tumor-infiltrating capability of NK cells with down-regulated NKG2D expression levels in a well-defined system. Tumor spheroids were utilized to monitor specific migratory capability and tumor chemoattraction of effector cells, especially NK cells, over well-defined time periods (71). However, only few reports describe the effector cell-mediated immunosurveillance toward tumor spheroids. Inactivation of NK cells by clustered Ewing's sarcoma cells and cytotoxicity of γδ+ T cells against pediatric liver tumor spheroids was demonstrated (72, 73). In our tumor models, we were able to assess the cetuximab-mediated tumor infiltration from sMICAinhibited NK cells toward corresponding primary HNSCC cell clusters and tumor spheroids that expressed high levels of HER1 and HER2. HNSCC spheroids expressed only low-to-moderate levels of membrane-bound MICA, but sMICA and TGF-β<sup>1</sup> release steadily increased in supernatants of cultured HNSCC tumor spheroids and reached saturation levels after a few days of cultivation. The accumulation of immunosuppressive factors in HNSCC spheroid supernatants supported the assumption that NKG2D-mediated cytotoxicity of co-cultured patient NK cells was diminished, which corroborates the negative effect toward NK cell-mediated immunosurveillance by these tumor spheroids as described previously (71). Importantly, we detected early effector-to-target interactions displayed in small HNSCC tumor cell clusters during the first hours of co-cultivation in the presence of cetuximab. Furthermore, we successfully demonstrated cetuximab-mediated tumor infiltrations and increased TNFα and IFN-γ secretions of sMICA-inhibited NK cells in associated HNSCC spheroids using 2D- and 3D-microscopy techniques. This supported results from monolayer cultures of different effector-to-target cell ratios co-incubated for previous cytotoxicity assays. Correspondingly, others evaluated the direct localization of fluorescently stained therapeutic antibody, cetuximab-IRDye800CW, in histologic sections with tonsil, tongue, and cutaneous squamous cell carcinoma (SCC) by fluorescence immunohistochemistry (74). Interestingly, the heterogeneous composition and tumor architecture in short-term culture of HNSCC tumor slices demonstrated a high diversity of individual responses to cetuximab, but the absence of any effector cell subset (75). It was recently suggested that drug resistance to anti-EGFR therapies in HNSCC is not affected by the hypoxic tumor microenvironment within the investigated tumors (76). Indeed, the inhibition of EGFR via cetuximab reduces angiogenesis via hypoxia-inducible factor-1α and Notch1 in HNSCC (77).

## CONCLUSION

In conclusion, our current results emphasize that cetuximab is able to neutralize negative effects of TIEMs. This was shown in a corresponding effector-target system of patient NK cells, PP containing immunosuppressive factors (sMICA and TGF-β1), and associated HNSCC tumors/tumor cells as targets. This is of clinical relevance as the shedding of soluble NKG2DLs combined with secretion of immunosuppressive cytokines may promote tumor progression and is predictive for a negative prognosis in tumor patients (78).

In summary, observations of our clinical phase I/II haploidentical NK cell study for adaptive immunotherapy (Clin-Gov-No-NCT01386619) (47, 79, 80) and our current results indicate that allogeneic NK cells in combination with tumor antigen-specific mAbs, especially cetuximab, might be an innovative approach to circumvent TIEM-derived limitations against patient NK cells. Therefore, it is necessary to monitor overexpressed tumor antigens, such as HER1, and immunosuppressive ligands and the polymorphic Fcγ receptor IIIa to regenerate the tumoricidal properties of NK cells for improved therapeutic benefit.

# ACKNOWLEDGMENTS

This project was supported by the Wilhelm Sander-Stiftung für Krebsforschung, the Frankfurter Stiftung für krebskranke Kinder, the Kind-Philipp-Stiftung, and the Integrated Research and Treatment Center Transplantation (IFB-Tx, Ref. No. 01EO1302) financed by the German Federal Ministry of Education and Research.

## REFERENCES


# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://journal.frontiersin.org/article/10.3389/fimmu.2015.00543

Figure S1 | Gating strategy. Flow cytometry-based cytotoxicity assay of NK cells against singularized HNSCC cells. NK and HNSCC cells were stained with CD9 FITC, CD56 PC7, and CD45 KO mAbs. 7-AAD (7-amino-actinomycin D) was used to discriminate non-viable cells in FCM. Blue: CD45+ NK cells; red: CD9+ HNSSC cells; and grey: 7-AAD+ effector and target cells. Dot plot areas were defined by the analysis of mono-cultured NK and HNSCC cells incubated for the identical time periods. Dot plot (WBC/MNC) displays an overview of all scatter events properties, and shows differentiations to non-specifically stained debris by low forward scatter signals. Plot (CAL) illustrates the events of the region "beads" along the time course to calibrate the events for "cells/μl" and detect even sample flow. Region (Target cells) is defined to include viable CD9+ HNSCC cells, and region (Effector cells) is defined to include all CD45+ NK cells. Plot ("Cell clusters") represents the viable effector-target–cell interactions. Plot ("Viability") is defined as a region to exclude the 7-AAD+ cells. These 7-AAD+ cells are not further presented in the following dot plots by using the characteristic signal when representing the 7-AAD fluorescence against the side scatter (SSC) properties.

Video S1 | Time-lapse imaging of PP (high sMICA and TGF-beta1 levels)-treated NK cells against corresponding HNSCC tumor spheroids over a time period of 24 h. These NK cells (smaller rounded effector cells) isolated from the same HNSCC patient showed a decreased tumor recognition against associated tumor spheroids via reduced migratory capability and a decreased cytotoxicity.


multiple myeloma effect: a therapeutic target for CT-011, a novel monoclonal anti-PD-1 antibody. *Blood* (2010) **116**(13):2286–94. doi:10.1182/ blood-2010-02-271874


**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 Klöss, Chambron, Gardlowski, Weil, Koch, Esser, Pogge von Strandmann, Morgan, Arseniev, Seitz and Köhl. 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.*

# NK Cell Subgroups, Phenotype, and Functions After Autologous Stem Cell Transplantation

*Benedikt Jacobs1,2,3 , Sara Tognarelli4,5 , Kerstin Poller3 , Peter Bader4,5 , Andreas Mackensen3 and Evelyn Ullrich3,4,5\**

*1Department of Cancer Immunology, Institute for Cancer Research, Oslo University Hospital, Radiumhospital, Oslo, Norway, <sup>2</sup> The KG Jebsen Center for Cancer Immunotherapy, Institute of Clinical Medicine, University of Oslo, Oslo, Norway, 3Department of Haematology and Oncology, University Hospital Erlangen, Erlangen, Germany, 4Department of Pediatric Stem Cell Transplantation and Immunology, Children's Hospital, Johann Wolfgang Goethe-University, Frankfurt, Germany, <sup>5</sup> LOEWE Center for Cell and Gene Therapy, Johann Wolfgang Goethe-University, Frankfurt, Germany*

#### *Edited by:*

*Francisco Borrego, BioCruces Health Research Institute and Cruces University Hospital, Spain*

#### *Reviewed by:*

*Roberto Biassoni, Istituto Giannina Gaslini, Italy Jacques Zimmer, Luxembourg Institute of Health, Luxembourg*

> *\*Correspondence: Evelyn Ullrich evelyn.ullrich@kgu.de*

#### *Specialty section:*

*This article was submitted to NK Cell Biology, a section of the journal Frontiers in Immunology*

*Received: 22 September 2015 Accepted: 02 November 2015 Published: 24 November 2015*

#### *Citation:*

*Jacobs B, Tognarelli S, Poller K, Bader P, Mackensen A and Ullrich E (2015) NK Cell Subgroups, Phenotype, and Functions After Autologous Stem Cell Transplantation. Front. Immunol. 6:583. doi: 10.3389/fimmu.2015.00583*

High-dose chemotherapy with consecutive autologous stem cell transplantation (auto-SCT) is a well-established treatment option for patients suffering from malignant lymphoma or multiple myeloma. Natural killer (NK) cells are an important part of the immune surveillance, and their cell number after autoSCT is predictive for progression-free and overall survival. To improve knowledge about the role of NK cells after autoSCT, we investigated different NK cell subgroups, their phenotype, and their functions in patients treated with autoSCT. Directly after leukocyte regeneration (>1000 leukocytes/μl) following autoSCT, CD56++ NK cells were the major NK cell subset. Surprisingly, these cells showed unusually high surface expression levels of CD57 and killer Ig-like receptors (KIRs) compared to expression levels before or at later time points after autoSCT. Moreover, these NK cells strongly upregulated KIR2DL2/3/S2 and KIR3DL1, whereas KIR2DL1/ S1 remained constant, indicating that this cell population arose from more immature NK cells instead of from activated mature ones. Remarkably, NK cells were already able to degranulate and produce IFN-γ and MIP-1β upon tumor interaction early after leukocyte regeneration. In conclusion, we describe an unusual upregulation of CD57 and KIRs on CD56++ NK cells shortly after autoSCT. Importantly, these NK cells were functionally competent upon tumor interaction at this early time point.

Keywords: NK cells, CD57, KIR, autologous stem cell transplantation, CD107a expression, IFN-γ production

# INTRODUCTION

Natural killer (NK) cells are an important part of the innate immune system and are able to kill virus-infected or malignantly transformed cells (1). Their important role in tumor surveillance has been demonstrated in many different tumor models (1). NK cell cytotoxicity is regulated by a diverse repertoire of inhibitory and activating receptors. Inhibitory receptors, such as killer Ig-like receptors (KIRs) and the C-type lectin-like receptor NKG2A, recognize

**Abbreviations:** HDC, high-dose chemotherapy; (auto)SCT, (autologous) stem cell transplantation; MM, multiple myeloma; NHL, non-Hodgkin lymphoma; TP1–3, time point 1–3; KIRs, killer Ig-like receptors; NK cell, natural killer cell.

different alleles of HLA molecules (HLA-A, B, and C by KIRs and HLA-E by NKG2A) on healthy cells. In contrast, many tumor cells downregulate their HLA molecules to evade T cell recognition, making them more susceptible to NK cell killing (2). Additionally, tumor cells may express stress-induced molecules, such as MHC I chain-related molecule A/B or UL-16-binding proteins, which are ligands for the activating NK cell receptor NKG2D (3, 4).

High-dose chemotherapy (HDC) with consecutive autologous stem cell transplantation (autoSCT) is an effective and wellestablished treatment option for patients suffering from multiple myeloma (MM) (5) or malignant lymphoma (6–9). Before treatment with the myeloablative chemotherapy, hematopoietic stem cells are collected from peripheral blood and frozen. Following HDC, these cells are thawed and given back to the patient in order to shorten the time of aplasia, thereby reducing the infection and blood transfusion rates.

Many reports have demonstrated the important role of the absolute lymphocyte count after HDC/autoSCT (10). It has been shown that an absolute lymphocyte count >500/μl is associated with improved overall and progression-free survival in patients with Hodgkin lymphoma (11), non-Hodgkin lymphoma (NHL) (12), acute myeloid leukemia (13), MM (12), and metastatic breast cancer (14). By analyzing the different lymphocyte subsets at day 15 following autoSCT, a clear correlation between improved overall survival and progression-free survival could only be found for NK cell counts >80/μl. No correlation was found for any other lymphocyte subset (15). In a more recent study, improved median overall and progression-free survival as well as the NK cell count at day 15 after HDC/autoSCT were all associated with an increased IL-15 concentration at day 15 of ≥76.5 pg/ml for NHL patients receiving HDC/autoSCT (16).

Because there is no information available regarding the detailed analysis of NK cell subsets or function early after HDC/ autoSCT, in our study, we prospectively investigated the major NK cell subsets directly after leukocyte recovery (leukocytes >1000/μl) and also at later time points after HDC/autoSCT in patients with different lymphoproliferative diseases. Moreover, we further analyzed the different NK cell subsets, evaluating their education and differentiation markers, as well as their functional properties, such as cytokine/chemokine production and degranulation capacity.

#### MATERIALS AND METHODS

#### Patients' Characteristics and Study Design

This study was carried out in accordance with the recommendations of the local ethics committee of the University of Erlangen, and all patients gave written informed consent in accordance with the Declaration of Helsinki. Patients who suffered from MM or malignant lymphoma and received HDC/autoSCT were included. Blood was taken from these patients at three different time points. Time point 1 (TP1) was before the start of the HDC and at least 3 weeks after the last chemotherapy. The second time point (TP2) was 1–2 days after leukocyte regeneration (>1000 leukocytes/μl) following autoSCT, and the third time point (TP3) was after at least 2 weeks following leukocyte recovery.

#### Reagents

For NK and K562 cell culture, we used full media containing RPMI 1640 media (Gibco®) supplemented with 10% FBS, MEM non-essential amino acids (1%), sodium pyruvate (1%), l-glutamine (1%; all from PanBiotech), and penicillin/streptomycin (1%; Thermo Fischer Scientific). For the washing steps, we used Dulbecco's phosphate-buffered saline (DPBS; Gibco®).

To analyze the different leukocyte subsets, CD3, CD14, CD16, CD19, CD45, and CD56 antibodies with different fluorochromes from Becton Dickinson (BD) were used. For detailed NK cell subset analyses, we used anti-KIR2D-, KIR3DL1/2-, and KIR2DL1/ S1-PE (Miltenyi), KIR2DL1-PerCP (R&D), KIR2DL2/3/S2-APC (Beckman Coulter), KIR3DL1 PE-Vio770 (Miltenyi), NKG2A FITC (Miltenyi), and CD57 APC (BD). For the KIR staining, the clones of the antibodies were selected according to Czaja et al. (17), and a sequential staining protocol was used as described by Beziat et al. (18). For intracellular staining, we used IFN-γ PE-Cy-7 and MIP-1β APC-H7 (BD). To exclude dead cells, 7-AAD (BD) for extracellular and Fixable Viability Dye eFluor® 520 (eBioscience) for intracellular staining were used.

## PBMC Preparation, Freezing, and Thawing

Blood samples were obtained from the patients at the indicated time points. PBMCs were isolated by performing a Ficoll density centrifugation of whole blood samples, and then the PBMCs were frozen (5 × 106 PBMCs/ml freezing media containing 90% FCS + 10% DMSO; from Sigma) until they were used.

For thawing, tubes were incubated at room temperature and gently thawed by re-suspending the cells in prewarmed full media (without FCS). The cells were washed twice and counted before being used for further analysis.

## Extra- and Intracellular Antibody Staining

For surface staining, the cells were incubated with different antibody cocktails for 10′ at 4°C; then, they were washed and either fixed in BD CytoFix™ solution or further processed for intracellular staining using the BD Cytofix/Cytoperm™ kit. Briefly, the cells were first incubated for 30′ at 4°C with Fixable Viability Dye eFluor® 520 (eBioscience), and then they were washed and fixed for 20′ at 4°C in 100 μl BD Cytofix™ solution. Subsequently, the cells were washed in BD Cytoperm™ solution and incubated with the indicated intracellular antibodies for 20′ at 4°C. Finally, the cells were washed, placed into BD CytoFix™ solution, and analyzed with a BD FACSCanto II™ or Canto10c™ using the FlowJo® software (FlowJo, LLC) was used to analyze the FACS data.

#### Functional Assays

After thawing, the cells were placed into full media supplemented with 100 IU IL-2/ml (Proleukin®, Novartis) in a 96-U well plate overnight. The cells were harvested, washed, counted, and incubated with K562 cells (ratio 1:1) in full media in a 96-U well plate for 4 h. CD107a APC (BD) was added at the start of the coculturing period and BD GolgiStop™ (BD) was added after 1 h for the rest of the incubation time.

#### Statistics

For statistical analysis, we used GraphPad Prism® software. In all graphs, the mean and SD were calculated and plotted. For comparison between matched samples, we used a Wilcoxon test, whereas for non-matched samples, we performed a Mann– Whitney test. Statistical significance is indicated with the *p*-values (\*<0.05; \*\*<0.01; \*\*\*<0.001; \*\*\*\*\*<0.0001).

# RESULTS

#### Patients' Characteristics and Leukocyte Subsets

Peripheral blood samples from 32 different patients collected at three specific time points (TP1–3, as described in Section "Materials and Methods") were available for the analysis of leukocyte subsets. The basic patient characteristics are summarized in **Table 1**. The ratio between male and female patients was approximately 2:1. Half of the patients suffered from MM. The average age at HDC/autoSCT was 56.7 years. The average time between SCT and TP2 was 11.6 days, whereas the time between collecting samples at TP2 and TP3 was 38.8 days. All three values were normally distributed.

Before the start of HDC/autoSCT at TP1, CD3<sup>+</sup>CD56<sup>−</sup> T cells and CD14<sup>+</sup> monocytes were the two major leukocyte subsets in all patients (**Figure 1A**; for gating strategy, see Figure S1 in Supplementary Material). While CD14<sup>+</sup> monocytes were the major subset early after leukocyte recovery after HDC/autoSCT at TP2 (**Figures 1A,B**), CD3+CD56− T cells were the major subset at TP3 (**Figures 1A,B**). In contrast, the NK cell percentages within the leukocyte population significantly decreased from TP1 to TP2 (*p*-value: 0.0175) but recovered to the initial value at TP3 (*p*-value TP2/TP3: 0.0263; TP1/3: 0.19; **Figures 1A,B**).

By correlating the NK cell dynamics at the three different time points with clinical data, we observed that the fold change of the NK cell percentage within the leukocyte population between TP1 and TP2 (ratio TP2/TP1) significantly differed between patients having a time period of ≤11 days between SCT and TP2 and those having a period of >11 days (*p*-value: 0.04). When the time period was >11 days, no decrease in the


NK cell percentage within the leukocyte population at TP2 was observed. Moreover, another significant difference was observed when comparing the fold change in the NK cell percentage between TP1 and TP2 in patients who were refractory/ recurrent or not at 1 year after SCT (*p*-value: 0.0258). Patients with recurrent or refractory disease did not have a decrease in their fold change ratio (TP2/TP1), while patients without recurrent/refractory disease did have a decrease at 1 year after SCT. Additionally, the fold increase in the NK cell percentage between TP2 and TP3 (TP3/TP2 ratio) was more pronounced when the time period between TP2 and TP3 was ≤38 days (*p*-value: 0.12; **Figure 1C**).

No differences were observed when analyzing the patients' age or hematological malignancies in relation to the fold changes of the NK cell percentages within the leukocyte population between the three different time points (Figure S2 in Supplementary Material).

# CD56++CD16−/<sup>+</sup> NK Cells are the Major Subset at Leukocyte Recovery

Next, we analyzed the different NK cell subsets based on their CD56 and CD16 expression. NK cells were divided into CD56++CD16<sup>−</sup> or CD16<sup>+</sup> and CD56<sup>+</sup>CD16++ NK cells (see Figure S1 in Supplementary Material). The CD56++CD16<sup>+</sup> population has been reported to be an intermediate state between CD56++CD16<sup>−</sup> and CD56+CD16++ NK cells (19, 20). The CD56+CD16− population was excluded, as it was shown that this population could be induced by cryopreservation (21).

At TP1, the major NK cell subset was the CD56<sup>+</sup>CD16++ NK cell population (71.86%), followed by the CD56++CD16<sup>+</sup> (17.71%) and the CD56++CD16<sup>−</sup> (10%) populations (**Figure 2A**). After leukocyte regeneration (TP2), the CD56<sup>+</sup>CD16++ cells significantly decreased (39.98%; *p*-value: <0.0001), whereas both CD56++ NK cell subsets significantly increased (CD16<sup>−</sup>: 22.85%, *p*-value: <0.0001; CD16<sup>+</sup>: 36.51%, *p*-value: <0.0001). At TP3, the levels of CD56<sup>+</sup>CD16++ NK cells increased again (64.85%; *p*-value: <0.0001) but remained reduced in contrast to the starting levels at TP1 (*p*-value: 0.0064). Conversely, the levels of CD56++CD16<sup>−</sup> (14.41%; *p*-value: 0.0006) and CD16<sup>+</sup> (20.66%; *p*-value: <0.0001) cells decreased but remained significantly elevated compared to the starting values [*p*-values: 0.0184 (CD16<sup>−</sup>) and 0.0428 (CD16<sup>+</sup>); **Figure 2B**].

The ratio between mature CD56<sup>+</sup> and more immature CD56++ NK cells, which is approximately 10 within healthy donors (data not shown), was already significantly reduced within our patients before HDC/autoSCT (TP1; ratio: 3.707) because all patients had previously received chemotherapy. Nevertheless, the ratio significantly dropped at TP2 (ratio: 1.157; *p*-value: <0.0001) and did not recover to initial values at TP3 [ratio: 2.49; *p*-value: 0.0001 (TP2–3); 0.0111 (TP1–3); **Figure 2C**].

The CD56+/CD56++ ratios at TP2 and TP3 were both independent of the time period between SCT and TP2 (≤11 days: 1.247 vs. >11 days: 1.041; *p*-value: 0.392) and between TP2 and TP3 (≤38 days: 2.59 vs. >38 days: 2.32; *p*-value: 0.504; **Figure 2D**). Within the group of MM patients, the CD56<sup>+</sup>/ CD56++ ratio was lower compared to lymphoma patients, but

decreased below the starting levels at TP3 (34.6%; *p*-value TP2/TP3: <0.0001; TP1/3: 0.0088). In contrast, the T cell percentages decreased from TP1 (37.89%) to TP2 (22.5%; *p*-value: <0.0001), but they increased again at TP3 (48.49%) above the initial values to become the major leukocyte subset at TP3 (*p*-value TP2/TP3: <0.0001; TP1/3: 0.0093). The NK cell percentages decreased from TP1 (8.94%) to TP2 (7.32%) but reached the initial levels again at TP3 (8.29%). (C) Patients with a time period of ≤11 days between SCT and TP2 had a decrease from TP1 to TP2 within their NK cell percentage (ratio TP2/TP1: 0.49), in contrast to patients with a time period >11 days (ratio TP2/TP1: 1.07). Similarly, patients with no recurrent/refractory disease 1 year after SCT had a decrease of their NK cell percentage at TP2 (ratio TP2/TP1: 0.66), as opposed to patients who were recurrent/refractory at 1 year (ratio TP2/TP1: 1.28). The increase of the NK cell percentage from TP2 to TP3 was more pronounced in patients with a time period of ≤38 days between TP2 and TP3 (ratio TP3/TP2: 3.73) than in patients with a time period of >38 days (ratio TP3/TP2: 1.97).

it was only significantly lower at TP2 (1.656 vs. 0.6578; *p*-value: 0.0019; **Figure 2D**). In contrast, the patients' age and relapse/ refractory status at 1 year after autoSCT seemed to have no impact on the CD56<sup>+</sup>/CD56++ ratio at any time point (Figure S3 in Supplementary Material).

# Increased Levels of CD57 and KIR Expression After Leukocyte Regeneration

Next, we analyzed the expression of markers for NK cell education and differentiation at the indicated time points.

As expected, the NKG2A expression on all NK cells increased from TP1 (67.63%) to TP2 (76.51%; *p*-value: 0.0179), and the percentage of NKG2A<sup>+</sup> NK cells remained elevated above the starting values until TP3 (77.67%; *p*-value: 0.0009). Further NK cell subset analyses revealed that the observed early increase in NKG2A at TP2 was potentially an effect of the elevated CD56++ population expressing higher levels of NKG2A than the CD56<sup>+</sup>CD16++ population because the NKG2A expression did not significantly differ within the distinct subsets between TP1 and TP2. In contrast, the percentage of NKG2A-expressing NK cells was significantly elevated in all subsets at TP3 compared to

starting levels (TP1: 71.86%). Moreover, the percentages of CD56++CD16−/+ NK cells were markedly increased at TP2 (CD16−: 22.85%; CD16+: 36.51%) and remained elevated at TP3 (CD16−: 14.41%; CD16+: 20.66%) compared to the TP1 values (CD16−: 10%; CD16+: 17.71%). (C) The ratio of CD56+/CD56++ NK cells at TP1 was already lower than in healthy controls (approx 10; data not shown), with a ratio of 3.707, and decreased further at TP2 (1.157). Although the ratio increased at TP3 (2.49), it was still lower compared to TP1 and healthy control samples. (D) There were no significant differences when analyzing the CD56+/CD56++ ratio at TP2 with regard to the time period between SCT and TP2 (≤11 days: 1.247; >11 days: 1.041) or at TP3 with regard to the period between TP2 and TP3 (≤38 days: 2.59; >38 days: 2.32). A significant difference of the CD56+/CD56++ ratio between lymphoma and myeloma patients was only observed at TP2 (lymphoma patients: 1.656; myeloma patients: 0.6578).

TP1 [*p*-value: 0.0054 (CD56++CD16<sup>−</sup>); 0.0259 (CD56++CD16<sup>+</sup>); 0.0031 (CD56<sup>+</sup>CD16++); **Figure 3A**].

CD57<sup>+</sup> NK cells, known to define terminally differentiated NK cells, significantly increased from TP1 (43.53%) to TP2 (56.66%; *p*-value: 0.0163) in all NK cells but decreased to the initial values at TP3 (39.13%; *p*-value: 0.4274). Surprisingly, CD57 expression was significantly increased within the CD56++CD16+/− population at TP2 (CD16<sup>−</sup>: 32.98%; CD16<sup>+</sup>: 57.78%, *p*-value: <0.0001) but decreased again from TP2 to TP3 (CD16<sup>−</sup>: 5.592%; CD16<sup>+</sup>: 19.06%, *p*-value: <0.0001). Nevertheless, the percentage of CD57<sup>+</sup> cells was still elevated in contrast to the starting values at TP1 [CD16<sup>−</sup>: 4.1%; CD16<sup>+</sup>: 15.58%, *p*-values: 0.0102 (CD16<sup>−</sup>); 0.0214 (CD16<sup>+</sup>)]. CD57 expression within the CD56<sup>+</sup>CD16++ population was also elevated at TP2 (72.15%; *p*-value: 0.0006) but to a much lesser extent (**Figure 3B**).

Strikingly, the percentage of KIR<sup>+</sup> NK cells remained constant over time (TP1: 42.45%; TP3: 44.09%; *p*-value: 0.463), even shortly after leukocyte regeneration (TP2: 40.77%; *p*-value: 0.4106). Of note, the KIR expression within the CD56++CD16<sup>+</sup>/<sup>−</sup> NK cell population was markedly increased at TP2 (CD16−:

19.35%; CD16<sup>+</sup>: 42.14%; *p*-value: <0.0001 for both populations) and remained elevated at TP3 (CD16<sup>−</sup>: 10.43%; CD16<sup>+</sup>: 28.14%) in comparison to the starting values at TP1 [CD16<sup>−</sup>: 7.82%; CD16<sup>+</sup>: 22.44%; *p*-values: 0.0083 (CD16<sup>−</sup>); 0.0015 (CD16<sup>+</sup>)]. KIR expression within the CD56<sup>+</sup>CD16++ population remained stable throughout all time points (TP1: 51.42%, TP2: 52.89%, and TP3: 55.23%; **Figure 3C**).

# Elevated CD57 and KIR Expression is Age Dependent

Additionally, we analyzed the influence of different clinical factors on the expression levels of NKG2A, CD57, and KIRs within the different NK cell subsets. As CD57 expression levels within healthy individuals are known to be age dependent (22), we compared CD57 expression within the younger (≤56 years) and older (>56 years) patient populations. Notably, there was a significant difference in the percentage of CD57<sup>+</sup> cells within the CD56++CD16<sup>−</sup> population at TP2 (≤56 years: 23.98%; >56 years: 42.8%; *p*-value: 0.0056), whereas this effect was not observed within the CD56++CD16<sup>+</sup> or the CD56<sup>+</sup>CD16++ population (**Figure 4A**).

Similar, by analyzing KIR expression within the CD56++CD16<sup>+</sup>/<sup>−</sup> populations, we observed a significant difference in their expression at TP2 between the two age groups [≤56 years: 14.74% (CD16<sup>−</sup>) and 37.29% (CD16<sup>+</sup>); >56 years: 24.39% (CD16<sup>−</sup>) and 47.44% (CD16<sup>+</sup>); *p*-values: 0.0088 (CD16<sup>−</sup>), 0.0317 (CD16<sup>+</sup>)]. No age-dependent difference was found for KIR expression within the CD56<sup>+</sup>CD16++ population at TP2 but at TP3 (≤56 years: 50.51%; >56 years: 60.37%; *p*-value: 0.0225; **Figure 4B**).

In contrast, there was no difference in NKG2A expression throughout all age groups, time points, or NK cell populations (**Figure 4C**).

We also addressed CD57, KIR, and NKG2A expressions in different subsets with regard to recurrent/refractory disease at 1 year after SCT and hematological malignancy. There were no differences in the expression of the three markers regarding the rate of relapsed/refractory disease throughout all NK cell subsets and time points, except for CD57 expression at TP1 within the CD56++CD16− NK cell population, which was higher in patients

(≤56 years: 70.25% vs. >56 years: 74.23%). (B) The percentage of KIR+ NK cells at TP2 differed significantly between the two age groups within the CD56++ subset (≤56 years: CD56++CD16−: 14.74% and CD56++CD16+: 37.29% vs. >56 years: CD56++CD16−: 24.39% and CD56++CD16+: 47.44%). A significant difference within the CD56+CD16++ subset was observed at TP3 (≤56 years: 50.51% vs. >56 years: 60.37%). (C) No difference was observed in NKG2A expression within the different NK subsets and age groups.

with no recurrent/refractory disease 1 year after SCT (*p*-value: 0.0083; Figure S4A in Supplementary Material).

Furthermore, patients with MM had higher NKG2A expression within the CD56++CD16<sup>−</sup>/<sup>+</sup> subset at TP2 compared to lymphoma patients. No further differences were observed (Figure S4B in Supplementary Material).

#### Detailed KIR Expression Analysis

We next performed an extended KIR analysis within the samples of five additional patients, who were not included into the original analysis group because we did not have a sample from TP3. We analyzed the expression levels and distribution of global KIR expression (anti-KIR2D and anti-KIR3DL1/2), as well as KIR2DL1/S1, KIR2DL2/3/S2, and KIR3DL1, within the different subsets at TP1 and TP2. Notably, global KIR expression levels were upregulated within both CD56++ NK cell populations at TP2 compared to TP1, whereas they remained stable within the CD56<sup>+</sup>CD16++ NK cell population, confirming the results from the former analyzed patient cohort (**Figures 5A–C**). Within the CD56<sup>+</sup>CD16++ population, no clear differences between the expression levels of the different KIR subsets were observed comparing TP1 and TP2 (**Figure 5C**). In contrast, both CD56++ NK cell subsets upregulated their KIR2DL2/3/S2 and KIR3DL1 expression levels from TP1 to TP2, whereas the KIR2DL1/S1 levels

remained stable between the two time points (**Figures 5A,B**). Moreover, the proportion of the different KIR subsets within the global KIR population changed within the CD56++CD16<sup>−</sup> population, with KIR2DL2/3/S2 being the dominant KIR subset at TP1 and KIR3DL1 being the dominant one at TP2 (**Figure 5A**; pie chart). Within the CD56++CD16<sup>+</sup> population, KIR2DL2/3/S2 was the dominant KIR population at TP1 and was the only one to increase from TP1 to TP2 (**Figure 5B**; pie chart).

# NK Cell Function is Preserved After Leukocyte Recovery

Finally, we analyzed the functions of the different NK cell subsets at the three time points before and after HDC/autoSCT. After an overnight incubation with low-dose IL-2 (100 IU/ml) and 4 h of coculture with K562 cells, we investigated the cytokine (IFN-γ) and chemokine (MIP-1β) productions, as well as the NK cell degranulation (CD107a expression) capacity (for gating strategy see Figure S5 in Supplementary Material). Due to very low cell numbers at TP2, functional analysis was only possible in a subset of all included patients (*n* = 17).

As expected, CD56++CD16<sup>−</sup> NK cells were the main subset to produce IFN-γ upon interaction with K562 cells at all three time points (**Figure 6A**). The percentage of IFN-γ-positive CD56++CD16<sup>−</sup> NK cells was slightly decreased at TP2 compared to TP1 but significantly increased from TP2 to TP3 (*p*-value: 0.0008). Similarly, MIP-1β- and CD107a-positive CD56++CD16<sup>−</sup> cells remained constant between TP1 and TP2, whereas their percentages increased from TP2 to TP3 [*p*-values: 0.0056 (MIP-1β) and 0.0232 (CD107a)].

Whereas IFN-γ production was only marginal at all three time points within the CD56++CD16<sup>+</sup> NK cell population (**Figure 6B**), MIP-1β- and CD107a-positive cells had similar percentages at TP1 and TP2. Both percentages significantly increased from TP2 to TP3 (*p*-value: 0.0079 for both).

FIGURE 6 | (A) The CD56++CD16− subset revealed an increase of IFN-γ-, MIP-1β-, and CD107a-positive cells between TP2 and TP3 [TP2: 4.9% (IFN-γ), 15.17% (MIP-1β), 34.37% (CD107a), TP3: 11.58% (IFN-γ), 22.58% (MIP-1β), 43.82% (CD107a)], while the percentages between TP1 and TP2 were similar [TP1: 7.92% (IFN-γ), 16.89% (MIP-1β), and 32.7% (CD107a)]. (B) IFN-γ-positive cells within the CD56++CD16+ population were only marginal at all three time points (TP1: 1.29%; TP2: 1.01%; TP3: 1.72%), whereas MIP-1β- and CD107a-positive cells increased from TP2 to TP3 [TP2: 13.26% (MIP-1β), 25.78% (CD107a); TP3: 18.37% (MIP-1β), 35.26% (CD107a)] but not from TP1 to TP2 [TP1: 14.79% (MIP-1β), 27.64% (CD107a)]. (C) No IFN-γ-positive cells were detectable within the CD56+CD16++ population at any of the three time points. MIP-1β-positive cells remained stable throughout all three time points (TP1: 8.93%; TP2: 10.06%; TP3: 10.18%) which is similar to CD107a-expressing cells (TP1: 14.26%; TP2: 16.47%; TP3: 19.33%). (D) Patients were grouped according to the duration between SCT and TP2 (≤11 vs. >11 days). While there were no differences between the percentage of CD107a-expressing cells within the CD56++CD16− population at TP1 and TP2, the percentage of CD107a-positive NK cells at TP2 was lower within patients with a longer duration between SCT and TP2 within the CD56++CD16<sup>+</sup> (≤11 days: 30.74%; >11 days: 16.68%) and the CD56+CD16++ (≤11 days: 19.34%; >11 days: 11.21%) populations.

Within the CD56<sup>+</sup>CD16++ NK cell subsets, the percentage of MIP-1β- and CD107a-positive NK cells after coincubation with K562 cells remained constant at all three time points, whereas no IFN-γ-positive NK cells were detected at any time point (**Figure 6C**).

Although the number of available patient samples was low, we tried to correlate the NK cell function values with clinical data. Remarkably, there was an impact of the duration from SCT to TP2 (≤11 vs. >11 days), as the percentage of CD107a-positive cells within the CD56++CD16<sup>+</sup> and CD56<sup>+</sup>CD16++ populations was significantly lower when the time period between SCT and TP2 exceeded 11 days [*p*-values: 0.0111 (CD56++CD16<sup>+</sup>); 0.027 (CD56+CD16++); **Figure 6D**]. In investigating age-dependent differences (≤56 vs. >56 years), we observed that patients older than 56 years tended to have slightly higher percentages of IFN-γ- and MIP-1β-positive CD56++CD16− cells at TP1, although the difference was not significant. Furthermore, no significant differences in the presence of MIP-1β- or CD107a-positive CD56++CD16<sup>+</sup> or CD56<sup>+</sup>CD16++ NK cells between the two age groups were observed at all three time points, although older patients tended to have lower CD107a expression within the CD56<sup>+</sup>CD16++ subset at TP3 (*p*-value: 0.074). Notably, the observed increase of CD107a-positive CD56<sup>+</sup>CD16++ cells from TP2 to TP3 was only present in younger but not older patients [≤56 years: 15.69% (TP2) and 24.05% (TP3), *p*-value: 0.1; >56 years: 17.16% (TP2) and 15.14% (TP3), p-value: 0.82; Figure S6 in Supplementary Material].

# DISCUSSION

In the setting of HDC/autoSCT, it has been demonstrated that a rapid NK cell recovery at 1 month after HDC/autoSCT is associated with a prolonged progression-free survival in MM (23) and NHL patients (16). In those studies, the absolute NK cell count (cells/μl) at 1 month or 15 days after HDC/autoSCT was investigated, whereas in our study, we analyzed the NK cell percentage within the leukocyte population in correlation with the day of leukocyte recovery following autoSCT.

Our data demonstrate that the percentage of NK cells within the leukocyte population decreased after leukocyte recovery but increased to the initial levels over time. Notably, when the time period between SCT and TP2 was >11 days, indicating a delay in leukocyte recovery, the decrease within the NK cell percentage was lower. This may be explained by the fact that leukocyte recovery (white blood cell count >1000/μl) after SCT is mainly due to the recovery of neutrophil granulocytes (10), and their recovery can be delayed in contrast to NK cell recovery (24). Moreover, the higher increase of the NK cell percentage from TP2 to TP3 in patients from whom the third blood sample was collected ≤38 days after TP2 indicates that the NK cell percentage within the leukocytes increases much more within the first month after leukocyte recovery and then decreases again. Similar results have been demonstrated by Rueff et al. by analyzing the absolute NK cell count numbers (cells/μl) 1, 3, 6, 12, and 24 months after SCT. Here, the NK cell numbers first increased 1 month after SCT, but then they decreased again until 6 months after SCT (23). The absence of a decrease in the NK cell percentage at TP2 within patients with recurrent/refractory disease at 1 year after SCT could be explained by the fact that most of these patients (4/6; 66.6%) had a time period of >11 days from SCT to TP2, unlike non-relapsing/refractory patients (8/21; 38%).

In line with other NK cell reconstitution studies after SCT (24–26), we observed elevated percentages of the more immature CD56++CD16<sup>+</sup>/<sup>−</sup> NK cell subsets shortly after SCT, decreasing only slowly at later time points. As the ratio between CD56<sup>+</sup>/ CD56++ NK cells did not differ with shorter (≤38 days) or longer (>38 days) time periods between TP2 and TP3, we assume that normalization of the NK cell subset distribution takes much longer than the recovery of the NK cell numbers. Similar observations have been made within patients receiving an allogeneic SCT after reduced-intensity conditioning (26). The conditioning and former treatment regimens could explain the different NKG2A<sup>+</sup> NK cells ratios between myeloma and lymphoma patients at TP2.

Moreover, the percentage of NKG2A<sup>+</sup> NK cells was increased after HDC/autoSCT and remained high even after several months, as it has been recently demonstrated by Pical-Izard et al. after allogeneic SCT (26). In contrast, we could demonstrate a highly significant increase of CD57<sup>+</sup> and KIR<sup>+</sup> NK cells, mainly within the CD56++CD16<sup>+</sup>/<sup>−</sup> subsets at TP2. This effect has not been described thus far within the literature because most of the studies have evaluated NK cell subsets 1 month after SCT (23–27). At this time point, the percentage of CD57<sup>+</sup> and KIR<sup>+</sup> NK cells had already decreased back toward normal levels within our study group. The prolonged immature phenotype in the Pical-Izard study may be attributed to GVHD prophylaxis, especially cyclosporine A (CSA). Vukicevic et al. investigate the NK cell phenotype at an equally early time point after allogeneic stem cell transplantation as we did but did not detect an upregulation of KIRs within the CD56++ subsets (28), which might be due to the allogeneic transplantation setting. Acquisition of CD57 and KIRs as well as downregulation of NKG2A has been demonstrated as signs of NK cell differentiation and maturation (22). It is known that CD57 expression increases with age within the CD56<sup>+</sup>CD16++ population (29), whereas we did not observe any age-related differences within this NK cell subset at any time point. Nevertheless, an age-dependent difference was observed at TP2 within the CD56++CD16<sup>−</sup> population.

Therefore, the question arises whether these CD56++ cells described here represent more mature new NK cells or are just activated "old" NK cells that might have increased their CD56 expression and lost CD16 on their surface. As this phenotype is reset after at least 2 weeks after leukocyte regeneration, one could argue that the phenotype shift is due to the cytokine milieu during HDC/autoSCT. Indeed, there have been several reports about increased cytokine concentrations during allogeneic and autologous SCT (16, 25, 30) shaping the NK cell phenotype. It has been reported that CD56<sup>+</sup>CD16++ NK cells are capable of up-regulating CD56 expression upon IL-15 (28) or IL-12 stimulation (31), of which IL-15 is known to be increased during HDC/autoSCT (16). Furthermore, different groups demonstrated a downregulation of CD16 by metalloproteinases, which can be induced by IL-2 (32, 33). In general, the combined effect of CD56 upregulation upon IL-15 stimulation and the loss of CD16 through IL-2-stimulated upregulation of metalloproteinases might result in the observed CD56++CD16<sup>+</sup>/<sup>−</sup> NK cell phenotype at TP2. Nevertheless, the CD56++ NK cells described in our study upregulated KIR3DL1 and KIR2DL2/3/S2, while KIR2DL1/S1 remained stable. KIR3DL1 and KIR2DL2/3 are the first KIRs expressed after SCT, whereas KIR2DL1 is upregulated quite late (34, 35). Therefore, we could assume that the KIR upregulation was due to the generation of "fresh/new" NK cells and was not due to a shift of "old" NK cells to a CD56++ phenotype because we should have observed no changes within the KIR subtype distribution between TP1 and TP2 if the cells were derived from the "old" NK population. Moreover, we observed a slight upregulation of CX3CR1 at TP2 (data not shown). CX3CR1 expression is associated with a more mature and differentiated NK cell phenotype within healthy donors (36). In contrast to the observed CD56 upregulation, which can be explained by IL-15 stimulation (28), CX3CR1 is known to be downregulated upon IL-15 stimulation (36), which contradicts the idea that CD56++ NK cells with a mature phenotype (CD57<sup>+</sup>, CX3CR1<sup>+</sup>, and KIR<sup>+</sup>) have arisen from CD56dim NK cells. In future studies, it would be very interesting to investigate which factors are responsible for this NK cell phenotype because protocols for inducing NK cell maturation and differentiation have yet to be optimized.

Most importantly, we analyzed NK cell functional activity directly after leukocyte recovery after HDC/autoSCT. Upon interaction with K562 tumor cells, the percentage of IFN-γ- and MIP-1β-positive CD56++CD16<sup>+</sup>/<sup>−</sup> NK cells did not differ between TP1 and TP2. This result demonstrates that NK cells are capable of recognizing tumor cells and inducing cytokine and chemokine production, even at a very early time point after HDC/autoSCT. Moreover, the degranulation capacity of the CD56<sup>+</sup>CD16++ NK cell subset, known to be mainly responsible for NK cell cytotoxicity (37), remained stable throughout the whole time period until TP3, indicating that these NK cells were able to kill tumor cells at an early time point after SCT. This finding is consistent with other studies in which patients received allogeneic SCT. For example, in the setting of HLA-matched SCT after reducedintensity conditioning, it has been demonstrated that the NK cell degranulation and chemokine production capacity was similar to healthy controls as early as 1 month after SCT. In contrast to our data, IFN-γ production upon interaction with K562 cells was significantly reduced after SCT compared to healthy donors (26). These differences might be due to the use of immunosuppressive drugs such as CSA because it has been demonstrated that CSA is able to reduce IFN-γ production upon target-cell interaction (38), although a recent report has failed to demonstrate such an effect (39). Furthermore, we compared IFN-γ production upon interaction with K562 cells before and after SCT and not directly with healthy control samples. Therefore, although we did not observe a significant decrease in IFN-γ-positive cells between TP1 and TP2, their percentage might be still significantly lower than in healthy controls. Additionally, we observed that the degranulation capacity was influenced by the time period between SCT and TP2 because patients with a time period >11 days had significantly reduced CD107a-positive CD56<sup>+</sup>CD16++ cells at TP2. Because the prolonged time from SCT to TP2 indicates a longer period for leukocyte recovery, this might give an explanation for the reduced degranulation capacity. Nevertheless, we could not discover a correlation between NK cell function at TP2 and the rate of recurrent/refractory disease at 1 year after SCT, which might be due to the low number of recurrent/refractory patients and the short follow-up period.

To the best of our knowledge, this is the first study investigating NK cell function at such an early time point after HDC/ autoSCT. We were able to demonstrate that NK cells were capable of cytokine/chemokine production and degranulation upon tumor cell interaction. Furthermore, we describe an unusual CD56++ NK cell population expressing high levels of CD57 and KIRs shortly after SCT. Further analysis and characterization of this population might reveal more details about how NK cell maturation and differentiation are regulated.

# AUTHOR CONTRIBUTIONS

All authors critically revised the work for important intellectual content and approved the final version of the manuscript. They all agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. BJ, ST, and KP were responsible for acquisition, analysis, and interpretation of the data. EU, PB, AM, and BJ were responsible for the concept and design of the work.

# ACKNOWLEDGMENTS

We thank Franziska Ganss for working with the patient samples during the first phase of the study. We are grateful to the Department of Haematology and Oncology physician team for collecting blood samples from the patients. The authors thank Becton Dickinson (BD) for providing the FACSCanto II™ and Canto10c™ Flow Cytometry Analyzers used in this study.

# FUNDING

Authors were supported by the German Cancer Aid (Max Eder Nachwuchsgruppe, Deutsche Krebshilfe; EU), the LOEWE Center for Cell and Gene Therapy Frankfurt (EU, ST, and PB) funded by the Hessian Ministry of Higher Education, Research and the Arts, Germany (III L 4-518/17.004) and by the "Alfred- und Angelika Gutermuth-Stiftung," Frankfurt, Germany (EU). BJ was funded by a Mildred Scheel postdoctoral scholarship from the Dr. Mildred Scheel Foundation for Cancer Research of the German Cancer Aid Organization. ST was founded by a GO-IN postdoctoral fellowship (PCOFUND-GA-2011-291776).

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at http://journal.frontiersin.org/article/10.3389/fimmu.2015.00583

# REFERENCES


analysis of KIR repertoires and cellular differentiation. *Eur J Immunol* (2014) **44**(7):2192–6. doi:10.1002/eji.201444464


for sequential acquisition of HLA-C-specific inhibitory killer Ig-like receptor. *J Immunol* (2007) **178**(6):3918–23. doi:10.4049/jimmunol.178.6.3918


peripheral blood and cord blood natural killer cells activated with interleukin-2. *Cytotherapy* (2014) **16**(10):1409–18. doi:10.1016/j.jcyt.2014.05.010

**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 Jacobs, Tognarelli, Poller, Bader, Mackensen and Ullrich. 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.*

# **Present and future of allogeneic natural killer cell therapy**

#### *Okjae Lim<sup>1</sup> , Mi Young Jung<sup>1</sup> , Yu Kyeong Hwang<sup>2</sup> and Eui-Cheol Shin<sup>3</sup> \**

*<sup>1</sup> Virology and Immunology Team, MOGAM Biotechnology Institute, Yongin, South Korea, <sup>2</sup> Cell Therapy Center, GreenCross LabCell, Yongin, South Korea, <sup>3</sup> Laboratory of Immunology and Infectious Diseases, Graduate School of Medical Science and Engineering, KAIST, Daejeon, South Korea*

Natural killer (NK) cells are innate lymphocytes that are capable of eliminating tumor cells and are therefore used for cancer therapy. Although many early investigators used autologous NK cells, including lymphokine-activated killer cells, the clinical efficacies were not satisfactory. Meanwhile, human leukocyte antigen (HLA)-haploidentical hematopoietic stem cell transplantation revealed the antitumor effect of allogeneic NK cells, and HLAhaploidentical, killer cell immunoglobulin-like receptor ligand-mismatched allogeneic NK cells are currently used for many protocols requiring NK cells. Moreover, allogeneic NK cells from non-HLA-related healthy donors have been recently used in cancer therapy. The use of allogeneic NK cells from non-HLA-related healthy donors allows the selection of donor NK cells with higher flexibility and to prepare expanded, cryopreserved NK cells for instant administration without delay for *ex vivo* expansion. In cancer therapy with allogeneic NK cells, optimal matching of donors and recipients is important to maximize the efficacy of the therapy. In this review, we summarize the present state of allogeneic NK cell therapy and its future directions.

#### *Edited by:*

*Francisco Borrego, Cruces University Hospital, Spain*

#### *Reviewed by:*

*Jacques Zimmer, Luxembourg Institute of Health, Luxembourg Michael G. Brown, University of Virginia School of Medicine, USA*

#### *\*Correspondence:*

*Eui-Cheol Shin, Laboratory of Immunology and Infectious Diseases, Graduate School of Medical Science and Engineering, KAIST, 291 Daehak-ro, Yuseong-gu, Daejeon 305-701, South Korea ecshin@kaist.ac.kr*

#### *Specialty section:*

*This article was submitted to NK Cell Biology, a section of the journal Frontiers in Immunology*

> *Received: 31 March 2015 Accepted: 18 May 2015 Published: 03 June 2015*

#### *Citation:*

*Lim O, Jung MY, Hwang YK and Shin E-C (2015) Present and future of allogeneic natural killer cell therapy. Front. Immunol. 6:286. doi: 10.3389/fimmu.2015.00286* **Keywords: natural killer cells, allogeneic, cancer immunotherapy, adoptive cell therapy, non-HLA-related donor**

# **Introduction**

Cancer is a major threat for humans worldwide, with approximately 14 million new cases and 8.2 million cancer-related deaths in 2012 (1). Although most common cancer treatments include surgery, chemotherapy, and radiotherapy, unsatisfactory cure rates require new therapeutic approaches, especially for refractory cancers. For this purpose, cancer immunotherapies with various cytokines, antibodies, and immune cells have been clinically applied to patients to encourage their own immune system to help fight the cancer (2).

Adoptive cellular immunotherapies have employed several types of immune cells, including dendritic cells (DCs), cytotoxic T lymphocytes (CTLs), lymphokine-activated killer (LAK) cells, cytokine-induced killer (CIK) cells, and natural killer (NK) cells. Although there has been recent progress in DC therapy and CTL therapy, clinical applications are somewhat limited because cancer antigens must first be characterized and autologous cells must be used. By contrast, LAK cells, CIK cells, and NK cells have antigen-independent cytolytic activity against tumor cells. In particular, NK cells can be used from not only autologous sources but also allogeneic sources and, recently, allogeneic NK cells have been employed more often in cancer treatment. Whereas autologous NK cells from cancer patients may have functional defects (3), allogeneic NK cells from healthy donors have normal function and can be safely administered to cancer patients (4). Allogeneic NK cell therapy is particularly beneficial because it can enhance the anti-cancer efficacy of NK cells via donor–recipient incompatibility in terms of killer cell immunoglobulin-like receptors (KIRs) on donor NK cells and major histocompatibility complex (MHC) class I on recipient tissues.

# **Biology of NK Cells and Their Receptors**

Natural killer cells are innate lymphocytes that provide a first line of defense against viral infections and cancer (5). Human NK cells are recognized as CD3*−*CD56<sup>+</sup> lymphocytes. They can be further subdivided into two subsets based on the surface expression level of CD56. The CD56dim population with low-density expression of CD56 comprises approximately 90% of human blood NK cells and has a potent cytotoxic function, whereas the CD56bright population (approximately 10% of blood NK cells) with highdensity expression of CD56 displays a potent cytokine producing capacity and has immunoregulatory functions (6). The CD56dim NK cell subset also expresses high levels of the Fc receptor for IgG (FcγRIII, CD16), which allows them to mediate antibodydependent cellular cytotoxicity (ADCC) (7). NK cells comprise 5–15% of circulating lymphocytes and are also found in peripheral tissues, including the liver, peritoneal cavity, and placenta. Activated NK cells are capable of extravasation and infiltration into tissues that contain pathogens or malignant cells while resting NK cells circulate in the blood (8).

The NK cell activity is regulated by signals from activating and inhibitory receptors (9, 10). The activating signal is mediated by several NK receptors including NKG2D and natural cytotoxicity receptors (NCRs) (9–11). By contrast, NK cell activity is suppressed by inhibitory receptors, including KIRs, which bind to human leukocyte antigen (HLA) class I molecules on target cells (9, 10, 12). NKG2A is also an important inhibitory receptor binding to non-classical HLA molecule, HLA-E (13). If target cells lose or downregulate HLA expression (14), the NK inhibitory signal is abrogated, allowing NK cells to become activated and kill malignant targets. However, NK cell function is impaired in cancer patients by various mechanisms, particularly in tumor microenvironment (15).

Although NK cell activity is determined by the summation of signals from activating and inhibitory receptors, the inhibitory signal through KIRs is a main regulator of NK cell function particularly in allogeneic settings. Inhibitory KIRs have long cytoplasmic tails containing two immunoreceptor tyrosinebased inhibition motifs (ITIMs). Each KIR has its cognate ligand and consists of two (KIR2DL) or three (KIR3DL) extracellular Ig-domains. KIR2DL1 and KIR2DL2/3 recognize group 2 HLA-C (called C2, Lys80) and group 1 HLA-C (called C1, Asn80), respectively. KIR3DL1 recognizes HLA-Bw4 (16). The KIR repertoire on human NK cells is randomly determined and independent of the number and allotype of HLA class I ligands (17).

# **Therapeutic Efficacy of Allogeneic NK Cells**

#### **Role of Allogeneic NK Cells in Hematopoietic Stem Cell Transplantation**

The antitumor activity of allogeneic NK cells has been demonstrated in the setting of hematopoietic stem cell transplantation (HSCT). Allogeneic HSCT is an established curative treatment for hematologic malignancies. In allogeneic HSCT, donor T cells contribute to graft-versus-host disease (GVHD) and graft-versustumor (GVT) effects (18). In T cell-depleted HSCT, however, donor NK cells are the major effector cells responsible for controlling residual cancer cells before T cell reconstitution (19, 20).

Natural killer cells are the first lymphoid population to recover after allogeneic HSCT. In the first month of transplantation, reconstituted NK cells represent the predominant lymphoid cells and play a crucial role in controlling the host immune system. Allogeneic NK cells prevent viral infections and restrain residual cancer cells in the early phase of transplantation (21). Of note, the GVT activity of donor NK cells is significantly improved when KIRs of donor and HLA class I of the recipient are incompatible, and consequently when inhibitory signals are absent, as observed in HLA-haploidentical HSCT (22). Therefore, increased GVT activity of NK cells with KIR-HLA incompatibility is the underlying rationale for the development of allogeneic NK cell therapy.

#### **Allogeneic NK Cell-Based Immunotherapy**

Following the discovery of inhibitory KIRs and the understanding that they play a role in preventing NK cell killing of self MHC class I-expressing tumor cells, investigators began to research the possibility of using allogeneic donor NK cells instead of autologous NK cells for cancer therapy. Several groups have infused activated, expanded donor NK cells to patients early after allogeneic HSCT to provide antitumor effects (23). In **Table 1**, clinical trials with allogeneic NK cells as therapeutics are summarized.

Allogeneic NK cells can be delivered either in a setting of HSCT or a non-HSCT setting. HSCT is a curative platform for many patients with hematologic malignancies. For patients lacking an HLA-identical donor and for those with progressive disease, the use of HLA-haploidentical family donors is increasingly considered to be a suitable alternative. Therefore, in most clinical trials using allogeneic NK cells, autologous or haploidentical HSCT are followed by NK cell infusion as therapeutics to protect relapse and delay recurrence. Several groups have explored the use of allogeneic NK cells in treating relapses of hematologic malignancies following HLA-haploidentical HSCT in clinical trials, and GVHD did not develop when allogeneic haploidentical NK cells were used (19, 24). In these studies, tumor responses were observed in some patients and overall rates of relapse were reduced. Notably, infusion of allogeneic NK cells can cause cancer regression even without allogeneic HSCT. The patients received allogeneic NK cells without HSCT following non-myeloablative chemotherapy. The chemotherapy pre-conditioning delayed the rejection of the transferred cells, and in some cases, the allogeneic NK population even expanded before being ultimately rejected (25, 26).

In a non-transplantation setting, Miller and colleagues were the first to establish the safety and efficacy of adoptive cellular transfer of HLA-haploidentical NK cells in patients with advanced cancer (27). In this study, 19 acute myeloid leukemia (AML) patients were given haploidentical NK cell infusions together with IL-2 and 5 patients achieved complete remission. Allogeneic NK cells with KIR-HLA mismatches between patients and donors exhibited greater tumor-killing activity without causing GVHD. Based on the success observed in AML, a number of clinical trials are being carried out to determine the feasibility and efficacy of allogeneic NK cell infusion for cancer treatment. Many of 15 ongoing clinical trials are oriented to hematological malignancies including

#### **TABLE 1 | Selected clinical trials with expanded allogeneic NK cells**.


*HSCT, hematopoietic stem cell transplantation; RIC, reduced-intensity conditioning; ALL, acute lymphocytic leukemia; AML, acute myeloid leukemia; TBI, total body irradiation.*

leukemia, multiple myeloma, and myelodysplastic/proliferative diseases. Additionally, clinical trials have shown that allogeneic NK cells play a therapeutic role in solid tumors (26, 28, 29). The clinical efficacy of expanded allogeneic NK cells was investigated in patients with recurrent metastatic breast and ovarian cancers in combination with a Hi-Cy/Flu preparative chemotherapy regimen (29). Adoptive transfer of *ex vivo-*expanded allogeneic NK cells was safe and effective in patients with advanced non-small cell lung cancer (26). These findings provided proof of concept that allogeneic NK cells could be effective not only in hematologic malignancy patients but also in solid tumor patients. Clinical trials are currently carried out in hepatocellular carcinoma and neuroblastoma (NCT02008929, NCT01807468).

Adoptive transfer of allogeneic NK cells that come from a totally unrelated donor has also been demonstrated to be safe without any significant side effects (NCT01212341). Allogeneic NK cell therapy is currently applied to patients with advanced hepatocellular carcinoma after curative resection (NCT02008929). In this clinical trial, *ex vivo*-expanded allogeneic NK cells were administered without combination with other therapeutic modalities to investigate the isolated effect of infused allogeneic NK cells.

Taken together, the clinical studies mentioned above demonstrated that the infusion of allogeneic NK cells after *ex vivo* expansion is largely safe and some responses appear encouraging.

# **Optimized Selection of Donors**

#### **Lessons from Allogeneic HSCT**

In T cell-depleted HSCT, donor NK cells are the major effector cells responsible for controlling residual cancer cells (19), and it has been shown that the KIR genotype of donors influences the outcome of HSCT (30). From the experience of allogeneic HSCT, we can learn how allogeneic NK cell donors are selected to maximize the antitumor activity of infused allogeneic NK cells.

There are two distinct types of KIR haplotypes: group A and group B. The KIR group B haplotype has more activating receptors than the KIR group A haplotype (31). According to the KIR genotype, all individuals can be divided into the A/A genotype (homozygous for A haplotypes) or the B/x genotype (having 1 or 2 B haplotypes). There have been reports that the donor KIR genotype influences outcomes of unrelated HSCT for acute hematological malignancies and that the B/x genotype confers significant survival benefit to patients (22, 32, 33). B/x donors are further differentiated on whether their B haplotype genes are in the centromeric or/and telomeric part. On the basis of this information, the KIR B-content score can be calculated from 0 to 4 (30, 34). High donor KIR B-content scores have been associated with a significantly reduced relapse in children after haploidentical HSCT for acute lymphocytic leukemia (ALL) (35), and donors with two or more B-content scores showed superior survival after unrelated HSCT for AML (27).

Incompatibility between KIRs of donors and HLAs of recipients is also an important factor. Considering that each KIR binds to specific HLA allotypes as an inhibitory ligand (e.g., KIR2DL1 to group 2 HLA-C, KIR2DL2/3 to group 1 HLA-C, and KIR3DL1 to HLA-Bw4), a recipient may lack specific HLA allotypes that inhibit donor NK cells. In this case, higher antitumor activity of donor NK cells is expected. Indeed, antitumor activity of donor NK cells is significantly improved when KIRs and HLAs are incompatible between donor and recipient (19, 24, 36).

In addition to the KIR genotype and incompatibility, actual expression of KIRs on NK cells needs to be considered for the best antitumor activity of allogeneic NK cells because the expression of KIRs occurs in stochastic combination (37). Antitumor activity is likely to be mediated by single-KIR<sup>+</sup> allogeneic NK cells not encountering any inhibitory signal from HLA molecules on recipient cells (38). Although NK cells are the first lymphoid population to recover after allogeneic HSCT (21), reconstitution of mature NK receptor repertoires requires at least 3 months (39). Importantly, during this period, donor-derived single-KIR<sup>+</sup> NK cells are not fully functional (38). In this aspect, infusion of single-KIR<sup>+</sup> mature NK cells selected for KIR-HLA mismatches might lead to better clinical outcomes. Currently, multicolor flow cytometry enables the examination of KIR expression in the NK cell population. The approach to generate GMP-grade single-KIR<sup>+</sup> NK cells (40) will allow customized allogeneic NK cell therapy.

#### **Sources of Allogeneic NK Cells**

To permit therapeutic use of allogeneic NK cells in clinical settings, a sufficient number of highly enriched NK cells must be obtained. The sources for allogeneic NK cells include peripheral blood mononuclear cells (PBMCs) collected by leukapheresis from healthy donors and umbilical cord blood (UCB).

Peripheral blood mononuclear cells collected by leukapheresis are generally utilized as a source of allogeneic NK cells. Various methods to obtain *ex vivo*-expanded, activated, and CD3<sup>+</sup> T cell-depleted NK cells have been well established in clinical scales and grades (41). Although those NK cells showed potent antitumor efficacy *in vitro* and *in vivo*, clinical outcomes were insufficient. The clinical results might be influenced by several factors including malignancy types and pre-conditioning treatment. As described above, the therapeutic efficacy of allogeneic NK cell therapy can be potentiated by optimal selection of NK cell donors in a non-HSCT setting. Since NK cells from haploidentical donors had been used in allogeneic HSCT settings, allogeneic NK cells were mostly obtained from haploidentical donors even in a non-HSCT setting. Recently, *ex vivo*-expanded, allogeneic NK cells from unrelated, random donors were successfully administered to patients with malignant lymphoma or advanced solid tumors in a phase 1 trial (NCT01212341) that has proceeded to a phase 2 trial of patients with hepatocellular carcinoma (NCT02008929). This strategy, which used unrelated NK donors, allowed free selection of the best donor in terms of donor KIR-recipient HLA incompatibility without limitation of small pools of related donors. Furthermore, the use of allogeneic NK cells from non-HLA-related healthy donors allows preparation of expanded, cryopreserved NK cells for instant administration without delay for *ex vivo* expansion.

Umbilical cord blood is another promising source of allogeneic NK cells. However, cytokine-based differentiation of CD34<sup>+</sup> hematopoietic stem and progenitor cells to NK cells needs to be carried out to obtain large numbers of functional NK cells from UCB (42). This process requires high-dose cytokine cocktails and delicate culture regimens that may result in low-cost effectiveness. Recently, an NK cell expansion method from UCB using artificial antigen presenting feeder cells was reported. NK cells expanded by this method showed *in vitro* cytotoxicity against various myeloma targets and *in vivo* antitumor activity in a mouse model of myeloma (43).

#### **Future Directions**

#### **Genetic Modification**

Genetic modification is a promising option for redirecting the function of various types of immune cells (44). Much work has been performed, particularly on genetically redirecting T cells against a range of tumor antigens. For example, T cells expressing chimeric antigen receptors (CARs) targeting CD19 antigens have been developed to treat B-cell-derived malignancy, and clinical trials are currently ongoing (45–47). The successful experience with CAR-expressing T cells in the treatment of hematological malignancies has prompted the development of CAR-expressing NK cells. NK cells are attractive for CAR expression because they have cytotoxic function and, unlike T cells, allogeneic NK cells do not cause GVHD.

As summarized in **Table 2**, two clinical trials are investigating the use of CAR-expressing allogeneic NK cells. The aim of both studies is to assess the safety, feasibility, and efficacy of expanded, activated, and CD19-redirected haploidentical NK cells in ALL patients who have persistent disease after intensive chemotherapy or HSCT (NCT00995137, NCT01974479). Further, other tumor antigens, such as CS1, CEA, CD138, and CD33, are targeted by CARs expressed by NK cells, although NK-92, YT, or NKL cell lines were used (48–51).

Genetic modification is also performed to express cytokine transgenes in NK cells. NK cell function could be enhanced by



*ALL, acute lymphocytic leukemia; B-NHL, B-cell non-Hodgkin lymphoma; B-ALL, B-cell acute lymphoblastic leukemia.*

expression of cytokines, such as IL-2 (57, 58), IL-12 (56, 59), and IL-15 (60–62). Cytokine expression enhances the activation of NK cells, survival and proliferation of NK cells, and accumulation of NK cells in tumor tissues. To improve the efficacy of NK cell therapy, genetic modification of NK cells is explored to express activating receptors, such as NKG2D (55).

#### **Therapeutic Regimens**

In allogeneic NK cell therapy, optimal therapeutic regimens for clinical applications should be considered because adoptively transferred NK cells not only target tumor cells but also interact with the immunological environment. To potentiate the therapeutic efficacy of allogeneic NK cells, proper strategies, including pre-conditioning or combination therapy, could be applied (34).

Upregulation of NKG2D ligands by spironolactone (63) or histone deacetylase inhibitors (64, 65) and upregulation of TRAIL-R2 by doxorubicin (66) result in enhanced antitumor efficacy of NK cells. Proteasome inhibitors also sensitize tumor cells to NK cellmediated killing via TRAIL and FasL pathways. In addition, c-kit tyrosine kinase inhibitor (67) and JAK inhibitors (68) increase the susceptibility of tumor cells to NK cytotoxicity and enhance antitumor responses by increased IFN-γ production from NK cells. However, protein kinase inhibitors should be used cautiously because some protein kinase inhibitors, such as sorafenib, inhibit the effector function of NK cells (69).

Immunomodulatory drugs can augment NK cell function. Lenalidomide enhances rituximab-induced killing of non-Hodgkin's lymphoma and B-cell chronic lymphocytic leukemia through NK cell and monocyte-mediated ADCC mechanisms (70). Combination therapy using IL-2 and anti-CD25 shows

#### **References**


anti-leukemic effects by depletion of regulatory T cells in addition to activation and expansion of NK cells (71). Alloferon, an immunomodulatory peptide, enhances the expression of NKactivating receptor 2B4 and granule exocytosis from NK cells against cancer cells (72).

Therapeutic antibodies can be combined with allogeneic NK cell therapy (73). Antibodies against tumor antigens (e.g., CD20 and CS1) can induce ADCC of NK cells (74, 75). Antibodies to activating NK receptors (e.g., 4-1BB, GITR, NKG2D, DNAM-1, and NCRs) can enhance NK activation (74, 76–79). In addition, inhibitory receptors (e.g., KIR2DL, PD-1, PD-L1, and NKG2A) can be blocked by antibodies (80–85). Bispecific and trispecific killer cell engagers directly activate NK cells through CD16 signaling and thus, induce cytotoxicity and cytokine production against tumor targets (86, 87).

#### **Conclusion**

Antitumor activity of allogeneic NK cells was first observed in a setting of HLA-haploidentical HSCT. Allogeneic NK cell therapy was tried mostly using HLA-haploidentical NK cells with or without allogeneic HSCT and, recently, allogeneic NK cells from unrelated, random donors have been used in a non-HSCT setting. The efficacy of allogeneic NK cell therapy can be enhanced by optimal donor selection in terms of the KIR genotype of donors and donor KIR-recipient MHC incompatibility. Furthermore, efficacy can be increased by genetic modification of NK cells and optimized therapeutic regimens. In the future, allogeneic NK cell therapy can be an effective therapeutic modality for cancer treatment.


T-cell-mediated killing by augmented TRAIL receptor signaling. *Int J Cancer* (2013) **133**(7):1643–52. doi:10.1002/ijc.28163


cellular cytotoxicity in a model of anti-lymphoma therapy. *J Immunol* (2008) **180**(9):6392–401. doi:10.4049/jimmunol.180.9.6392


**Conflict of Interest Statement:** Yu Kyeong Hwang is a current employee of Green-Cross LabCell. The other co-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 Lim, Jung, Hwang and Shin. 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.*

# **Alloreactive natural killer cells for the treatment of acute myeloid leukemia: from stem cell transplantation to adoptive immunotherapy**

*Loredana Ruggeri <sup>1</sup> , Sarah Parisi <sup>2</sup> , Elena Urbani <sup>1</sup> and Antonio Curti <sup>2</sup> \**

*<sup>1</sup> Department of Medicine, Division of Hematology and Clinical Immunology, Ospedale Santa Maria della Misericordia, University of Perugia, Perugia, Italy, <sup>2</sup> Department of Experimental, Diagnostic and Specialty Medicine, Institute of Hematology "L. and A. Seràgnoli", S. Orsola-Malpighi Hospital, University of Bologna, Bologna, Italy*

#### *Edited by:*

*Rafael Solana, University of Cordoba, Spain*

#### *Reviewed by:*

*Hugh Thomson Reyburn, Spanish National Research Council, Spain Raquel Tarazona, University of Extremadura, Spain*

#### *\*Correspondence:*

*Antonio Curti, Department of Experimental, Diagnostic and Specialty Medicine, Institute of Hematology "L. and A. Seràgnoli", S. Orsola-Malpighi Hospital, University of Bologna, Via Massarenti 9, Bologna 40138, Italy antonio.curti2@unibo.it*

#### *Specialty section:*

*This article was submitted to NK Cell Biology, a section of the journal Frontiers in Immunology*

> *Received: 30 June 2015 Accepted: 03 September 2015 Published: 15 October 2015*

#### *Citation:*

*Ruggeri L, Parisi S, Urbani E and Curti A (2015) Alloreactive natural killer cells for the treatment of acute myeloid leukemia: from stem cell transplantation to adoptive immunotherapy. Front. Immunol. 6:479. doi: 10.3389/fimmu.2015.00479* Natural killer (NK) cells express activating and inhibitory receptors, which recognize MHC class-I alleles, termed "Killer cell Immunoglobulin-like Receptors" (KIRs). Preclinical and clinical data from haploidentical T-cell-depleted stem cell transplantation have demonstrated that alloreactive KIR-L mismatched NK cells play a major role as effectors against acute myeloid leukemia (AML). Outside the transplantation setting, several reports have proven the safety and feasibility of NK cell infusion in AML patients and, in some cases, provided evidence that transferred NK cells are functionally alloreactive and may have a role in disease control. The aim of the present work is to briefly summarize the most recent advances in the field by moving from the first preclinical and clinical demonstration of donor NK alloreactivity in the transplantation setting to the most recent attempts at exploiting the use of alloreactive NK cell infusion as a means of adoptive immunotherapy against AML. Altogether, these data highlight the pivotal role of NK cells for the development of novel immunological approaches in the clinical management of AML.

**Keywords: natural killer cells, acute myeloid leukemia, stem cell transplantation, immunotherapy, alloreactivity**

# **Introduction**

The clinical management of acute myeloid leukemia (AML) relies on aggressive chemotherapy, followed by allogeneic stem cell transplantation (SCT). Although the post-chemotherapy complete remission (CR) rate ranges from 60 to 85% in younger patients, the disease relapse is still very high, thus reducing overall survival (OS) to 40%. Of note, the prognosis of elderly patients is particularly poor with an OS of about 10%. Such a particular dismal clinical outcome is due to an increase in unfavorable biological features, which reduces the CR rate and, whenever CR is obtained, to the inability to undergo post-CR consolidation programs, including SCT, due to co-morbidities. In the attempt to improve AML clinical outcome, novel regimens and targeted therapies have been proposed in the last few years but the clinical results have proven limited. In particular, a minimal residual disease (MRD), resistant to further treatments, often persists after induction chemotherapy. In that context, the use of an immunological approach to target MRD may significantly impact on the eradication of disease. The proof-of-principle of the capacity of immune cells to eradicate MRD derives from the results of allogeneic SCT, which clearly represents an option for relapse prevention. In particular, the critical role of natural killer (NK) cells as key players in AML prevention and eradication has been clearly established, especially in the context of haploidentical SCT. However, the SCT approach has important limitations and is not applicable to all patients. For these reasons, it is conceivable to exploit the anti-leukemia potential of NK cells outside the transplantation setting as adoptive immunotherapy.

# **NK Cells: Biological Pills**

Within allogeneic SCT, donor lymphocytes recognize and destroy the recipient's residual leukemic cells. The demonstration that this process, known as graft versus leukemia (GvL) effect, plays a major role in the therapeutic effect of SCT has led to the development of novel strategies of adoptive immunotherapy before and after SCT (1). Although most of the data refer to allogeneic T cells as GvL mediators, it is known that other subsets of circulating lymphocytes, such as NK cells, may significantly act as effector cells against leukemia in the post-transplantation setting. NK cells are defined by the expression of CD56 and CD16 and by the absence of the T-cell marker, CD3. NK cells are involved in the innate immune response and in cancer immunosurveillance, where they kill transformed tumors in a major histocompatibility complex (MHC)-unrestricted manner (2). NK cells originate from the bone marrow (BM) and then home to secondary lymphoid tissues. They account for 10–15% of peripheral blood lymphocytes and their activity depends on the expression on their surface of several activating and inhibitory receptors that recognize MHC class-I molecules (**Figure 1A**); the most notable inhibitory receptors are named killer cell immunoglobulin-like receptors (KIRs), which recognize allotypic determinants within certain groups of HLA class-I alleles. The lack of expression of the specific HLA class-I allele by allogeneic target cells allows KIRs to sense the absence of the self class-I KIR-ligand (KIR-L), thus mediating NK alloreactivity. Indeed, the engagement of these NK cell receptors results in stimulation or inhibition of NK cell effector function. KIR genes are closely packed in the leukocyte receptor complex on chromosome 19q13.4 and are inherited as haplotypes. Two distinct KIR haplotypes have been identified. Group A haplotypes encode inhibitory KIRs and have a fixed gene content, whereas group B haplotypes are variable in number and combination. NK cells become functionally competent only after they encounter self-HLA molecules during a process named licensing or NK cell education. About 10–20% of NK cells remain unlicensed and are hypo-responsive. Recently, other inhibitory receptors on NK cells have been identified, such as CD94/NKG2A receptors, that recognize a non-classic MHC class-I molecule (HLA-E). CD94/NKG2A continuously recycles from the cell surface through endosomal compartments, thus facilitating its inhibitory capacity (3). NK cells can also express activating forms of KIRs and the activating receptor CD94/NKG2C that interact with the same HLA molecules as their inhibitory counterparts. Other activating receptors include natural cytotoxicity receptors (NKp46, NKp30, NKp44), DNAM-1 that interacts with CD112 and CD155, and NKG2D, which recognizes ligands, upregulated during cellular stress, such as tumor transformation and viral infections (4). In addition, CD16 (FcgammaRIIIA) triggers antibody-dependent cellular cytotoxicity on opsonized target

Thirteen AML patients, five with active disease, two in molecular relapse, and six in morphological complete remission (CR) were treated with alloreactive NK cells, after fludarabine/cyclophosphamide immunosuppressive chemotherapy. Only one of the five patients with active disease achieved transient CR, whereas the other four patients had no clinical benefit. On the contrary, five out of eight patients showed a response, which in some cases was long-lasting CR [adapted from Curti et al. (5)].

cells, including tumor cells. Integrins also play a central role in mediating adhesion to target cells and degranulation (4).

# **NK Cell Alloreactivity in Allogeneic SCT**

#### **Haploidentical Hematopoietic SCT**

Although autologous NK cell activity is usually impaired in cancer patients (6), allogeneic alloreactive NK cells from healthy donors have been shown to exert important effector cell function and could be safely infused in cancer patients without side effects (7). In particular, within the setting of allogeneic SCT, the KIR-L incompatibility between donor and recipient in the GvL direction has been demonstrated to enhance the anti-cancer efficacy of NK cells, thus providing a novel and successful platform for anti-tumor immunotherapy. In the setting of T-cell-depleted haploidentical KIR-L mismatched SCT, we demonstrated that NK cell alloreactivity mediates a powerful and protective GvL effect, which is dissociated from Graft versus Host Disease (GvHD) (8). Indeed, early after T-cell-depleted transplantation, the reconstituted NK cells represent the predominant lymphoid cell population and the hematopoietic stem cells gave origin to an NK cell repertoire, which is identical to the donor one (8–10). Of note, the GvL activity of donor NK cells is triggered when the donor's KIRs and the recipient's HLA class-I molecules are incompatible, and consequently when inhibitory signals in the recipient are lacking. Preliminary studies in preclinical models demonstrated that human alloreactive NK clones, infused into NOD/SCID mice, previously engrafted with human AML cells, are capable of clearing leukemia cells and improving survival. These preclinical data on the impact of donor NK cell alloreactivity were confirmed in a series of more than 112 AML patients transplanted with haploidentical donors. Patients were divided in two subgroups, according to KIR-L incompatibility in the GvH direction. In the group with KIR-L incompatibility a significantly reduced relapse rate was observed. Moreover, 5-year event-free survival was 5% in contrast to 60% observed in the group without KIR-L incompatibility, thus demonstrating that KIR-L incompatibility is the only independent predictive factor for survival in AML patients. Furthermore, alloreactive mismatched NK cells facilitated hematopoietic engraftment after infusion of haploidentical stem cells and inhibited GvHD by killing host antigen-presenting cells (9). Of note, in our adult cohort of transplanted patients, NK cell alloreactivity was not efficient in preventing disease relapse of acute lymphoblastic leukemia (ALL). In contrast, in a group of pediatric leukemia patients, including ALL, undergoing haploidentical SCT, Pende et al. showed the beneficial role of activating KIRs and donor NK cell alloreactivity (11). Other clinical studies support the protective role of alloreactive NK cells in the haploidentical SCT setting. Stern et al. showed the results of a phase II multicenter study in which purified NK cells were administered pre-emptively in recipients of T-cell-depleted haploidentical SCT. Sixteen young patients with high-risk leukemia or highly malignant solid tumors were included in this protocol and received NK-donor lymphocyte infusions (DLI) on days 40 and 100 after transplantation. This study demonstrated the feasibility of the procedure. However, as a consequence of contaminating T cells in NK-DLI cell preparation, the trial showed a high incidence of acute GvHD, whereas the anti-leukemic activity appeared to be very limited (12). The use of allogeneic donor NK cells instead of autologous NK cells for cancer therapy has been recently reported. Several groups have explored allogeneic NK cells for the treatment of relapse following HLA-haploidentical SCT. Interestingly, GvHD did not occur. Some groups have infused activated, expanded donor NK cells in patients early after allogeneic SCT. **Table 1** summarizes clinical trials with allogeneic NK cells as therapeutics. In these studies, clinical responses were observed in some patients and overall rates of relapse were reduced. For patients lacking an HLA-identical donor and for those with progressive disease, the use of HLA-haploidentical family donors is now increasingly considered to be a suitable alternative. However, many different protocols of T-cell-repleted haploidentical transplantation are ongoing and few data are available on the role of NK alloreactive donors in the presence of T cells in the graft and/or when GvHD prophylaxis is administered. The results of these studies are highly warranted to better elucidate the possible competition of NK and T cells and the role of immune suppressive drugs on NK cells *in vivo*.

#### **Unrelated and Matched SCT**

Unlike haploidentical SCT, the role of NK cell alloreactivity in the field of unrelated SCT is controversial, even though several studies have already investigated this setting (**Table 2**). Some years ago Giebel et al. conducted a study involving 130 patients with hematological malignancies who underwent allogeneic SCT and received Cyclosporine, ATG and short-term methotrexate as GvHD prophylaxis. With a median follow-up of 4.5 years, the OS was 87% in patients with a KIR mismatch in the donor direction versus 48% in non-KIR-mismatched patients; disease-free survival (DFS) was 87% in the first group compared with 39% in the second one. Transplant-related mortality was 6% in the KIR-mismatched patients and 40% in non-mismatched patients (13). These results were not confirmed in studies published by other centers (14, 15), which showed a detrimental effect of KIR-L incompatibility, correlated with HLA mismatching. These controversial data demonstrated that the role of NK cells remains unclear in the setting of unrelated SCT. Several factors, such as post-transplantation immunosuppressive therapies, T-cell depletion, different stem cell sources and doses, may impact in this patient setting (13–15). In a group of donor-recipient pairs missing an inhibitory KIR-L, a beneficial role of alloreactive NK cells, transiently and randomly originated from donor stem cells, was observed (16). These cells expressed the inhibitory single KIR receptor that could not be blocked by the host cells. However, these alloreactive NK cells were not functional, thus corroborating


#### **TABLE 1 | Clinical trials with expanded allogeneic NK cells in haploidentical SCT**.

**TABLE 2 | The most relevant papers reporting the impact of KIR-L mismatch in unrelated SCT**.


*<sup>a</sup>Trend, P-value between 0.05 and 0.09.*

*<sup>b</sup>GvHD grade II–IV.*

*<sup>c</sup>GvHD grade III–IV.*

the notion that NK cells must be educated and consequently armed by the presence of the appropriate inhibitory KIR-L (17). New data have been provided on the possible role of activating KIRs which are present on KIR B haplotypes. Cooley et al. showed that B haplotype, which is present in 60% of donors, is fundamental in preventing relapse while NK cell alloreactivity does not influence the outcome of a very large cohort of unrelated transplants (18). However, in the setting of T-cell-depleted haploidentical transplants, the presence of KIR B haplotypes is associated with reduced infection-related mortality in the group of patients transplanted from NK alloreactive donors without any impact on relapse (19).

# **Alloreactive NK Cells as Adoptive Immunotherapy**

Natural killer cells have already been used as a means of adoptive immunotherapy beside the SCT setting (16, 20). These studies reported on the trafficking and body distribution of infused NK cells. Based on these preliminary data, NK cell selection for immunotherapy has recently been developed at clinical level (21). In 2005, Miller et al. published the results of a seminal study in which up to 1.5 *<sup>×</sup>* <sup>10</sup><sup>7</sup> /haploidentical NK cells/kg were safely infused in AML and cancer patients following Fludarabine/Cyclophosphamide (Flu/Cy) immunosuppressive chemotherapy; in this study some clinical responses without GvHD were observed. Circulating haploidentical NK cells were found up to 28 days after infusion, especially when exogenous interleukin (IL)-2 was given. *In vivo* expansion of NK cells was correlated with a high IL-15 serum concentration. In particular, 19 poor risk AML patients, together with 10 metastatic melanoma patients and 13 metastatic renal cell carcinoma patients received a cell population enriched in NK cells. Five out of 19 AML patients achieved CR, NK cell adoptive immunotherapy was well tolerated and no hematological toxicity was recorded. The maximum tolerated dose of NK cells was not achieved and GvHD was not observed despite the relatively high number of infused haploidentical T cells. However, it should be noted that NK cells were only partially purified after a single round of depletion of CD3<sup>+</sup> cells which resulted in less than a 2 log reduction of T cells (21). A group of 10 low-risk pediatric AML patients were treated with haploidentical KIR–HLA mismatched NK infusion. All patients were alive at the 2-year follow-up. As compared to the adult trial by Miller's group, the median number of infused NK cells was significantly higher and NK cells were processed to obtain a highly purified cell population (22). We reported the results of a trial of NK cell-based adoptive immunotherapy in 13 AML patients, 5 with active disease, 2 in molecular relapse, and 6 in morphological CR. The median age was 62 years (range 53–73). Highly purified CD56+CD3- NK cells from haploidentical KIR-ligand mismatched donors were infused after fludarabine/cyclophosphamide immunosuppressive chemotherapy. No signs of GvHD and/or NK cell-related toxicity were reported. As expected, patients with active disease had no clinical benefit. Interestingly, both patients in early molecular relapse achieved CR and three patients in CR were disease-free after a follow-up of 34, 32, and 18 months of follow up (**Figure 1B**). Infused NK cells were detected in the peripheral blood of all evaluable patients and in the BM in some cases. Importantly, infused NK cells were demonstrated *ex vivo* to be alloreactive by killing *in vitro* the recipient's cells, including leukemia (5). Several biological factors, both of recipient and donor origin, may be implicated in the therapeutic effect of NK cells after infusion into AML patients. Miller and collaborators recently reported on the critical impact of some components of the recipient immune response on the anti-leukemia activity of infused NK cells (23). In particular, they reported NK cell expansion correlates with the post-chemotherapy serum concentrations of some cytokines, such as IL-15 and IL-35, and the number of T regulatory cells (Tregs) critically influences the capacity of infused NK cells to expand and to kill AML cells. The clinical relevance of these findings is supported by a better DFS in patients undergoing NK immunotherapy and depleted of Tregs. As for the donor, it would be interesting to test whether the composition of the donor NK cell population may be correlated to a different clinical outcome. In particular, the frequency and the function of alloreactive NK cells may impact on the anti-leukemia capacity of infused NK cells. Moreover, the presence within the graft of different subsets of CD56<sup>+</sup> cells, other than "classic" NK cells, should be evaluated in more detail and, possibly, correlated with the response to NK therapy. The manipulation of donor NK cell graft may represent an interesting approach to increase NK cell activity before infusion. This is a critical point since NK cell-based immunotherapy may be significantly hampered by the transient effector function of infused NK cells. In particular, *in vitro* priming with cytokines, such as IL-12, IL-18, and IL-15, has been reported. This treatment results in the expansion of the memory-like NK cell subset with enhanced functional properties (24) and more prolonged persistence in the host. (25). Recently, in a mouse tumor model a cytokine-based treatment resulted in enhanced anti-tumor activity via the reversal of NK cell anergy, which occurs in the presence of MHC-deficient tumors (26). Since HLA-loss is known to be a fundamental immune escape means for adaptive T-cell-mediated immune response (27), preventing NK-cell anergy may have important clinical implications for cancer immunotherapy.

# **Conclusions**

Alloreactive NK cells have been emerging as a potent effector cell population against AML. The demonstration of a significant clinical activity of alloreactive purified NK cell infusion outside the transplantation setting represents a proof-of-principle for such an anti-leukemia effect, which has been clearly established in the context of haploidentical SCT. Altogether, these data highlight the pivotal role of NK cells for the development of novel immunological approaches in the clinical management of AML. Nevertheless, several biological issues still require full elucidation. In particular, a more in-depth evaluation of the impact that recipient- and donor-derived factors may have in influencing *in vivo* NK cell

# **References**


activity is an important point. The design of future NK cell-based clinical trials, both in the SCT and adoptive immunotherapy settings, should include a correlation between clinical results and biological outputs. Indeed, correlative biological studies may make it possible to identify those subsets of patients, who may really benefit from NK immunotherapy, in the attempt to tailor both pharmacological and immunological therapies to the patients' characteristics.

cell transplantation using unrelated donors. *Blood* (2004) **103**:2860–1. doi:10.1182/blood-2003-11-3893


**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 Ruggeri, Parisi, Urbani and Curti. 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.*

# **Effector functions of natural killer cell subsets in the control of hematological malignancies**

*Angela Gismondi 1,2 \*, Helena Stabile<sup>1</sup> , Paolo Nisti <sup>1</sup> and Angela Santoni 1,3 \**

*<sup>1</sup> Department of Molecular Medicine, Istituto Pasteur-Fondazione Cenci Bolognetti, Sapienza University of Rome, Rome, Italy, 2 Eleonora Lorillard Spencer Cenci Foundation, Rome, Italy, <sup>3</sup> Italian Institute of Technology, Rome, Italy*

Treatment of hematological malignant disorders has been improved over the last years, but high relapse rate mainly attributable to the presence of minimal residual disease still persists. Therefore, it is of great interest to explore novel therapeutic strategies to obtain long-term remission. Immune effector cells, and especially natural killer (NK) cells, play a crucial role in the control of hematological malignancies. In this regard, the efficiency of allogeneic stem cell transplantation clearly depends on the immune-mediated graft versus leukemia effect without the risk of inducing graft versus host disease. Alloreactive donor NK cells generated following hematopoietic stem cell transplantation ameliorate the outcome of leukemia patients; in addition, *in vivo* transfer of *in vitro* expanded NK cells represents a crucial tool for leukemia treatment. To improve NK cell effector functions against resistant leukemia cells, novel immunotherapeutic strategies are oriented to the identification, isolation, expansion, and administration of particular NK cell subsets endowed with multifunctional anti-tumor potential and tropism toward tumor sites. Moreover, the relationship between the emergence and persistence of distinct NK cell subsets during post-graft reconstitution and the maintenance of a remission state is still rather unclear.

#### *Edited by:*

*Francisco Borrego, Cruces University Hospital, Spain*

#### *Reviewed by:*

*Kerry S. Campbell, Fox Chase Cancer Center, USA Michael R. Verneris, University of Minnesota, USA*

#### *\*Correspondence:*

*Angela Gismondi angela.gismondi@uniroma1.it; Angela Santoni angela.santoni@uniroma1.it*

#### *Specialty section:*

*This article was submitted to NK Cell Biology, a section of the journal Frontiers in Immunology*

> *Received: 20 July 2015 Accepted: 23 October 2015 Published: 05 November 2015*

#### *Citation:*

*Gismondi A, Stabile H, Nisti P and Santoni A (2015) Effector functions of natural killer cell subsets in the control of hematological malignancies. Front. Immunol. 6:567. doi: 10.3389/fimmu.2015.00567* **Keywords: NK cell subsets, hematological malignancies, HSCT, NK cell therapy, NK cells**

# **INTRODUCTION**

Natural killer (NK) cells belong to innate lymphocytes that play an important role in the early phase of immune defense against microbial infections, and tumor growth and dissemination. They represent a population of highly specialized large granular lymphocytes capable of mediating cytotoxic activity and endowed with the ability to release cytokines and chemokines when properly activated by target cells or pro-inflammatory stimuli (1–3). During microbial infection or tumor growth and invasion, NK cells can be rapidly recruited to and accumulate in the parenchyma of injured organs, contributing to the elimination of infected or transformed cells as well as to the recruitment and activation of other immune cells (4, 5). Thus, NK cells are important effectors in the early phase of innate immune responses and play a crucial role as immune regulators of adaptive immunity (6).

Activation of NK cell functional program results from a delicate balance of signals initiated by a complex receptor system formed by both inhibitory and activating receptors. These receptors are acquired during differentiation, and are oligoclonally distributed on mature NK cells; in most instances, the inhibitory signals override the triggering ones (2).

The inhibitory receptors that mainly bind to MHC class I molecules are grouped into two classes: the killer cell immunoglobulin-like receptor (KIR) family that includes receptors for human leukocyte antigen (HLA)-A, -B, and -C group of alleles, and the C-type lectin receptors, such as CD94/NKG2A that binds to the non-classical HLA class I molecule, HLA-E.

Both the inhibitory receptor families have also an activating counterpart with similar specificity but different ligand affinity, showing inhibitory receptors greater ligand affinity with respect to their activating counterpart. The MHC I activating receptors possess a short intracellular domain and associate with a transducing chain that initiates the activating signaling pathway when engaged by ligands (7–9). The human KIR family consists of 13 genes and 2 pseudogenes, and displays a high degree of diversity that arises from both variability in KIR content and allelic polymorphisms. KIR genes are variably inherited by individuals and expressed by NK cells in an oligoclonal manner.

Among non-MHC I NK cell activating receptors, the best studied is the low-affinity Fc-γ receptor IIIA (CD16) involved in the NK cell-mediated antibody-dependent cellular cytotoxicity (ADCC) (10). Another important activating receptor is NKG2D that binds to self-molecules undergoing up-regulation on stressed, infected, or damaged cells belonging to MIC and ULBP families (11). In addition, NK cell activating receptors also include NKp44, NKp46, and NKp30 Ig-like molecules, collectively termed natural cytotoxicity receptors (NCR), and DNAM-1 (CD226) that cooperatively triggers natural killing (12–14).

Activating and inhibitory receptors are acquired during NK cell differentiation and activation, and are selectively expressed on distinct NK cell subsets. Recognition of MHC I receptors during NK cell development is critical for the acquisition of the functional competence through a process defined as NK cell education or licensing (15, 16). Thus, based on the receptor repertoire and expression levels, phenotypically distinct mature NK cell populations have been identified and suggested to represent specialized subsets mediating different functions and endowed with distinct migratory properties (17).

Natural killer cell differentiation primarily occurs in the bone marrow (BM), although NK cell progenitors can undergo final maturation also in the periphery, and the existence of a thymic pathway of NK cell differentiation has been also described in mice (18–20).

Fully mature, NK cells mainly circulate in the peripheral blood (PB) but they can be also found in several lymphoid and nonlymphoid organs, such as spleen, tonsils, lymph nodes, liver, intestine, lungs, and uterus (1, 21–24). PB NK cells represent about 5–20% of total lymphocytes.

Two major human NK cell subsets, namely, CD56highCD16+/*<sup>−</sup>* and CD56lowCD16high, can be distinguished in the PB based on the expression levels of the low-affinity Fc-receptor γ IIIA (CD16) and the neural cell adhesion molecule (NCAM, CD56). CD56highCD16+/*<sup>−</sup>* NK cells primarily secrete immunoregulatory cytokines, whereas the CD56lowCD16high NK cell subset is the major killer population mediating both natural cytotoxic activity and ADCC. It is still matter of debate whether these subsets are functionally distinct terminally differentiated NK cells or NK cells at a different stage of maturation (17, 25). A sequential relation between these two major NK cell subsets has been recently reported in that it has been shown that CD56high NK cells have longer telomeres than CD56low NK cells and can differentiate into CD56low in humanized mice in the presence of human IL-15, thus suggesting that they represent a more immature stage (26–28).

Recently, authors identified a distinct CD56low NK cell subset based on CD16 expression levels in the BM and PB of healthy children and acute lymphoblastic leukemia (ALL) pediatric patients (**Figure 1**). The CD56lowCD16low NK cells are more prominent in the BM, and in this organ their frequency further increases in ALL children. Both BM and PB CD56lowCD16low NK cells release IFNγ upon IL-12 plus IL-15 stimulation, and are the major killer population against K562 erythroleukemia cells. However, unlike healthy donors, BM and PB CD56lowCD16low NK cells from ALL children poorly degranulate in response to K562 target cell stimulation. Interestingly, using PB NK cell subsets from two haploidentical HSC donors as source of effector cells and the leukemic blasts of the corresponding recipients as targets, CD56lowCD16low NK cells are the unique population capable of killing leukemic blasts (29).

Overall, our findings suggest that CD56lowCD16low NK cells represent an intermediate state between CD56high and CD56lowCD16high NK cells. However, the lower levels of CD16 may also imply that CD56lowCD16low NK cells represent a post-activation stage, as a disintegrin and metalloproteinase-17 (ADAM-17)-dependent CD16 shedding can occur following NK cell activation (30, 31). In this regard, in accordance with the increased number of CD56lowCD16low NK cells in leukemic children, our preliminary data indicate that ADAM-17 is significantly more abundant in their BM plasma as compared to healthy donors (32).

## **NK CELLS FOR HEMATOLOGICAL CANCER THERAPY: ALLOREACTIVE NK CELLS**

The ability of NK cells to kill leukemic cells in mice without T-cell involvement and prior sensitization was first reported in 1975 (33, 34). Thereafter, growing evidence showed that NK cells preferentially lyse target cells expressing lower or aberrant MHC class I molecules. Based on this evidence, Kärre et al. formulated the missing self hypothesis, arguing that NK cells survey the body for the expression of self-MHC class I molecules and destroy cells on which they are missing (35). Only at the beginning of the 1990s, MHC I inhibitory receptors were discovered in the mouse and humans, and they were shown to deliver negative signals to the NK cells, thus, preventing their cytotoxic activity (2). Accordingly, the use of human NK cell clones revealed that NK cells are unable to lyse autologous normal cells when they express inhibitory receptors for at least one self class I allele. Subsequent findings indicated that NK cells are endowed with alloreactivity and can lyse target cells lacking MHC I molecules (36). This notion has had important clinical implications.

The NK cell therapeutic potential was clearly demonstrated when a potent anti-leukemic effect was observed in patients

with acute myeloid leukemia (AML) undergoing mismatched/ haploidentical hematopoietic stem cell transplantation (HSCT). This protocol of transplantation relied on the generation of alloreactive NK cells with a KIR repertoire unable to bind to host MHC class I molecules, and was associated with a 65% probability of disease-free survival, decreased incidence of relapse, and no increased incidence of graft versus host disease (GVHD) in T-cell-depleted transplants (37). However, patients with KIRligand incompatibility can be still at high risk of GVHD as Tcell alloreactivity may dominate NK cell alloreactivity in minimally T-cell-depleted grafts and in T-cell-repleted transplants (38). The AML-specific effect in unrelated donor transplantation was observed only under particular conditions, including infusion of high doses of stem cells, almost complete depletion of T-cells, no post-grafting immunosuppression, and donors selected for the perfect mismatch at HLA loci (39).

The KIR genes are polymorphic and are organized into two broad haplotypes: group A KIR haplotype, which encodes mainly inhibitory receptors (KIR2DL2/3, KIR3DL2/3, KIR2DL4, KIR3DL1) and one activating receptor (KIR2DS4), and group B KIR haplotype, which encodes both inhibitory (KIR2DL2/3, KIR3DL2/3, KIR2DL5B/A, KIR2DL4, KIR3DL1) and several activating KIR (KIR2DS1, KIR2DS2, KIR2DS3, KIR2DS5, KIR3DS1) (**Table 1**) (40, 41).

Although the initial findings outlined the importance of inhibitory KIRs on the outcome of HSCT, several later studies also focused attention on the influence of activating KIRs. Thus, AML patients transplanted with HSC from donor with alloreactive NK cells carrying KIR2DS1 displayed a lower rate of leukemia relapse as compared to those carrying KIR3DS1 who had no effect on leukemic disease but showed a reduced risk of infection-related mortality (42). Similarly, Mancusi et al. reported that transplantation from donors with KIR2DS1 and/or KIR3DS1 resulted in reduced non-relapse mortality and improved survival (43).

The KIR protective effects were dependent on high levels of HLAC ligands and was restricted to donors with HLA-C1/C1 or HLA C1/C2, whereas it was lost if the donors were HLA C2/C2. This is probably due to the high levels of activating HLA C2 ligands that could induce NK cell tolerance, thus impairing their anti-leukemic effector functions (42).

Variation of the KIR gene family and the impact of KIR-ligand mismatch on the outcome of HSCT have been also addressed.



*Gray and white box contain telomeric and centrometric KIRs respectively; C1/c2 denotes a strong reaction with C1 and weaker cross-reaction with C2; ? denotes uncertainty.*

Kröger et al. analyzed the KIR haplotypes in leukemic patients who received T-cell-depleted unrelated HSCT. They observed that only patients that have received transplants from donors carrying group A haplotype or a small number of activating KIR genes, exhibit reduced relapse and increased disease-free survival. This effect was observed only for AML/myelodysplastic syndrome and to a lesser extent for chronic myeloid leukemia, whereas no effect was evidentiated for ALL (44). The influence of donor and recipient KIR genotype on the outcome of HSCT between HLAmatched siblings was also reported by McQueen et al. Transplants were divided in four groups according to the combination of A and B KIR genotype in the donor and recipient. Better survival was found to be associated with the donor lacking and the recipient having group B KIR haplotype. When haplotype B was present in the donor and absent in the recipient, increased relapse and acute GVHD was observed only if recipient and donor were homogyzous for HLAC1 KIR ligand and lacked HLAC2 ligand.

These findings could be attributable to the presence of activating KIRs in the graft, with a preferential promotion of GVHD but not GVL response, as the activating KIRs on grafted NK or T cells might cause host alloaggression and impaired reconstitution of a responsive immune system (45). Conversely, a multicenter analysis demonstrated a significant and substantial survival benefit for AML patients receiving grafts from unrelated donors having 1 or 2 KIR B haplotypes, thus providing evidence that donors with KIR B haplotype should be used preferentially in HLA-matched or HLA-mismatched unrelated donor transplantation (46). In addition, it is not the presence of the activating B haplotype by itself, but rather the presence of three particular donor genes (2DL5A, 2DS1, and 3DS1) within the B haplotype that is associated with reduced relapse (47). Moreover, AML patients transplanted with HSC from donor with alloreactive NK cells carrying KIR2DS1 and/or KIR3DS1 also display a reduced mortality related to infections, and thus a better event-free survival (42, 43).

Although many reports demonstrate a therapeutic role for alloreactive NK cells in AML, evidence are also available on the importance of KIR-HLA matching. In this regard, Farag et al. analyzed the outcome of 1571 unrelated donor–recipient transplanted patients with myeloid malignancies by comparing donor–recipient pairs, such as HLA-A, -B, -C, and -DRB1 matched, KIR-ligand-mismatched, and HLA-B and/or-Cmismatched but KIR-ligand-matched, and reported that treatment-related mortality, treatment failure, and overall mortality were lowest only after matched transplantation (48).

No clear benefit has been observed for haploidentical HSCT in adult ALL patients with respect to AML patients (37, 44). However, successful results have been obtained in children with ALL transplanted with haploidentical T cell depleted HSC (49). A possible explanation for the distinct role of alloreactive NK cells in pediatric versus adult ALL transplanted patients stems from a recent report demostrating a differential activating ligand repertoire on the leukemic cells. Pediatric ALL blasts exhibited higher expression levels of both the DNAM-1 ligand Nec-2, and the NKG2D ligands ULBP-1 and ULBP-3, as compared to adult ALL blasts (50).

# **NK CELL RECONSTITUTION AFTER HSCT AND ADOPTIVE INFUSION OF NK CELLS**

After HSCT, the first class of lymphocytes which reconstitute in the PB are NK cells that precede T cell reconstitution. The earliest reconstituted NK cells exhibit a CD56high phenotype and express high levels of NKG2A/CD94 and lower levels of inhibitory KIR (51, 52). The analysis of the functional ability of NK cells reconstituted after adult unrelated donor or umbilical cord blood grafting reveal that NK cells from T cell-depleted transplant recipients without immunosuppression, display poor degranulating ability, whereas degranulation is normal or increased in patients undergoing T-cell-repleted transplants and receiving immunosuppression. Based on this observation, a role for T cells in NK cell education and KIR acquisition has been suggested (53). However, another study on the comparison of NK cell reconstitution in T cell-repleted and T cell-depleted HLA-matched sibling, HSCT, indicates that functional recovery of both uneducated and educated NK cells is similar in T cell-depleted and T cell-repleted settings. Moreover, NKG2A<sup>+</sup> donor NK cells are predominant early after transplantation before expression of KIR, whereas the NKG2A*−* NK cells expressing KIR for non-self ligand, remain tolerant in both settings, suggesting that NK cell subsets expressing inhibitory receptors for non-self HLA class I molecules remain hyporesponsive after HLA-matched HSCT (54).

Overall, the analysis of the NK cell repertoire and functional ability at different times after transplantation reveal a marked hyporesponsiveness of NK cells early after transplantation (55).

Moreover, cytokine producing and degranulating abilities are not co-expressed in reconstituting NK cells after allo-HSCT, as target cell-induced IFNγ production is markedly diminished in all transplant settings, whereas the cytotoxic activity is impaired only in T cell-depleted HSCT. The decreased ability to produce IFNγ is rapidly reverted by exposure to low dose of IL-15 suggesting a potential therapeutic role for this cytokine by enhancing NK cell protective ability against infection and relapse (56).

In accordance with these findings, our preliminary results on the CD56lowCD16low NK cells indicate that like CD56high NK cells this subset is present in the PB and BM already at 1 month post-T cell-depleted HSCT, and their number is increased with respect to healthy individuals. However, unlike healthy donors, no differences in CD56lowCD16low NK cell distribution between the two tissue compartments during the first 6 months after HSCT are observed. In transplanted patients, CD56lowCD16low NK cells produce higher levels of IFNγ after IL-12 plus IL-15 stimulation, but they are still the only NK cell degranulating subset when challenged with K562 cells, although the extent of degranulation is lower than that of healthy controls (Stabile et al. unpublished observations).

Adoptive transfer of NK cells have been also considered a promising therapeutic option in the treatment of hematological malignancies, especially in T-cell-depleted haplo-SCT setting because their potent GVL effect (55, 56).

The early clinical trials based on adoptive cell transfer, utilized lymphokine-activated killer (LAK) cells generated from autologous PBMCs cultured *in vitro* with high doses IL-2 for 3–7 days in order to induce anti-tumor killer cells that mainly consisted of NK cells (57–59). Systemic high doses of IL-2 were also administered in order to activate the autologous NK cells *in vivo*; however, in this case, severe toxicity occurred due to capillary leak syndrome induced by IL-2 (59, 60). Subcutaneous administration of low doses of IL-2 alone or in combination with LAK cells gave encouraging results only in patients with melanoma and renal carcinoma (61, 62). Although these approaches augmented *in vivo* activity of NK cells, no consistent efficacy of autologous NK-cell therapy could be detected in patients with other cancer histotypes, including hematological malignancies (62).

The failure of this kind of immunotherapy has been attributed to downregulation of NK cell activity by KIR engagement by self-MHC (37), competition with recipient's lymphocytes for cytokines and space, chronic immunosuppression induced by tumor and/or expansion of Treg cells by IL-2 (63, 64). However, by analyzing cancer patients (metastatic melanoma, renal cell carcinoma, refractory Hodgkin's disease, and refractory AML) subjected to adoptive transfer of human NK cells from

#### **REFERENCES**


haploidentical-related donors, Miller and collaborators demonstrated that transferred NK cells can be expanded *in vivo* and that expansion is dependent on the more intense cyclophosphamide/fludarabine chemotherapy regimen that induces lymphopenia and high endogenous concentrations of IL-15, which are not observed when lower doses of chemotherapy are administered. More importantly, 5 of 19 poor-prognosis AML patients achieved complete remission after haploidentical NK cell therapy, with a significant higher complete remission rate when KIR-ligand mismatched donors were used (65). In accordance with the presence of high concentrations of IL-15 mature NK cells transferred in high-risk AML patients undergoing haploHSCT, were found to proliferate *in vivo* during the early days after haplo HSCT even in the absence of exogenous IL-2 administration, and this resulted in relative low patient relapse rate (66). Moreover, IL-15 together with IL-12 and IL-18 was reported to increase the expression of high affinity IL-2 receptor that was associated with increased NK cell survival, proliferation, and effector function, thus leading to propose immunotherapeutic strategies based on short cytokine preactivation of NK cell before adoptive transfer and followed by low doses of IL-2 therapy (67, 68).

The existence of NK cell subsets with distinct phenotype, functional ability, and adhesion and chemotactic properties that drive their tropism to different tissue compartments, strongly suggests that NK cell-based therapies still require a better identification of NK cell subsets endowed with optimal anti-tumor potential and tropism to tumor sites, to achieve optimal clinical benefit.

### **CONCLUSION**

Authors suppose that the identification and characterization of multifunctional NK cell subsets, which can be rapidly mobilized in the PB and with strong ability to migrate to tumor sites, would provide new insights on the role played by NK cells under pathological conditions and, more importantly, would allow the design of new approaches of adoptive immunotherapy to treat patients with NK cell-susceptible hematological malignancies. Further studies would also clarify the relationship between emergence and persistence of distinct NK cell subsets during post-graft reconstitution and the maintenance of a state of remission.

#### **ACKNOWLEDGMENTS**

This work was supported by grants from the Italian Association for Cancer Research (AIRC: project #16014 and AIRC 5xmille: project #9962), Istituto Pasteur-Fondazione Cenci Bolognetti and Ministero dell'Istruzione, dell'Università e della Ricerca (Centri di Eccellenza BEMM, PRIN: project #PRIN20103 FMJEN and PRIN 2010 C2LKKJ–003, FIRB-MIUR, 60%) and from the Italian Institute of Technology (A2 project).


lymphokine-activated killer cells and recombinant interleukin-2 to patients with metastatic cancer. *N Engl J Med* (1985) **313**:1485–92. doi:10.1056/ NEJM198512053132327


**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 Gismondi, Stabile, Nisti and Santoni. 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.*

# **Increased NK cell maturation in patients with acute myeloid leukemia**

*Anne-Sophie Chretien<sup>1</sup> \*, Samuel Granjeaud<sup>2</sup>† , Françoise Gondois-Rey 1,3† , Samia Harbi <sup>4</sup> , Florence Orlanducci <sup>2</sup> , Didier Blaise1,4, Norbert Vey 1,5, Christine Arnoulet 1,6 , Cyril Fauriat <sup>1</sup> and Daniel Olive1,2*

*<sup>1</sup> Centre de Cancérologie de Marseille, Team Immunity and Cancer, INSERM, U1068, Institut Paoli-Calmettes, Aix-Marseille Université, UM 105, CNRS, UMR7258, Marseille, France, <sup>2</sup> Centre de Cancérologie de Marseille, Systems Biology Platform, INSERM, U1068, Institut Paoli-Calmettes, Aix-Marseille Université, UM 105, CNRS, UMR7258, Marseille, France, <sup>3</sup> Centre de Cancérologie de Marseille, Plateforme d'Immunomonitoring en Cancérologie, INSERM, U1068, Institut Paoli-Calmettes, Aix-Marseille Université, UM 105, CNRS, UMR7258, Marseille, France, <sup>4</sup> Hematology and Transplant and Cellular Therapy Department, Institut Paoli-Calmettes, Marseille, France, <sup>5</sup> Hematology Department, Institut Paoli-Calmettes, Marseille, France, <sup>6</sup> Biopathology Department, Institut Paoli Calmettes, Marseille, France*

#### *Edited by:*

*Raquel Tarazona, University of Extremadura, Spain*

#### *Reviewed by:*

*Evelyn Ullrich, Goethe University Frankfurt, Germany Nadia Guerra, Imperial College London, UK*

#### *\*Correspondence:*

*Anne-Sophie Chretien anne-sophie.chretien@inserm.fr †Samuel Granjeaud and Françoise Gondois-Rey have contributed equally to this work.*

#### *Specialty section:*

*This article was submitted to NK Cell Biology, a section of the journal Frontiers in Immunology*

> *Received: 16 July 2015 Accepted: 23 October 2015 Published: 06 November 2015*

#### *Citation:*

*Chretien A-S, Granjeaud S, Gondois-Rey F, Harbi S, Orlanducci F, Blaise D, Vey N, Arnoulet C, Fauriat C and Olive D (2015) Increased NK cell maturation in patients with acute myeloid leukemia. Front. Immunol. 6:564. doi: 10.3389/fimmu.2015.00564* Understanding immune alterations in cancer patients is a major challenge and requires precise phenotypic study of immune subsets. Improvement of knowledge regarding the biology of natural killer (NK) cells and technical advances leads to the generation of high dimensional dataset. High dimensional flow cytometry requires tools adapted to complex dataset analyses. This study presents an example of NK cell maturation analysis in Healthy Volunteers (HV) and patients with Acute Myeloid Leukemia (AML) with an automated procedure using the FLOCK algorithm. This procedure enabled to automatically identify NK cell subsets according to maturation profiles, with 2D mapping of a four-dimensional dataset. Differences were highlighted in AML patients compared to HV, with an overall increase of NK maturation. Among patients, a strong heterogeneity in NK cell maturation defined three distinct profiles. Overall, automatic gating with FLOCK algorithm is a recent procedure, which enables fast and reliable identification of cell populations from highdimensional cytometry data. Such tools are necessary for immune subset characterization and standardization of data analyses. This tool is adapted to new immune cell subsets discovery, and may lead to a better knowledge of NK cell defects in cancer patients. Overall, 2D mapping of NK maturation profiles enabled fast and reliable identification of NK cell subsets.

**Keywords: AML, NK maturation, automated gating, FLOCK algorithm, multidimensional flow cytometry**

# **INTRODUCTION**

Natural Killer (NK) cells are immune effectors that play a key role in tumor rejection, with an ability to detect and lyse tumor cells without prior stimulation (1–3). Their fundamental role in anti-tumor immune response has been widely demonstrated in both solid tumors and malignant hemopathies, and parameters linked to NK cell activation can either be prognostic factors (4–9) or predictive markers of response to chemotherapy or radiotherapy (10, 11). Thus, monitoring NK cell parameters seems to be an important point to stratify patients at diagnosis and to assess NK cell response during the course of treatment. For such applications, NK cell alterations in cancer patients need to be further described in order to dissect mechanisms involved and define the relevant therapeutic strategies based on NK restoration (12, 13).

In addition to classical NK activating receptors, maturation is fundamental for triggering immune response while maintaining self tolerance (14). NK cell maturation and activation are intrinsically linked (14). Therefore, this point is probably of primary importance for NK cell reactivity in the context of malignancies. Recent studies highlight increasing number of markers that define NK cell subsets according to maturation parameters. In mice, some parameters appear as relevant markers to define NK cell clusters according to maturation process, such as CD16 or CD11b, CD27, and Mac-1, which define NK subsets with progressive acquisition of NK cell effector functions (14–17). In Humans, four parameters further define NK cell subsets according to the expression of NKG2A, KIR, CD57, and CD56 (16, 18). NK cells initially differentiate from immature CD56bright to CD56dull phenotype, with different functions, including cytotoxicity, cytokine production, and migratory capacities (14, 19, 20). Subsequently, NK cells lose expression of NKG2A, and sequentially express CD57 and KIR. Accordingly, five states of maturation stages are defined according to expression of these markers.

To date, an increasing number of NK cell activation and maturation markers have been described (17, 21). The improvement of knowledge regarding the biology of NK cells led to an increase of markers required to phenotypically and functionally characterize these cells (21). Technological advances in the field of flow and mass cytometry led to the development of complex panels to study NK cells, with subsequent generation of high dimensional dataset (21). In this context, manual processing of the data presents many limitations. First, generation of gates in two dimensions is time-consuming, and subject to operator subjectivity (22). Hence, gating strategy can impact on results and conclusions when multiple gates are drawn [(22); Gondois-Rey et al. manuscript in preparation]. Second, the high number of results generated is sometimes hard to interpret and summarize, in particular when large cohorts of patients are analyzed. Another problem is the global comprehension of a complex system and interpretation of results when conclusions are drawn parameter by parameter. New tools are then required to address these problems. Recently, algorithms for automatic gating and 2D mapping of high dimensional dataset have been developed, such as Spade (23), viSNE (24), flowClust (22), or FLOCK (25, 26). These algorithms combined to classification methods enable the visualization of multiple parameters and summarize information. These approaches are of particular importance to enable data visualization, in particular in the context of study of complex systems such as immunity. The present study is an example of application of NK maturation profiling in Healthy Volunteers (HV) and patients with Acute Myeloid Leukemia (AML) using automated analysis of flow cytometry data.

#### **PATIENTS AND METHODS**

# **Patients and Healthy Volunteers**

Fresh peripheral blood samples were prospectively collected from AML patients (*N* = 18) at diagnosis before induction chemotherapy and from aged-matched healthy volunteers (*N* = 18). All participants gave written informed consent in accordance with the Declaration of Helsinki. Patients above 65 years at diagnosis were excluded. The entire research procedure was approved by the ethical review board (Institut Paoli-Calmettes Marseille, France). **Table 1** lists the baseline characteristics of patients.

#### **Flow Cytometry**

A FACS Canto II (BD Biosciences, San Jose, CA, USA) was used for flow cytometry. NK cells from whole blood EDTA were immunostained with Krome Orange-conjugated anti-CD45, Phycoerythrin cyanin 7 (PC7)-conjugated anti-CD3, allophycocyanin (APC)-conjugated anti-CD56, fluorescein isothiocyanate (FITC) conjugated anti-CD158b1b2j, FITC-conjugated anti-CD158a,h (further referred to as KIR), APC-alexafluor 750 (APC AF 750) conjugated anti-CD159 (NKG2A), pacific blue-conjugated anti-CD57. All the antibodies used in the study were a kind gift of Beckman Coulter, Marseille, France. Red blood cells were lysed with BD FACS Lysing solution (BD Biosciences) before data acquisition.

## **Cluster Identification Procedure**

FCS files were read, compensated, transformed, and exported using flowCore (R, Bioconductor) (27). FLOCK algorithm was then applied to each exported data file (26). Resulting gated data were imported with R. Centers of populations were extracted, assembled, and exported as a unique tabulated text file using R. MeV permitted heatmap visualization and hierarchical clustering (28). Centers were clustered using euclidean distance. The tree of centers was cut at a threshold that results in clusters with homogeneous mean fluorescence intensity (MFI) profile. Those clear populations were annotated using expert knowledge. The automated gating and cluster identification procedure was described by Gondois-Rey et al. (manuscript in preparation).

## **Statistical Analyses**

Statistical analyses were performed using GraphPad Prism (GraphPad Software, San Diego, CA, USA). Differences in the distribution of continuous variables between categories were analyzed by either Mann–Whitney test (for comparison of two groups) or Kruskal–Wallis with Dunns' post test (comparison of three or more groups). Statistical significance was set at *P <* 0.05.

# **RESULTS**

## **Automatic Gating with FLOCK Provides Reliable Results**

Multiparametric flow analysis of NK cell subsets has become complex over recent years with the identification of several subsets of NK cells depending on classical activating receptors, inhibitory receptors, and maturation markers. More importantly, it has become evident that rare subsets may be under or overestimated during manual analysis by different investigators and different software. We have, therefore, searched for a more reliable tool to provide unbiased analysis of NK cell subsets in the peripheral blood of AML patients. First, PBMCs from healthy donors (*N* = 18) and AML patients at diagnosis (*N* = 18) were isolated and stained for CD56, CD3, KIR (see PATIENTS AND METH-ODS), NKG2A, and CD57. Then CD3*−*CD56<sup>+</sup> live NK cells were

#### **TABLE 1 | Patients characteristics**.


initially manually pre-gated, exported, and then analyzed with FLOCK algorithm.

Results obtained with manual gating were compared with results obtained with FLOCK. Annotated clusters were merged and graphically compared to the equivalent subsets obtained with manual gating (**Figure 1A**). Frequencies of FLOCK-gated and manually gated subsets of NK cells in HV samples with respect to CD56, CD57, KIR expression were comparable (**Figure 1B**). Thus, for each sample, proportions of NK cells within the different clusters with the two approaches were found to be fully consistent.

# **Automated Gating with FLOCK Algorithm Evidences NK Subpopulations**

Natural killer cell maturation profiles in HV and AML patients were defined according to FLOCK output. For cluster annotation of FLOCK output data, we used an unsupervised hierarchical clustering with MeV (**Figures 2A,B**). Overall, the procedure enabled identification of five subsets of NK cells based on the expression of CD56, CD57, KIR, and NKG2A in both patients and healthy volunteers. NK cell differentiate from CD56bright to CD56dim phenotype. CD56bright phenotype then defines the most immature subset of circulating NK cells. In CD56dim NK cells, expression of NKG2A, KIR, and CD57 define several maturation stages (18). Automatic gating procedure with FLOCK enable identification of these different subsets of NK cells, with CD56bright NK cells, and among CD56dim cells, four subsets defined by the positivity or the negativity of KIR and CD57 (**Figures 2A,B**). In accordance with previous studies, KIR positivity inversely correlates with NKG2A expression, in both HV and AML patients (18, 29). We then considered the repartition of NK cells within the different clusters. On average, the percentage of CD56bright cells was found to be significantly lower in AML patients compared to HV (1.3 *±* 3.2% vs. 6.4 *±* 5.8%, respectively; *P* = 0.001) (**Figures 2C,D**).

Overall, 2D mapping of NK maturation profiles enabled the visualization of high dimensional dataset as well as fast and reliable identification of NK cell subsets. With this unsupervised automated gating of NK cells with four parameters, the algorithm was able to distinguish all the NK subsets that were previously described in the literature, but did not identify any new population. Notably, NKG2A was not informative in NK cell cluster definition by the algorithm.

# **AML Patients Present Distinct Maturation Profiles**

Patients and HV were clustered according to the percentages of NK cells represented in the CD56bright, KIR*−*/CD57*−*, KIR+/CD57*−*, KIR*−*/CD57+, KIR+/CD57<sup>+</sup> clusters with MeV using unsupervised hierarchical clustering (HClust, Pearson correlation) (**Figure 3**). This representation allowed defining three distinct groups of patients. The first group of patients (*N* = 7) presented a NK cell maturation profile with 50% (range: 40–67%) NK cells in the cluster KIR*−*/CD57*−* cluster. Considering the repartition of NK cells within the different clusters, there was no significant difference between this group and HV. The second group of patients (*N* = 5) presented an intermediate maturation

profile, with 43% (range: 30–52%) NK cells in the KIR*−*/CD57*−* cluster and 37% (range: 28–48%) NK cells in the KIR+/CD57<sup>+</sup> cluster. For this group, the proportion of NK cells in the CD56bright cluster was significantly lower than HV (*P* = 0.05). Of note, the apparent high frequency of cells in the cluster KIR+/CD57<sup>+</sup> was not significantly different from HV. The third group of patients (*N* = 6) presented an hyper-maturation profile, with proportions of NK cells of 13% (range: 7–24%) NK cells in the KIR*−*/CD57*−* cluster and 58% (range: 34–83%) NK cells in the KIR+/CD57<sup>+</sup> cluster. For this group, the proportion of NK cells in the KIR*−*/CD57*−* cluster was significantly lower than HV (*P <* 0.05) whereas the proportion of NK cells in the KIR+/CD57<sup>+</sup> cluster was significantly higher (*P <* 0.01).

In conclusion, we observed that NK cells in AML patients display marked differences compared to HV, with a strong interindividual variability, defining three distinct groups of patients according to NK maturation profiles.

# **DISCUSSION**

Accumulating evidence highlights NK cell parameters as potential prognostic factors in cancer patients, which provides a strong

rationale for developing therapeutic strategies aiming at restoring NK cell functions (4–9). However, reaching this point warrants better characterization of NK cell alterations in cancer patients as well as elucidation of the mechanisms involved (30, 31).

Among important parameters involved in NK cell functions, the maturation process is of particular importance; since, depending on the maturation stage NK cells will gain or lose important functions, such as migration capacities, effector functions, response to cytokines, proliferative capacities, IFN-γ production, or cytotoxic activity (8, 14, 32). All these functions are absolutely required for a functional effect against tumor cells. Thus alteration of the maturation process is likely to

according to the percentage of NK cells represented in the CD56bright, KIR*−*/CD57*−*, KIR+/CD57*−*, KIR*−*/CD57+, KIR+/CD57<sup>+</sup> clusters using hierarchical clustering (HClust, Pearson correlation). This second step of clusterization enabled to define three distinct groups of patients. The frequency of NK cells in each subset for each individual is presented in the right panel. The dashed lines represent the mean frequencies of NK subpopulations in HV and in the three groups of patients.

impact NK cell functions, with direct consequences on patients' survival (33).

Natural killer cell maturation is a multistep process marked by differential expression of many markers, among which CD56, NKG2A, KIR, and CD57 are of particular importance (18). NK cell subsets can also be defined according to the expression of CD16 and CD56. For instance, it has been described discrete stages of NK cell differentiation. First, CD56bright NK cells expressing low levels of CD16 correspond to a transition between early immature CD56bright CD16*<sup>−</sup>* NK cells and CD56dim CD16<sup>+</sup> NK cells (34). Variations of NK cells in these different compartments have been described in several clinical conditions such as HIV infection and in aging (13). In addition, another NK cell population of CD56*−*CD16<sup>+</sup> cells has been described and found expanded in particular pathological conditions such as HIV or hepatitis C virus infection (35, 36). Although these discrete stages have been evidenced, the functions of these cells remain elusive, particularly in the context of AML.

Whether NK maturation is impacted by the close proximity with leukemic blasts is an important question. Under physiological conditions, circulating NK cells differentiate from CD56bright to CD56dim phenotype. Then NK cells lose NKG2A expression, and gain KIR expression. CD57 is acquired at later stages of differentiation, and defines a subset of NK cells with low proliferative capacities and high cytotoxic potential (16, 18). In our study, we show that NK cells in AML patients present marked differences compared to Healthy Volunteers. The proportion of CD57<sup>+</sup> NK cells is increased in one-third of patients, at the expense of less mature NK subsets, with a drastic decrease of immature NK cells. Although CD57<sup>+</sup> NK cells have been described as the most cytotoxic subset of NK cells, in the context of AML, we still need to check whether these cells display efficient cytotoxic activity on cancer target cells. In addition, the impact of these extreme maturation profiles on clinical outcome warrants further exploration on a larger cohort of patients.

Natural killer cell maturation has been studied in human breast and lung cancer (37, 38). In contrast to our study, tumorinfiltrating NK cells display an immature phenotype, with high percentages of CD56bright NK cells compared to healthy tissues. A notable difference in our study is that we analyzed peripheral blood cells, whereas the previously cited studies were done with infiltrating NK cells. Mamessier et al. hypothesized that NK cells were de-differentiated rather than immature cells; this could explain the high proportion of CD56bright in tumor tissue, without direct impact on central NK maturation or migration of the most immature cells on the tumor site. One additional difference in the context of AML is that NK cells maturate in close contact with tumor cells. This could explain the high proportion of highly maturated NK cells. However, some authors also described CD56dimCD57<sup>+</sup> enrichment in tumor-infiltrated lymph nodes in patients with metastatic melanoma, with significant impact on patients' survival (33).

Technical advances in flow and mass cytometry now enable the dissection of NK cell biology with high precision, with the subsequent need for tools adapted to the analysis of datasets with unprecedented dimensionality (21, 25, 39). In our study, we used an automated procedure using the FLOCK algorithm and hierarchical clustering, which enabled unsupervised identification of NK subsets and patients profiling based on NK parameters. Automatic gating algorithms are powerful and reliable tools adapted to high dimensional dataset analysis (25, 40) with potential limitations highlighted in the case of rare populations (40, 41). In the case of immunomonitoring studies on large cohorts of patients, the development of such approaches is of primary importance for data analysis standardization. First, the high number of subjects included in these studies requires automated gating in order to reduce the time of analysis. Second, visualization of all the clusters allows fast and unsupervised identification of cell populations. Moreover, the hierarchical classification of patients according to maturation profiles enables the discovery of distinct patterns or specific subgroups among patients. The clinical consequences of such observations should be evaluated on larger cohorts of patients. Considering the potential impact of NK maturation on clinical outcome, NK cell maturation profiling might be informative in prognostic immune signatures and may find applications in patients' stratification at diagnosis.

# **AUTHOR CONTRIBUTIONS**

A-SC: design of the work, data analysis and interpretation, statistical analyses and interface with biological findings, redaction of the article, revisions and final approval of the version to be published, agreement of all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. SG: data interpretation, statistical analyses and interface with biological findings, drafting and revisions of the work, final approval of the version to be published, agreement of all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. FG-R: data interpretation, statistical analyses and interface with biological findings, drafting and revisions of the work, final approval of the version to be published, agreement of all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. SH: data acquisition and interpretation, revisions of the work, final approval of the version to be published, agreement of all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. FO: data analysis and interpretation, drafting of the work and final approval of the version to be published, agreement of all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. DB: conception and design of the work, revisions of the work and final approval of the version to be published, agreement of all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. NV: conception and design of the work, revisions of the work and final approval of the version to be published, agreement of all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. CA: conception and design of the work, revisions of the work and final approval of the version to be published, agreement of all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. CF: conception and design of the work, data analysis and interpretation, article redaction, revisions of the work and final approval of the version to be published, agreement of all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. DO: conception and design of the work, interpretation of data for the work, revisions of the work and final approval of the version to be published, agreement of all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

# **ACKNOWLEDGMENTS**

The authors thank Beckman Coulter for their technical advices and for providing the antibodies used in this study. The authors

# **REFERENCES**


thank the CRCM the Immunomonitoring platform and the CRCM Systems Biology platform for their valued contributions to this work.

# **FUNDING**

This work has been financially supported by the INCa and the Canceropole PACA.

the human CD56dimCD16+ NK-cell subset. *Blood* (2010) **116**(19):3865–74. doi:10.1182/blood-2010-04-282301


**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 Chretien, Granjeaud, Gondois-Rey, Harbi, Orlanducci, Blaise, Vey, Arnoulet, Fauriat and Olive. 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.*

# Natural Killer Cell Recognition of Melanoma: New Clues for a More Effective Immunotherapy

*Raquel Tarazona1 \*, Esther Duran2 and Rafael Solana3*

*<sup>1</sup> Immunology Unit, University of Extremadura, Caceres, Spain, 2Histology and Pathology Unit, Faculty of Veterinary Medicine, University of Extremadura, Caceres, Spain, 3 Immunology Unit, Instituto Maimónides de Investigación Biomédica de Córdoba (IMIBIC), Reina Sofia University Hospital, University of Cordoba, Cordoba, Spain*

Natural killer (NK) cells participate in the early immune response against melanoma and also contribute to the development of an adequate adaptive immune response by their crosstalk with dendritic cells and cytokine secretion. Melanoma resistance to conventional therapies together with its high immunogenicity justifies the development of novel therapies aimed to stimulate effective immune responses against melanoma. However, melanoma cells frequently escape to CD8 T cell recognition by the down-regulation of major histocompatibility complex (MHC) class I molecules. In this scenario, NK cells emerge as potential candidates for melanoma immunotherapy due to their capacity to recognize and destroy melanoma cells expressing low levels of MHC class I molecules. In addition, the possibility to combine immune checkpoint blockade with other NK cell potentiating strategies (e.g., cytokine induction of activating receptors) has opened new perspectives in the potential use of adoptive NK cell-based immunotherapy in melanoma.

#### *Edited by:*

*Lutz Walter, Leibniz-Institute for Primate Research, Germany*

#### *Reviewed by:*

*Evelyn Ullrich, Goethe University Frankfurt, Germany Ralf Dressel, University Medical Center Göttingen, Germany*

#### *\*Correspondence:*

*Raquel Tarazona rtarazon@unex.es*

#### *Specialty section:*

*This article was submitted to NK Cell Biology, a section of the journal Frontiers in Immunology*

*Received: 23 October 2015 Accepted: 14 December 2015 Published: 07 January 2016*

#### *Citation:*

*Tarazona R, Duran E and Solana R (2016) Natural Killer Cell Recognition of Melanoma: New Clues for a More Effective Immunotherapy. Front. Immunol. 6:649. doi: 10.3389/fimmu.2015.00649*

Keywords: melanoma, immunotherapy, natural killer cells, adoptive transfer, checkpoint blockade

# INTRODUCTION

Melanoma is largely resistant to current therapies as chemotherapy and radiotherapy (1) and consequently remains as an important cause of mortality mainly in Caucasians. Metastatic melanoma is highly aggressive constituting the most lethal skin cancer (2). Despite the different approaches developed for primary prevention of melanoma, its incidence rate continues increasing in many countries (3).

It has been postulated that melanoma ability of inducing an immune response contributes to patient survival. Thus, melanoma is usually highly immunogenic and induces cytotoxic T cell (CTL)-mediated immune responses. Tumor infiltrating lymphocytes (TILs) have been identified in melanoma lesions usually associated with spontaneous tumor regression and favorable prognostic in primary melanoma (4).

Innate immune responses against melanoma have also been described. Natural killer (NK) cells constitute the first line of defense against transformed cells as tumors or virus-infected cells. *In vitro* experiments have established that NK cells can recognize and destroy melanoma cell lines (5–7). The role of NK cells against melanoma *in vivo* has been demonstrated in murine models (8), and it is also supported by the observation of NK cell alterations (e.g., down-regulation of activating receptors or NK cell exhaustion) in melanoma patients (9, 10) suggesting the development of escape mechanisms to evade NK cell-mediated destruction of melanoma cells.

It is well known that age affects both adaptive and innate immune responses against tumors (11–14). The hypothesis of immunosurveillance against melanoma is further sustained by the recent finding that elderly melanoma patients had a higher incidence of melanoma-related mortality than younger patients in spite of the lower incidence of sentinel lymph node metastasis (15).

Altogether, these characteristics of melanoma reinforce the previous consideration of melanoma as a suitable model for studying tumor immunity. Here, we review the current state of knowledge on NK cell-mediated recognition and lysis of melanoma cells and the up to date immunotherapeutic strategies against melanoma based on NK cells.

## NK Cell-Mediated Anti-Melanoma Responses

The key role played by NK cells as a first line of defense against tumors has been established in hematological malignancies based on the graft-versus-leukemia effect (16–18). However, their role against solid tumors such as melanoma is less recognized. It has been reported that NK cells contribute to melanoma surveillance *in vivo* (19–21). NK cells can actively participate in the initial phase of tumor development and may control metastasis, but the direct action of NK cells against tumor tissue is not well known. NK cells may contribute to cancer elimination not only by the lysis of tumor cells but also by the secretion of cytokines and the promotion of antigen-presenting cell maturation contributing to the adaptive immune response (22–24).

Natural killer cells express several activating receptors that after cross-linking with their respective ligands trigger NK cell degranulation releasing their cytotoxic granule content leading to target cell apoptosis (**Figure 1A**). Research during the last decade has highlighted that several activating receptors are involved in NK cell recognition of tumor cells (6, 25). The existence of diverse ligand–receptor interactions is relevant in melanoma recognition since it has been demonstrated that melanoma cells express a variety of ligands for different NK cell-activating receptors (7). It has been postulated that the integration of multiple activating signals may overcome the inhibitory signals mediated by major histocompatibility complex (MHC) class I-specific inhibitory receptors (25, 26). In addition, different ligands may interact with the same activating receptor as occur for NKG2D ligands (MICA/B and ULBPs) (27) and DNAM-1 ligands [CD112, also named Nectin-2, and CD155 that is considered the poliovirus receptor (PVR)] contributing together to NK cell activation (28). Recently, the family of receptors that bind nectin and nectin-like proteins has expanded. It has been described that some of these activating receptors have an inhibitory counterpart that compete for the same ligands. For instance, the activating DNAM-1 and the inhibitory T cell immunoreceptor with immunoglobulin and ITIM domains (TIGIT) compete for the same ligand (CD155) on the target cells, regulating NK cell activation (29). The receptor TACTILE (CD96) also binds CD155 and may inhibit cytokine secretion in mice (30, 31), although its role in human NK cell function remains unclear. Other receptor for nectin-like proteins is CRTAM that is expressed on NK cells and CD8 T cells upon

activation and binds nectin-like 2 promoting adhesion to target cells (32).

A characteristic that makes melanoma a prototype for the study of NK cell-mediated tumor destruction is the fact that melanoma cells frequently show altered expression of MHC class I molecules (33). Diminished expression of MHC class I molecules makes melanoma cells unaffected by CTLs but facilitate NK cell killing (34). The altered MHC class I phenotypes on tumor cells can be classified as reversible ("soft lesions") when the MHC class I expression can be recovered or upregulated after cytokine treatment or irreversible ("hard lesions") when the molecular defect is structural and cannot be recovered such as loss of heterozygosity due to mutations on β2 microglobulin (34). Thus, the molecular mechanisms involved in the down-regulation or loss of MHC class I molecules in tumor cells have an impact on tumor development and in CTL-based immunotherapy efficacy. In experimental and clinical models, tumor regression has been associated with reversible MHC class I alterations whereas irreversible alterations were linked with tumor progression (34–36).

Mature NK cells express CD16 (FcγR-III) that mediates antibody-dependent cell cytotoxicity (ADCC) representing an effective mechanism of lysis of antibody-coated target cells. However, it has been described that NK cell activation is associated with metalloproteinase-mediated cleavage of CD16 molecules. The treatment with metalloproteinase inhibitors prevented CD16 down-regulation and increased NK cell polyfunctionality (cytokine production and degranulation). The use of metalloproteinase inhibitors in monoclonal antibody (mAb) based immunotherapy is proposed to benefit cancer patients (37).

# Melanoma Cells Express Ligands for NK Cell-Activating Receptors

We have previously analyzed a large panel of melanoma cell lines from the "European Searchable Tumor Cell Line and Data Bank" (ESTDAB, http://www.ebi.ac.uk/ipd/estdab/) and "Outcome and impact of specific treatment in European research on melanoma" (OISTER, QLG1-CT-2002-00668) projects demonstrating a high expression of ligands for NK-cell activating receptors on these cell lines. A high percentage of melanoma cell lines expressed ligands for NKG2D (85%) and DNAM-1 (95%)-activating receptors. The expression of MICA/B on melanoma cell lines prevailed over ULBP expression (7). Several studies have analyzed the expression of NKG2D ligands on melanoma specimens by immunohistochemistry showing a high heterogeneity. MICA/B expression was observed at a higher frequency than ULBP2 on melanoma metastasis (38). The analysis of MICA expression on melanoma lesions revealed a higher expression in primary melanoma than in metastatic melanoma (39, 40). The pattern of expression was not homogeneous, and interestingly, in some patients, a preferential staining was observed at the invasive front (38). Regarding DNAM-1 ligands, CD155 was found to be expressed in the majority of melanoma cell lines analyzed in contrast with the 26% of melanoma cell lines expressing CD112 (7). The expression of CD155 on melanoma specimens and melanoma cell lines also showed a stronger expression on metastatic melanoma compared to primary melanoma (41).

The identification of cellular ligands for the natural cytotoxicity receptors (NCRs) NKp30, NKp44, and NKp46 has remained elusive until recently. The use of chimera proteins constructed using the extracellular domain of NKp30, NKp44, or NKp46 fused to the Fc immunoglobulin domain (NCR-Fc) or to an amino-terminal isoleucine zipper (NCR-ILZ) allowed to analyze the expression of NCR ligands on tumor cells. A high variability in the binding of NCR chimeras to melanoma cells was observed with melanoma cell lines expressing ligands for NKp30 and NKp44 but not for NKp46 (6, 42) and other cell lines expressing ligands for NKp46 (43). The study of melanoma lesions in patients with metastatic tumors identified NKp44 ligands in all melanoma samples analyzed and NKp30 ligands in the majority of samples, whereas the expression of NKp46 ligands was null (44). The expression of NCR ligands was also analyzed on melanoma cells from lymph nodes and paired samples obtained from skin metastasis. Melanoma cells from lymph nodes showed staining with NKp44-Fc and NKp46-Fc chimeras and were more susceptible to NK cell-mediated lysis than melanoma cells from skin metastasis that had low or negative staining with NCR-Fc (6). These differences probably represent different stages of the disease. Thus, it has been proposed that in early stages, melanoma cells overexpress NCR ligands and during melanoma progression NCR ligand expression is down-regulated (6, 43) representing an immunoescape mechanism used by melanoma.

Recently, several cellular ligands for NCRs have been identified. NKp30 recognizes B7-H6 that has been found expressed on melanoma cell lines (45), human leukocyte antigen (HLA)-Bassociated transcript 3 (BAT3) (46), and CMV pp65 tegument protein (47). The proliferating cell nuclear antigen (PCNA) has been recognized as a NKp44 ligand (48, 49). In contrast, cellular ligands for NKp46 remain elusive. The characteristics of NCR ligands identified so far suggest that these receptors may recognize damage-associated molecular patterns related to cellular stress (e.g., tumor transformation or infection) (50). *In vitro* receptor blocking experiments showing NCR-mediated lysis of melanoma cell lines further support the role of this receptor family in the control of melanoma (5, 6, 51).

# NK Cell–Melanoma Interaction

Natural killer cell recognition and lysis of melanoma cells involve different receptor–ligand interactions including NKG2D-, DNAM-1-, and NCRs-activating receptors. The expression pattern of ligands for activating receptors on melanoma and the expression of MHC class I molecules recognized by inhibitory receptors will determine the activation of NK cells (**Figures 1A– C**). As indicated before, NK cell lysis of melanoma cells may depend on the disease stage and the anatomical location due to the differential expression of ligands (6, 52). Antibody blocking experiments have demonstrated that usually melanoma cell lysis requires signaling through several activating receptors (25, 52).

The role of NKG2D in NK cell recognition and lysis of melanoma cells has been extensively discussed. Whereas, NKG2D is clearly involved in the lysis of melanoma cells expressing high levels of NKG2D ligands, and NCRs and DNAM-1 are the receptors involved in the elimination of melanoma cells with low expression of ligands for NKG2D. Thus, it has been described that NCRs and DNAM-1 cooperation is frequently involved in the lysis of melanoma cells both in humans and in mice (6, 53). The participation of several activating receptors in the activation of NK cells against melanoma contributes to the effective NK cellmediated lysis of these cells (**Figure 1**).

The majority of studies analyzing effector–target interactions in melanoma are performed using cell lines cultured as monolayer or in suspension testing ligand expression correlation with CTL- or NK cell-susceptibility to lysis. Recently, the use of three-dimensional (3D) cell culture systems has been proposed for the analysis of melanoma interaction with lymphocytes. Thus, melanoma cells grown in 3D architecture showed lower recognition by melanoma-specific CTLs compared to those melanoma cells growing in 2D monolayers. It has been proposed that culture in 3D affects the expression of molecules involved in melanoma recognition by CTLs (54, 55). We can speculate that 3D culture also alter the expression of ligands for NK-cell activating receptors increasing melanoma resistance to NK cell lysis in a similar way as occurs in melanoma tissue.

An expansion of highly cytotoxic CD57<sup>+</sup> NK cells has been found in tumor-infiltrating lymph nodes in melanoma patients. Their potential use as a source of cytotoxic NK cells for adoptive immunotherapy is discussed (56). The expansion of highly mature CD57<sup>+</sup> NK cells has been observed in CMV-seropositive individuals, and it is further increased by age (11, 12). These cells represent highly differentiated NK cells with low proliferative capacity and high cytotoxicity. Although these cells have a lower expression of NKp30 and NKp46 (57–59), the expression of the activating receptors DNAM-1 and NKG2C is increased on the CD57+ subpopulation of CD56dimCD16+ NK cells in CMVseropositive young donors, but it is reduced in the old individuals (59). These changes in the expression of cytotoxicity activating receptors may have functional relevance not only against CMV infection but also against other age-associated diseases as cancer. Thus, the potential use of CD57+ NK cells in melanoma immunotherapy requires a detailed analysis of their cytotoxic capacity and the expression of activating receptors since it depends on other factors as CMV latent infection and age (11, 12, 59).

### Checkpoints in NK Cell Activation

Natural killer cell activation depends on a tune balance mediated by inhibitory and activating signals transmitted through surface receptors upon contact with their respective ligands. In this process, the interaction between MHC class I molecules on target cells and MHC class I-specific inhibitory receptors on NK cells represents a major checkpoint regulating NK cell functions (60). Killer cell immunoglobulin-like receptors (KIR) are a family of highly polymorphic receptors that recognize MHC class I molecules. Inhibitory and activating KIRs have been described. KIRs govern NK cell education and function and inhibitory KIR– HLA interactions may be associated with failed tumor immunosurveillance mediated by NK cells (61). NKG2A, an inhibitory C-type lectin-like receptor, forms heterodimers with CD94 and recognizes HLA-E molecules (62–64). The immunoglobulin-like transcript-2 (ILT-2) specific for HLA-G is also expressed by NK cells. It has been observed an inverse correlation between ILT-2 expression on T cells and clinical response in melanoma patients treated with oncolytic virus immunotherapy (65).

The discovery of inhibitory receptor-recognizing ligands other than MHC class I molecules such as TIGIT or the programed cell death-1 (PD-1) molecules constitute novel checkpoints in NK cell activation that requires further consideration (22, 31, 66–68). The PD-1/PD-L1 axis has been described as a checkpoint that regulates NK cell functions in tumor-bearing mice. Thus, blockade of PD-1/PD-L1 in nude mice resulted in anti-metastatic effect supporting the role of PD-1 on NK cell function (69).

Together with the expression level of MHC class I molecules on melanoma cells and the expression of MHC class I-specific inhibitory receptors on autologous NK cells, the expression of activating receptors on NK cells, and their ligands on melanoma are key actors in the final balance leading to an effective NK cell activation (9, 25).

# MELANOMA ESCAPE MECHANISMS TO AVOID NK CELL CYTOTOXICITY

Immune evasion by tumor cells through the down-regulation of MHC class I molecules to avoid CD8 T cell recognition constitutes a well-known mechanism used by melanoma (33). Melanoma loss of MHC class I expression increases its susceptibility to NK cells. As indicated above, the altered expression of HLA class I antigens is frequently found in melanoma (33), and several studies have shown that melanoma cells evolve down-regulating class I antigens to avoid being recognized by CD8<sup>+</sup> T cells (34, 36). However, the analysis of the HLA class I antigen alterations in melanoma cell lines from ESTDAB showed that the most frequently observed phenotype is the down-regulation of HLA-B locus that is reversible after treatment with IFN-γ whereas the total lack of expression as a consequence of gene mutations or deletions leading to HLA heavy chain or β2m deficiency is only found in a minor group of samples (33). The bidirectional interaction between NK cells and melanoma cells induces changes in both effector and target cells (**Figure 1D**). It has been shown that melanoma immunoediting by NK cells make melanoma cells resistant to NK cell-mediated killing by increasing the expression of HLA class I molecules (70) and that blockade of HLA antigens with mAbs results in increased NK cell-mediated killing, indicating that HLA antigens expressed on melanoma cells interact with NK-inhibitory receptors avoiding NK cytotoxicity (71).

It has been also proposed that NK cell-mediated immunosurveillance against melanoma can generate immunoselection of melanoma cell variants with low expression of ligands for activating receptors that are resistant to NK cells (72). Thus, MICA and NCR ligand expression is lower in metastatic melanoma compared to primary melanoma lesions (6, 43). Shedding of soluble ligands for activating receptors constitutes another mechanism used frequently by melanoma cells to escape to the action of effector cells (25). Soluble NKG2D ligands MICA and ULBP2 are released by melanoma cells and can down-regulate the expression of NKG2D on effector cells. Thus, soluble ULBP2 was associated with lower survival in melanoma patients (38). NKG2D ligands can be released by ADAM protease-mediated shedding or secreted in exosomes with different functional outcomes (73). Shedding of B7-H6, a ligand for NKp30, by tumor cells has been recently described (74) also contributing to tumor escape from NK cells.

The down-regulation of NK cell-activating receptors has been described as an additional mechanism that contributes to tumor escape in cancer patients (25, 75–77). Thus, the decreased expression of NKp30 on NK cells from metastatic melanoma patients was associated with a reduced ability to kill melanoma cells (44). NK cells in stage IV melanoma patients displayed low levels of activating receptors that correlated with lower survival (20). IFN-γ released by NK cells induces indoleamine 2,3-dioxygenase (IDO) expression and prostaglandin E2 (PGE2) production by melanoma cells that inhibit NK cell function by down-regulating the expression of NKp30- and NKG2D-activating receptors further contributing to melanoma escape (78, 79).

T cell immunoreceptor with immunoglobulin and ITIM domains signaling after interaction with its ligands suppresses NK cell production of IFN-γ (67). In advanced melanoma patients, CD112 and CD155 were found upregulated in melanoma cells. In these patients, the expression of TIGIT either on CD8<sup>+</sup> T cells or NK cells did not show significant differences compared with healthy donors whereas the expression of DNAM-1 on CD8<sup>+</sup> T cells was down-regulated (66). These results suggest that inhibitory signaling through TIGIT can contribute to immune escape in melanoma.

Finally, suppression of NK cells by factors or cytokines secreted either by tumor cells or other cells in the tumor microenvironment such as myeloid derived suppressor cells (MDSCs) or macrophages can also contribute to immunoescape of cytotoxic cells (22).

All these mechanism together may contribute to the alterations of NK cell phenotype and function described in cancer patients.

# NK CELL-BASED IMMUNOTHERAPY IN MELANOMA

Different strategies of melanoma immunotherapy developed during the last decade focused on the use of checkpoints inhibitors or immune modulators, oncolytic virus therapy, cancer vaccines, adoptive T cell, and NK cell therapies and the use of cytokines (80). Many of those clinical trials are currently underway and include combined therapies. Here, we described those strategies focused on NK cell-mediated activation against melanoma or those immunotherapies that, although are not specifically directed to enhance NK cell function, may favor NK cell activation (**Table 1**).

#### Modulation of NK Cell Responses

There are different strategies to exploit the possibility to modulate NK cells in melanoma immunotherapy. The use of new forms of cytokine therapies or mAbs against tumor antigens can directly contribute to enhance NK cytotoxicity whereas immune checkpoints regulators constitute a novel immunotherapy strategy to modulate immune responses through their interaction with inhibitory receptors on immune cells.

#### Cytokines

Different cytokines have demonstrated a role in tumor immunity. Two cytokines have been approved by the Food and Drug Administration (FDA) for melanoma treatment as single agent: high doses of IL-2 for metastatic melanoma and IFN-α for the adjuvant therapy of Stage III melanoma based on the results obtained in clinical trials using high doses of IL-2 in metastatic melanoma patients (81) and IFN-α that demonstrated a significant benefit in relapse-free and overall survival of high-risk melanoma patients (82). Novel strategies have been developed such as bifunctional molecules consisting in cytokines fused to antibodies that allow the targeted delivery of the cytokines or the expression of cytokines in viral vectors or irradiated tumor cells for their use as vaccines. In addition, cytokines such as IL-2 or IL-15 are also used for the *in vitro* expansion of NK cells and T cells for adoptive transfer (83).

#### Checkpoint Blockade

As indicated before, one of the major checkpoints in NK cell activation is mediated by MHC class I-specific inhibitory receptors interacting with their ligands on target cells. Thus, blockade of this checkpoint constitutes an emerging area of research. Two NK cell checkpoint inhibitors lirilumab (anti-KIR mAb) and IPH2201 (anti-NKG2A mAb) are currently under revision. A safety study to analyze anti-KIR mAb in combination with ipilimumab (anti-CTLA4) (NCT01750580) is completed and a Phase I clinical trial of anti-KIR mAb in combination with anti-PD-1 is still recruiting patients (NCT01714739). IL-18 secretion by tumor cells upregulates PD-1 on NK cells (84). It has been shown that IL-18 secreted by tumor cells could elicits an expansion of NK cells overexpressing PD-L1 with immunoablative functions by reducing the number of mature NK cells and dendritic cells (DC) in a PD-L1-mediated manner, at least in the B16F10 melanoma model in mice (85). It has been suggested that the use of anti-IL-18 neutralizing antibodies in combination with anti-PD-1 mAb (nivolumimab) may bypass NK cell inhibition by PD-1 (22). Blocking several immune checkpoints can achieve synergistic anti-tumor effect with therapeutic benefits.

The clinical efficacy and pharmacological activity of anti-NKG2A mAb IPH2201 are going to be analyzed in clinical trials currently recruiting patients with squamous cell carcinoma of the oral cavity for an efficacy study of pre-operative use of IPH2201 (NCT02331875) or for a dose-ranging study of patients with high grade serious carcinoma of ovarian, fallopian tubes, or peritoneal origin (NCT02459301). The results of these trials may open new perspectives for melanoma treatment.


Increased tumor sensitivity to NK cells has been observed after treatment with proteasome inhibitors, doxorubicin or histone deacetylase inhibitors that upregulates the expression of NKG2D ligands, the secretion of proinflammatory cytokines, or the expression of TNF receptors. However, when combining these therapies with NK cell adoptive transfer, a strict control of NK cell function should be taken into account (22). In addition to the checkpoint blockade exerted by mAbs directed to receptors on cytotoxic cells or their ligands on tumors, mAbs may also act through ADCC or by redirected lysis of target cells.

#### Bispecific Killer Engagers

Novel strategies are in progress aimed to redirect NK cell cytotoxicity by CD16-directed bispecific and trispecific killer engagers (BiKEs and TriKEs respectively) constructed using one (BiKEs) or two (TriKEs) variable single-chain fragments against tumor-associated antigens. BiKEs and TriKEs trigger NK cell activation through CD16 (86). When combined with an inhibitor of ADAM17 to prevent CD16 shedding after NK cell activation, an enhancement of tumor cell lysis was observed (37, 87). The use of CD16-directed BiKEs has been limited so far to malignant hematological diseases.

## Adoptive NK Cell Therapy in Melanoma Patients

Optimal adoptive cancer immunotherapy should link both innate and adaptive immune responses. NK cells may contribute to the adaptive immune responses by favoring DC maturation and priming of T cells. The bidirectional crosstalk between NK cells and DC was demonstrated for the first time by Gerosa et al. in 2002 (88). NK cells activated by IL-2 or by mature DC directly induced DC maturation and enhanced DC ability to stimulate naïve CD4<sup>+</sup> T cells. These effects were cell contact dependent, and IFN-γ and TNF secreted by NK cells also contributed to DC maturation (88). The interaction of NK cells and DC in the tumor microenvironment has shown to play a pivotal role in the induction of tumor-specific immune responses. However, tumor-induced immunosuppressive environment can deregulate the interactions of NK cells with DC (89, 90). Co-culture of DCs and lymphokine-activated killer (LAK) cells resulted in NK cell activation associated with enhanced inflammatory cytokine production and lysis of melanoma cells. LAK cell-mediated induction of DCs maturation has a significant effect on priming of anti-tumor CTLs (91).

#### Autologous NK Cells

Lymphokine-activated killer cells were used for the first time in melanoma patients by Roserberg et al. (92) showing complete remission in one patient with metastatic melanoma that lasted at least 10 months after combined therapy (LAK and IL-2).

Clinical trials of adoptive NK cell-based immunotherapy against melanoma are very limited. A Phase II trial (NCT00328861) completed in 2009 combined autologous NK cells with intravenous (i.v.) IL-2 and chemotherapy. Although no clinical effect was observed, the transferred NK cells persisted in the peripheral blood from 14 weeks to several months suggesting that combined therapy with antibodies could be beneficial (93).

Another trial using autologous NK cells combined with the proteasome inhibitor bortezomib is ongoing (NCT00720785). The use of bortezomib has been related to the upregulation of NKG2D ligands on tumor cells that may promote NK cell recognition and lysis of tumor cells (22).

Because, the expression of activating receptors on NK cells from tumor-bearing patients is frequently found down-regulated, the efficacy of autologous NK cells expanded *in vitro* is limited by the activating receptor phenotype of expanded NK cells that should be taken into consideration.

#### Allogeneic NK Cells

Few clinical trials using allogeneic NK cells for melanoma treatment have been reported usually combined with chemotherapy. It has been shown that NK cell activation of activating receptors together with administration of anti-tumor antibodies have substantial anti-cancer effects supporting that the combination of allogeneic NK cells and antibody therapy can be an efficient strategy in clinical trials (94). A phase I trial using allogeneic NK cells in 10 metastatic melanoma patients showed successful engraftment of NK cells. Four melanoma patients demonstrated stable disease after the first cell infusion but the disease progressed few weeks after a second infusion of NK cells. In the same trial, 5 of 19 poor prognosis AML patients achieved complete remission after NK cell infusion showing best results when KIR ligand mismatched donors were used (95). The role of haploidentical NK cell transfer was analyzed in a clinical trial (NCT00846833) in patients with refractory or relapsed malignant melanoma. A recent study analyzed the adoptive transfer of mismatched lymphocytes activated *in vitro* with recombinant human IL-2 (NCT00855452) for the induction of graft-versus-tumor effect in metastatic solid tumors including melanoma. The results of these trials have not yet been published.

#### Adoptive Transfer of NK Cell Lines

The difficulties of expanding large numbers of clinical grade NK cells (96) together with the lower transduction efficacy of primary NK cells are major limiting factors for their clinical application compared to NK cell lines. Further developments of viral vectors such as the alpharetroviral platform are required to fully exploit NK cells in cancer immunotherapy (97). It has been postulated that the use of NK cell lines that can be easily expanded *in vitro* could facilitate the development and standardization of protocols for the use of NK cells in therapy. The human NK cell line NK-92 (98) represents an alternative to donor-derived peripheral NK cells since it can be maintained *in vitro* and expanded to large numbers under good manufacturing practice (GMP) conditions for immunotherapy (99). The NK-92 cell line was evaluated in a Phase I trial in one metastatic melanoma patient that showed a minor response (100). The toxicity was low and this cell line was approved by the FDA for the treatment of melanoma. The possibility of engineered NK cell lines to express chimeric receptors has been also considered (101).

#### Chimeric Antigen Receptor-Modified NK Cells

A strategy to redirect NK cell cytotoxicity against melanoma is the use of chimeric antigen receptor (CAR)-modified NK cells. CARs consist of an external domain that specifically recognizes a given tumor antigen, linked with one or more intracellular signaling domains that trigger cytotoxic cell activation. NK cell lines, peripheral blood NK cells, and NK cells derived from human pluripotent stem cells can be engineered to express CARs. These CAR-transduced NK cells can specifically recognize and kill a variety of tumor targets expressing the surface target antigen [for review in Ref. (101, 102)]. It has been shown that the CAR-transduced NK92 cell line, NK-92MI-GPA7-zeta can recognize the melanoma-associated gp100 peptide in the context of HLA-A2, showing redirected killing of melanoma cell lines and primary melanoma (103). These results support the use of CAR engineering to redirect the specificity of NK cells to augment their cytotoxicity against tumors including refractory melanoma cells.

# CONCLUSION

Stimulation of the immune system has been considered a possible therapy for melanoma for many years. Experimental and clinical efforts have focused in exploring possibilities to use different elements of the adaptive and innate immune responses to control and eliminate melanoma cells. However, the heterogeneity of these tumors makes necessary a detailed analysis of the possible interactions between the melanoma and the immune system cells. NK cells are undoubted components within the anti-melanoma immunotherapy arsenal. The potential efficacy of NK cell-based immunotherapy in melanoma patients will rely on melanoma phenotype (expression of ligands for activating receptors and low expression of MHC class I molecules for the use of autologous NK cells), NK cell status (no exhausted, no senescent), NK cell

# REFERENCES


phenotype (high level of NKG2D, NCRs and DNAM-1; CD16 expression for ADCC), microenvironment (proinflammatory versus inhibitory), NK cell crosstalk with other cell types (e.g., DCs, macrophages, MDSCs). The better understanding of the interactions between NK cells and melanoma will open the possibility to use combined strategies of checkpoints blockade and cytokine or activating receptor stimulation to enhance autologous NK cell cytotoxic capacity. These strategies should also be considered to modulate NK cell functionality in protocols of adoptive therapy against melanoma using autologous, allogeneic, or engineered *ex vivo-*expanded NK cells.

# AUTHOR CONTRIBUTIONS

RT and RS designed the manuscript. RT, ED, and RS contributed to the writting and revised the manuscript.

## ACKNOWLEDGMENTS

We apologize to our colleagues whose work was not cited due to space limitations. This work was supported by grants SAF2009- 09711 and SAF2013-46161-R (to RT) from the Ministry of Economy and Competitiveness of Spain, PS09/00723 and PI13/02691 (to RS) from Spanish Ministry of Health and CTS-208 from Junta de Andalucia (to RS) and grants to INPATT research group (GRU10104 and GR15183) and PRI09A029 from Junta de Extremadura and University of Extremadura (to RT and ED) cofinanced by European Regional Development Funds (FEDER). This work was also supported by contracts QLRT-2001- 00668 (Outcome and Impact of Specific Treatment in European Research on Melanoma, OISTER) and LABPOLE project from the Ministry of Economy and Competitiveness cofinanced by FEDER.

costimulatory receptors involved in cell-mediated cytotoxicity in a large panel of human melanoma cell lines. *Cancer Immunol Immunother* (2009) **58**:1517–26. doi:10.1007/s00262-009-0682-y


in human melanoma cell lines from ESTDAB cell bank. *Cancer Immunol Immunother* (2009) **58**:1507–15. doi:10.1007/s00262-009-0701-z


**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 Tarazona, Duran and Solana. 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.*