Metabolic Requirements of NK Cell Responses to Viral Infection

Washington University in St. Louis Washington University Open Scholarship

Arts & Sciences Electronic Theses and Dissertations Arts & Sciences

Spring 5-15-2020

Annelise Yoo Mah-Som Washington University in St. Louis

Follow this and additional works at: https://openscholarship.wustl.edu/art_sci_etds

Part of the Allergy and Immunology Commons, Immunology and Infectious Disease Commons, and the Medical Immunology Commons

Recommended Citation Mah-Som, Annelise Yoo, "Metabolic Requirements of NK Cell Responses to Viral Infection" (2020). Arts & Sciences Electronic Theses and Dissertations. 2216. https://openscholarship.wustl.edu/art_sci_etds/2216

This Dissertation is brought to you for free and open access by the Arts & Sciences at Washington University Open Scholarship. It has been accepted for inclusion in Arts & Sciences Electronic Theses and Dissertations by an authorized administrator of Washington University Open Scholarship. For more information, please contact [email protected]. WASHINGTON UNIVERSITY IN ST. LOUIS

Division of Biology and Biomedical Sciences Immunology

Dissertation Examination Committee: Megan Cooper, Chair Takeshi Egawa Todd Fehniger Anthony French Deborah Lenschow Wayne Yokoyama

Metabolic Requirements of NK Cell Responses to Viral Infection by Annelise Yoo Mah-Som

A dissertation presented to The Graduate School of Washington University in partial fulfillment of the requirements for the degree of Doctor of Philosophy

May 2020 St. Louis, Missouri

Table of Contents List of Figures ...... iv List of Abbreviations ...... v Acknowledgments ...... vi Abstract ...... viii Chapter 1: Introduction ...... 1 1.1 Natural Killer Cells and Their Functions ...... 3 1.2 Metabolic Regulation of Immune Cells ...... 7 1.2.1 Major Metabolic Pathways ...... 7 1.2.2 Metabolic Regulation in T cells ...... 11 1.2.3 Metabolic Regulation in Macrophages ...... 13 1.2.4 Metabolic Regulation in Dendritic Cells ...... 16 1.2.5 Metabolic Regulation in B cells ...... 17 1.2.6 Metabolic Regulation in Neutrophils ...... 19 1.3 Metabolic Regulation in NK cells ...... 20 1.4 Conclusions and Aims ...... 28 1.5 References ...... 31 Chapter 2: Glycolytic Requirement for NK Cell Cytotoxicity and Cytomegalovirus Control ..... 38 2.1 Abstract ...... 39 2.2 Introduction ...... 39 2.3 Results ...... 42 2.4 Discussion ...... 58 2.5 Materials and Methods ...... 63 2.6 Supplementary Figures ...... 69 2.7 References ...... 73 Chapter 3: Cox10 is Required for Expansion of Ly49H+ NK cells During Murine Cytomegalovirus Infection ...... 78 3.1 Abstract ...... 79 3.2 Introduction ...... 80 3.3 Results ...... 82 3.4 Discussion ...... 92

3.5 Materials and Methods ...... 96 3.6 Supplementary Figures ...... 101 3.7 References ...... 102 Chapter 4: Conclusions and Future Directions ...... 105 4.1 Glycolytic Metabolism in NK Cell Antiviral Function ...... 106 4.2 Oxidative Metabolism in NK Cell Antiviral Function ...... 108 4.3 NK Cells in the Immunometabolism Paradigm ...... 110 4.4 Future Directions ...... 112 4.5 References ...... 116

iii

List of Figures Figure 1.1: NK cell functions during MCMV infection...... 6 Figure 1.2: Major metabolic pathways and programs in immune cells...... 8 Figure 1.3: Metabolic regulators...... 9 Figure 1.4: Mitochondrial electron transport chain...... 10 Figure 1.5: Increased rates of metabolism with prolonged IL-15 exposure but not short-term activation...... 22 Figure. 1.6: Naive NK cells require OXPHOS for receptor-stimulated IFN-g production, but activation with IL-15 eliminates this requirement...... 23 Figure 2.1: Glucose metabolism blockade during activation decreases NK cell proliferation and cytotoxicity...... 43 Figure 2.2: In vitro treatment with 2DG decreases NK cell production of granzyme B and alters actin accumulation at the immunological synapse...... 45 Figure 2.3: 2DG treatment in vivo causes decreased target clearance during MCMV infection..47 Figure 2.4: 2DG treatment in vivo confers MCMV susceptibility...... 50 Figure 2.5: NK activation with ALT-803 rescues MCMV susceptibility caused by 2DG treatment...... 52 Figure 2.6: ALT-803 rescue of 2DG-treated mice requires NK cells, and ALT-803 priming can eliminate susceptibility to MCMV caused by mTOR inhibition...... 54 Figure 2.7: 2DG decreases human NK cell cytotoxicity...... 56 Figure 2.8: ALT-803 treatment of CMV reactivation in a post-hematopoietic cell transplant patient...... 58 S. Figure 2.1: Effects of 1mM 2DG culture on NK cell metabolism and killing...... 69 S. Figure 2.2: T cells upregulate Gzmb and proliferate, but are not affected by 2DG treatment..70 S. Figure 2.3: IFN-γ production by NK cells during MCMV infection is not inhibited by 2DG..71 S. Figure 2.4: MCMV susceptibility is dependent on the dose of 2DG and the inoculum of MCMV...... 71 S. Figure 2.5: The effects of ALT-803 are not mediated through CD8+ T cells...... 72 Figure 3.1: Cox10-deficient NK cells develop normally...... 83 Figure 3.2: Cox10-deficient NK cells have normal function and energy production at baseline..84 Figure 3.3: Cox10-deficiency leads to decreased expansion of the Ly49H+ population of NK cells during MCMV infection...... 87 Figure 3.4: Cox10 deficiency also causes NK cell proliferation defects with Ly49H stimulation in vitro...... 89 Figure 3.5: Increased MCMV susceptibility in Ncr1-Cox10D/D mice...... 90 Figure 3.6: Lack of proliferation of Cox10-deficient NK cells leads to few MCMV- adapted NK cells...... 91 S. Figure 3.1: Normal development of Cox10-deficient NK cells...... 101

List of Abbreviations

2DG – 2-deoxy-ᴅ-glucose LEM – lymphocyte expansion molecule 7AAD – 7-aminoactinomycin D Ly49H – Klra8, killer cell lectin-like Actb – beta actin receptor, subfamily A, member 8 AKT – AKT serine/threonine kinase LPS – lipopolysaccharide AMPK – AMP-activated protein kinase MCMV– murine cytomegalovirus AnnV – annexin V MFI – mean fluorescence intensity ANOVA – analysis of variance moDC – monocyte-derived DC ATP – adenosine triphosphate MTOC – microtubule organizing complex BMDC – bone marrow-derived DC mTOR – mammalian target of rapamycin CD – cluster of differentiation mTORC1 – mTOR complex I cDC – conventional dendritic cell NAD+/NADH – nicotinamide adenine CFSE – carboxyfluorescein succinimidyl dinucleotide (oxidized & reduced) ester NK – natural killer CoA – coenzyme A NKR – NK cell (activating) receptor CoQ – coenzyme Q NLRP3 – NOD-like receptor family pyrin COX – cytochrome c oxidase domain containing 3 Cox10 – COX subunit 10 NO – nitric oxide CSR – class switch recombination OCR – oxygen consumption rate DC – dendritic cell OXPHOS – oxidative phosphorylation ECAR – extracellular acidification rate PPARg – peroxisome proliferator-activated ER – endoplasmic reticulum receptor gamma ETC – electron transport chain PCG-1a – PPARg coactivator 1-alpha FAO – fatty acid oxidation PKCb – protein kinase C beta FAS – fatty acid synthesis PFU – plaque forming units FCCP – trifluoromethoxy carbonylcyanide PBMC – peripheral blood mononuclear cells phenylhydrazone PFKFB1 – phosphofructokinase-2/fructose- GAPDH – glyceraldehyde 3-phosphate 2,6-bisophosphatase 1 dehydrogenase PI3K – phosphoinositide 3-kinase GC – germinal center Poly(I:C) – polyinosinic:polycytidylic acid GSK-3 – glycogen synthase kinase 3 RISP – Rieske iron-sulfur protein (Complex GVHD – graft-versus-host disease III subunit) Gzmb – granzyme B ROS – reactive oxygen species HCMV – human cytomegalovirus SEM – standard error of measurement HCT – hematopoietic cell transplant TCA – tricarboxylic acid HIF-1a – hypoxia-inducible factor 1 alpha TEM – transmission electron microscopy HSV – herpes simplex virus Th1, Th2, Th17 – T helper type 1, 2, or 17 i.p. – intraperitoneal TLR – toll-like receptor Ie1 – intermediate early gene 1 Treg – regulatory T cell IFN-g – interferon gamma u-PFK2 – ubiquitous phosphofructokinase-2 IL – interleukin YFP – yellow fluorescent protein IS – immunological synapse

Acknowledgments

I could not have done the following work without the gracious help of many people. First and foremost, my PI, Megan Cooper. Thank you for reading 10 drafts of every manuscript, for your tolerance of my working hours and questionable work-life balance, for directing my scattered focus, and for trying to push me out of the nest when all I’d like to do is hear another of your brilliant experimental ideas that always work despite my initial doubts.

I am equally indebted to our technicians, Molly Keppel and Nermina Saucier. Thank you for putting up with my constant questions, for your moral support, for handing me mice and picking up the slack when I drop the ball, and for your super useful life advice. Without you two, I couldn’t have gotten even a fraction of this work done. I’d also like to thank our previous lab members:

Tiphanie Vogel, for being a role model as both a scientist and a clinician, and Emily Moore, for breathing life into our human memory NK cell experiments and being still inspired to pursue a career in research after my long-winded ramblings about metabolic regulation.

Thank you also to Tony French, my unofficial co-PI, and the French lab for all your help understanding and performing MCMV experiments, and for being available when I need help or reagents. A special thank you to Joshua Alinger for being a fellow MD/PhD student, talking about the science late at night, and bringing life to the 6th floor.

I also owe a significant debt to Emily Mace, the Mace and Orange labs, and Sandeep

Tripathy for bringing me to Houston to learn fluorescence microscopy and monopolize the scope time for 3 weeks.

A huge thank you to my parents, Michael Mah and Min Yoo, who from a young age inspired and challenged me academically, who suggested the MD/PhD pathway, and who are

always there to support me even when I’m bad at asking for help. Soon, I’ll be graduating for the last time (hopefully), so thank you for equipping me to launch into this new stage of adulthood.

And last but not least, a huge hug for my husband, Avik Som. Thank you for reading all of my drafts and listening to all my presentations, and dealing with the poor grace with which I accept criticism. Thank you for being there late at night, for dealing with me making fun of you for the number of mice you engineers consider adequate n, for asking me to talk about my research, and for believing in me as a scientist as well as a human being. I love you.

For the more formal acknowledgements:

For our JCI Insight manuscript, we thank B. Parikh, J. Gunesch, T. Kumar, C. Luke and G.

Silverman for experimental advice, and H. Gu and M. Wallendorf for statistical analysis, and T.

Vogel for helpful scientific discussion.

For our work with Cox10-deficient mice, we thank V. Sexl for the Cre mice, W. Beatty and the Microbiology microscopy core for help with TEM, and K. Toth and T. Harmon for helpful scientific discussion.

Work in the Cooper laboratory was supported by the NIH/NIAID (1K08AI085030 and

R01AI127752), the Rheumatology Research Foundation, the Children’s Discovery Institute and

St. Louis Children’s Hospital, and the American Association of Immunologists. A.Y.M. was supported by NIH training grant T32 GM07200 and NIAID fellowship 1F30AI129110.

Annelise Yoo Mah-Som

Washington University in St. Louis

May 2020

vii

ABSTRACT OF THE DISSERTATION

Metabolic Requirements of NK Cell Responses to Viral Infection

Annelise Yoo Mah-Som

Doctor of Philosophy in Biology and Biomedical Sciences

Immunology

Washington University in St. Louis, 2020

Associate Professor Megan A. Cooper, Chair

In recent years, the field of immunometabolism – the study of how specific changes in cellular metabolism regulate the function of diverse immune cell types—has grown exponentially. Several in vitro studies have examined the metabolic regulation of natural killer (NK) cells, which are first responders for viral infection and malignant transformation; however, much less is known regarding the role of metabolism in directing NK cell responses in vivo, such as during viral infection. In order to examine how NK cell antiviral function is regulated in vivo, we used a well- characterized infection with murine cytomegalovirus (MCMV) to assess NK cell cytokine production, proliferation, and killing. Two treatments were used to manipulate major metabolic pathways: a drug model to examine the role of glycolytic metabolism, and a genetic model to examine the role of oxidative metabolism. We found that treatment with the glucose metabolism inhibitor 2-deoxy-ᴅ-glucose (2DG) impaired both mouse and human NK cell cytotoxicity following priming in vitro. Similarly, MCMV-infected mice treated with 2DG had impaired clearance of NK-specific targets in vivo, which was associated with higher viral burdens and increased susceptibility to infection on the C57BL/6 background. These effects could be reversed

viii

by pre-treatment with IL-15, which alters NK cell metabolism and metabolic requirements for activation. To interrogate the role of oxidative metabolism on NK cell function, we examined responses to MCMV in mice with an NK-specific deletion of Cox10, a nuclear gene encoding a protein required for functional assembly of mitochondrial electron transport chain Complex IV.

Cox10-deficient NK cells developed normally, retaining normal function, energy production, and oxygen consumption at baseline. However, they had a defect in proliferation when activated through the Ly49H receptor in vitro and in vivo. Ly49H recognizes an MCMV-encoded protein and mediates preferential expansion of the Ly49H+ NK cell population. The expansion defect in

Cox10-deficient NK cells was associated with decreased mitochondrial production of reactive oxygen species and increased susceptibility to MCMV infection. Thus, we found that uninhibited glucose metabolism is required by NK cells for maximal cytotoxic function, and that the oxidative pathway gene Cox10 is required by NK cells to proliferate optimally in response to MCMV infection. Our findings emphasize that metabolic pathways have specific interactions with different immune functions, and further develop methods to investigate metabolic regulation of immune cells in physiological contexts.

Chapter 1

Introduction

Part of this chapter is adapted from a manuscript published in Critical Reviews in Immunology:

Mah AY, Cooper MA. Metabolic Regulation of Natural Killer Cell IFN-g Production. Crit Rev Immunol. 2016:36(2):131-147.

In the last decade, there has been a resurgence of interest amongst immunologists in the key role that metabolism plays in regulating immune cell function. In 1958, Otto Warburg described but dismissed his observation that proliferating leukocytes use aerobic glycolysis rather than the more energy-efficient process of oxidative phosphorylation, thinking his findings were experimental artifact. Since then, immunologists have discovered that this “metabolic reprogramming” of activated leukocytes toward aerobic glycolysis is critical for immune cell function, and have delved into the complex but critical interplay between the immune system and metabolic pathways. The study of “immunometabolism” aims to untangle how activation and other canonical immunological signals alter cell metabolism, and how metabolic intermediates and pathways are able to regulate immune cell effector function in turn.

We are primarily interested in natural killer (NK) cells, which are cytotoxic cells of the innate immune response. The immune regulation of these cells is especially interesting because they have characteristics of both the innate and adaptive immune systems. They are functionally and developmentally similar to CD8+ T cells, but can act in a shorter time span, meaning that their metabolic demands are intrinsically different from T cells. While our understanding of T cells is one of the most advanced in the immunometabolism field, few laboratories are actively engaged in studying NK cell metabolic regulation. Given their key function in early antiviral and antitumor immunity, as well as growing investigation into their therapeutic use, a greater understanding of

NK cell regulation through their metabolism is needed to be able to manipulate NK cell function.

Also, given the difficulties in manipulating metabolism in vivo, more methods need to be developed in order to interrogate NK cell function in physiological contexts such as during viral infection. Thus, we set out to investigate the role of two core metabolic pathways—glycolysis and oxidative phosphorylation—in regulating NK cell function, especially antiviral function in vivo.

In this dissertation, I will introduce NK cells and their major functions, how other immune cells are regulated by metabolism, and what we know so far of how NK cells are metabolically regulated. Chapter 2 presents our findings using the drug 2-deoxy-D-glucose to investigate the role of glycolysis on NK cell antiviral function in vitro and in vivo, with an emphasis on its role in priming cells for cytotoxic function. Chapter 3 presents our ongoing work using a genetic model of Cox10 deletion to examine how NK cell response to viral infection, and formation of NK cell

“memory” is reliant on mitochondrial oxidative metabolism. Finally, I will conclude by discussing the implications of our findings for the field of immunometabolism and suggesting future directions for related research.

1.1 Natural Killer Cells and Their Functions

Natural killer (NK) cells are lymphocytes that comprise approximately 10-15% of human peripheral blood mononuclear cells and 2-3% of splenocytes in mice. They are early responders to viral infection and malignant transformation, and are critical for the response to these threats before

T cells become active. They share their role as cytotoxic and interferon gamma (IFN-γ)-producing cells with CD8+ T cells. However, unlike T cells which mount an antigen-specific response a week after priming, NK cells are antigen-independent, recogne infected and transformed cells through germline-encoded NK receptors (NKRs), and kill them within minutes to hours (1-4). NK cells can be activated by three complementary and often overlapping routes: 1) engagement of NKRs that recognize stress- and virally-induced ligands on target cells in the absence of a strong inhibitory receptor signal, 2) recognition of antibody-coated target cells through the activating Fc receptor CD16 (antibody-dependent cellular cytotoxicity), and/or 3) activation by innate and adaptive immune cytokines including IL-1, IL-2, IL-12, IL-15, and IL-18 (3, 5). NK cell activation

can lead to three major responses: production of cytokines, killing of infected cells, and proliferation. In this section, we emphasize NK cell regulation in response to infection with murine cytomegalovirus (MCMV), which we use to interrogate NK cell function in vivo (Fig. 1.1).

Regulation of NK cell cytokine production

NK cells produce several cytokines and chemokines that coordinate innate and adaptive immune responses, and are the major source of the immunostimulatory and immunomodulatory cytokine IFN-γ in the first hours of infection (2). Experimentally, NK cell IFN-γ production can be induced by brief incubation (4-24h) with purified cytokines or cytokine-producing antigen presenting cells, co-culture with tumor cells expressing activating ligands, or culture with plate- or bead-bound antibodies against activating NKRs. The Ifng locus is open and available for transcription factor binding in NK cells, allowing for constitutive transcription as well as rapid upregulation of expression upon activation (6-8). NK cell activation leads to translation of the pre- existing pool of Ifng mRNA, the translation of which is regulated by binding of microRNAs, long noncoding RNAs, and regulatory proteins to the untranslated region (9). During infection with

MCMV, IFN-g production is stimulated by the systemic release of inflammatory cytokines such as IL-12 and peaks around 36-48 hours post-infection (Fig. 1.1)(10, 11).

Regulation of NK cell killing

NK cell killing is a highly-regulated process, reviewed in (12). The first step is forming an

“immunological synapse” with the target cell. Target recognition is regulated by the balance between signaling through activating NKRs recognizing stress-induced or pathogen-encoded ligands, and inhibitory receptors recognizing self-MHC. Downregulation of self-MHC due to malignancy or viral infection makes cells more susceptible to killing by NK cells. Next, NK cells must polarize and traffic their cytotoxic granules toward the target, which requires coordination of

the microtubule and actin networks of the cell. NK cell granules contain cytotoxic proteins such as perforin and granzymes, which can enter target cells and cause apoptosis; components of these granules are upregulated by stimuli such as IL-15 to boost NK cell cytotoxicity (13). Finally, granules are released onto the target, and the NK cell detaches from the dying cell to seek new targets. Disruptions to the signaling and/or cytoskeletal movements required for directed target cell killing effectively inhibits NK cell killing (14). During MCMV infection, one ligand that induces killing of infected cells is the MCMV-encoded protein m157, which is recognized by the

NKR Ly49H (15). This receptor is present in mice of the C57BL/6 genotype and allows NK cells to specifically recognize MCMV-infected cells, enhancing their resistance to the virus (15). Killing of infected cells occurs early during infection, but cytotoxic function is boosted as infection continues by increasing levels of cytotoxic proteins like granzyme B (Gzmb) (Fig. 1.1)(13).

Regulation of NK cell proliferation

NK cell proliferation can occur through both NKR and cytokine stimulation. In vitro NK cell expansion was first described in response to high concentrations of IL-2, creating highly activated “lymphokine-activated killer cells” (16). IL-15, which is similar to IL-2 but more NK- specific, also causes proliferation and activation at high doses in vitro and in vivo (17, 18).

Additionally, exposure to the inflammatory cytokines IL-12 and IL-18 causes subsequent proliferation by NK cells after adoptive transfer into mice (19). The extent of NK cell proliferation can be controlled by the concentration of cytokine.

Activating receptor-driven proliferation has been described in vitro with stimulation through ligation of NKRs such as NK1.1 and Ly49H, in conjunction with low levels of supportive cytokines (20, 21). In vivo proliferation driven by Ly49H is well-studied; although inflammatory cytokines induce low-level proliferation during MCMV infection, Ly49H+ NK cells preferentially

proliferate at higher rates than Ly49H- NK cells in response to exposure to m157 (Fig. 1.1) (11).

Finally, NK cells have been described to expand homeostatically in lymphopenic mice, although the exact mechanism for this expansion is not completely understood (22, 23).

Activation Antiviral functions Inflammatory cytokines (IL-12, IL-18) IFN-g production

NK DC cell MCMV 36-48 hours (intraperitoneal infection) post-infection

Viral molecular Killing patterns sensed of infected cells

m157 MCMV- Ly49H+ MCMV- m157 Host cell infected NK cell infected Ly49H

Continuous Recognition of Cytotoxic viral antigen granules Proliferation m157 of Ly49H+ population MCMV- Ly49H+ infected NK cell Ly49H

Ly49H- Ly49H- NK cell NK cell

4-7 days post-infection

Time

Figure 1.1. NK cell functions during MCMV infection. After intraperitoneal injection of MCMV, infected cells express the MCMV-encoded protein m157. Dendritic cells (DCs) sense pathogen-associated molecular patterns and produce systemic inflammatory cytokines such as IL-12 and IL-18, which non-specifically activate NK cells to produce IFN-g at 36-48 hours post-infection. The m157 ligand is recognized by a subset of NK cells that express the Ly49H activating receptor, leading to preferential killing of infected cells by Ly49H+ NK cells throughout the course of infection, and preferential expansion of the Ly49H+ NK cell population between 4-7 days post-infection.

In summary, NK cells are critical for the initial response to viral infection and malignant transformation. They can be stimulated by activating receptors or by inflammatory cytokines, and in turn produce cytokines and kill non-self or compromised cells. Certain types of stimulation can also lead to proliferation of mature NK cells. We aim to understand how known forms of regulation

interact with and are regulated by metabolism. MCMV infection induces all of these responses and provides a well-characterized, clinically-relevant system to study metabolism.

1.2 Metabolic Regulation of Immune Cells

The field of immunometabolism examines how changes in metabolic pathway usage regulate and are regulated by immune cell activation and function. Significant progress has been made in elucidating the importance of metabolism for the development, differentiation, activation, effector function, and/or memory formation of murine B & T lymphocytes, macrophages, dendritic cells (DCs), and even neutrophils. Here, we will review the regulation of non-NK immune cell types by glycolysis and oxidative phosphorylation in order to describe possible mechanisms by which NK cells might be regulated by metabolism.

1.2.1 Major Metabolic Pathways

The two main energy-generating pathways in a cell, glycolysis and oxidative phosphorylation (OXPHOS), are often used in concert in resting cells, but can be regulated independently to fuel specific functions (Fig. 1.2). The main fuel for most cells is glucose, which is processed by glycolysis into pyruvate, high-energy electrons carried by NADH, and 2 molecules of ATP. The rate of glycolysis is regulated by diverse signals including the cell’s energy balance, the abundance of downstream intermediates, and cell signaling events (24). Several metabolic regulators, including mTOR, HIF-1α, and c-myc (Fig. 1.3), increase the rate of glycolysis through upregulation of glycolytic enzymes and/or glucose transporters (25). Aerobic glycolysis, or fermentation of pyruvate into lactate in the presence of oxygen, is a distinguishing feature of proliferating cells first described in cancer cells and later in leukocytes (26, 27). In these cells, 7

fermentation is required to restore the pool of NAD+ from NADH so glycolysis can continue.

Glucose can be diverted from glycolysis into the pentose phosphate pathway for regeneration of the reducing agent NADPH and nucleotide synthesis (28), or glycosylation of proteins for their correct folding, function, and expression (29).

Figure 1.2: Major metabolic pathways and programs in immune cells. (Left) Naïve immune cells use glucose to fuel glycolysis, producing ATP and pyruvate. Pyruvate is converted into acetyl-CoA, enters into the mitochondria, and is processed by the TCA cycle into high-energy electrons. Carriers bring these electrons to the electron transport chain for OXPHOS, which requires oxygen to synthesize large amounts of ATP. (Center) In activated effector cells, e.g. Th1 and inflammatory macrophages, glycolysis is highly upregulated without concomitant increase in OXPHOS, leading to the production of lactate. Metabolites are shunted toward nucleotide synthesis in T cells, ROS/nitric oxide in macrophages, fatty acid synthesis in DCs, and antibody production in plasma cells. Depleted TCA intermediates can be replaced by glutamine. OXPHOS may function at a low level in activated effector cells. (Right) Tolerogenic macrophages, regulatory T cells, and memory T cells downregulate glycolysis, instead oxidizing lipids to fuel the TCA cycle. Increased mitochondrial mass leads to greater respiratory capacity and increased OXPHOS.

Pyruvate from glycolysis is converted into acetyl-CoA to enter into the tricarboxylic acid

(TCA) cycle (Fig. 1.2). The TCA cycle occurs in the mitochondria and processes acetyl-CoA into

CO2 and high-energy electrons. It contains multiple entry and exit points into other metabolic pathways, and can be fueled by glycolysis or fatty acid oxidation (FAO). TCA intermediates such as citrate can be used for several biosynthetic reactions, including fatty acid and amino acid synthesis, and can be replaced using glutamine catabolism (30). Shunt of TCA intermediates

toward biosynthesis and away from OXPHOS is critical in fueling various immune cell functions.

Acetyl-CoA is required for acetylation, which can also regulate metabolism directly by promoting degradation of metabolic enzymes, as well as indirectly affect the transcription of many genes through acetylation of histones, which allows greater chromatin accessibility (29).

Metabolic Name Function regulator Pro-glycolytic Kinase; increased protein synthesis, upregulation of Mammalian target of mTOR glycolytic enzymes, support of cell growth and rapamycin proliferation, cytoskeletal motility Hypoxia-inducible factor 1 Transcription factor; stabilization leads to increase in HIF-1a alpha glycolytic enzymes toward fermentation Transcription factor; upregulates protein synthesis, c-myc increases proliferation Bifunctional enzyme, regulates the rate of glycolysis. u- Phosphofructokinase- uPFK2/PFKFB1 PFK2 encourages a high rate of glycolysis, while 2/fructose-2,6-biphosphatase PFKFB1 leads to slower rates. Pro-oxidative Kinase; senses energy level by AMP/ATP ratio and AMPK AMP-activated protein kinase stimulates FAO Peroxisome proliferator- Nuclear lipid receptor; adipocyte differentiation, PPARg activated receptor gamma increases fatty acid storage and insulin sensitivity Peroxisome proliferator- Transcriptional coactivator; stimulates mitochondrial PCG-1a activated receptor gamma biogenesis coactivator 1 alpha Figure 1.3. Metabolic regulators. A selected list of metabolic regulators important in upregulating glycolysis or oxidative metabolism and a brief description of their function.

Conversion of the high-energy electrons produced by glycolysis, FAO, and the TCA cycle into energy in the form of ATP occurs during OXPHOS, which takes place in the mitochondria

(Fig. 1.2). A series of enzyme complexes in the inner mitochondrial membrane, the mitochondrial electron transport chain (ETC)(Fig. 1.4), converts these electrons into a H+ gradient that fuels ATP synthesis (31). Complexes I and II accept electrons from carriers such as NADH and transport them through Complex III to Complex IV, which fixes them to O2 to produce H2O. The energy stored in these electrons is used by Complexes I, III, and IV to transport H+ ions across the inner mitochondrial membrane to form a chemiosmotic gradient, that is then drained by Complex IV 9

(ATP synthetase) for energy to produce ATP. OXPHOS of electrons from processing one molecule of glucose produces around 30 ATP molecules, making it much more efficient than anaerobic glycolysis. Reactive oxygen species (ROS) are also generated, especially by Complexes I and III, and can cause phosphorylation-like post-translational modifications that modify signaling cascades in the cell (32-34). Excess ROS induce damage and the oxidative stress response. Major upregulators of oxidative function include AMPK, PPARg, and PGC-1a (Fig. 1.3)(35, 36).

Figure 1.4. Mitochondrial electron transport chain. High energy electrons are carried by NADH to Complex I, and are passed through Complex II, coenzyme Q (CoQ), Complex III, and cytochrome c, before they are fixed to oxygen in Complex IV. Transfer of electron generates reactive oxygen species (ROS) from Complexes I and III, and powers hydrogen ion transport. Transport of hydrogen causes a gradient across the inner mitochondrial membrane that is converted to energy in the form of ATP by Complex V.

In summary, the two energy-generating pathways—glycolysis and oxidative phosphorylation, linked by the TCA cycle—are the center of cellular metabolism. It has been described in multiple immune cell types that glycolytic and oxidative metabolism are counter- regulated to produce inflammatory versus tolerogenic/memory cell types, respectively (Fig. 1.2).

However, the reasons for the preference for one pathway over another, and the exceptions to this generalization, differ between cell types. Findings in these diverse immune cells provide a basis for understanding how NK cells might be regulated by metabolism.

1.2.2 Metabolic Regulation in T cells

T cells, especially CD8+ T cells, are developmentally and functionally related to NK cells.

Many of the same regulatory mechanisms exist in both cell types, although NK cells activate much more quickly than T cells and form less durable memory, which creates a different set of metabolic requirements for function. Given those caveats, T cell immunometabolism provides a basis for our understanding of NK cell metabolism.

T cell activation induces aerobic glycolysis and releases enzymatic repression of IFN-γ

Naïve murine T cells are relatively metabolically quiescent, and primarily rely upon glucose or fatty acids to fuel energy production through OXPHOS (25). T cells upregulate aerobic glycolysis following activation through their T cell receptor (CD3) and CD28 (37-39). Although aerobic glycolysis is not strictly required for T cell activation (40), it is thought to assist in the rapid proliferation of T cells by shunting glucose toward production of biosynthetic precursors needed for proliferation, including nucleotide precursors, amino acids, and fatty acids for lipid membranes (41, 42). Activating signals such as IL-2 promote the switch to glycolysis by upregulating mTOR, c-myc, and other metabolic regulators (25). Increased glycolytic flux also encourages differentiation into inflammatory effector cells, i.e. Th1, Th2, or Th17 cells, while blockade of glycolysis or mTOR promotes regulatory T cell (Treg) differentiation (25).

It was recently shown that the glycolytic enzyme GAPDH directly affects IFN-γ expression in murine CD8+ T cells by binding to and preventing translation of IFN-γ transcript (40). Increased glycolysis after activation sequesters GAPDH, allowing for efficient translation of IFN-γ. Thus, increasing the rate of glycolysis not only enhances the ability of activated T cells to proliferate, but also directly controls production of the IFN-γ protein.

OXPHOS is required for T cell signaling, and is enhanced in regulatory and memory T cells

OXPHOS is important for T cell activation via generation of ROS. TCR signaling induces mitochondrial and cytoplasmic ROS (32, 33), which can activate several signaling pathways (34).

Loss of the Complex III subunit RISP, and the associated decrease in ROS production, inhibits proliferation of murine CD8+ cells (32). On the other hand, overexpression of the ETC assembly factor LEM increases ROS and enhances effector functions, including cytokine production and proliferation (43).

The ability to upregulate OXPHOS and decrease glycolysis is also important for the generation of Tregs and memory T cells (25, 44). In fact, inhibition of the pro-glycolytic regulator mTOR in mice causes differentiation of Tregs in the absence of Treg-polarizing cytokines (45).

However, mTOR-driven lipid biosynthesis is required for some aspects of normal Treg function, such as expression of immune checkpoint blockade receptors (46). Thus, generation of regulatory

T cells involves downregulation of glycolytic pathways that would lead to more inflammatory phenotypes, but Treg function requires balanced usage of different metabolic pathways.

Peripherally-generated regulatory T cells are also responsive to metabolites produced from the diet and microbiota as part of their role in maintaining oral tolerance (44), meaning their regulation by metabolism is especially complex.

Memory CD4+ and CD8+ T cells have greater mitochondrial mass and are highly oxidative, which is thought to support longevity and enhanced function during reactivation (47-49).

Disruption of FAO in murine T cells decreases the number of memory CD8+ T cells, their survival, and their ability to produce IFN-γ upon re-stimulation (50), while suppression of mTOR and glycolytic metabolism encourages development of memory (51). Increased expression of LEM and the associated increase in oxidative function also enhances production of memory T cells (43).

This is not to say that memory T cells only rely on oxidative metabolism—re-activation of memory

CD8+ T cells is associated with elevated glycolytic rates compared to naïve T cells (52).

In summary, work in T cells has shown that T cell activation induces aerobic glycolysis and accessory pathways leading to the production of macromolecules for proliferation. This increase in glycolysis also releases GAPDH-mediated suppression of IFN-g production, providing a fascinating example of direct regulation of immune function by metabolic enzymes. On the other hand, generation of regulatory and memory T cells is promoted by oxidative metabolism, and the enhanced lifespan and increased reactivity of memory T cells is linked to an increased number of mitochondria. Whether NK cell IFN-g production is regulated by GAPDH, and whether changes similar to those in T cells fuel MCMV-adapted “memory” NK cells, are currently unknown or under investigation.

1.2.3 Metabolic Regulation in Macrophages

Macrophages are non-proliferative phagocytes. Their metabolism is relatively well- studied, and provides a model for regulation of innate immunity by metabolism. Macrophages and

T cells both adhere to the core paradigm in immunometabolism: that inflammation is associated with aerobic glycolysis and tolerance is associated with oxidative metabolism.

Inflammatory macrophages use aerobic glycolysis to produce NO and other defense molecules

Initial differentiation of circulating monocytes into macrophages is correlated with becoming more oxidative and less glycolytic, and macrophages remain oxidative at baseline (53,

54). Similar to T cells, proinflammatory (classically-activated or M1) macrophages switch to aerobic glycolysis upon activation with LPS (53, 55). This metabolic reprogramming is mediated by increasing pro-glycolytic regulators such as mTOR and u-PFK2 (“fast glycolysis” enzyme),

decreasing signaling of the energy sensor AMPK, and activating shunt pathways (Fig. 1.3). Rather than use TCA intermediates for production of ATP through OXPHOS or to build biomass for proliferation, as macrophages do not proliferate, LPS-activated macrophages use TCA intermediates to produce pathogen defense molecules (56, 57). Specifically, citrate is used to produce NADPH and arachidonic acid, which are converted into nitric oxide (NO), ROS, and prostaglandins (56). NO directly inhibits the ETC Complexes I and IV to further decrease the rate of OXPHOS (58). Additionally, secretion of the inflammatory cytokine IL-1b is tied to the rate of glycolysis on two levels. First, production of the metabolite itaconate by activated macrophages inhibits Complex II and increases the level of succinate (59, 60). Succinate then stabilizes HIF-

1a, which increases the rate of glycolysis, increases mitochondrial ROS production, and enhances pro-IL-1b transcription (61, 62). Second, splicing of pro-IL-1b protein into its active form by the

NLRP3 inflammasome requires glycolysis in an unclear mechanism (63).

After activation, macrophages appear to shift back to oxidative metabolism, possibly as a mechanism to limit inflammation. Work in the human monocyte cell line THP-1 has shown that, after activation with LPS, THP-1 cells are highly glycolytic during early responses, then return to primarily using oxidative metabolism. This was linked to changes in the NAD+/NADH redox balance governed by activity of sirtuins that sense NAD+ levels and alter HIF-1a and PCG-1a activity to control glycolytic and fatty acid metabolism, respectively (64). Whether this return to normal metabolic state after activation occurs in monocytes and macrophages in vivo is not known.

Tolerogenic macrophages maintain oxidative metabolism

While inflammatory macrophages have “broken” their TCA cycles in order produce reactive oxygen and nitrogen species, tolerogenic (alternatively-activated or M2) macrophages produced by exposure to IL-4 and IL-13 have intact TCA cycles and higher rates of OXPHOS than

LPS-activated macrophages (53, 65). Compared to inflammatory macrophages, opposite regulatory pathways are induced in tolerogenic macrophages: AMPK rather than mTOR, HIF-2a rather than HIF-1a, PFKFB1 (“slow glycolysis” enzyme) rather than u-PFK2, and Arginase-1 rather than nitric oxide synthase (both compete for arginine) (65, 66). This is reflected functionally in decreased production of IL-1b and NO, and increased production of the tolerogenic cytokine

IL-10. It appears that tolerogenic macrophages, like regulatory T cells, have significantly downregulated their glycolytic rate and rely primarily on oxidative metabolism.

“Trained immunity” is associated with aerobic glycolysis

While innate immunity is not typically considered to have memory, several memory-like features have been described that boost function in innate cells upon re-exposure to a pathogen.

For example, human monocytes exposed to b-glucan for 24 hours, then rested for seven days, have greatly increased cytokine production when exposed to various pathogenic ligands. “Trained immunity” is associated with downregulation of mitochondrial function and increased levels of aerobic glycolysis (67), and is linked to mTOR and HIF-1a signaling as well as epigenetic changes. This is very different from T cell memory, which is associated with an oxidative state that likely allows T cell memory to persist for much longer than trained immunity.

In summary, LPS induces inflammatory macrophages which undergo aerobic glycolysis in order to use TCA intermediates for the production of reactive oxygen and nitrogen species, which in turn inhibit OXPHOS. On the other hand, IL-4 induces tolerogenic macrophages that maintain oxidative function. Unlike T cell memory, macrophage “trained immunity” is associated with a glycolytic phenotype and lower mitochondrial function. It is not known whether NK cell activation and “memory” are more similar metabolically to macrophage rapid activation and “trained immunity,” or to T cell activation (linked to proliferation) and classical memory.

1.2.4 Metabolic Regulation in Dendritic Cells

Dendritic cells (DCs) present antigen to T cells and are major producers of inflammatory cytokines, many of which NK cells respond to. As NK cells are a major producer of the cytokine

IFN-g early in viral infection, it is possible that some of the metabolic changes required to generate cytokines in DCs are present in more subtle forms in NK cells.

Increased glycolytic flux fuels fatty acid synthesis in activated dendritic cells

In general, DCs use fatty acids to fuel OXPHOS at rest, but increase glucose consumption and glycolytic rates within minutes of activation through TLRs (68-70). Increased glycolytic flux is thought to aid rapid cytokine synthesis by increasing citrate levels for fatty acid synthesis (FAS), which fuels expansion of the endoplasmic reticulum (ER) and Golgi complex membranes (53, 71).

Studies have also shown that proper antigen presentation requires appropriate control of lipid metabolism and autophagy, though lipid accumulation by DCs can be activating or suppressing depending on the context (35, 72). It is thought that the initial increase in glycolysis during activation is dependent on an unusual pathway of AKT signaling, and only later are more conventional mTOR and HIF-1a pathways engaged to maintain elevated glycolysis (35). mTOR signaling has also shown to be required for production of several immunogenic as well as tolerogenic cytokines in DCs. Functionally, metabolic changes divide DC activation into early and late states, and different types of DCs have different requirements.

Metabolism of activated DCs varies depending on source and function

DCs can differentiate from several types of precursors, with a range of ability to activate the immune system (immunogenicity). After prolonged activation (12h), murine bone marrow- derived DCs (BMDCs) rely exclusively on aerobic glycolysis, as BMDC production of NO inhibits

OXPHOS (73). Although less dramatically than BMDCs, murine conventional DCs (cDCs) also

upregulate glycolysis and decrease OXPHOS upon activation (70), though they do not produce

NO and so maintain some OXPHOS function. On the other hand, human monocyte-derived DCs

(moDCs) have increased mitochondrial and oxidative function through increased expression of

PPARg and PGC-1a (35, 74). The findings in cytokine-secreting plasmacytoid dendritic cells

(pDCs) are mixed. Human pDCs stimulated with toll-like receptor 7 (TLR7) agonists and RNA viruses upregulate glycolysis during activation (75), while pDCs stimulated through TLR9 primarily use FAO and OXPHOS for function (76).

DC immunogenicity can be downregulated by metabolites acquired through the diet or microbiota, similar to sensing by regulatory T cells. Butyrate, oxysterols, ATP, and lactate all decrease DC immunogenicity via decreased antigen presentation, decreased production of inflammatory cytokines, and increased production of anti-inflammatory cytokines like IL-10 (35).

In this way, DCs tune the immune response based on the extracellular metabolic environment.

In summary, DCs upregulate glycolysis after TLR activation to fuel fatty acid synthesis.

Fatty acids are used to expand organelles used for secretion of cytokines, and affect regulation of antigen presentation. Different types of DCs use a range of metabolic strategies, from NO- producing DCs that rely completely on aerobic glycolysis to more oxidative moDCs. Whether or not NK cell cytokine synthesis, especially in activated NK cells that are primed for increased cytokine production, is associated with expansion of the secretory organelles is unknown.

1.2.5 Metabolic Regulation in B cells

While B cells are similar in many ways to T cells, antibody-producing plasma cells are a unique case. The exaggerated phenotype of plasma cells displays some of the metabolic adaptations required for massive protein production, which may be present in less extreme form in NK cells as they produce granule proteins and cytokines.

B cell activation upregulates both glycolysis and OXPHOS

Naïve B cells are metabolically quiescent, which is maintained by GSK3 signaling (77).

Their activation through the B cell receptor is accompanied by increases in both glycolysis and

OXPHOS, with less aerobic glycolysis than in T cells (78-80). Upregulation of glycolysis has been linked to signaling through protein kinase C beta (PKCb) (81) and mTOR. Activated B cells can become short-lived plasma cells that secrete antibody immediately, or migrate to the germinal center (GC) to undergo class switch recombination (CSR) and somatic hypermutation, eventually forming memory B cells or long-lived plasma cells. The GC is a resource-limited space in which

B cells rapidly proliferate, providing unique metabolic challenges. Low oxygen availability may combine with c-myc and HIF-1a activation to drive aerobic glycolysis and increase fuel uptake for proliferation (80). However, GC B cells are also noted to have higher mitochondrial mass than naïve B cells (77). B cell fate upon exit from the GC has been correlated with the amount of mitochondria and ROS production, as well as the amount of mTOR signaling, with memory cells generated from the more mitochondrially-active subset (82).

Plasma cells fuel antibody production using fatty acids for secretory organelles and glucose for glycolysation

An extraordinary metabolic transformation occurs when B cells become plasma cells, which secrete thousands of antibodies per second. Transformation from an activated B cell into a plasma cell requires growing the ER and Golgi membranes to support antibody production, which requires a switch to fatty acid synthesis pathways as in DCs (83). Interestingly, OXPHOS appears to be maintained, suggesting that citrate (the major fuel for FAS) is replenished by glycolysis and glutaminolysis (84). There is some evidence that plasma cells in the gut also take up fatty acids from the environment, which may assist in maintaining OXPHOS or in expanding membranes

(85). Mature plasma cells have also greatly increased their uptake of amino acids and signaling through mTOR to support protein synthesis. Plasma cells have an unusual requirement for glucose, some of which fuels energy production but most of which appears to fuel antibody glycosylation

(86). Increased mitochondrial function and the ability to use glucose to fuel OXPHOS is associated with the increased longevity of long-lived plasma cells compared to short-lived plasma cells (86).

In summary, B cells upregulate both glycolysis and OXPHOS after activation, but becoming an antibody-secreting plasma cell requires a completely different metabolic profile dedicated to supporting antibody production. Like plasma cells, NK cells may require glucose metabolism for glycosylation of proteins, although at lower levels than in plasma cells.

1.2.6 Metabolic Regulation in Neutrophils

While neutrophils share few similarities with NK cells, they are included here to provide an example of a extreme metabolic phenotype. Neutrophils are the first and primary defense against extracellular pathogens, which they kill with a respiratory burst of damaging reactive oxygen and nitrogen species. Neutrophils are extremely short-lived cells with relatively few mitochondria with low function, and thus mainly use aerobic glycolysis for their function (87).

Mitochondria do not participate in neutrophil respiratory burst or phagocytosis, but are likely involved in neutrophil trafficking and apoptosis (88, 89). The requirement for aerobic glycolysis in neutrophils is unclear, but may be related to the need to function in hypoxic environments or specialization of the mitochondria to induce apoptosis. Neutrophils are able to cast neutrophil extracellular traps (NETs) of their DNA into the extracellular space, and this process requires glycolysis, glutamine, and cytoplasmic (not mitochondrial) ROS (90, 91). In summary, preliminary work on neutrophils has found that they almost exclusively use aerobic glycolysis for

their survival and function. One hypothesis is that reliance on aerobic glycolysis decreases longevity due to the metabolic strain it places on the cell; this is likely the case in neutrophils, and may be relevant in NK cells.

1.3 Metabolic Regulation in NK cells

Compared to what is known in T cells, macrophages, and DCs, less is known about metabolic regulation in NK cells. While they are most developmentally and functionally similar to CD8+ T cells and thus might be expected to be similarly metabolically regulated, they also share some features of the myeloid cells of the innate immune system in their quick reaction time from activation to effector function. We are particularly interested in the role of glycolytic and oxidative metabolism in regulating NK cell function.

NK cell development requires mTOR signaling

Several recent in vitro and in vivo studies have found that IL-15 signals through

PI3KàAKTàmTOR to allow for murine NK cell development and survival, as well as for priming NK cells for enhanced activity (92-94). mTOR activity is required for proliferation of immature NK cells in wildtype mice, so mice with NK-specific deletion of Mtor have very few circulating NK cells. Any surviving NK cells are immature, with decreased proliferative capability and slower/lower production of Gzmb (92). While the role of mTOR signaling in NK development has been described, it is not well understood which of these changes are due to mTOR’s alteration of the rate of glycolysis compared to its other effects on cell growth; this is an area in which more targeted inhibition of glycolytic pathways may provide additional insight. While mTOR is required for development, as NK cells mature, they become more quiescent and reliant on oxidative

metabolism. Naïve murine splenic NK cells rely on glucose-driven OXPHOS at baseline, similar to resting T cells (95).

IFN-γ production by naïve NK cells requires glycolysis and OXPHOS if activated by receptors, but not cytokines

Stimulation for 6 hours with either cytokines or NKR ligation is not accompanied by significant changes in flux through the glycolytic or OXPHOS pathways, as demonstrated by extracellular flux assays of glycolysis (measured as extracellular acidification rate, ECAR) and

OXPHOS (measured as oxygen consumption rate, OCR) (Fig. 1.5)(95, 96). However, flux through glycolysis and OXPHOS is required for NK cell IFN-γ production within this time period.

Interestingly, there is a difference in the metabolic requirements when NK cells are stimulated through NKRs versus cytokines (95). IFN-γ production stimulated through the NK1.1 or Ly49D receptors is almost completely blocked by glucose deprivation or OXPHOS inhibition (Fig. 1.6).

Figure 1.5: Increased rates of metabolism with prolonged IL-15 exposure but not short-term activation. Short- term activation (6 hours) by NKRs or cytokines does not change the rates of glycolysis or OXPHOS, despite production of IFN-γ and degranulation. Culture with IL-15 for at least 18 hours in one studies and 3 days in another increases both glycolysis and OXPHOS, with a greater enhancement of glycolysis. These IL-15 “primed” NK cells also have different metabolic requirements for IFN-g production compared to naïve NK cells.

However, IFN-γ produced downstream of IL-12 + IL-18 signaling is not susceptible to inhibition of OXPHOS (Fig. 1.6), or deprivation of most metabolic fuels (glucose, glutamine, fatty acids)(95). Thus, IFN-γ production induced by IL-12+IL-18 appears to be relatively metabolism- independent, and these findings suggest there might be significant flexibility in the fuels that NK cells utilize to assist with production of IFN-γ. This is in contrast to T cells, in which glycolysis is required for IFN-γ production in response to both receptor stimulation and IL-12+IL-18 (40, 97).

It is not clear why stimulation by NKRs versus cytokines leads to dramatically differential metabolic requirements in NK cells, and this is an area of active investigation in our lab. The requirement for OXPHOS in NKR-activated NK cells is unlikely to be due to a lack of ROS, as supplementation with H2O2 did not rescue IFN-γ production under OXPHOS inhibition (95).

Additional studies, both in vitro and in vivo, are needed to find the differences between NKR and cytokine activation pathways that allow one, but not the other, to be regulated metabolically.

To summarize, naïve murine NK cells use OXPHOS, consistent with findings in other resting immune cell types, and do not require significant changes in metabolism during short-term function. However, activation of NK cells is regulated by metabolism in a stimuli-specific manner.

Why NKR activation requires flux through glycolysis and OXPHOS, while cytokine activation does not, is not known and likely represents a novel mechanism for metabolic control of immune function. Examining metabolic regulation of other NK cell functions that can be induced by both

NKR and cytokine signaling, such as proliferation, will help us better understand this phenomenon.

IL-15 priming changes the basic metabolism of NK cells

We and others have observed significant upregulation of glycolysis and OXPHOS after culture in IL-15, which primes the cell for increased function (92, 95). IL-15 is produced by

Figure. 1.6. Naive NK cells require OXPHOS for receptor-stimulated IFN-g production, but activation with IL- 15 eliminates this requirement. Fresh murine splenic NK cells (top and middle) or IL-15-activated “primed” NK cells (bottom, 72hr stimulation with 100ng/ml IL-15) were activated with IL-12/18 (top) or plate-bound anti-NK1.1 (center and bottom) in the presence or absence of the ATP synthetase inhibitor oligomycin (100nM). NK cell production of IFN-g protein was measured after 6 hours. % of IFN-g+ NK cells shown. multiple cell types and supports NK cell differentiation, survival, proliferation, and cytotoxicity

(98, 99). At low doses (1-10ng/ml), IL-15 promotes the survival of NK cells in vitro, while high dose (100ng/ml) IL-15 induces proliferation and increased cytotoxicity, but does not cause degranulation or IFN-g production (95). After 18h culture with high-dose IL-15, NK cells significantly increase maximal respiratory capacity and basal OCR and ECAR (92). Longer (72-

120h) culture in high-dose IL-15 increases both ECAR and OCR, with a decreased OCR:ECAR ratio, suggesting a greater reliance on glycolysis (Fig. 1.5)(92, 95). This parallels findings in T cells, in which several days of activation induces a transition to aerobic glycolysis.

Upregulation of glycolysis after activation in T cells has been attributed to the activity of master metabolic regulators such as c-myc, HIF-1a, and mTOR (25). Out of these, the role of 23

mTOR signaling in NK cell function has recently to attention. In vitro studies using the mTOR inhibitor rapamycin have shown that mTOR signaling downstream of IL-15 is required for optimal

NK cell IFN-g and Gzmb production, as well as NK cell proliferation (93). In vivo studies examining the role of mTOR are covered in the next section, but show that mTOR is necessary for normal NK cell antiviral function.

Extended activation with IL-15 alters not only the baseline metabolism of NK cells, but also their metabolic requirements for IFN-γ production. While fresh murine NK cells require both

OXPHOS and glycolysis for NKR-stimulated IFN-γ production, NK cells primed for at least three days in high-dose IL-15 produce IFN-γ in response to receptor stimulation even when OXPHOS is inhibited (Fig. 1.5)(95). This effect was not seen in T cells, in which inhibition of glycolysis decreased IFN-γ production in T cells stimulated for three days with CD3 + CD28 or IL-12 + IL-

18. (40, 97). These IL-15-primed NK cells are also significantly less susceptible to glucose deprivation. This “metabolic priming” was not apparent at <72 hours of stimulation or with low- dose IL-15, suggesting both time and dose-dependent effects of IL-15 priming. While high doses of IL-15 cause robust NK cell proliferation, metabolic priming is not dependent on proliferation, as IFN-γ production in both highly- and poorly-proliferated NK cells was similarly resistant to

OXPHOS inhibition (95). On the other hand, an alternate priming protocol using intermediate- dose IL-15 (50ng/mL) for seven days found that IFN-γ production in response to 18h stimulation with IL-2 + IL-12 is significantly impaired in the presence of glycolytic or mTOR inhibitors (93).

These results suggest fundamental differences between metabolic regulation T cell and NK cell of

IFN-g production, and that metabolic requirements may shift depending on the type or length of stimulation.

Together, these findings demonstrate that short-term activation of NK cells is not

accompanied by changes in metabolic flux, but priming with high-dose IL-15 leads to upregulation of metabolism, specifically enhanced glycolysis. Investigating different routes of activation is important to gain insight into how NK cells increase glycolysis and OXPHOS, and what benefits this confers. The observation that prolonged activation significantly alters NK cell metabolism is particularly relevant to experiments in the literature using IL-2 or IL-15 in vitro-expanded NK cells. Caution should be used in interpreting findings from in vitro expanded NK cells and extension of those findings to naïve NK cells in vivo that have not been highly primed.

Additionally, in order to understand how IL-15 alters NK cell metabolism, we need to understand how IL-15 reverses metabolic requirements for short-term NK cell activation. Whether this is due to increased fuel flexibility, or use of activation pathways that are not metabolically-regulated, is unknown.

Manipulation of murine NK cell metabolism in vivo

While it is simpler to examine metabolism in vitro, some work has been done examining the effect of decreased mTOR signaling on NK cell function in vivo using both drug treatment and genetic models. Treatment of mice with inhibitors of mTOR or glycolysis in vivo significantly decreases NK cell size, nutrient uptake capacity, and production of IFN-γ 24 hours after activation by injection of poly(I:C) (93). In an NK-specific model of Mtor deletion, there is a significant decrease in mature, peripheral NK cells despite normal bone marrow numbers (92). Remaining

NK cells from these mice are immature, likely due to an mTOR-dependent proliferative stage during NK cell development. Mtor-deficient NK cells produce IFN-γ normally to short-term NKR and IL-12 + IL-18 stimulation, but have some defects in poly(I:C) priming of NKR activation.

Interestingly, NK cells genetically lacking mTOR have no significant differences in their ability to produce IFN-γ in response to MCMV infection at day 2 post-infection (92). This is in contrast

to NK cells inhibited by injection of rapamycin, which have decreased IFN-γ production at 1.5 days post-MCMV infection (94). It is difficult to compare these two studies, since Mtor-deficient

NK cells are immature and present at a much lower frequency. These studies also highlight the potential for discrepancies between drug-based, non-specific inhibition strategies and genetic models; differences may also be due to differences in the kinetics of mTOR inhibition, or off- target effects of drug treatment.

In summary, much of the in vivo work in mouse models has focused on the key role of mTOR signaling in NK cell development, function, and responses to MCMV. Although mTOR signaling upregulates glycolytic metabolism, mTOR signaling and glycolysis likely have separate but overlapping roles in regulating NK cell function. This possibility deserves further investigation. Careful experimentation may reveal that for some of these functions, upregulation of glycolysis may be sufficient to cause the changes attributed to mTOR signaling, while for other functions, additional metabolic regulators or pathways may be required. Further work needs to be done to better understand the role of both glycolytic and oxidative metabolism during viral infection, and better models that are more specific and allow maturation of NK cells need to be developed to do it.

Metabolic requirements in human NK cells

Ideally, the knowledge we gain through murine models of NK cell regulation are applicable to human cells. A recent study demonstrated that human peripheral blood NK cells have metabolic requirements similar to murine NK cells, despite differences between human and murine NK cells.

One key distinction is that the expression of CD56 in humans is linked to maturation, and distinguishes less mature, cytokine-producing CD56bright NK cells from more mature, cytotoxic, and abundant CD56dim NK cells (100). Sometimes the metabolic phenotype and requirements

differed between the two groups. CD56bright NK cells were more glycolytic and had increased metabolism after stimulation compared to CD56dim cells, suggesting either developmentally- regulated differences in metabolism, or perhaps that cells more focused on cytokine synthesis require increased metabolic capacity compared to cells focused on killing.

Similar to murine NK cells at rest and after IL-15 priming, human NK cells have relatively low metabolism at rest, but significantly increase glycolysis and OXPHOS after 18h of stimulation with IL-2 or IL-12 + IL-15 (101). Inhibition of mTOR in human NK cells significantly inhibits the ability of IL-2 alone to upregulate glycolysis, but glycolytic upregulation with IL-12+IL-15 is unchanged (101). The fact that mTOR is required for IL-2- but not IL-12+IL-15-driven glycolysis in human NK cells is intriguing since IL-2 and IL-15 use the same signaling receptors. These results suggest that IL-12 signaling might drive increased glycolysis in an mTOR-independent fashion.

Human NK cells also have specific metabolic requirements for IFN-γ production.

Inhibition of OXPHOS using oligomycin decreases IFN-γ production in response to IL-12+IL-15

(101), but has no effect on the small quantity of IFN-g produced in response to IL-2 alone (95).

The opposite phenotype is true with mTOR inhibition. Limiting the rate of glycolysis by culturing cells in the presence of galactose rather than glucose leads to decreased IFN-γ production by

CD56bright NK cells in response to IL-12+IL-15 (101), although the effect of direct glycolytic inhibition is unknown. Other functions of NK cell activation, such as degranulation or upregulation of CD69 and GzmB, are not affected by rapamycin, galactose, or oligomycin, with the exception that oligomycin inhibits degranulation in the more cytotoxic CD56dim NK cells. It is not yet clear why each combination of stimulation + function + CD56 phenotype has different metabolic requirements.

In summary, work examining metabolic regulation in human NK cells has just begun. IFN- g production is regulated by glycolysis, OXPHOS, and mTOR signaling, as in murine NK cells.

Metabolic differences in CD56dim and CD56bright populations reflect their disparate roles as cytotoxic vs. cytokine-producing subsets, but additional investigation is required understand how differences in metabolism between the two groups fuels differential function.

1.4 Conclusions and Aims

Several recent studies have demonstrated that glycolysis, OXPHOS, and mTOR signaling are important for mouse and human NK cell activity. In general, metabolic “reprogramming” upon activation to use aerobic glycolysis is associated with production of biomass for proliferation and shunting of glycolytic and TCA intermediates toward biosynthetic pathways for specific effector functions. On the other hand, downregulation of glycolysis and reliance on oxidative metabolism, with high mitochondrial function, is associated with long-lived tolerogenic and memory phenotypes. However, the interactions between metabolic pathways and effector functions are complex, and do not always follow this paradigm. Metabolic requirements for NK cell function are dependent on the specific stimuli used, the length of stimulation, and the activation status of the NK cell.

Our major interest is in understanding how NK cells are regulated by metabolism. There is a significant body of work examining the role of metabolism in regulating NK cell IFN-g production, but less examining NK cell cytotoxicity and proliferation. While work is proceeding in examining NK cell metabolism in a variety of culture systems, it has been technically challenging to investigate NK cell metabolism in vivo. Cell microenvironments are more difficult to control within a mouse, and drug treatments affect more than just NK cells. On the other hand, 28

genetic knockout models can lead to loss of viable NK cells to study. Since the study of metabolism in vivo is necessary to investigate more complex functions such as response to viral infection, and due to the non-physiological conditions of most in vitro assays, more in vivo research is needed.

In my dissertation work, I develop both a drug treatment and a genetic model to manipulate metabolism in NK cells, with the hope that characterization of these models will be a starting point for more tools for in vivo manipulation of metabolism.

I am focused on the two central energy-generating pathways of the cell: glycolysis and oxidative phosphorylation. These have been shown to be counter-regulated in other cell types, especially CD8+ T cells which are closely related to NK cells. However, NK cells are different than T cells—they activate more quickly, and are not built for longevity. Thus, I was interested in how NK cells might diverge from the paradigm found in T cells, and what clues that might reveal on the properties that make NK cells unique. While recent work has shown that mTOR plays a critical role in NK cell development, activation, and response to viral infection, it is unclear if this is due to mTOR’s upregulation of glycolysis or its other functions. While some work has been done examining the roles of glycolysis and OXPHOS in NK cell function, this body of work attempts to both deepen and widen our understanding of how these two central pathways affect

NK cell activation. In summary, I am asking, “How are NK cell antiviral functions regulated by glycolysis and oxidative phosphorylation in vivo?” We hypothesize that:

1. Upregulation of glycolysis is required for maximal NK cell responses during MCMV

infection, and

2. Intact oxidative metabolism is required for NK cell antiviral functions that are not

glycolysis-dependent, including formation of “memory” NK cells.

Application of our findings has the potential to inform development of new protocols for

NK cell immunotherapy that take into account the metabolic changes NK cells undergo during activation. For example, modulation of nutrients or use of inhibitors might enhance formation of activated NK cells for cancer adoptive immunotherapy. Continued investigation of the metabolic regulation of NK cells will provide valuable insight into regulation of NK cell effector functions, and the metabolic programs essential for activation of different immune cell types.

1.5 References

1. Biron, C. A., and L. Brossay. 2001. NK cells and NKT cells in innate defense against viral infections. Curr Opin Immunol 13: 458-464. 2. French, A. R., and W. M. Yokoyama. 2003. Natural killer cells and viral infections. Curr Opin Immunol 15: 45-51. 3. Vivier, E., E. Tomasello, M. Baratin, T. Walzer, and S. Ugolini. 2008. Functions of natural killer cells. Nat Immunol 9: 503-510. 4. Morvan, M. G., and L. L. Lanier. 2016. NK cells and cancer: you can teach innate cells new tricks. Nat Rev Cancer 16: 7-19. 5. Marcais, A., S. Viel, M. Grau, T. Henry, J. Marvel, and T. Walzer. 2013. Regulation of mouse NK cell development and function by cytokines. Front Immunol 4: 450. 6. Chang, S., and T. M. Aune. 2005. Histone hyperacetylated domains across the Ifng gene region in natural killer cells and T cells. Proc Natl Acad Sci 102: 17095-17100. 7. Stetson, D. B., M. Mohrs, R. L. Reinhardt, J. L. Baron, Z. E. Wang, L. Gapin, M. Kronenberg, and R. M. Locksley. 2003. Constitutive cytokine mRNAs mark natural killer (NK) and NK T cells poised for rapid effector function. J Exp Med 198: 1069-1076. 8. Tato, C. M., G. A. Martins, F. A. High, C. B. DiCioccio, S. L. Reiner, and C. A. Hunter. 2004. Cutting Edge: Innate Production of IFN-γ by NK Cells Is Independent of Epigenetic Modification of the IFN-γ Promoter. J Immunol 173: 1514-1517. 9. Mah, A. Y., and M. A. Cooper. 2016. Metabolic Regulation of Natural Killer Cell IFN- gamma Production. Crit Rev Immunol 36: 131-147. 10. Orange, J. S., and C. A. Biron. 1996. An Absolute and Restricted Requirement for IL-12 in Natural Killer Cell IFN-y Production and Antiviral Defense. J Immunol. 11. Dokun, A. O., S. Kim, H. R. Smith, H. S. Kang, D. T. Chu, and W. M. Yokoyama. 2001. Specific and nonspecific NK cell activation during virus infection. Nat Immunol 2: 951- 956. 12. Mace, E. M., P. Dongre, H. T. Hsu, P. Sinha, A. M. James, S. S. Mann, L. R. Forbes, L. B. Watkin, and J. S. Orange. 2014. Cell biological steps and checkpoints in accessing NK cell cytotoxicity. Immunol Cell Biol 92: 245-255. 13. Fehniger, T. A., S. F. Cai, X. Cao, A. J. Bredemeyer, R. M. Presti, A. R. French, and T. J. Ley. 2007. Acquisition of murine NK cell cytotoxicity requires the translation of a pre- existing pool of granzyme B and perforin mRNAs. Immunity 26: 798-811. 14. Orange, J. S. 2013. Natural killer cell deficiency. The Journal of allergy and clinical immunology 132: 515-525; quiz 526. 15. Smith, H. R., J. W. Heusel, I. K. Mehta, S. Kim, B. G. Dorner, O. V. Naidenko, K. Iizuka, H. Furukawa, D. L. Beckman, J. T. Pingel, A. A. Scalzo, D. H. Fremont, and W. M. Yokoyama. 2002. Recognition of a virus-encoded ligand by a natural killer cell activation receptor. Proc Natl Acad Sci U S A 99: 8826-8831. 16. Grimm, E. A., A. Mazumder, H. Z. Zhang, and S. A. Rosenberg. 1982. Lymphokine- activated killer cell phenomenon. Lysis of natural killer-resistant fresh solid tumor cells by interleukin 2-activated autologous human peripheral blood lymphocytes. J Exp Med 155: 1823-1841. 17. Carson, W. E., J. G. Giri, M. J. Lindemann, M. L. Linett, M. Ahdieh, R. Paxton, D. Anderson, J. Eisenmann, K. Grabstein, and M. A. Caligiuri. 1994. Interleukin (IL)-15 Is a 31

Novel Cytokine That Activates Human Natural-Killer-Cells Via Components of the IL-2 Receptor. J Exp Med 180: 1395-1403. 18. Kim, P. S., A. R. Kwilas, W. Xu, S. Alter, E. K. Jeng, H. C. Wong, J. Schlom, and J. W. Hodge. 2016. IL-15 superagonist/IL-15RalphaSushi-Fc fusion complex (IL-15SA/IL- 15RalphaSu-Fc; ALT-803) markedly enhances specific subpopulations of NK and memory CD8+ T cells, and mediates potent anti-tumor activity against murine breast and colon carcinomas. Oncotarget 7: 16130-16145. 19. Cooper, M. A., J. M. Elliott, P. A. Keyel, L. Yang, J. A. Carrero, and W. M. Yokoyama. 2009. Cytokine-induced memory-like natural killer cells. Proc Natl Acad Sci U S A 106: 1915-1919. 20. Reichlin, A., and W. M. Yokoyama. 1998. Natural killer cell proliferation induced by anti- NK1.1 and IL-2. Immunol Cell Biol 76: 143-152. 21. French, A. R., H. Sjolin, S. Kim, R. Koka, L. Yang, D. A. Young, C. Cerboni, E. Tomasello, A. Ma, E. Vivier, K. Karre, and W. M. Yokoyama. 2006. DAP12 signaling directly augments proproliferative cytokine stimulation of NK cells during viral infections. J Immunol 177: 4981-4990. 22. Prlic, M., B. R. Blazar, M. A. Farrar, and S. C. Jameson. 2003. In vivo survival and homeostatic proliferation of natural killer cells. J Exp Med 197: 967-976. 23. Jamieson, A. M., P. Isnard, J. R. Dorfman, M. C. Coles, and D. H. Raulet. 2004. Turnover and proliferation of NK cells in steady state and lymphopenic conditions. J Immunol 172: 864-870. 24. Berg, J., J. Tymoczko, and L. Stryer. 2002. Glycolysis and Gluconeogenesis. In Biochemistry, 5th ed ed. W H Freeman, New York. 25. Buck, M. D., D. O'Sullivan, and E. L. Pearce. 2015. T cell metabolism drives immunity. J Exp Med 212: 1345-1360. 26. Warburg, O. 1956. On the Origin of Cancer Cells. Science 123: 309-314. 27. Warburg, O., K. Gawehn, and A. W. Geissler. 1958. [Metabolism of leukocytes]. Zeitschrift fur Naturforschung. Teil B: Chemie, Biochemie, Biophysik, Biologie 13b: 515- 516. 28. Berg, J., J. Tymoczko, and L. Stryer. 2002. The Calvin Cycle and the Pentose Phosphate Pathway. In Biochemistry, 5th ed ed. W H Freeman, New York. 29. Wellen, K. E., and C. B. Thompson. 2012. A two-way street: reciprocal regulation of metabolism and signalling. Nat Rev Mol Cell Biol 13: 270-276. 30. Berg, J., J. Tymoczko, and L. Stryer. 2002. The Citric Acid Cycle. In Biochemistry, 5th ed ed. W H Freeman, New York. 31. Berg, J., J. Tymoczko, and L. Stryer. 2002. Oxidative Phosphorylation. In Biochemistry, 5th ed ed. W H Freeman, New York. 32. Sena, Laura A., S. Li, A. Jairaman, M. Prakriya, T. Ezponda, David A. Hildeman, C.-R. Wang, Paul T. Schumacker, Jonathan D. Licht, H. Perlman, Paul J. Bryce, and Navdeep S. Chandel. 2013. Mitochondria Are Required for Antigen-Specific T Cell Activation through Reactive Oxygen Species Signaling. Immunity 38: 225-236. 33. Belikov, A. V., B. Schraven, and L. Simeoni. 2015. T cells and reactive oxygen species. J Biomed Sci 22: 85. 34. Simeoni, L., and I. Bogeski. 2015. Redox regulation of T-cell receptor signaling. In Biol Chem. 555-568. 35. Pearce, E. J., and B. Everts. 2015. Dendritic cell metabolism. Nat Rev Immunol 15: 18-29. 32

36. Hardie, D. G. 2011. AMP-activated protein kinase: an energy sensor that regulates all aspects of cell function. Genes Dev 25: 1895-1908. 37. Frauwirth, K. A., and C. B. Thompson. 2004. Regulation of T Lymphocyte Metabolism. J Immunol 172: 4661-4665. 38. Fox, C. J., P. S. Hammerman, and C. B. Thompson. 2005. Fuel feeds function: energy metabolism and the T-cell response. Nat Rev Immunol 5: 844-852. 39. Pearce, E. L., M. C. Poffenberger, C. H. Chang, and R. G. Jones. 2013. Fueling immunity: insights into metabolism and lymphocyte function. Science 342: 1242454. 40. Chang, C. H., J. D. Curtis, L. B. Maggi, Jr., B. Faubert, A. V. Villarino, D. O'Sullivan, S. C. Huang, G. J. van der Windt, J. Blagih, J. Qiu, J. D. Weber, E. J. Pearce, R. G. Jones, and E. L. Pearce. 2013. Posttranscriptional control of T cell effector function by aerobic glycolysis. Cell 153: 1239-1251. 41. Lunt, S. Y., and M. G. Vander Heiden. 2011. Aerobic glycolysis: meeting the metabolic requirements of cell proliferation. Annu Rev Cell Dev Biol 27: 441-464. 42. Vander Heiden, M. G., L. C. Cantley, and C. B. Thompson. 2009. Understanding the Warburg effect: the metabolic requirements of cell proliferation. Science 324: 1029-1033. 43. Okoye, I., L. Wang, K. Pallmer, K. Richter, T. Ichimura, R. Haas, J. Crouse, O. Choi, D. Heathcote, E. Lovo, C. Mauro, R. Abdi, A. Oxenius, S. Rutschmann, and P. G. Ashton- Rickardt. 2015. The protein LEM promotes CD8+ T cell immunity through effects on mitochondrial respiration. Science 348: 995-1001. 44. Zeng, H., and H. Chi. 2015. Metabolic control of regulatory T cell development and function. Trends Immunol 36: 3-12. 45. Delgoffe, G. M., T. P. Kole, Y. Zheng, P. E. Zarek, K. L. Matthews, B. Xiao, P. F. Worley, S. C. Kozma, and J. D. Powell. 2009. The mTOR kinase differentially regulates effector and regulatory T cell lineage commitment. Immunity 30: 832-844. 46. Zeng, H., K. Yang, C. Cloer, G. Neale, P. Vogel, and H. Chi. 2013. mTORC1 couples immune signals and metabolic programming to establish T(reg)-cell function. Nature 499: 485-490. 47. van der Windt, G. J., B. Everts, C. H. Chang, J. D. Curtis, T. C. Freitas, E. Amiel, E. J. Pearce, and E. L. Pearce. 2012. Mitochondrial respiratory capacity is a critical regulator of CD8+ T cell memory development. Immunity 36: 68-78. 48. van der Windt, G. J. W., D. O’Sullivan, B. Everts, S. C.-C. Huang, M. D. Buck, J. D. Curtis, C.-H. Chang, A. M. Smith, T. Ai, B. Faubert, R. G. Jones, E. J. Pearce, and E. L. Pearce. 2013. CD8 memory T cells have a bioenergetic advantage that underlies their rapid recall ability. Proc Natl Acad Sci 110: 14336-14341. 49. Dimeloe, S., M. Mehling, C. Frick, J. Loeliger, G. R. Bantug, U. Sauder, M. Fischer, R. Belle, L. Develioglu, S. Tay, A. Langenkamp, and C. Hess. 2016. The Immune-Metabolic Basis of Effector Memory CD4+ T Cell Function under Hypoxic Conditions. J Immunol 196: 106-114. 50. Pearce, E. L., M. C. Walsh, P. J. Cejas, G. M. Harms, H. Shen, L. S. Wang, R. G. Jones, and Y. Choi. 2009. Enhancing CD8 T-cell memory by modulating fatty acid metabolism. Nature 460: 103-107. 51. Sukumar, M., J. Liu, Y. Ji, M. Subramanian, J. G. Crompton, Z. Yu, R. Roychoudhuri, D. C. Palmer, P. Muranski, E. D. Karoly, R. P. Mohney, C. A. Klebanoff, A. Lal, T. Finkel, N. P. Restifo, and L. Gattinoni. 2013. Inhibiting glycolytic metabolism enhances CD8+ T cell memory and antitumor function. The Journal of clinical investigation 123: 4479-4488. 33

52. Gubser, P. M., G. R. Bantug, L. Razik, M. Fischer, S. Dimeloe, G. Hoenger, B. Durovic, A. Jauch, and C. Hess. 2013. Rapid effector function of memory CD8+ T cells requires an immediate-early glycolytic switch. Nat Immunol 14: 1064-1072. 53. O'Neill, L. A., and E. J. Pearce. 2016. Immunometabolism governs dendritic cell and macrophage function. J Exp Med 213: 15-23. 54. Saeed, S., J. Quintin, H. H. Kerstens, N. A. Rao, A. Aghajanirefah, F. Matarese, S. C. Cheng, J. Ratter, K. Berentsen, M. A. van der Ent, N. Sharifi, E. M. Janssen-Megens, M. Ter Huurne, A. Mandoli, T. van Schaik, A. Ng, F. Burden, K. Downes, M. Frontini, V. Kumar, E. J. Giamarellos-Bourboulis, W. H. Ouwehand, J. W. van der Meer, L. A. Joosten, C. Wijmenga, J. H. Martens, R. J. Xavier, C. Logie, M. G. Netea, and H. G. Stunnenberg. 2014. Epigenetic programming of monocyte-to-macrophage differentiation and trained innate immunity. Science 345: 1251086. 55. Galván-Peña, S., and L. A. J. O'Neill. 2014. Metabolic reprogramming in macrophage polarization. Front Immunol 5. 56. Infantino, V., P. Convertini, L. Cucci, Maria A. Panaro, Maria A. Di Noia, R. Calvello, F. Palmieri, and V. Iacobazzi. 2011. The mitochondrial citrate carrier: a new player in inflammation. Biochem J 438: 433-436. 57. Galván-Peña, S., and L. A. J. O’Neill. 2014. Metabolic Reprograming in Macrophage Polarization. Frontiers in Immunology 5: 420. 58. Clementi, E., G. C. Brown, M. Feelisch, and S. Moncada. 1998. Persistent inhibition of cell respiration by nitric oxide: crucial role of S-nitrosylation of mitochondrial complex I and protective action of glutathione. Proc Natl Acad Sci U S A 95: 7631-7636. 59. Lampropoulou, V., A. Sergushichev, M. Bambouskova, S. Nair, E. E. Vincent, E. Loginicheva, L. Cervantes-Barragan, X. Ma, S. C. Huang, T. Griss, C. J. Weinheimer, S. Khader, G. J. Randolph, E. J. Pearce, R. G. Jones, A. Diwan, M. S. Diamond, and M. N. Artyomov. 2016. Itaconate Links Inhibition of Succinate Dehydrogenase with Macrophage Metabolic Remodeling and Regulation of Inflammation. Cell Metab 24: 158- 166. 60. Cordes, T., M. Wallace, A. Michelucci, A. S. Divakaruni, S. C. Sapcariu, C. Sousa, H. Koseki, P. Cabrales, A. N. Murphy, K. Hiller, and C. M. Metallo. 2016. Immunoresponsive Gene 1 and Itaconate Inhibit Succinate Dehydrogenase to Modulate Intracellular Succinate Levels. J Biol Chem 291: 14274-14284. 61. Tannahill, G. M., A. M. Curtis, J. Adamik, E. M. Palsson-McDermott, A. F. McGettrick, G. Goel, C. Frezza, N. J. Bernard, B. Kelly, N. H. Foley, L. Zheng, A. Gardet, Z. Tong, S. S. Jany, S. C. Corr, M. Haneklaus, B. E. Caffrey, K. Pierce, S. Walmsley, F. C. Beasley, E. Cummins, V. Nizet, M. Whyte, C. T. Taylor, H. Lin, S. L. Masters, E. Gottlieb, V. P. Kelly, C. Clish, P. E. Auron, R. J. Xavier, and L. A. O'Neill. 2013. Succinate is an inflammatory signal that induces IL-1beta through HIF-1alpha. Nature 496: 238-242. 62. Mills, E. L., B. Kelly, A. Logan, A. S. H. Costa, M. Varma, C. E. Bryant, P. Tourlomousis, J. H. M. Dabritz, E. Gottlieb, I. Latorre, S. C. Corr, G. McManus, D. Ryan, H. T. Jacobs, M. Szibor, R. J. Xavier, T. Braun, C. Frezza, M. P. Murphy, and L. A. O'Neill. 2016. Succinate Dehydrogenase Supports Metabolic Repurposing of Mitochondria to Drive Inflammatory Macrophages. Cell 167: 457-470 e413. 63. Moon, J.-S., S. Hisata, M.-A. Park, Gina M. DeNicola, Stefan W. Ryter, K. Nakahira, and Augustine M. K. Choi. 2015. mTORC1-Induced HK1-Dependent Glycolysis Regulates NLRP3 Inflammasome Activation. Cell Rep 12: 102-115. 34

64. Liu, T. F., V. T. Vachharajani, B. K. Yoza, and C. E. McCall. 2012. NAD+-dependent sirtuin 1 and 6 proteins coordinate a switch from glucose to fatty acid oxidation during the acute inflammatory response. J Biol Chem 287: 25758-25769. 65. Jha, A., S.-C. Huang, A. Sergushichev, V. Lampropoulou, Y. Ivanova, E. Loginicheva, K. Chmielewski, K M. Stewart, J. Ashall, B. Everts, E. Pearce, E. Driggers, and M. Artyomov. Network Integration of Parallel Metabolic and Transcriptional Data Reveals Metabolic Modules that Regulate Macrophage Polarization. Immunity 42: 419-430. 66. Kelly, B., and L. A. O'Neill. 2015. Metabolic reprogramming in macrophages and dendritic cells in innate immunity. Cell Res 25: 771-784. 67. Cheng, S. C., J. Quintin, R. A. Cramer, K. M. Shepardson, S. Saeed, V. Kumar, E. J. Giamarellos-Bourboulis, J. H. Martens, N. A. Rao, A. Aghajanirefah, G. R. Manjeri, Y. Li, D. C. Ifrim, R. J. Arts, B. M. van der Veer, P. M. Deen, C. Logie, L. A. O'Neill, P. Willems, F. L. van de Veerdonk, J. W. van der Meer, A. Ng, L. A. Joosten, C. Wijmenga, H. G. Stunnenberg, R. J. Xavier, and M. G. Netea. 2014. mTOR- and HIF-1alpha- mediated aerobic glycolysis as metabolic basis for trained immunity. Science 345: 1250684. 68. Krawczyk, C. M., T. Holowka, J. Sun, J. Blagih, E. Amiel, R. J. DeBerardinis, J. R. Cross, E. Jung, C. B. Thompson, R. G. Jones, and E. J. Pearce. 2010. Toll-like receptor–induced changes in glycolytic metabolism regulate dendritic cell activation. Blood 115: 4742-4749. 69. Jantsch, J., D. Chakravortty, N. Turza, A. T. Prechtel, B. Buchholz, R. G. Gerlach, M. Volke, J. Glasner, C. Warnecke, M. S. Wiesener, K. U. Eckardt, A. Steinkasserer, M. Hensel, and C. Willam. 2008. Hypoxia and hypoxia-inducible factor-1 alpha modulate lipopolysaccharide-induced dendritic cell activation and function. J Immunol 180: 4697- 4705. 70. Everts, B., E. Amiel, S. C.-C. Huang, A. M. Smith, C.-H. Chang, W. Y. Lam, V. Redmann, T. C. Freitas, J. Blagih, G. J. W. van der Windt, M. N. Artyomov, R. G. Jones, E. L. Pearce, and E. J. Pearce. 2014. TLR-driven early glycolytic reprogramming via the kinases TBK1-IKK[epsiv] supports the anabolic demands of dendritic cell activation. Nat Immunol 15: 323-332. 71. Rehman, A., K. C. Hemmert, A. Ochi, M. Jamal, J. R. Henning, R. Barilla, J. P. Quesada, C. P. Zambirinis, K. Tang, M. Ego-Osuala, R. S. Rao, S. Greco, M. Deutsch, S. Narayan, H. L. Pachter, C. S. Graffeo, D. Acehan, and G. Miller. 2013. Role of fatty-acid synthesis in dendritic cell generation and function. J Immunol 190: 4640-4649. 72. Ramakrishnan, R., V. A. Tyurin, F. Veglia, T. Condamine, A. Amoscato, D. Mohammadyani, J. J. Johnson, L. M. Zhang, J. Klein-Seetharaman, E. Celis, V. E. Kagan, and D. I. Gabrilovich. 2014. Oxidized lipids block antigen cross-presentation by dendritic cells in cancer. J Immunol 192: 2920-2931. 73. Everts, B., E. Amiel, G. J. van der Windt, T. C. Freitas, R. Chott, K. E. Yarasheski, E. L. Pearce, and E. J. Pearce. 2012. Commitment to glycolysis sustains survival of NO- producing inflammatory dendritic cells. Blood 120: 1422-1431. 74. Zaccagnino, P., M. Saltarella, S. Maiorano, A. Gaballo, G. Santoro, B. Nico, M. Lorusso, and A. Del Prete. 2012. An active mitochondrial biogenesis occurs during dendritic cell differentiation. Int J Biochem Cell Biol 44: 1962-1969. 75. Bajwa, G., R. J. DeBerardinis, B. Shao, B. Hall, J. D. Farrar, and M. A. Gill. 2016. Cutting Edge: Critical Role of Glycolysis in Human Plasmacytoid Dendritic Cell Antiviral Responses. J Immunol 196: 2004-2009. 35

76. Wu, D., D. E. Sanin, B. Everts, Q. Chen, J. Qiu, M. D. Buck, A. Patterson, A. M. Smith, C. H. Chang, Z. Liu, M. N. Artyomov, E. L. Pearce, M. Cella, and E. J. Pearce. 2016. Type 1 Interferons Induce Changes in Core Metabolism that Are Critical for Immune Function. Immunity 44: 1325-1336. 77. Jellusova, J., M. H. Cato, J. R. Apgar, P. Ramezani-Rad, C. R. Leung, C. Chen, A. D. Richardson, E. M. Conner, R. J. Benschop, J. R. Woodgett, and R. C. Rickert. 2017. Gsk3 is a metabolic checkpoint regulator in B cells. Nat Immunol 18: 303-312. 78. Doughty, C. A., B. F. Bleiman, D. J. Wagner, F. J. Dufort, J. M. Mataraza, M. F. Roberts, and T. C. Chiles. 2006. Antigen receptor-mediated changes in glucose metabolism in B lymphocytes: role of phosphatidylinositol 3-kinase signaling in the glycolytic control of growth. Blood 107: 4458-4465. 79. Caro-Maldonado, A., R. Wang, A. G. Nichols, M. Kuraoka, S. Milasta, L. D. Sun, A. L. Gavin, E. D. Abel, G. Kelsoe, D. R. Green, and J. C. Rathmell. 2014. Metabolic reprogramming is required for antibody production that is suppressed in anergic but exaggerated in chronically BAFF-exposed B cells. J Immunol 192: 3626-3636. 80. Boothby, M., and R. C. Rickert. 2017. Metabolic Regulation of the Immune Humoral Response. Immunity 46: 743-755. 81. Blair, D., F. J. Dufort, and T. C. Chiles. 2012. Protein kinase Cbeta is critical for the metabolic switch to glycolysis following B-cell antigen receptor engagement. Biochem J 448: 165-169. 82. Jang, K. J., H. Mano, K. Aoki, T. Hayashi, A. Muto, Y. Nambu, K. Takahashi, K. Itoh, S. Taketani, S. L. Nutt, K. Igarashi, A. Shimizu, and M. Sugai. 2015. Mitochondrial function provides instructive signals for activation-induced B-cell fates. Nat Commun 6: 6750. 83. Dufort, F. J., M. R. Gumina, N. L. Ta, Y. Tao, S. A. Heyse, D. A. Scott, A. D. Richardson, T. N. Seyfried, and T. C. Chiles. 2014. Glucose-dependent de novo lipogenesis in B lymphocytes: a requirement for atp-citrate lyase in lipopolysaccharide-induced differentiation. J Biol Chem 289: 7011-7024. 84. Lam, W. Y., and D. Bhattacharya. 2018. Metabolic Links between Plasma Cell Survival, Secretion, and Stress. Trends Immunol 39: 19-27. 85. Kim, M., Y. Qie, J. Park, and C. H. Kim. 2016. Gut Microbial Metabolites Fuel Host Antibody Responses. Cell Host Microbe 20: 202-214. 86. Lam, W. Y., A. M. Becker, K. M. Kennerly, R. Wong, J. D. Curtis, E. M. Llufrio, K. S. McCommis, J. Fahrmann, H. A. Pizzato, R. M. Nunley, J. Lee, M. J. Wolfgang, G. J. Patti, B. N. Finck, E. L. Pearce, and D. Bhattacharya. 2016. Mitochondrial Pyruvate Import Promotes Long-Term Survival of Antibody-Secreting Plasma Cells. Immunity 45: 60-73. 87. Borregaard, N., and T. Herlin. 1982. Energy metabolism of human neutrophils during phagocytosis. The Journal of clinical investigation 70: 550-557. 88. Fossati, G., D. A. Moulding, D. G. Spiller, R. J. Moots, M. R. H. White, and S. W. Edwards. 2003. The Mitochondrial Network of Human Neutrophils: Role in Chemotaxis, Phagocytosis, Respiratory Burst Activation, and Commitment to Apoptosis. The Journal of Immunology 170: 1964-1972. 89. Maianski, N. A., J. Geissler, S. M. Srinivasula, E. S. Alnemri, D. Roos, and T. W. Kuijpers. 2004. Functional characterization of mitochondria in neutrophils: a role restricted to apoptosis. Cell death and differentiation 11: 143-153.

90. Rodriguez-Espinosa, O., O. Rojas-Espinosa, M. M. Moreno-Altamirano, E. O. Lopez- Villegas, and F. J. Sanchez-Garcia. 2015. Metabolic requirements for neutrophil extracellular traps formation. Immunology 145: 213-224. 91. Kirchner, T., S. Moller, M. Klinger, W. Solbach, T. Laskay, and M. Behnen. 2012. The impact of various reactive oxygen species on the formation of neutrophil extracellular traps. Mediators Inflamm 2012: 849136. 92. Marcais, A., J. Cherfils-Vicini, C. Viant, S. Degouve, S. Viel, A. Fenis, J. Rabilloud, K. Mayol, A. Tavares, J. Bienvenu, Y. G. Gangloff, E. Gilson, E. Vivier, and T. Walzer. 2014. The metabolic checkpoint kinase mTOR is essential for IL-15 signaling during the development and activation of NK cells. Nat Immunol 15: 749-757. 93. Donnelly, R. P., R. M. Loftus, S. E. Keating, K. T. Liou, C. A. Biron, C. M. Gardiner, and D. K. Finlay. 2014. mTORC1-dependent metabolic reprogramming is a prerequisite for NK cell effector function. J Immunol 193: 4477-4484. 94. Nandagopal, N., A. K. Ali, A. K. Komal, and S. H. Lee. 2014. The Critical Role of IL-15- PI3K-mTOR Pathway in Natural Killer Cell Effector Functions. Front Immunol 5: 187. 95. Keppel, M. P., N. Saucier, A. Y. Mah, T. P. Vogel, and M. A. Cooper. 2015. Activation- specific metabolic requirements for NK Cell IFN-gamma production. J Immunol 194: 1954-1962. 96. Keppel, M. P., and M. A. Cooper. 2016. Assessment of NK Cell Metabolism. Methods in molecular biology (Clifton, N.J.) 1441: 27-42. 97. Cham, C. M., and T. F. Gajewski. 2005. Glucose Availability Regulates IFN- Production and p70S6 Kinase Activation in CD8+ Effector T Cells. J Immunol 174: 4670-4677. 98. Carson, W. E., J. G. Giri, M. J. Lindemann, M. L. Linett, M. Ahdieh, R. Paxton, D. Anderson, J. Eisenmann, K. Grabstein, and M. A. Caligiuri. 1994. Interleukin (IL) 15 is a novel cytokine that activates human natural killer cells via components of the IL-2 receptor. J Exp Med 180: 1395-1403. 99. Becknell, B., and M. A. Caligiuri. 2005. Interleukin-2, Interleukin-15, and Their Roles in Human Natural Killer Cells. In Adv Immunol. W. A. Frderick, ed. Academic Press. 209- 239. 100. Freud, A. G., and M. A. Caligiuri. 2006. Human natural killer cell development. Immunol Rev 214: 56-72. 101. Keating, S. E., V. Zaiatz-Bittencourt, R. M. Loftus, C. Keane, K. Brennan, D. K. Finlay, and C. M. Gardiner. 2016. Metabolic Reprogramming Supports IFN-γ Production by CD56bright NK Cells. Journal of immunology.

Chapter 2

Glycolytic Requirement for NK Cell Cytotoxicity and Cytomegalovirus Control

This chapter is adapted from a manuscript published in JCI Insight:

Mah AY, Rashidi A, Keppel MP, Saucier N, Moore EK, Alinger JB, Tripathy SK, Agarwal SK, Jeng EK, Wong HC, Miller JS, Fehniger TA, Mace EM, French AR, Cooper MA. Glycolytic requirement for NK cell cytotoxicity and cytomegalovirus control. JCI Insight. 2017 Dec 7;2(23):95128. doi: 10.1172/jci.insight.95128.

Author contributions: A.R. and J.S.M. collected clinical information for patient treatment with ALT-803, and performed and analyzed experiments using the patient’s cells.

2.1 Abstract

Natural killer (NK) cell activation has been shown to be metabolically regulated in vitro; however, the role of metabolism during in vivo NK cell responses to infection is unknown. We examined the role of glycolysis in NK cell function during murine cytomegalovirus (MCMV) infection and the ability of IL-15 to prime NK cells during CMV infection. The glucose metabolism inhibitor 2-deoxy-ᴅ-glucose (2DG) impaired both mouse and human NK cell cytotoxicity following priming in vitro. Similarly, MCMV-infected mice treated with 2DG had impaired clearance of NK-specific targets in vivo, which was associated with higher viral burdens and susceptibility to infection on the C57BL/6 background. IL-15 priming is known to alter NK cell metabolism and metabolic requirements for activation. Treatment with the IL-15 superagonist

ALT-803 rescued mice from otherwise lethal infection in an NK-dependent manner. Consistent with this, treatment of a patient with ALT-803 for recurrent CMV reactivation post-hematopoietic cell transplant was associated with clearance of viremia. These studies demonstrate that NK cell- mediated control of viral infection requires glucose metabolism, and that IL-15 treatment in vivo can reduce this requirement and may be effective as an antiviral therapy.

2.2 Introduction

It is now recognized that metabolism plays a key role in the activation and specialized functions of immune cells. Cells fuel ATP production through two major metabolic pathways: glycolysis, the metabolic breakdown of glucose, and oxidative phosphorylation, the synthesis of

ATP in the presence of oxygen. For T cells, macrophages, and DCs, inflammatory activation is associated with “metabolic reprogramming” from a relatively metabolically-quiescent to a highly glycolytic state (1-4). These metabolic changes are thought to assist the generation of biosynthetic

precursors for effector functions. For example, high glycolytic flux can produce nucleotide precursors for cell division in T cells (5), citrate for production of reactive oxygen species in macrophages (6), and lipid precursors for endoplasmic reticulum membrane synthesis in DCs (7).

Glycolytic enzymes have also been shown to directly regulate cytokine production and gene expression in T cells (8, 9). Thus, in many immune cell types, metabolic regulation equips activated cells to carry out specialized functions.

NK cells are innate cytotoxic lymphocytes that are the first responders against virally- infected and tumor cells. They respond within hours of stimulation by releasing cytokines, such as

IFN-γ, and killing target cells. Stimulation with cytokines for longer periods, including inflammatory cytokines such as IL-12 and IL-15, significantly enhances NK cell functions. Recent studies have demonstrated that certain NK cell responses require metabolic cues. Naïve NK cells rely on glucose-fueled oxidative phosphorylation, while expansion with IL-15 elevates metabolic rates with a preference toward glycolysis, similar to changes seen during the activation of CD8+ T cells (10, 11). Several studies show that NK cell production of the cytokine IFN-γ is regulated by metabolism (10, 12-14). We previously demonstrated that in freshly-isolated murine NK cells,

IFN-γ production after short-term receptor activation is highly dependent on glycolytic and oxidative metabolism, while IFN-γ production stimulated by cytokine activation (IL-12 + IL-18) is independent of both metabolic pathways. We found that receptor activation depends less on these metabolic pathways after priming with high-dose IL-15 (10). Others have found that after expansion in IL-15 and stimulation with IL-2 + IL-12, highly-activated “lymphokine-activated killer” (LAK) cells require glycolysis for maximal production of IFN-γ and the cytotoxic granule protein granzyme B (Gzmb) (12). Studies in human NK cells show a similar pattern—IFN-γ and

Gzmb production downstream of certain stimuli is sensitive to metabolic inhibition (13). Together,

these studies demonstrate that NK cells have specific metabolic requirements for IFN-γ and Gzmb production. Few studies have directly examined the metabolic regulation of other NK cell functions such as target killing and proliferation. Here, we investigate the role of glycolysis in NK cell cytotoxicity in vitro and in vivo in response to viral infection.

NK cells are particularly important for antiviral responses against cytomegalovirus. In response to human cytomegalovirus (HCMV), NK cells bearing the activating NKG2C receptor rapidly expand and persist (15, 16). These cells, termed CMV-adapted NK cells, display enhanced responses, particularly when triggered by antibodies through their Fc receptor (17-20). While the adaptive T cell response to HCMV is required for long-term immunity, evidence from a B and T cell-deficient patient demonstrates that the expansion of NKG2C-positive NK cells alone can control HCMV infection (21).

Acute infection with murine cytomegalovirus (MCMV) in C57BL/6 mice is well described to be controlled by NK cells (22-26). During the first 2 days of infection, NK cells are activated by circulating cytokines and respond by producing a burst of immunomodulatory cytokines, most notably IFN-γ (27). After day 2, expression of the virally-encoded ligand m157 on the surface of infected cells leads to the expansion of Ly49H+ NK cells which specifically recognize and kill m157-expressing cells (28-31). Thus, this murine system can be used to examine NK cell activation, proliferation, and killing of infected cells in vivo, and is a model for human infection.

In the current study, we found that NK cell proliferation and cytotoxicity following IL-15 priming in vitro were impaired when NK cells were cultured with the glucose metabolism inhibitor

2-deoxy-ᴅ-glucose (2DG), which blocks the first two enzymes of glycolysis. Glycolytic inhibition decreased expression of cytotoxic proteins and altered the adhesion of NK:target conjugates. In vivo, treatment of MCMV-resistant C57BL/6 mice with 2DG caused significant susceptibility to

infection and was associated with impaired NK-mediated clearance of m157+ targets. Similar susceptibility to infection was also observed when animals were treated with rapamycin, a drug that inhibits the mTOR pathway. Priming with the IL-15 superagonist ALT-803 rescued 2DG- or rapamycin- treated animals from lethal infection. 2DG also decreased human NK cell cytotoxicity, and here we describe a hematopoietic cell transplant (HCT) patient with recurrent CMV reactivation treated with ALT-803. Similar to findings in our murine model, ALT-803 treatment in this patient was associated with clearance of viremia, suggesting therapeutic potential for IL-15 agonists in CMV.

2.3 Results

Inhibiting glucose metabolism with 2DG decreases NK cell proliferation and cytotoxicity in vitro

Murine NK cells require activation, such as priming with IL-15, in order to upregulate cytotoxic effector molecules, such as Gzmb, and effectively kill target cells (32). To investigate the requirement for glucose metabolism in NK cytotoxicity, NK cells were purified from wildtype

C57BL/6 mice and plated in 100ng/mL IL-15 with or without 1mM 2DG for 72 hours. This dose of 2DG maintained NK cell viability, with minimal differences between treated and untreated cells

(Supplementary Figure 2.1A). Extracellular flux assays confirmed that 2DG treatment significantly decreased NK cell glycolysis, as measured by the extracellular acidification rate

(ECAR) (Figure 2.1A). We previously demonstrated that naïve murine NK cells primarily use glucose-based oxidative phosphorylation (10). Consistent with this, 2DG treatment also decreased the rate of oxygen consumption (OCR) (Figure 2.1A). 2DG significantly inhibited IL-15-induced proliferation (Figure 2.1B), leading to a reduction in the number of cells recovered at the end of the incubation (Supplementary Figure 2.1B).

Figure 2.1. Glucose metabolism blockade during activation decreases NK cell proliferation and cytotoxicity. Purified NK cells were cultured for 72 hours in 100ng/mL IL-15 without (black) or with 1mM 2-deoxy-ᴅ-glucose (2DG, green). A) Representative extracellular flux assay showing glycolytic rate as estimated by extracellular acidification rate (ECAR); glycolytic stress test shows baseline ECAR in glucose-free media and after addition of glucose, maximal ECAR after addition of oligomycin, and non-glycolytic acidification after 2DG. Summary data shows ECAR, oxygen consumption rate (OCR), and the OCR/ECAR ratio normalized to control (n=4/group in 3 separate experiments, ratio paired t test performed on raw data). B) Representative histogram shows CFSE dilution of proliferating NK cells. Summary graph shows the percentage of NK cells proliferated (n=5 experiments, paired t test). C) NK cells activated in IL-15 +/- 2DG were assessed for their ability to kill Ba/F3-m157 target cells at different effector:target ratios (ratios represent Ly49H+ NK cells:targets), n=8/group, 4 experiments, 2-way ANOVA. Data is given as mean +/- SEM. ns=not significant (p≥0.05), *=p<0.05, ***=p<0.001. The ability of NK cells to kill target cells after priming with IL-15 +/- 2DG was measured at various effector-to-target ratios in complete media. Ba/F3-m157 cells are murine pro-B cells ectopically expressing the MCMV-encoded protein m157 that is recognized by Ly49H+ NK cells, and killing of Ba/F3-m157 is mediated solely by Ly49H-expressing NK cells (29, 33, 34). Ly49H+

NK cells killing of Ba/F3-m157 targets was significantly impaired when activated in the presence of 2DG (Figure 2.1C). NK cell killing of MHC class I-deficient RMA-S tumor cells (35, 36) was also significantly impaired by 2DG. Interestingly, killing of YAC-1 tumor cells, which are primarily recognized by NKG2D (37-39), was not impaired by 2DG (Supplementary Figure 2.1C).

These results suggest that glycolytic inhibition caused target-specific deficits in NK cell killing, possibly based on different mechanisms of target cell recognition and NK activation.

Decreased killing by 2DG-treated NK cells is associated with low Gzmb and variable actin accumulation during target conjugation

We investigated several mechanisms by which 2DG might decrease NK cell cytotoxicity.

First, IL-15 increases levels of cytotoxic granule proteins in NK cells (32). Treatment with 2DG reduced expression of Gzmb (Figure 2.2A), suggesting one potential mechanism for decreased NK cell cytotoxicity. Next, we investigated the dynamics of NK cell recognition and killing of target cells. Target killing is a tightly-orchestrated event comprised of multiple steps including recognition of the target cell, adhesion and formation of an immunological synapse (IS), anchoring to the target with filamentous F-actin, convergence of granules to the microtubule organizing center (MTOC), polarization and trafficking of the MTOC towards the IS, degranulation, and finally detachment (40). By flow cytometry, 2DG did not significantly decrease NK cell recognition of potential targets, as demonstrated by normal conjugate formation between 2DG- treated NK cells and Ba/F3-m157 targets at a variety of timepoints (Figure 2.2B). Confocal microscopy of co-cultured NK cells and Ba/F3-m157 targets also demonstrated conjugate formation by 2DG-treated NK cells (Figure 2.2C). In some conjugates in both treatment groups,

Ly49H was detected at the synapse (Figure 2.2D). However, actin accumulation at the IS was highly variable among 2DG-treated NK cells, indicating that 2DG may impair proper adhesion to the target (Figure 2.2C, E). Compared to 2DG-treated NK cells, most control NK cells had neutral or positive actin accumulation, indicating abnormal anchoring to the target and commitment to killing (Figure 2.2E). However, 2DG did not affect polarization of the MTOC, as measured by the distance of the MTOC to the IS (Figure 2.2F), suggesting that once NK cells committed to killing,

Figure 2.2. In vitro treatment with 2DG decreases NK cell production of granzyme B and alters actin accumulation at the immunological synapse. Purified NK cells were cultured for 72 hours in 100ng/mL IL-15 with or without 1mM 2-deoxy-ᴅ-glucose (2DG). A) Representative flow histogram shows Gzmb levels in naïve (filled) and activated NK cells cultured without (black) or with 1mM 2DG (green). Summary data is normalized to MFI of naïve NK cells (n=9/group, 5 separate experiments, ratio paired t test performed on raw data). Error bars denote SEM. B) NK cells were incubated at a 1:1 ratio with Ba/F3-m157 target cells for various times; conjugates were detected by flow cytometry. Representative flow cytometry shows CFSE+, CellTrace Violet (CTV)+ conjugates, and conjugation curve shows percentage of CFSE+ cells conjugated (n=2-3 technical replicates, error bars indicate SEM, representative of 3 separate experiments). C-F) NK cells were incubated with Ba/F3-m157 cells for 40 minutes at a 2:1 ratio. C-D) Fluorescence microscopy of NK:Ba/F3-m157 conjugates show convergence of granules (green = Gzmb, white = Prf), Ly49H accumulation at the synapse (yellow), polarization of the microtubule organizing center (MTOC, blue) toward the synapse, and actin accumulation (red). Scale bar = 2µm. Representative images for the range of control and 2DG- treated NK cells are shown. E) Quantification of actin accumulation by area * intensity of synapse - (NK + target), (n=25-27 conjugates/group, 4 experiments, distribution analyzed by Kolmogorov-Smirnov test). F) Quantification of MTOC-synapse distance in micrometers, not significant by Welch’s t test for means or Kolmogorov-Smirnov test (n=58-60 conjugates/group, 5 experiments). For (E-F), the gray box displays median, 25th, and 75th percentiles. ns=not significant (p≥0.05), *=p<0.05, ***=p<0.001.

MTOC polarization was not impaired. Thus, these data demonstrate that inhibition of glucose metabolism interfered with NK cell priming, leading to decreased levels of Gzmb protein production and altered actin accumulation at the IS.

2DG-treated, MCMV-infected mice have impaired target clearance day 2 post-infection

To examine the effects of glycolytic inhibition on NK cell function in vivo, we treated

MCMV-infected mice with 2DG. Female WT mice were infected intraperitoneally with 1 x 105

PFU of MCMV (~40% LD50) (41) and injected daily with either 19mg (1g/kg) of 2DG or vehicle

(PBS). To determine whether 2DG administration altered NK cell cytotoxicity, in vivo target clearance was assessed. Syngeneic WT splenocytes (non-targets) and transgenic splenocytes expressing the MCMV-encoded activating ligand m157, which is only recognized by Ly49H+ NK cells, were differentially labeled and adoptively transferred into infected mice 2 days after infection (42). Clearance of targets in 2DG-treated MCMV-infected mice was significantly lower than in PBS-treated MCMV-infected mice, and was similar to baseline clearance in mock-infected mice (Figure 2.3A). This finding suggests that 2DG treatment impaired NK cell activation with

MCMV infection and subsequent killing of target cells. The defect in target clearance in 2DG- treated mice was not due to differences in NK cell numbers, as there were no significant differences in the number or percentage of NK cells (Figure 2.3B), or in Ly49H expression on NK cells (not shown), in the spleens or livers at this early timepoint. To determine whether NK cells had a defect in cytotoxic capacity, intracellular Gzmb and degranulation, as measured by cell-surface expression of CD107a, were tested. NK cells rapidly upregulate Gzmb and other cytotoxic proteins after infection with MCMV, and levels of Gzmb at day 2 in NK cells from infected mice were significantly higher than in naïve mice. However, there were no differences between 2DG- and

PBS-treated mice in the percentage of Gzmb+ NK cells, or in the amount of Gzmb protein as

Figure 2.3. 2DG treatment in vivo causes decreased target clearance during MCMV infection. 2-deoxy-ᴅ- glucose (2DG)- or PBS-treated mice were analyzed on day 2 after infection with murine cytomegalovirus (MCMV). A) Equal numbers of labeled non-target splenocytes and target m157-transgenic splenocytes were injected i.v. and target clearance in the spleen was measured 4 hours later (n=7-12 individuals/group, 2 separate experiments). B) The number and percentage of NK cells isolated from the spleen or liver (n=7-8/group, 2 experiments). C) Flow plots and mean fluorescence intensity of granzyme B measured by flow cytometry in Ly49H+ (solid) and Ly49H- (hatched) NK cells from the spleen or liver (n=4/group, representative of 2 experiments). D) CD107a expression on splenocytes (n=7-8/group, 2 experiments). All statistical analyses are 1-way ANOVA with Tukey’s multiple comparison test, ns=not significant (p≥0.05), *=p<0.05, **=p<0.01, ***=p<0.001.

measured by mean fluorescence intensity (MFI) (Figure 2.3C). NK degranulation measured by cell-surface level of CD107a stained directly ex vivo was equivalent between 2DG-treated and

PBS control mice (Figure 2.3D). Similarly, no differences in T cell numbers, Gzmb expression, or degranulation were detected (Supplementary Figure 2.2A-C). These results suggest that NK cells from 2DG-treated mice retained cytotoxic machinery, but were unable to mount productive target cell killing.

2DG-treated mice produce normal levels of IFN-g during MCMV infection

NK cell production of IFN-γ peaks at 1.5 days post-MCMV infection in response to systemic release of inflammatory cytokines, including IL-12 made by dendritic cells (43). While others have observed that IFN-γ production by LAK cells in vitro or by poly(I:C)-stimulated NK cells in vivo is susceptible to inhibition by 2DG (12), we previously demonstrated in vitro that naïve NK cell production of cytokine-driven IFN-γ is not inhibited by 2DG or by culture in glucose-free media (10). In agreement with our previous in vitro studies, there was no difference in IFN-γ production during MCMV infection with 2DG treatment (Supplementary Figure 2.3).

These data also suggest an intact innate immune cytokine response to MCMV in the context of systemic 2DG injection, since early NK cell IFN-γ production is dependent on cytokines produced by antigen-presenting cells.

2DG treatment causes susceptibility to MCMV infection.

To test whether impaired NK cell clearance of targets correlates with susceptibility to infection, MCMV copy number was assessed in the spleen and liver. 2DG treatment was associated with higher viral burdens as early as day 2 post-infection in the spleen (Figure 2.4A), and by day 4 there was significant loss of viral control, evidenced by high viral copy number in both the spleen and liver (Figure 2.4B). Data on organ-specific control of MCMV is mixed, but

one hypothesis is that NK cell IFN-γ is more important for control of MCMV in the liver, while

NK cell cytotoxicity and Ly49H are responsible for control of infection in the spleen (44-47), which is consistent with our findings.

By day 4 post-infection, 2DG treatment and high viral burden led to a decrease in the total number of splenocytes and liver lymphocytes (Figure 2.4C), with correspondingly decreased numbers of NK cells. There were no differences in the numbers of lymphocytes or NK cells with

2DG treatment alone, and cell loss was specifically associated with uncontrolled MCMV infection in 2DG-treated mice. Low NK cell numbers at this timepoint were not due to decreased NK cell proliferation, since the percentage of NK cells with BrdU incorporation at day 4 post-infection was similar with and without 2DG treatment (Figure 2.4D). T cell numbers and BrdU incorporation were also similar with 2DG treatment (Supplementary Figure 2.2D, E). Thus, 2DG did not affect in vivo receptor-mediated proliferation of Ly49H+ NK cells, but did inhibit in vitro IL-15- stimulated proliferation.

Loss of viral control with 2DG treatment was associated with significant mortality, with

80% of normally-resistant female B6 mice dying between days 5 and 6 post-infection (Figure

2.4E). 2DG treatment alone without infection had no effect on survival. Mortality was dependent on both the dosage of virus and drug, as lower doses of MCMV or 2DG did not lead to mortality

(Supplementary Figure 2.4). Together, these data show that glucose metabolism was crucial in the clearance of NK-dependent viral infection.

IL-15 can rescue MCMV-infected, 2DG-treated mice

In vitro, culture with IL-15 enhances NK cell metabolism, preferentially glycolysis, and confers some resistance to the effects of metabolic inhibition; specifically, IFN-γ production in

Figure 2.4. 2DG treatment in vivo confers MCMV susceptibility. Female WT C57BL/6 mice were infected with 1 x 105 PFU murine cytomegalovirus (MCMV) and treated daily with 2-deoxy-ᴅ-glucose (2DG). Viral copy number by quantitative PCR from the spleen and liver measured on (A) day 2 or (B) day 4 post-infection and is shown as copies of MCMV ie1 gene / copies of β-actin, x 1000 (n=3-8 individuals/group, 2 experiments, 1-way ANOVA on log-transformed data). C) Total numbers of lymphocytes and NK cells from the spleen or liver at day 4 post-infection. D) Representative flow cytometry and summary data showing the percentage of BrdU+ Ly49H+ NK cells (open) and Ly49H- NK cells (hatched), assessed 3 hours after BrdU injection. For (C) and (D), n=7-8/group, 2 experiments, 1- way ANOVA with Tukey’s multiple comparison test. E) Survival of uninfected mice compared to infected mice treated with 2DG or PBS (n=20/group, 4 experiments, log-rank Mantel-Cox test). ns = not significant (p≥0.05), *=p<0.05, ***=p<0.001.

glucose-free media was partially restored if NK cells were primed with IL-15 (10). Since we found that NK cell cytotoxicity was susceptible to glycolytic inhibition, we tested whether IL-15 could mitigate NK cell dependence on glucose. Thus, we tested whether IL-15 treatment could protect

2DG-treated mice from MCMV infection. Mice were treated with the human IL-15 superagonist complex ALT-803, which is composed of IL-15 with the N72D mutation and IL-15Ra fused to the Fc of IgG1. ALT-803 has been shown to activate NK cells and CD8+ T cells, and exhibited anti-tumor and anti-retroviral activities in various experimental animal models (48-57). Mice were treated with 5µg of ALT-803 at 3 and 1 days prior to infection, which increases the number and percentage of NK cells in the spleen by 2- to 10-fold on the day of infection (not shown). ALT-

803 treatment significantly increased the baseline and maximal glycolytic and oxidative capacity of NK cells, but did not alter the ratio of oxidative phosphorylation to glycolysis (OCR/ECAR), suggesting that ALT-803-activated NK cells had high metabolic potential and activity but did not especially favor glycolysis (Figure 2.5A). This is in contrast to IL-15 treatment in vitro, which we previously demonstrated decreased the OCR/ECAR ratio (10).

Treatment with ALT-803 rescued the deficit in NK cell target clearance at day 2 post- infection in MCMV-infected, 2DG-treated mice (Figure 2.5B). ALT-803 treatment increased the total number of splenic NK cells and Gzmb expression, but neither of these was affected by 2DG

(Figure 2.5C, D). There was no difference in splenic MCMV copy number at day 2 post-infection between 2DG-treated and PBS control mice that had been pretreated with ALT-803 (Figure 2.5E), suggesting enhanced viral control as compared to 2DG treatment alone (Figure 2.4A). However, liver MCMV copy number was increased with 2DG and ALT-803 treatment as compared to ALT-

803 alone (Figure 2.5E). The difference in liver MCMV burden appears to be due to decreased

Figure 2.5. NK activation with ALT-803 rescues MCMV susceptibility caused by 2DG treatment. A) Female wildtype C57BL/6 mice were treated with 5 μg ALT-803 twice, NK cells were purified from their spleens, and a glycolytic stress test was performed. Representative graph showing baseline, normal (+glucose), and maximal (+oligomycin) extracellular acidification rate (ECAR) in control vs. ALT-803-treated mice. Average ECAR, oxygen consumption rate (OCR), and OCR/ECAR ratio are shown after glucose addition (n=3 individuals/group, 2 separate experiments). B-E) ALT-803 treated mice were infected with 1 x 105 PFU murine cytomegalovirus (MCMV) and treated with 2-deoxy-ᴅ-glucose (2DG) daily. B) Clearance of m157-transgenic targets from the spleens of infected and uninfected mice +/- ALT-803 and 2DG on day 2 post-infection (n=5/group, representative of 2 experiments, 1- way ANOVA). C) The number and percentage of NK cells isolated from the spleens or livers of mice in (B). D) Mean fluorescence intensity of granzyme B in all NK cells from (B). E) Viral copy number measured by quantitative PCR from the spleens and livers collected in b), shown as copies of MCMV ie1 gene / copies of β-actin, x 1000. For (C- E), n=10/group, 2 experiments, 1-way ANOVA, using log-transformed data for (E). F) Mice were given an additional dose of ALT-803 2 days after infection and followed for 10 days. Survival of mice with no treatment (gray dashed), 2DG treatment (black), ALT-803 treatment (blue), or ALT-803 with 2DG (green) (n=10/group, 2 experiments, Log- rank Mantel-Cox test). ns = not significant (p≥0.05), **=p<0.01, ***=p<0.001.

MCMV copy number in the liver of ALT-803-treated mice when compared to those with MCMV infection alone (Figure 2.4A), and loss of this protective effect with 2DG treatment. To assess survival, mice were monitored and received an additional dose of ALT-803 at day 2 post-infection.

Mice receiving ALT-803 were completely protected from MCMV infection with 2DG treatment

(Figure 2.5F). Thus, treatment with ALT-803 rescued both the target clearance and survival deficits induced by 2DG.

IL-15 rescue is dependent on NK cells

To determine whether enhanced target clearance and survival with ALT-803 were NK- dependent, NK cells or CD8+ T cells, the two major cell types responsive to IL-15, were antibody- depleted (a-NK1.1 or a-CD8b). Mice depleted of NK or CD8+ cells, or control-treated animals, received ALT-803 prior to MCMV infection with 2DG treatment (Figure 2.6A, Supplementary

Figure 2.5A). NK depletion completely abrogated m157-transgenic target cell clearance and significantly increased MCMV copy number at day 2 post-infection (Figure 2.6B, C). Target cell clearance and MCMV copy number were not affected by CD8 depletion (Supplementary Figure

2.5B, C), suggesting that bystander activation of CD8+ memory T cells by ALT-803 did not affect control of acute MCMV infection in naïve mice. NK-depleted mice had significantly increased mortality (Figure 2.6D), while CD8+ T cell depletion had no effect on survival (Supplementary

Figure 2.5D), demonstrating that the survival of ALT-803–treated mice is dependent on NK cells.

NK depletion did not result in the 80-100% lethality of 2DG treatment alone, indicating that ALT-

803 may have some antiviral effects on NK1.1- cells, e.g. intra-epithelial lymphocytes in the gut that can respond to IL-15, or DCs that produce and display IL-15 (58). It is also possible that NK cells remained or developed after antibody depletion treatment due to IL-15’s strong survival and differentiation signaling; while there were no detectable NK cells in the blood or spleens of the

majority of mice (data not shown), the efficacy of depletion in other organs was not evaluated.

Figure 2.6. ALT-803 rescue of 2DG-treated mice requires NK cells, and ALT-803 priming can eliminate susceptibility to MCMV caused by mTOR inhibition. A) Mice were depleted of NK cells using α-NK1.1 at day -6 and day 0, or treated with IgG2a, given two doses of ALT-803, and infected with 1 x 105 PFU murine cytomegalovirus (MCMV). 2-deoxy-ᴅ-glucose (2DG) was given every 24 hours, and mice were euthanized on day 2 for assessment of (B) clearance of m157-transgenic targets from the spleen and (C) viral copy number from the spleen and liver measured by quantitative PCR and log-transformed (n=10 individuals/group, 2 separate experiments, t test). D) To assess susceptibility, an additional NK depletion was given at day +3 (red, n=30); control mice were treated with IgG2a (black, n=23). Mice were followed for 10 days (3 separate experiments, log-rank Mantel-Cox test). E) Mice were infected with 1 x 105 PFU MCMV, given daily i.p. injections of 1.5mg/kg rapamycin (green) or vehicle (black), and monitored for susceptibility (n=10/group, 1 experiment). F) ALT-803 was given to MCMV-infected, rapamycin- treated mice on the same schedule as for 2DG-treated mice, with doses at -3, -1, and +2 days relative to infection (n=10/group, 1 experiment). ns = not significant (p≥0.05), **=p<0.01, ***=p<0.001. mTOR inhibition leads to a similar survival defect with MCMV infection, which is rescued with

IL-15 priming

High-dose IL-15 signaling has been shown to lead to mTOR signaling, and mTOR activity is linked to the upregulation of glycolysis (11, 59, 60), so we tested whether inhibition of mTORC1 54

led to susceptibility to MCMV infection similar to 2DG treatment. Based on our finding that glycolytic inhibition impaired NK cell responses and led to mortality with MCMV, we hypothesized that mTOR signaling is required for NK cell protective effects during MCMV, as was suggested by a prior study demonstrating increased viral copy number in rapamycin-treated

MCMV-infected mice (59). Indeed, daily injection with 1.5 mg/kg rapamycin, an inhibitor of mTOR complex 1 (mTORC1), during MCMV infection led to susceptibility similar to 2DG treatment, with death 5-6 days post-infection (Figure 2.6E). Also similar to 2DG, IL-15 priming with ALT-803 was protective and completely abrogated the effects of rapamycin during MCMV infection (Figure 2.6F). This suggests that while IL-15 signaling has been described to upregulate mTOR and glycolysis, continued signaling through mTOR or maintenance of glycolysis is not required after IL-15-priming to sustain NK cell effector function.

Human NK cell cytotoxicity is also reduced by treatment with 2DG.

Relatively little is known about the effects of metabolism on human NK cell function, particularly cytotoxicity. To investigate whether human NK cells have a glycolytic requirement for cytotoxicity similar to the murine system, human NK cells were activated overnight with IL-

15 in the presence of varying doses of 2DG, then used in killing assays with K562 targets. Human

NK cells were more resistant to 2DG than murine NK cells, but killing of K562 target cells was significantly decreased when cells were primed with IL-15 in the presence of 20-40mM 2DG

(Figure 2.7A). Actin accumulation and MTOC polarization were measured and there were no significant defects in either (Figure 2.7B, C). Thus, both human and murine NK cell killing can be dampened by inhibition of glucose metabolism, though it is unclear if there are similar mechanisms regulating this effect.

Figure 2.7. 2DG decreases human NK cell cytotoxicity. Human NK cells were isolated from healthy donors, enriched using RosetteSep, and cryopreserved. Cells were later thawed and activated in 100ng/mL human IL-15 overnight in various concentrations of 2DG. A) NK cells were washed and co-cultured with K562 targets for 4 hours at various E:T ratios. 2 representative killing assays are shown (n=technical duplicate, representative of 6 donors across 4 experiments). B-C) NK cells from control and 20-40mM 2DG were cultured at a 2:1 ratio with K562 cells, and imaged after 40 minutes. B) Quantification of actin accumulation by area * intensity of synapse - (NK + target), (n=41-43 conjugates/group, 1 experiment). C) Quantification of MTOC-synapse distance in micrometers (n=67- 61conjugates/group, 2 experiments). Gray boxes show median, 25th, and 75th percentiles. Neither (B) nor (C) was significant by t test or Kolmogorov-Smirnov test (ns, p≥0.05). Treatment of CMV reactivation in a patient with an IL-15 superagonist

Our results demonstrating that IL-15 treatment leads to enhanced protection against

MCMV suggest that this cytokine may have therapeutic potential in patients. A 23-year-old CMV- seropositive male with dyskeratosis congenita and aplastic anemia received a matched sibling allogeneic hematopoietic cell transplant (HCT) from his CMV-seronegative brother with full engraftment by 9 months (see Methods for case presentation). This patient experienced five episodes of CMV reactivation post-transplant, which manifested as viremia without organ disease

(Figure 2.8A). Three episodes of CMV viremia within the 9 months post-HCT improved with antiviral therapy. A fourth episode at 18 months post-HCT resolved with CMVpp65-specific T cell adoptive transfer.

At 51 months post-HCT, the patient again experienced CMV reactivation. Because of recurrent CMV viremia, this patient was treated with compassionate-use ALT-803. Absolute numbers of peripheral blood NK cells were within normal limits at the time of ALT-803 administration. The patient received 4 weekly doses followed by 2 biweekly doses of ALT-803 at

6 µg/kg per dose. No other anti-viral therapy was given before or during the time of ALT-803 administration. Within 2 months, CMV viremia was below the limit of detection (Figure 2.8B), and CMV levels remained low for 7 weeks after ALT-803 therapy without additional antiviral therapy. Unfortunately, this patient experienced a recurrence of viremia when immunosuppressive therapy was required for treatment of post-transplant lymphoproliferative disorder.

NK cell function in this patient before treatment and 7 or 14 days after each dose of ALT-803 was evaluated. The patient had an elevated proportion of CD57+NKG2C+ NK cells before treatment

(~50%), consistent with the presence of CMV-adapted NK cells typically seen after CMV infection. This population of CMV-adapted NK cells was not altered by treatment with ALT-803

(data not shown). NK cell IFN-γ production after overnight stimulation of cryopreserved peripheral blood mononuclear cells (PBMC) with IL-12 + IL-18 increased following ALT-803 treatment (Figure 2.8C). Although the absolute IFN-γ response was lower in the patient’s cells than in PBMC from healthy controls treated with IL-12 + IL-18 (mean ± standard deviation: 44.69

± 2.96% IFN-γ-producing NK cells, n=8 controls), the increased responsiveness in this patient suggests that NK cell function was augmented by ALT-803 therapy.

Figure 2.8. ALT-803 treatment of CMV reactivation in a post-hematopoietic cell transplant patient. A) The patient’s course of cytomegalovirus (CMV) reactivation post-hematopoietic cell transplant (HCT), with periods of viremia detectable by quantitative PCR assay shown as orange bars. Post-transplant immunosuppression is shown as a blue bar. Years 1-4 post-HCT marked. B) During reactivation event #5, CMV levels responded to treatment with ALT-803 as indicated by arrows. Points below limit of detection (dashed line) at 137 copies/mL are rounded up. C) To assess IFN-γ production by the patient NK cells, PBMC were collected and cryopreserved before ALT-803 treatment or 7-14 days after the listed ALT-803 dose. PBMC were thawed and stimulated with 10ng/mL IL-12 and 100ng/mL IL-18 for 16 hours, then washed and plated for an additional 5 hours (n=4 technical replicates, statistical significance in a 1-way ANOVA compared to pre dose 1 shown above relevant column). *=p<0.05, ***=p<0.001. 2.4 Discussion

Here we examine the role of glycolysis in NK cell functions in vitro and in vivo. We found that the glucose metabolism inhibitor 2DG prevented IL-15-induced NK cell proliferation and priming of cytotoxicity in vitro. Using MCMV as an NK-dependent infection model, we demonstrate that administration of 2DG in vivo impaired NK-mediated target clearance and significantly increased susceptibility of mice to infection. These defects could be rescued in an

NK-dependent manner with the IL-15 superagonist ALT-803. Treatment with an mTORC1 inhibitor led to a similar susceptibility to infection, which was also rescued by ALT-803 priming, suggested that IL-15-primed NK cells have mTOR-independent sustained effector function. In a patient, treatment with ALT-803 also boosted NK cell function and was associated with clearance of CMV viremia, suggesting that targeting NK cell function in humans can enhance CMV control.

Our in vitro findings show that culture in 2DG blocks the ability of IL-15 to activate NK cell functions including proliferation, Gzmb production, and acquisition of enhanced cytotoxic functions. Cytotoxic defects were target cell-line specific, suggesting that that certain pathways of

NK cell activation may be able to overcome the dampened activation state of 2DG-treated NK cells, or that certain pathways require glycolytic metabolites for signaling. Possible mechanisms for 2DG’s effects on cytotoxicity include downregulation of Gzmb, leading to less cytotoxic granules, and altered actin accumulation at the immunological synapse. Actin accumulation is crucial for NK cell cytotoxicity, as evidenced by diseases associated with defects in actin accumulation such as Wiskott-Aldrich Syndrome, where NK cells fail to form conjugates or kill target cells (61). Although conjugation to target cells and MTOC polarization towards the synapse appeared normal in 2DG-treated NK cells, altered actin accumulation could be indicative of a less stable conjugation that could lead to slower or less effective killing, or could interfere with release of target cells and thus interfere with serial killing by NK cells. Interestingly, Wiskott-Aldrich NK cell defects can be reversed by IL-2 administration, as IL-2 activates an alternate actin polymerizing pathway (62, 63). Since IL-15 and IL-2 are closely related, this may provide a potential mechanism for IL-15 rescue of 2DG-induced deficits.

The observed target clearance defect caused by 2DG treatment resulted in higher viral replication, overwhelming infection, and mortality. MCMV infection has a steep LD50 curve (41), and loss of control of infection with 2DG treatment phenocopies mortality observed in mice depleted of Ly49H+ NK cells or total NK cells (24). Notably, not all NK cell functions were affected by 2DG, with intact IFN-g production in vitro (10) and in vivo, and proliferation in vivo.

Normal proliferation was somewhat surprising, as proliferating lymphocytes have been well- described to rely on aerobic glycolysis (64, 65). This may be due to non-IL-15 signals in MCMV-

infected mice, such as the presence of additional cytokines or activating receptor ligands, that could allow NK cells to proliferate despite glycolytic inhibition. The underlying mechanism for this glycolytic requirement for NK cell cytotoxicity can be partially explained by some of the findings here. 2DG decreases Gzmb upregulation in response to IL-2 + IL-12 (12) or IL-15 stimulation in vitro, and after poly(I:C) injection in vivo (12). While 2DG did not appear to inhibit

Gzmb upregulation in response to MCMV infection, this could be explained by different kinetics of Gzmb accumulation in response to different in vivo stimuli, or accumulation of Gzmb in 2DG- treated NK cells that are not actively degranulating and killing target cells. It is possible that killing defects in vivo are caused by changes in actin dynamics rather than Gzmb. Although actin accumulation defects did not alter conjugation in vitro, this may be due to the ready abundance and proximity of target cells. In vivo actin dynamics are crucial in NK cell trafficking and minor defects in the ability to polymerize actin could lead to decreased numbers of NK cells at sites of infection.

Although 2DG administration is confounded by systemic effects, prior model systems have demonstrated that 2DG does not lead to generalized susceptibility to infection, as 2DG treatment did not increase susceptibility to Listeria infection (66) and decreased virulence of other viruses, including HSV (67-69). Combined with our evidence that Ly49H-specific target clearance is decreased with 2DG and that deficient target clearance and survival can be rescued through IL-15, there is clear evidence of NK-specific defects with 2DG. However, NK cell-specific genetic knockouts of metabolic enzymes are necessary to interrogate the role of metabolism in NK cell functions in vivo.

While NK cells require glycolysis during viral infection for optimal killing, pre-treatment with IL-15 releases NK cells from their reliance on glucose metabolism, similar to what we

previously observed in vitro for IFN-γ (10). ALT-803 injection altered the metabolic rate of NK cells leading to increased glycolysis and oxidative phosphorylation, and preserved NK cell- mediated target clearance in the presence of 2DG. Mice pre-treated with ALT-803 were protected from MCMV infection when treated with either a glycolytic inhibitor or an inhibitor of mTORC1, suggesting that these pathways are not critical for IL-15-primed NK cell antiviral responses.

The underlying mechanism for IL-15 reversing NK cell dependence on glucose metabolism and mTOR activation is unknown. IL-15 is well described to promote NK cell differentiation, expansion, Gzmb expression, cytokine production, and cytotoxicity (32, 70). We have previously shown that IL-15 upregulates NK cell metabolic capacity in vitro, especially glycolysis (10), and here we demonstrate that ALT-803 significantly increases metabolism in vivo, suggesting that IL-

15 fundamentally alters cellular metabolic requirements. This altered metabolic state could lead to more flexible fuel usage by NK cells, allowing alternate fuels such as fatty acids, glutamine, or pyruvate to fuel NK cell function if glucose metabolism is limited. Alternately, we now have evidence that both cytokine-activated IFN-γ production (10) and clearance of target cells during

MCMV infection become less metabolically-dependent with IL-15 pre-treatment, suggesting that

IL-15 increases the range of stimuli and functions that are metabolically-independent. Recent studies have demonstrated a role for mTORC1 in IL-15 priming and NK cell responses to MCMV. mTORC1 activates glycolysis through HIF-1a, c-myc, and AKT signaling, in addition to orchestrating broader cell growth programs (71). High-dose IL-15 signaling has been shown to be dependent on mTORC1 (11, 59), which likely contributes to IL-15-induced upregulation of glycolysis. However, here we demonstrate for the first time that after IL-15 priming, mTORC1 signaling is not required for in vivo NK cell function, as shown by complete protection against an

NK-dependent infection, MCMV, in rapamycin-treated mice pre-treated with ALT-803. These

findings distinguish the effects of mTORC1 signaling on priming versus maintenance of NK cell effector function.

The concept of augmenting the antiviral effects of NK cells with IL-15 agonists may be particularly relevant in patients post-HCT and/or patients receiving immunosuppression targeting

T cell responses. Following HCT, NK cells are the first lymphocyte population to emerge from the new bone marrow and thus represent an important population to target for anti-leukemic and infection responses (72). In some cases, NK cells are reconstituted after HCT but are less functional for months, with reduced capacity for IFN-γ production and cytotoxicity (73, 74). Here, we show that treatment with an IL-15 superagonist significantly boosted NK cell antiviral responses in mice, and in one patient was associated with clearance of CMV infection in the absence of other antiviral therapy. Although limited conclusions can be drawn from a single patient, these studies suggest that clinical investigation of IL-15 agonists for treatment of refractory viral infections, particularly CMV, is warranted. Here, we demonstrate that IL-15 activation of NK cells decreases NK cell reliance on glucose metabolism and mTOR signaling in vivo, and may be a strategy to enhance NK cell function in metabolically challenging environments such as inflammatory and tumor microenvironments in patients.

In summary, our findings suggest that limiting glucose metabolism impairs NK cell killing and host antiviral responses to CMV, which can be rescued with IL-15-based priming. Such therapies may be particularly beneficial for patients after HCT, who may have normal numbers of

NK cells but poor NK cell function. Additional investigations of the mechanisms by which metabolism alters NK function will be important for targeting these innate immune lymphocytes for immunotherapy, and may provide insight into the metabolic regulation of killing in other cytotoxic lymphocytes.

2.5 Materials and Methods

Mice and MCMV infection

All mice were maintained in specific pathogen-free conditions, used between 6-12 weeks of age, and euthanized in accordance with institutional guidelines. C57BL/6 mice were purchased from Charles River Laboratories or from Jackson, and m157-transgenic mice on the C57BL/6 background (42) were bred in-house. 7-11 week old C57BL/6 female mice were injected i.p. with

1 x 105 PFU of Smith strain salivary gland MCMV (ATCC, prepared in BALB/c mice), or media as a control. 19mg of 2DG was given i.p. after infection and every 24 hours (~1g/kg per mouse), or PBS as a control. For proliferation assays, mice were injected with 2mg BrdU (BD Biosciences) i.p. 3 hours before harvest. For IL-15 superagonist treatment, 5µg ALT-803 (Altor BioScience) was injected i.p. 3 days and 1 day before MCMV infection, and 2 days after infection. For NK and

CD8 depletion, 100µg of a-NK1.1 (PK136, BioXcell), 250µg a-CD8b (53-5.8, BioXcell), or

100µg murine IgG2a (C1.18.4, BioXcell) was administered i.p. 6 days before MCMV infection, the day of infection, and 3 days after infection. Mice were bled 24h before infection or spleens were examined at time of euthanasia to confirm NK or CD8+ T cell depletion by loss of CD3-

NKp46+DX5+ or CD3+CD8a+ cells (not shown). Rapamycin was solubilized in 4% EtOH, 5%

Tween 80, 5% PEG400, and water, and 30µg (~1.5mg/kg) or an equivalent amount of vehicle was injected i.p. daily during MCMV infection.

In vitro assays

Primary splenocytes from female C57BL/6 mice were cultured in RPMI 1640 (Mediatech) supplemented with 10% FBS (Sigma-Aldrich), L-glutamine to a final concentration of 4mM

(Sigma), 50µM 2-ME (Sigma), and antibiotics (penicillin/streptomycin). NK cells were enriched from spleens using negative-selection magnetic bead kits (Miltenyi Biotec and StemCell

Technologies). Purity ranged from 55-90% at time of plating, and 77-99% after 72h activation.

For IL-15 activation, purified NK cells were cultured in 100ng/mL IL-15 (PeproTech) for 72 hours with or without 1mM 2DG. For proliferation assays, purified NK cells were stained with 1µM

CFSE before culture.

For mouse NK cell killing assays, NK cells were washed and plated in V-bottom plates at estimated 10:1, 5:1, and 1:1 NK:target ratios with 1 x 104 CellTrace-labeled target cells, including

RMA-S (from W. Yokoyama), Ba/F3-m157 (from W. Yokoyama (29)), and YAC-1 (from ATCC).

Cell lines tested negative for mycoplasma. Effector:target ratios were calculated based on NK cell purity by flow cytometry. After 4 hours, cells were stained with 7-AAD for 15-30 minutes before flow cytometry, and specific lysis was calculated by subtracting the percentage of 7-AAD+ target cells from wells without effectors. For mouse NK cell conjugation assays, NK cells were labeled with CFSE and Ba/F3-m157 target cells with CellTrace Violet (CTV). Cells were spun down at a

1:1 NK:target ratio, incubated in 37C for 15, 30, or 120min, and then vortexed gently and fixed in

1% paraformaldehyde. Percent conjugated NK cells was calculated by calculating the percentage of CFSE+CTV+ cells compared to all CFSE+ cells.

Human NK cells from healthy donors were enriched using RosetteSep (StemCell), For killing assays, previously cryopreserved NK cells were plated in 100ng/mL IL-15 (Miltenyi) and varying concentrations of 2DG for 22-26 hours, followed by incubation with labeled K562 targets for 4 hours at the indicated E:T ratios. Cells were then stained with 7-AAD and specific lysis calculated. For conjugation assays, NK cells were incubated with IL-15 and 2DG for 22-26 hours, then co-cultured with K562 targets at a 2:1 ratio for 40-50 minutes. To assess patient NK cell function, cryopreserved PBMC from serial timepoints were thawed, incubated for 16 h with human

IL-12 (10 ng/mL) and IL-18 (100 ng/mL), washed, plated in replicates of 4 per sample, and incubated for an additional 5 hours prior to intracellular staining for IFN-γ protein.

Flow cytometry

Cell viability was measured by live/dead stain (Zombie Yellow, BioLegend), followed by blocking with anti-FcγRIII (2.4G2) prior to surface staining. Cells were fixed with 1% paraformaldehyde for surface analysis, or fixed with Cytofix/Cytoperm (BD Bioscience) for intracellular staining. For intranuclear staining, cells were fixed in Cytofix/Cytoperm and

Cytoperm Plus buffer (BD) and DNAse treated. Data was acquired on a Cytek-modified FACScan

(BD and Cytek) or LSRFortessa (BD); data was analyzed using FlowJo 7.6.5. Geometric mean fluorescence intensity and % proliferated were calculated by FlowJo. The following fluorochrome- conjugated antibodies were used: CD3ε (145-2C11, BD/BioLegend), CD4 (GK1.5, Leinco), CD8α

(53-6.7, BD), CD11b (M1/70, BD), CD11c (HL3, BD), CD19 (eBio1D3, eBiosciences), IFN-γ

(XMG1.2, BioLegend), Ly49H (3D10, BioLegend), NK1.1 (PK136, BD), NKp46 (29A1.4, BD),

Gzmb (GB12, Invitrogen/Life Technologies), and BrdU (BrdU kit, BD). CD107a was stained by incubating splenocytes with a-CD107a (1D4B, BD) and GolgiStop (monensin, BD) for 4 hours at

37˚C.

Confocal microscopy

Purified NK cells and Ba/F3-m157 or K562 targets were co-cultured at a 2:1 ratio, incubated for 20 minutes at 37C, applied to poly-L-lysine-coated coverslips, and incubated for an additional 20 minutes. Slides were fixed in Cytofix/Cytoperm (BD) + 0.1% Triton X-100 (Sigma), then stained with the following antibodies: Ly49H (3D10, BioLegend), Biotin-Tubulin (236-

10501, ThermoFischer), Perforin (dG9, BioLegend), Gzmb (GB11, BioLegend), Phalloidin

(A12380, ThermoFischer), and streptavidin secondary antibody (S11222, ThermoFischer). Images

were acquired on a Leica SP8 laser scanning confocal microscope equipped with a 100X 1.45 NA objective. Excitation was by tunable white light laser and emission was detected by HyD detector

(Leica). Data were acquired using Leica LASAF and analyzed in Volocity 6.1.1 (Perkin Elmer) and 6.3 (Quorum Technologies). Measurement of MTOC-synapse distance was single-blinded.

Actin accumulation measurements were unblinded, and consisted of the total area * average intensity of three 1µm squares spaced evenly or where there was sufficient actin staining. Actin accumulation was calculated by area * intensity of synapse – (NK + target) (75).

Extracellular flux

Extracellular flux assays were performed using a XF96 Analyzer (Seahorse Bioscience).

Purified splenic NK cells were cultured for 72 hours in 100ng/mL IL-15 with or without 1mM

2DG, or NK cells were purified from ALT-803- or control-treated mice. Cells were then washed and plated in at least duplicate for extracellular flux analysis of oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) as previously described (10, 76). Glycolysis stress tests were performed in glucose-free DMEM supplemented with L-glutamine, sodium pyruvate, and

5% dialyzed FBS. Drug concentrations during the assay were as follows: glucose 10mM, oligomycin 100nM, 2DG 100mM.

MCMV quantitative PCR

5mm3 tissue pieces were minced and placed in Gentra Puregene Cell Lysis Buffer

(Qiagen), and DNA extracted following the manufacturer’s recommended protocol and diluted to

20-50ng/ml. MCMV immediate-early gene 1 (ie1)-specific primers and probe were used to determine viral copy number as described (47), and normalized to b-actin (Actb). Reactions were run in duplicate in on a 7500 Fast Real-Time PCR instrument (Applied Biosystems) with an absolute standard curve. The limit of reliable detection using the standard was 100 copies of

MCMV, and all values below 100 were rounded up. Graphs represent [(copies MCMV ie1)/(copies

Actb)] x 1000. Statistics were performed on log-transformed data.

In vivo target clearance assay

Splenocytes were harvested from C57BL/6 mice and C57BL/6 m157-transgenic mice, stained with 10µM or 1µM CFSE, and mixed together in a ~1:1 ratio. 7-20 x 106 cells were injected by tail vein into mice that had been infected with MCMV two days prior and treated daily with

2DG or PBS. Mice were treated with 2DG, but not ALT-803, prior to the clearance assay. The remaining labeled cells in the spleen were analyzed 4 hours after transfer. Percent target clearance was calculated by comparing the ratio of m157-transgenic splenocytes (m157tg, targets) to WT splenocytes (non-targets) recovered to that originally injected: 1 – (%m157tg/%WT cells at 4 hr)/(%m157tg/%WT cells input) x 100.

Patient case presentation

A 23-year-old CMV-seropositive male with dyskeratosis congenita and severe aplastic anemia received a matched sibling allogeneic hematopoietic cell transplant (HCT) from his CMV- seronegative brother. Conditioning was reduced intensity (fludarabine 40 mg/m2/day x 4, cyclophosphamide 50 mg/kg x1, total body irradiation 2 Gy, and alemtuzumab 0.2 mg/kg/d x 5).

Graft source was peripheral blood (9.3 x 106 CD34+ cells/kg), and the donor was his healthy brother (CMV seronegative). Graft-versus-host disease (GVHD) prophylaxis was with cyclosporine and mycophenolate mofetil and viral prophylaxis with high-dose acyclovir (800 mg

5 times daily) for ~25 days. Engraftment was complete as documented on day +28, and immunosuppressive medications for GVHD prophylaxis were discontinued 9 months post- transplant. The patient developed no acute or chronic GVHD, relapse, or any major bacterial or fungal infection. However, the course of transplant was complicated by recurrent CMV

reactivation without organ disease. Within the first 100 days post-HCT, he had one CMV reactivation episode treated successfully with valganciclovir and the viremia resolved after 21 days. The second episode occurred 4 months after HCT and resolved after 3 months of treatment with valganciclovir followed by foscarnet. The third episode occurred 9 months post-HCT and resolved after 6 months of treatment using maribavir (clinical trial NCT01611974) followed by valganciclovir. The fourth episode occurred 18 months post-HCT and resolved after 3 months with treatment using CMVpp65-specific T cell adoptive transfer (3 doses; NCT02136797 sponsored by

Atara Biotherapeutics). This viral remission was durable and lasted for 31 months, at which time

(51 months post-HCT) the patient experienced another episode of reactivation. Mutation testing for antiviral resistance was negative.

Statistics

Statistics calculated in Prism 7 include mean, SEM, median, 25th/75th percentiles, student’s t-test, Log-rank Mantel-Cox test, ordinary one-way or two-way ANOVA. All treatment groups used in an experiment were included in the ANOVA analysis, including controls; in non-survival experiments, mice that appeared uninfected were excluded from analysis. Tukey’s test and Sidak’s test were used to correct for multiple comparisons, and adjusted p values are shown. The

Kolmogorov-Smirnov test was calculated in Statistical Analysis System (SAS Institute) v9.4. P values represent 2-tailed probability, a=0.05. Data are presented with data averaged from each mouse or experiment as a separate point, and/or as mean +/- SEM.

Animal study approval

All animal studies were performed at Washington University or Baylor College of

Medicine using an Animal Studies Committee (ASC) approved protocol, and conducted in accordance with ASC regulations.

Human study approval

Written informed consent was obtained from the patient presented here in agreement with the guidelines of the University of Minnesota Human Research Protection Program. FDA approval for compassionate use of ALT-803 was obtained (IND #132197).

2.6 Supplementary Figures

Supplementary Figure 2.1. Effects of 1mM 2DG culture on NK cell metabolism and killing. Purified NK cells were cultured in 100ng/mL IL-15 with 1mM 2-deoxy-ᴅ-glucose (2DG) for 72 hours. A) Percentage of cells alive by trypan blue staining and B) percentage of NK cells recovered after 72-hour culture out of total number of NK cells plated (n=11 individuals, 7 separate experiments, paired t test). C) Specific lysis of RMA-S and YAC-1 target cells after culture with IL-15-activated NK cells that had been cultured in normal media (black) or 1mM 2DG (green) (n=4 individuals, 1 experiment, 2-way ANOVA, error bars denote SEM). ns = not significant (p≥0.05), **=p<0.01.

Supplementary Figure 2.2. T cells upregulate Gzmb and proliferate, but are not affected by 2DG treatment. The mice from Figure 4 (C57BL/6 females infected with 1 x 105 PFU murine cytomegalovirus and treated with 1g/kg 2-deoxy-ᴅ-glucose daily) were assessed for T cell function. A) The proportion of CD19+ B cells, CD3+CD4+ T cells, and CD3+CD8+ T cells in the spleen at day 2 post-infection (n=4/group, representative of 2 experiments). CD3+NK1.1- cells were assessed using flow cytometry for (B) degranulation without targets by CD107a staining and (C) Gzmb expression. D) Proportion of CD19+ B cells, CD3+CD4+ T cells, and CD3+CD8+ T cells in the spleen at day 4 post- infection (n=3-4/group, representative of 2 experiments). C) BrdU incorporation 3 hours after injection (n=7-8 individuals/group, 2 experiments) on day 4 post-infection. No significant differences were found (p≥0.05) by 1-way ANOVA.

Supplementary Figure 2.3. IFN-γ production by NK cells during MCMV infection is not inhibited by 2DG. Mice were infected with 1 x 105 PFU murine cytomegalovirus (MCMV) and injected with 1g/kg 2-deoxy-ᴅ-glucose (2DG) or PBS. Mice were sacked at 36 hours post-infection for flow cytometry (n=4-5 individuals/group, 2 separate experiments). Error bars show SEM. ns = not significant.

Supplementary Figure 2.4. MCMV susceptibility is dependent on the dose of 2DG and the inoculum of MCMV. 8-week-old female C57BL/6 mice were infected with murine cytomegalovirus (MCMV) and injected daily with 2- deoxy-ᴅ-glucose (2DG). Either the MCMV inoculum or 2DG dosage was varied. Survival is shown for 10 days, with the same mice for 1 x 105 pfu MCMV + 1000mg/kg 2DG (black) shown in both graphs (n=5/group, 1 experiment, log-rank Mantel Cox Test). ns = not significant (p≥0.05), *** = p<0.001.

Supplementary Figure 2.5. The effects of ALT-803 are not mediated through CD8+ T cells. A) Mice were depleted of CD8+ T cells using α-CD8b, or were treated with IgG2a as a control. Mice were then given two doses of 5µg ALT-803, infected with 1 x 105 PFU murine cytomegalovirus (MCMV), and injected with 1g/kg 2-deoxy-ᴅ- glucose (2DG) daily. B) Clearance of m157-transgenic targets from the spleen and (C) copy number of MCMV in the spleen and liver were assessed on day 2 (n=9-10 individuals/group, 1 experiment, copy number data log-transformed, analysis by t test). D) Mice were given an additional dose of ALT-803 and either α-CD8b (green, n=10) or IgG2a (black, n=5), and followed for 14 days (1 experiment, log-rank Mantel-Cox test). ns=not significant (p≥0.05).

2.7 References

1. Fox CJ, Hammerman PS, and Thompson CB. Fuel feeds function: energy metabolism and the T-cell response. Nat Rev Immunol. 2005;5(11):844-852. 2. Buck MD, O'Sullivan D, and Pearce EL. T cell metabolism drives immunity. J Exp Med. 2015;212(9):1345-1360. 3. Galvan-Pena S, and O'Neill LA. Metabolic reprograming in macrophage polarization. Front Immunol. 2014;5:420. 4. Pearce EJ, and Everts B. Dendritic cell metabolism. Nat Rev Immunol. 2015;15(1):18-29. 5. Lunt SY, and Vander Heiden MG. Aerobic glycolysis: meeting the metabolic requirements of cell proliferation. Annu Rev Cell Dev Biol. 2011;27(441-464). 6. Jha AK, Huang SC, Sergushichev A, et al. Network integration of parallel metabolic and transcriptional data reveals metabolic modules that regulate macrophage polarization. Immunity. 2015;42(3):419-430. 7. Everts B, Amiel E, Huang SC-C, et al. TLR-driven early glycolytic reprogramming via the kinases TBK1-IKK[epsiv] supports the anabolic demands of dendritic cell activation. Nat Immunol. 2014;15(4):323-332. 8. Chang CH, Curtis JD, Maggi LB, Jr., et al. Posttranscriptional control of T cell effector function by aerobic glycolysis. Cell. 2013;153(6):1239-1251. 9. De Rosa V, Galgani M, Porcellini A, et al. Glycolysis controls the induction of human regulatory T cells by modulating the expression of FOXP3 exon 2 splicing variants. Nat Immunol. 2015;16(11):1174-1184. 10. Keppel MP, Saucier N, Mah AY, Vogel TP, and Cooper MA. Activation-specific metabolic requirements for NK Cell IFN-gamma production. J Immunol. 2015;194(4):1954-1962. 11. Marcais A, Cherfils-Vicini J, Viant C, et al. The metabolic checkpoint kinase mTOR is essential for IL-15 signaling during the development and activation of NK cells. Nat Immunol. 2014;15(8):749-757. 12. Donnelly RP, Loftus RM, Keating SE, et al. mTORC1-dependent metabolic reprogramming is a prerequisite for NK cell effector function. J Immunol. 2014;193(9):4477-4484. 13. Keating SE, Zaiatz-Bittencourt V, Loftus RM, et al. Metabolic Reprogramming Supports IFN-γ Production by CD56bright NK Cells. J Immunol. 2016;196(6):2552-2560. 14. Mah AY, and Cooper MA. Metabolic Regulation of Natural Killer Cell IFN-gamma Production. Crit Rev Immunol. 2016;36(2):131-147. 15. Rolle A, and Brodin P. Immune Adaptation to Environmental Influence: The Case of NK Cells and HCMV. Trends Immunol. 2016;37(3):233-243. 16. López-Botet M, Muntasell A, and Vilches C. The CD94/NKG2C+ NK-cell subset on the edge of innate and adaptive immunity to human cytomegalovirus infection. Semin Immunol. 2014;26(2):145-151. 17. Wu Z, Sinzger C, Frascaroli G, et al. Human cytomegalovirus-induced NKG2C(hi) CD57(hi) natural killer cells are effectors dependent on humoral antiviral immunity. J Virol. 2013;87(13):7717-7725. 18. Zhang T, Scott JM, Hwang I, and Kim S. Cutting edge: antibody-dependent memory-like NK cells distinguished by FcRgamma deficiency. J Immunol. 2013;190(4):1402-1406.

19. Costa-Garcia M, Vera A, Moraru M, et al. Antibody-mediated response of NKG2Cbright NK cells against human cytomegalovirus. J Immunol. 2015;194(6):2715-2724. 20. Lee J, Zhang T, Hwang I, et al. Epigenetic modification and antibody-dependent expansion of memory-like NK cells in human cytomegalovirus-infected individuals. Immunity. 2015;42(3):431-442. 21. Kuijpers TW, Baars PA, Dantin C, et al. Human NK cells can control CMV infection in the absence of T cells. Blood. 2008;112(3):914. 22. Bukowski JF, Woda BA, Habu S, Okumura K, and Welsh RM. Natural killer cell depletion enhances virus synthesis and virus-induced hepatitis in vivo. J Immunol. 1983;131(3):1531-1538. 23. Scalzo AA, and et al. The effect of the Cmv-1 resistance gene, which is linked to the natural killer cell gene complex, is mediated by natural killer cells. J Immunol. 1992;149(2):581-589. 24. Brown MG, Dokun AO, Heusel JW, et al. Vital involvement of a natural killer cell activation receptor in resistance to viral infection. Science. 2001;292(5518):934-937. 25. Daniels KA, Devora G, Lai WC, et al. Murine Cytomegalovirus Is Regulated by a Discrete Subset of Natural Killer Cells Reactive with Monoclonal Antibody to Ly49h. J Exp Med. 2001;194(1):29-44. 26. Lee S-H, Girard S, Macina D, et al. Susceptibility to mouse cytomegalovirus is associated with deletion of an activating natural killer cell receptor of the C-type lectin superfamily. Nat Genet. 2001;28(1):42-45. 27. Orange JS, and Biron CA. Characterization of early IL-12, IFN-alphabeta, and TNF effects on antiviral state and NK cell responses during murine cytomegalovirus infection. J Immunol. 1996;156(12):4746-4756. 28. Dokun AO, Kim S, Smith HR, et al. Specific and nonspecific NK cell activation during virus infection. Nat Immunol. 2001;2(10):951-956. 29. Smith HR, Heusel JW, Mehta IK, et al. Recognition of a virus-encoded ligand by a natural killer cell activation receptor. Proc Natl Acad Sci U S A. 2002;99(13):8826-8831. 30. Arase H, Mocarski ES, Campbell AE, Hill AB, and Lanier LL. Direct Recognition of Cytomegalovirus by Activating and Inhibitory NK Cell Receptors. Science. 2002;296(5571):1323-1326. 31. French AR, Sjolin H, Kim S, et al. DAP12 signaling directly augments proproliferative cytokine stimulation of NK cells during viral infections. J Immunol. 2006;177(8):4981- 4990. 32. Fehniger TA, Cai SF, Cao X, et al. Acquisition of murine NK cell cytotoxicity requires the translation of a pre-existing pool of granzyme B and perforin mRNAs. Immunity. 2007;26(6):798-811. 33. Orr MT, Sun JC, Hesslein DGT, et al. Ly49H signaling through DAP10 is essential for optimal natural killer cell responses to mouse cytomegalovirus infection. J Exp Med. 2009;206(4):807-817. 34. Berry R, Ng N, Saunders PM, et al. Targeting of a natural killer cell receptor family by a viral immunoevasin. Nat Immunol. 2013;14(7):699-705. 35. Ljunggren HG, and Karre K. Host resistance directed selectively against H-2-deficient lymphoma variants. Analysis of the mechanism. J Exp Med. 1985;162(6):1745-1759.

36. Karre K, Ljunggren HG, Piontek G, and Kiessling R. Selective rejection of H-2-deficient lymphoma variants suggests alternative immune defence strategy. Nature. 1986;319(6055):675-678. 37. Diefenbach A, Jamieson AM, Liu SD, Shastri N, and Raulet DH. Ligands for the murine NKG2D receptor: expression by tumor cells and activation of NK cells and macrophages. Nat Immunol. 2000;1(2):119-126. 38. Jamieson AM, Diefenbach A, McMahon CW, et al. The role of the NKG2D immunoreceptor in immune cell activation and natural killing. Immunity. 2002;17(1):19- 29. 39. Guerra N, Tan YX, Joncker NT, et al. NKG2D-deficient mice are defective in tumor surveillance in models of spontaneous malignancy. Immunity. 2008;28(4):571-580. 40. Mace EM, Dongre P, Hsu HT, et al. Cell biological steps and checkpoints in accessing NK cell cytotoxicity. Immunol Cell Biol. 2014;92(3):245-255. 41. Geurs TL, Hill EB, Lippold DM, and French AR. Sex differences in murine susceptibility to systemic viral infections. J Autoimmun. 2012;38(2-3):J245-253. 42. Tripathy SK, Keyel PA, Yang L, et al. Continuous engagement of a self-specific activation receptor induces NK cell tolerance. J Exp Med. 2008;205(8):1829-1841. 43. Dalod M, Salazar-Mather TP, Malmgaard L, et al. Interferon alpha/beta and interleukin 12 responses to viral infections: pathways regulating dendritic cell cytokine expression in vivo. J Exp Med. 2002;195(4):517-528. 44. Tay CH, and Welsh RM. Distinct organ-dependent mechanisms for the control of murine cytomegalovirus infection by natural killer cells. J Virol. 1997;71(1):267-275. 45. Loh J, Chu DT, O'Guin AK, Yokoyama WM, and Virgin HWt. Natural killer cells utilize both perforin and gamma interferon to regulate murine cytomegalovirus infection in the spleen and liver. J Virol. 2005;79(1):661-667. 46. Fodil N, Langlais D, Moussa P, et al. Specific dysregulation of IFNgamma production by natural killer cells confers susceptibility to viral infection. PLoS Pathog. 2014;10(12):e1004511. 47. Parikh BA, Piersma SJ, Pak-Wittel MA, et al. Dual Requirement of Cytokine and Activation Receptor Triggering for Cytotoxic Control of Murine Cytomegalovirus by NK Cells. PLoS Pathog. 2015;11(12):e1005323. 48. Xu W, Jones M, Liu B, et al. Efficacy and mechanism-of-action of a novel superagonist interleukin-15: interleukin-15 receptor alphaSu/Fc fusion complex in syngeneic murine models of multiple myeloma. Cancer Res. 2013;73(10):3075-3086. 49. Gomes-Giacoia E, Miyake M, Goodison S, et al. Intravesical ALT-803 and BCG treatment reduces tumor burden in a carcinogen induced bladder cancer rat model; a role for cytokine production and NK cell expansion. PLoS ONE. 2014;9(6):e96705. 50. Rosario M, Liu B, Kong L, et al. The IL-15-Based ALT-803 Complex Enhances FcgammaRIIIa-Triggered NK Cell Responses and In Vivo Clearance of B Cell Lymphomas. Clin Cancer Res. 2016;22(3):596-608. 51. Kim PS, Kwilas AR, Xu W, et al. IL-15 superagonist/IL-15RalphaSushi-Fc fusion complex (IL-15SA/IL-15RalphaSu-Fc; ALT-803) markedly enhances specific subpopulations of NK and memory CD8+ T cells, and mediates potent anti-tumor activity against murine breast and colon carcinomas. Oncotarget. 2016;7(13):16130-16145.

52. Mathios D, Park CK, Marcus WD, et al. Therapeutic administration of IL-15 superagonist complex ALT-803 leads to long-term survival and durable antitumor immune response in a murine glioblastoma model. Int J Cancer. 2016;138(1):187-194. 53. Basher F, Jeng EK, Wong H, and Wu J. Cooperative therapeutic anti-tumor effect of IL- 15 agonist ALT-803 and co-targeting soluble NKG2D ligand sMIC. Oncotarget. 2016;7(1):814-830. 54. Allegrezza MJ, Rutkowski MR, Stephen TL, et al. IL15 Agonists Overcome the Immunosuppressive Effects of MEK Inhibitors. Cancer Res. 2016;76(9):2561-2572. 55. Felices M, Chu S, Kodal B, et al. IL-15 super-agonist (ALT-803) enhances natural killer (NK) cell function against ovarian cancer. Gynecol Oncol. 2017;145(3):453-461. 56. Seay K, Church C, Zheng JH, et al. In Vivo Activation of Human NK Cells by Treatment with an Interleukin-15 Superagonist Potently Inhibits Acute In Vivo HIV-1 Infection in Humanized Mice. J Virol. 2015;89(12):6264-6274. 57. Jones RB, Mueller S, O'Connor R, et al. A Subset of Latency-Reversing Agents Expose HIV-Infected Resting CD4+ T-Cells to Recognition by Cytotoxic T-Lymphocytes. PLoS Pathog. 2016;12(4):e1005545. 58. Schluns KS, Nowak EC, Cabrera-Hernandez A, et al. Distinct cell types control lymphoid subset development by means of IL-15 and IL-15 receptor alpha expression. Proc Natl Acad Sci U S A. 2004;101(15):5616-5621. 59. Nandagopal N, Ali AK, Komal AK, and Lee SH. The Critical Role of IL-15-PI3K- mTOR Pathway in Natural Killer Cell Effector Functions. Front Immunol. 2014;5:187. 60. Duvel K, Yecies JL, Menon S, et al. Activation of a metabolic gene regulatory network downstream of mTOR complex 1. Mol Cell. 2010;39(2):171-183. 61. Orange JS, Ramesh N, Remold-O'Donnell E, et al. Wiskott-Aldrich syndrome protein is required for NK cell cytotoxicity and colocalizes with actin to NK cell-activating immunologic synapses. Proc Natl Acad Sci U S A. 2002;99(17):11351-11356. 62. Gismondi A, Cifaldi L, Mazza C, et al. Impaired natural and CD16-mediated NK cell cytotoxicity in patients with WAS and XLT: ability of IL-2 to correct NK cell functional defect. Blood. 2004;104(2):436-443. 63. Orange JS, Roy-Ghanta S, Mace EM, et al. IL-2 induces a WAVE2-dependent pathway for actin reorganization that enables WASp-independent human NK cell function. Journal of Clinical Investigation. 2011;121(4):1535-1548. 64. Vander Heiden MG, Cantley LC, and Thompson CB. Understanding the Warburg effect: the metabolic requirements of cell proliferation. Science. 2009;324(5930):1029-1033. 65. Warburg O, Gawehn K, and Geissler AW. [Metabolism of leukocytes]. Z Naturforsch B. 1958;13b(8):515-516. 66. Miller ES, Bates RA, Koebel DA, Fuchs BB, and Sonnenfeld G. 2-deoxy-D-glucose- induced metabolic stress enhances resistance to Listeria monocytogenes infection in mice. Physiol Behav. 1998;65(3):535-543. 67. Kilbourne ED. Inhibition of Influenza Virus Multiplication with a Glucose Antimetabolite (2-deoxy-D-glucose). Nature. 1959;183(4656):271-272. 68. Schnitzer TJ, Hodes DS, Gerin J, Camargo E, and Chanock RM. Effect of 2-deoxy-D- glucose and glucosamine on the growth and functions of respiratory syncytial and parainfluenza 3 viruses. Virology. 1975;67(1):306-309. 69. Courtney RJ, Steiner SM, and Benyesh-Melnick M. Effects of 2-deoxy-D-glucose on herpes simplex virus replication. Virology. 1973;52(2):447-455. 76

70. Marcais A, Viel S, Grau M, et al. Regulation of mouse NK cell development and function by cytokines. Front Immunol. 2013;4:450. 71. Weichhart T, Hengstschlager M, and Linke M. Regulation of innate immune cell function by mTOR. Nat Rev Immunol. 2015;15(10):599-614. 72. Ullah MA, Hill GR, and Tey SK. Functional Reconstitution of Natural Killer Cells in Allogeneic Hematopoietic Stem Cell Transplantation. Front Immunol. 2016;7:144. 73. Foley B, Cooley S, Verneris MR, et al. NK cell education after allogeneic transplantation: dissociation between recovery of cytokine-producing and cytotoxic functions. Blood. 2011;118(10):2784-2792. 74. Pical-Izard C, Crocchiolo R, Granjeaud S, et al. Reconstitution of natural killer cells in HLA-matched HSCT after reduced-intensity conditioning: impact on clinical outcome. Biol Blood Marrow Transplant. 2015;21(3):429-439. 75. Banerjee PP, and Orange JS. Quantitative measurement of F-actin accumulation at the NK cell immunological synapse. J Immunol Methods. 2010;355(1-2):1-13. 76. Keppel MP, and Cooper MA. Assessment of NK Cell Metabolism. Methods Mol Biol. 2016;1441:27-42.

Chapter 3

Cox10 is Required for Expansion of Ly49H+ NK cells During Murine Cytomegalovirus Infection

3.1 Abstract

Studies in diverse cell types have rediscovered the importance of basic cellular metabolism in regulating immune responses. For example, glycolytic and oxidative metabolism are counter- regulated in T cells to develop inflammatory effectors vs. memory or regulatory cells, respectively.

NK cells, which are first responders for viral infection and malignant transformation, may be similarly regulated. We recently showed that NK cell cytotoxic activity during infection with murine cytomegalovirus (MCMV) required glucose metabolism, whereas other NK cell responses such as cytokine production did not. We hypothesized that oxidative metabolism would be required for some of these glycolysis-independent responses. To test this, we examined responses to MCMV in mice with an NK-specific deletion of Cox10, which is required for assembly of electron transport chain complex IV. These “Ncr1-Cox10D/D” mice had normal NK cell development and numbers, without evidence of metabolic defects at baseline. However, Ncr1-

Cox10 D/D mice were more susceptible to MCMV infection, with impaired expansion of MCMV- specific Ly49H+ NK cells and failure to generate “memory” NK cells. Similar defects in expansion were found when Cox10-deficient NK cells were proliferated in vitro with Ba/F3-m157 cells.

Decreased Ly49H-driven proliferation in vitro was associated with decreased levels of mitochondrial reactive oxygen species, without a significant increase in the rate of apoptosis of

Cox10-deficient NK cells. Thus, we find that Cox10 is required by NK cells to expand in response to Ly49H stimulation during MCMV infection. This is the first study directly examining the role of oxidative metabolism in NK cell functions in vivo.

3.2 Introduction

Work in multiple immune cell types has found that metabolic pathways produce signals that regulate immunological function, and vice versa. As a general principle, two of the most important metabolic pathways in a cell—glycolysis and oxidative phosphorylation (OXPHOS)— are counter-regulated to produce either inflammatory and proliferative functions, or memory and regulatory functions, respectively (1-4). However, oxidative metabolism is critical in many cellular functions outside of this generalization. OXPHOS uses fuel from various sources, including glycolysis, fatty acid b-oxidation, and the tricarboxylic acid cycle, to produce energy in the form of ATP for the cell. Additionally, OXPHOS is a major intracellular source of reactive oxygen species (ROS), which can assist in signal transduction or cause damage if uncontrolled (5-7). We are interested in understanding how natural killer (NK) cells, innate lymphocytes that are critical in the early response to viral infection and malignant transformation, are regulated by the balance of glycolytic and oxidative metabolism.

Our previous work has found that murine NK cells primarily use glucose-driven OXPHOS at rest, similarly to other naïve immune cell types (1, 3). In vitro, both murine and human NK cell production of IFN-g can be inhibited by oligomycin, an ATP synthase inhibitor (8, 9); however, murine NK cells are only susceptible to oligomycin if stimulated through NK activating receptors and not by inflammatory cytokines, indicating that different stimuli may have different metabolic requirements (8). Compared to these findings in vitro, relatively little work has examined the role of oxidative metabolism in regulating NK cell function in vivo. Using a model of glycolytic inhibition by injection of 2-deoxy-D-glucose in vivo, we found that even though aerobic glycolysis is required for inflammatory activation and proliferation of other cell types (1, 2), 2DG treatment did not affect cytokine production or proliferation during viral infection. Rather, glycolytic 80

impairment led to decreased NK cell cytotoxicity (10). Based on these findings, we speculate that

NK cell proliferation in response to viral infection might be dependent on oxidative metabolism rather than aerobic glycolysis. In order to test this hypothesis, we used an NK-specific model of genetic Cox10 deficiency.

COX10 is a protoheme farnesyltransferase, a subunit required for assembly of cytochrome c oxidase (COX), otherwise known as Complex IV of the mitochondrial electron transport chain

(ETC). COX accepts electrons from the other complexes and fixes them to O2, making it a rate- limiting step of OXPHOS (11). Additionally, its interaction with cytochrome c gives it an important role in regulating apoptosis. Defects in Complex IV, including Cox10 deficiency, are known to cause mitochondrial disease in humans, manifested as progressive degradation of neurons, muscle, kidneys, and/or liver (11-13). Similar findings appear in tissue-specific mouse models of Cox10 deficiency (14-16). Cox10-deficient T cells had higher rates of apoptosis after activation than wildtype T cells, leading to increased susceptibility to infection (17). Thus, COX10 appears to play a role in vivo in regulating immune cell function. We generated mice with NK- specific Cox10 deficiency that allowed us to interrogate the role of oxidative metabolism in a cell- specific manner in vivo.

In order to examine NK cell antiviral function, we used the well-characterized murine cytomegalovirus (MCMV) infection model. In the first 48 hours, systemic inflammatory cytokines activate NK cells, leading to their priming and production of interferon gamma (IFN-g) (18, 19).

NK cells bearing the Ly49H receptor (~50% in naïve mice) specifically respond to the presence of the MCMV-encoded protein m157 on infected cells, leading to enhanced killing of infected cells and preferential proliferation of the Ly49H+ population that peaks around day 7 post-infection

(18-21). As infection wanes, these Ly49H+ NK cells form a “memory” population with enhanced

functionality that can persist for months (22). Thus, MCMV infection provides a model to study multiple NK cell functions, including proliferation and NK cell “memory”.

In this study, we characterized a mouse with an NK-specific defect in oxidative metabolism, specifically loss of a subunit of cytochrome c oxidase (COX10). Cox10-deficient NK cells developed normally and have normal energy production at baseline. However, Cox10 was required for optimal expansion of the Ly49H+ subset in response to MCMV infection. In modeling the proliferation response using m157-expressing cells in vitro, we found that Cox10-deficient NK cells did not have significantly higher rates of apoptosis, but did have decreased production of mitochondrial ROS. Deficient proliferation of Cox10-deficient NK cells during MCMV infection in vivo was linked to increased susceptibility at higher doses of MCMV and failure to generate

“memory” NK cells.

3.3 Results

Cox10-deficient NK cells develop normally

In order to study the effects of oxidative metabolism and mitochondrial health on NK cell function, we generated an NK-specific model of Cox10 deletion (Ncr1-Cox10D/D). Cre recombinase was driven by the Ncr1 (NKp46) gene in mice, combined with a loxp-stop-loxp yellow fluorescent protein (YFP) reporter for Cre activity, and a floxed Cox10 allele (14). Cox10-deficient NK cells appeared at normal proportions and numbers in the blood, bone marrow, liver, lungs, lymph nodes, and spleen (Fig. 3.1A, Supplementary Fig. 3.1A) of adult mice. Additionally, the proportion of

NK cells that were YFP+ in Ncr1-Cox10D/D mice was normal, indicating that Cre induction (and therefore deletion of Cox10) did not lead to NK cell loss (Fig. 3.1B). We confirmed deletion of

Cox10 expression at the mRNA level in YFP+ NK cells by quantitative real-time PCR (Fig. 3.1C).

A) B o n e M a r r o w L iv e r L u n g L y m p h N o d e S p le e n B lo o d n s n s

1 .0 0 .4 n s 0 .8 n s 0 .0 2 5 n s 4 n s 8 0 0

)

6 d

0 0 .8 0 .0 2 0 o

1 0 .3 0 .6 3 6 0 0

(

0 .6 0 .0 1 5

s l

l 0 .2 0 .4 2 4 0 0

c 0 .4 0 .0 1 0

K N

N 0 .1 0 .2 1 2 0 0 #

0 .2 0 .0 0 5 #

0 .0 0 .0 0 .0 0 .0 0 0 0 0

/D /D /D /D /D /D T D T D T D T D T D T D W 0 W 0 W 0 W 0 W 0 W 0 1 1 1 1 1 1 x x x x x x o o o o o o C C C C C C B) S p le e n C) C o x 1 0 e x p r e s s io n

1 0 0 n s 2 .0 ) s l l 8 0 e e 1 .5 c g

n K 6 0 a N h (

c 1 .0

+ NK1.1 d

P 4 0 l o 81.6% F F Y

0 .5 2 0 %

0 0 .0

/D /D /D YFP (Cre reporter) T D T T D D W 0 + W - W 0 0 1 1 1 x P P x x o F F o o C Y Y + C - C P P F F Y Y B o n e M a r r o w L iv e r L u n g

D) 1 0 0 n s n s n s s l l 8 0 e

Maturity stages c 6 0 K N

f 4 0 o

2 3 2 0 %

0 1 2 3 4 1 2 3 4 1 2 3 4 1 4 L y m p h N o d e S p le e n B lo o d

CD27 1 0 0 n s n s n s s l l 8 0

CD11b e c 6 0 K N

f 4 0 o

2 0 %

0 1 2 3 4 1 2 3 4 1 2 3 4 E) NKp46 NKG2D CD49b CD122 KLRG1

0 0 00 00 0 2 3 4 5 2 3 4 5 22 33 44 55 22 33 44 55 2 3 4 5 0 10 10 10 10 0 10 10 10 10 0 10 10 10 10 0 10 10 10 10 DNAM-1 0 10NKG2A/C/E10 10 10 0 10 10Ly49C/I10 10 0 10 10 Ly49D10 10 Ly49H Ly49D NKG2D DX5 CD122Ly49H KLRG1 WT Cox10D/D

0 0 00 00 0 0 2 2 3 3 4 4 55 22 33 4 5 22 33 44 555 2 3 4 5 2 3 4 5 0 0 1010 1010 1010 1010 00 1010 1010 10 10 00 1010 1010 1010 101010 0 10 10 10 10 0 10 10 10 10

KLRG1KLRG1 NKG2ACENKG2ACE LY49C/ILY49C/INKG2D Ly49D Ly49H Figure 3.1. Cox10-deficient NK cells develop normally. Organs were collected from 8-12 week old mice of both sexes from Ncr1-Cox10D/D (orange square) and Ncr1-YFP reporter (WT, black circle) strains. Bone marrow (2 femur, 2 tibia) n=7 individuals/group, 3 separate experiments. Liver n=7, 3 exp. Lung n=8-9, 3 exp. Lymph node (average of axillary & inguinal) n=8-9, 3 exp. Spleen, n=10-11, 3 exp. Blood n=10, 4 exp. A) # of NK cells was calculated by %NK1.1+CD3- cells x # of cells per organ. B) The Ncr1 gene (NKp46) drives Cre expression, turning on a YFP reporter. Representative flow and summary data showing the induction of YFP (n=13/group, 6 separate experiments). C) NK cells were enriched from spleens and sorted by YFP expression. mRNA was extracted from cells and expression of Cox10 was analyzed by quantitative real-time PCR. Fold change by DDCt normalized to YFP+ WT levels in two separate experiments. D) Representative flow cytometry show maturity staging by CD11b x CD27 staining, and summary data from YFP+ NK cells from various organs (n=5-6, 2 exp). E) Representative histograms show expression of NK receptors in YFP+ NK cells. Statistics by unpaired t-test, error bars show SEM. ns, not significant (p≥0.05). 83

NK cell development in Ncr1-Cox10D/D mice was normal despite loss of the Cox10 gene.

Cox10-deficient NK cells showed normal maturity by CD11b and CD27 staining (Fig. 3.1D,

Supplementary Fig. 3.1B) (23), indicating that Cox10 is not required for development of mature

NK cells. In order to examine their reactivity, we assessed receptor expression on Cox10-deficient

NK cells. Normal levels were detected of the NK activating receptors NKp46, NKG2D,

NKG2C/E, Ly49D, and Ly49H, the inhibitory receptors Ly49C/I and NKG2A, and co-receptors

DNAM-1 and CD49b (DX5) (Fig. 3.1E, Supplementary Fig. 3.1C). Expression of CD122, the IL-

2/IL-15 receptor beta chain, was also the same between Cox10-deficient and control NK cells (Fig.

3.1E, S3.1C). Based on our panel, Cox10-deficient NK cells expressed NK receptors at the same level as wildtype NK cells, and were thus likely to recognize the same stimuli.

Cox10-deficient NK cells have normal mitochondrial function at baseline

Next, we examined the mitochondrial health and energy production capacity of Cox10- deficient NK cells at baseline. Transmission electron microscopy (TEM) showed that naïve Cox10- deficient NK cells had similar numbers of mitochondria per cell in each slice, which appeared to have normal “orthodox” ultrastructure (24). The folds of cristae were visible in most mitochondria, with the majority of the organelle’s volume taken up by the matrix (Fig. 3.2A). We did not find any significant defects in oxygen consumption by Cox10-deficient NK cells in naïve cells. There were no changes in baseline or maximal oxygen consumption rate as measured by extracellular flux analysis (Fig. 3.2B), although this analysis was done on bulk NK cells without selection for

Cre induction using the YFP reporter. These findings indicate that Cox10 deficiency does not detectably affect mitochondrial metabolism of naïve NK cells.

We also tested energy production and IFN-g production in Cox10-deficient NK cells responding to short-term stimulation, as we have previously shown that blockade of OXPHOS

using the ATP synthetase inhibitor oligomycin prevents NK cell activation and IFN-g production

(8). Unlike NK cells treatmed with oligomycin, Cox10-deficient NK cells produced normal levels of IFN-g with short-term stimulation through both activating receptors (NK1.1, Ly49H) and inflammatory cytokines (IL-12 + IL-18, IL-12 + IL-15) (Fig. 3.2C). This was associated with normal ATP production at baseline (IgG control) and after stimulation, although these measurements included ~30% YFP- (Cre-) NK cells (Fig. 3.2D). As resting NK cells primarily produce ATP through oxidative phosphorylation (8), this indicates that oxidative metabolism is intact enough for normal energy production in Cox10-deficient NK cells, and normal activation and IFN-g production after activation with either receptor and cytokine stimuli.

Figure 3.2. Cox10-deficient NK cells have normal function and energy production at baseline. A) NK cells were enriched and sorted from Ncr1-Cox10D/D and WT mice, then prepared for TEM. Representative images show similar mitochondrial morphology. B) Purified NK cells were analyzed by extracellular flux assay; technical duplicates (mean±SEM) from one of three mitochondrial stress test assays. Baseline oxygen consumption rate (OCR) in NK cells is similar to maximal OCR (after oligomycin & FCCP); black circle = wildtype, orange square = Cox10-deficient. C) Splenocytes were stimulated for 6 hours with various stimuli, and IFN-g was measured by intracellular flow cytometry. N=4-6 individuals/group, 3 separate experiments, except Ly49H (n=2-4, single exp). D) Purified NK cells were plated with various stimuli for 6h, before harvest for a luminescence-based ATP assay. Readings were corrected by the cell count (technical duplicates, 2 experiments plus single replicate from a third experiment.) Ns=not significant (p≥0.05) by t-test or 2-way ANOVA. 85

Cox10-deficient Ly49H+ NK cells expand less during MCMV infection in vivo

To examine the function of Cox10-deficient NK cells in vivo, Ncr1-Cox10D/D mice were infected with MCMV. In a wildtype mouse, the Ly49H+ population of NK cells expands from

~50% to ~90% of NK cells over the course of MCMV infection (Supplementary Fig. 3.1C, Fig.

3.3A). On day 5 post-infection, Ncr1-Cox10D/D mice had significantly fewer Ly49H+ NK cells than control mice, both in proportion and absolute number (Fig 3.3A, B). For example, in the spleens of female mice, 62.4±4.7% of NK cells were Ly49H+ in Ncr1-Cox10D/D mice compared to

81.5±3.8% in WT. Additionally, Ncr1-Cox10D/D mice had fewer overall NK cells in the spleen than wild type mice (Fig. 3.3C). This was primarily due to loss of Ly49H+ cells, as there was no significant difference in the number of Ly49H- cells (Fig. 3.3B).

We asked if this was due to a proliferative defect, or increased death of Cox10-deficient NK cells.

In both male and female mice, preferential Ly49H-driven proliferation measured by BrdU uptake during MCMV infection was lost, with Ly49H- NK cells taking up the same amount of BrdU as

Ly49H+ NK cells in Ncr1-Cox10D/D mice (Fig 3.3D). Only in male mice, which were infected at a higher dosage, was BrdU uptake decreased overall in Cox10-deficient NK cells compared to wildtype (Fig. 3.3D). The proportion of cells staining for Annexin V (apoptotic) and/or 7AAD

(dead) was not significantly elevated in proliferating Cox10-deficient NK cells (Fig. 3.3E) or in bulk splenocytes (Fig. 3.3F). Although dying and dead cells are cleared quickly in vivo, these preliminary results suggest that decreased expansion of Ly49H+ NK cells in the absence of Cox10 was more likely due to a proliferative defect rather than death of the proliferating cells.

In vitro model of Ly49H-driven proliferation also requires Cox10

In order to model MCMV infection in vitro, we used the pro-B cell Ba/F3 line expressing the m157 ligand recognized by Ly49H (25). Cox10-deficient and control NK cells were primed

Figure 3.3. Cox10-deficiency leads to decreased expansion of the Ly49H+ population of NK cells during MCMV infection. Mice were infected i.p. with MCMV and assessed at day 5 post-infection. Black circle, WT (Ncr1-YFP); orange square, Ncr1-Cox10D/D. A-C) Females infected at 5e4 PFU, n=8-9 individuals/group in 2 separate experiments; males at 7.5e4-1.25e5 PFU, n=7 in 2 exp. A) % of Ly49H+ cells amongst YFP+ (Cre-induced) NK cells was measured in the spleen and blood by flow cytometry. The number of B) Ly49H+ NK cells and C) NK cells per spleen. D) BrdU was given i.p. 3 hours before spleens were harvested for analysis at day 5 post-infection. The percentage of YFP+ NK cells that had incorporated BrdU was measured by intranuclear flow cytometry (females at 5e4 PFU, n=4-5 individuals/group; males at 7.5e4 PFU, n=4/group; single experiments). Rate of apoptosis as measured in E) Ly49H+, YFP+ splenic NK cells or F) all splenocytes by staining with Annexin V (AnnV) and/or 7AAD. Representative flow plots and summary data shown (females at 5.4-7.5e4 PFU, n=7-9/group, 2 experiments; males at 7.5e4 PFU, n=4/group, single experiment). Statistics by t-test or 2-way ANOVA, ns = p≥0.05, ** = p<0.01, *** = p<0.001.

with 10ng/mL IL-15 for two days to mimic inflammatory conditions early during infection, and then co-cultured with Ba/F3-m157 cells for three additional days to mimic viral proliferative stimuli. During Ba/F3-m157 co-culture, Ly49H+ NK cells that recognize the m157 ligand preferentially expand. As controls, cells were exposed to 10ng/mL or 100ng/mL IL-15 for five days for minimal and maximal proliferation. We found that Cox10-deficient Ly49H+ NK cells proliferated significantly less in response to Ba/F3-m157 stimulation than their wildtype counterparts (Fig. 3.4A). On the other hand, there was no pronounced difference in proliferation of cells that had been stimulated by 100ng/mL IL-15 (Fig 3.4B).

Based on these findings, we explored possible mechanisms for the observed proliferation defect. Cox10-deficient T cells had increased rates of apoptosis after activation through the T cell receptor (17). This did not appear to be the case in NK cells, as there were no significant differences between Cox10-deficient and wildtype NK cells in the proportion of apoptotic and dead by

Annexin and 7AAD staining (Fig. 3.4C). Cox10-deficient NK cells did have decreased levels of mitochondrial reactive oxygen species (ROS) as noted by MitoSOX staining after proliferation

(Fig. 3.4D). However, Cox10-deficient NK cells could proliferate in response to high-dose IL-15 despite decreased levels of ROS (Fig. 3.4D), suggesting that mitochondrial ROS may not be required for all NK cell proliferation. In summary, our in vitro MCMV model using two day IL-

15 priming + three day Ba/F3-m157 proliferation revealed proliferation defects in Cox10-deficient

NK cells that were not present in NK cells activated with short-term stimulation or high-dose IL-

15, and were associated with normal survival and decreased mitochondrial ROS levels.

Mice 1 & 4 72h Mice 2 & 5

10ng (filled) vs. 100ng/mL IL-15 72h All YFP+

Ly49H- (filled) vs. Ly49H+

120h

Mice 1 & 4 72h Mice 2 & 5

10ng (filled) vs. 100ng/mL IL-15 120h All YFP+ 10ng (filled) vs. 100ng/mL IL-15 72h All YFP+ 1 0 0 * A) d e n s BaF3-m157 t 8 0 a

Open: Ly49H+ r

e 6 0 f i

Filled: Ly49H- l Ly49H- (filled) W T + vs. Ly49H+ o L y 4 9 H

r 4 0 - Ly49H- (filled) P D /D L y 4 9 H vs.WT Ly49H+ 2 0 C o x 1 0 Cox10D/D % 0 L y 4 9 H + - + - CellTrace Violet n s B) 120h 1 0 0

IL-15 d e

t 8 0 a

Open: 100ng/mL r

e 6 0 f

Filled: 10ng/mL i n s 10ng (filled) vs. l o

100ng/mL IL-15 120h r 4 0 All YFP+ P 2 0 % 0 1 0 1 0 0 CellTrace Violet IL -1 5 (n g /m L ) Ly49H- (filled) vs. Ly49H+ C) D) 6 0 d ) 2 0 0 0 + a e D d I A

F s A 4 0 1 5 0 0 l l 7 W T M

r c D / D X

o C o x 1 0

K O 1 0 0 0 + S N V

2 0 o

t n f i n o

M 5 0 0 A ( %

0 0 IL -1 5 (n g /m L ) 1 0 1 0 1 0 0 IL -1 5 (n g /m L ) 1 0 1 0 1 0 0 B a /F 3 -m 1 5 7 - + - B a /F 3 -m 1 5 7 - + - Figure 3.4. Cox10 deficiency also causes NK cell proliferation defects with Ly49H stimulation in vitro. Splenocytes from and Ncr1-Cox10D/D and Ncr1-YFP (WT) mice were cultured in A) 10ng/mL IL-15 for 2 days, then BaF3-m157 cells were added at a 1:10 to 1:5 Ba/F3-m157:splenocyte ratio for the next 3 days, or B) IL-15 at 10ng/mL or 100ng/mL for 5 days. Proliferation after 5 days was measured by % of cells diluting CellTrace Violet. Representative flow cytometry and summary data shown (n=4-6 individuals/group in 3 separate experiments). C) Samples from B were stained for Annexin V (AnnV) or 7AAD to assess for apoptotic and dead cells, respectively; dead cells include cells staining with either dye. Representative data shows technical triplicates from one of two experiments. D) Samples from B, gated on YFP+ NK cells, were stained with MitoSOX for reactive oxygen species; representative data show technical triplicates from one of three experiments. ns=not significant (p≥0.05), *=p<0.05 by matched 2-way ANOVA. Cox10 deficiency leads to increased susceptibility to MCMV

Next, we asked whether the defect in expansion of Cox10-deficient Ly49H+ NK cells was associated in vivo with changes in viral control and mouse survival. Previous studies have shown that female mice are more susceptible to MCMV than male mice (26), so we infected each gender with different doses. We found that, in males infected with a low dose of MCMV (7.5e4 PFU,

D/D ~11% of male LD50)(26), Ncr1-Cox10 mice had significantly higher viral copy numbers in the spleen and liver (Fig. 3.5A). This difference was not significant in females infected with a low 89

dose of MCMV (5e4 PFU, ~19% of female LD50)(Fig. 3.5B). However, female mice infected with a higher dose (2e5 PFU, ~74% of LD50) did have a survival defect in a single experiment, with

33% mortality in Ncr1-Cox10D/D (Fig. 3.5C). These data suggest that loss of Ly49H-driven expansion of Cox10-deficient NK cells may cause increased susceptibility to infection in a dose- or sex-specific manner.

A) Males B) Females C)

S p le e n L iv e r S p le e n L iv e r 1 0 0

1 0 4 1 0 4 * * * 5 5 l 1 0 1 0 a

3 8 0 3 v

i 0

0 n s

v 1

* * 1

4 4 u x 3 1 0

3 x 1 0 6 0

1 0 1 0 s

b n s

t t

n c c 3 3

1 0 1 0 e

A 4 0

A / c 2 2 / W T

1 0 1 0 r

1 1

e D /D e

e 2 2

i i

P C o x 1 0

1 0 1 0 2 0

s s

e 1 1

e i

1 0 1 0 i 1 1 p

p 1 0 1 0 0

o o

C 0 2 4 6 8 1 0 C 1 0 0 1 0 0 1 0 0 1 0 0 D a y s p o s t in fe c tio n /D /D /D /D T D T D T D T D W 0 W 0 W 0 W 0 1 1 1 1 x x x x o o o o C C C C Figure 3.5. Increased MCMV susceptibility in Ncr1-Cox10D/D mice. (A-B) Mice were infected with i.p. with MCMV, and at day 5 post-infection spleen and liver were processed for DNA extraction. MCMV levels were assessed by qPCR for MCMV gene ie1, and normalized to b-actin levels. A) Male mice were infected with 7.5e4 PFU (n=4- 9/group, single experiment). Controls are both Ncr1-WT and littermate Cre- Cox10flox. B) Female mice infected with 5e4 PFU MCMV (n=8-9 individuals/group, 2 separate experiments). Copy number data was analyzed by t-test on log- transformed data, SEM shown. C) For susceptibility, female mice were infected with 2e5 PFU MCMV i.p. and followed for 10 days (n=6 individuals/group, single experiment). Formation of “memory” NK cells decreased after adoptive transfer of Cox10-deficient NK cells

Since we found that Ncr1-Cox10D/D mice are more susceptible to infection, we wanted to rule out the possibility that higher viral burdens or decreased cytokine production were contributing to the apparent defect in Ly49H+ NK cell expansion in Ncr1-Cox10D/D mice. To do this, we used an adoptive transfer model used to examine formation of Ly49H+ “memory” NK cells during MCMV infection (22). Cox10-deficient and control NK cells were co-transferred into

Ly49H-deficient Bxd8 mice that were then infected with MCMV. The peak of expansion of

Ly49H+ NK cells occurs at day 7 post-infection, after which they contract to form a “memory” population (22). Using this system, we tracked the expansion of the Ly49H+ population during

MCMV, using YFP to distinguish between Cox10-deficient (YFP+) and control (YFP-) NK cells

(Fig. 3.6A). We found that Cox10-deficient NK cells hardly proliferated in competition with control NK cells. Cox10-deficient and wildtype NK cells were adoptively transferred at a 1:1 ratio, but after 7 days of infection, Cox10-deficient NK cells comprised less than 10% of all Ly49H+ NK cells (Fig. 3.6B). In fact, virus-driven expansion was not detected in Cox10-deficient Ly49H+ NK cells (Fig. 3.6C). Thus, co-transfer shows that Cox10-deficient and WT NK cells have a cell- intrinsic defect in proliferation that is not due to cell-extrinsic factors such as differences in viral titer.

WT(YFP-) A) Ly49H-deficient B6.Bxd8 Bleed every 7 days Adoptive transfer 10-15 x 106 splenocytes/type Examine Ly49H+ population for YFP expression

Ncr1-Cox10D/D (YFP+) Infect with 2000pfu MCMV

B) C) C o m p e titiv e e x p a n s io n S p le e n L y 4 9 H + N K c e ll e x p a n s io n S p le e n

1 0 0 * * *

) 8

s l

W T l W T e K 8 0

D / D c D / D

C o x 1 0 6 C o x 1 0

K H 6 0 N U n in fe c te d

U n in fe c te d W T

y U n in fe c te d 4 L

4 0

% f

D / D (

o C o x 1 0

+ * H % 2 0 2 * * * * * * * * * 9 4 * * * 0 y L * * * * * * 0 7 1 4 2 1 2 8 0 0 7 1 4 2 1 2 8 D a y s p o s t-in fe c tio n D a y s p o s t-in fe c tio n

Figure 3.6. Lack of proliferation of Cox10-deficient NK cells leads to few MCMV-adapted “memory” NK cells. A) Ncr1-Cox10D/D splenocytes (>80% NK cells were YFP+) were mixed with splenocytes from littermate Cox10D/D (YFP-) mice and adoptively transferred to B6.Bxd8 mice, which lack Klra8, the Ly49H receptor. The next day, B6.Bxd8 mice were infected i.p. with 2000 PFU MCMV. Blood was acquired by cheek bleed every 7 days after infection, out to day 28, and the YFP+ to YFP- ratio of Ly49H+ NK cells was assessed. Mice were sacrificed at day 28, for which spleen and blood data were averaged. B) Competition assay showing the preferential expansion of wildtype (YFP-) NK cells and loss of Cox10-deficient (YFP+) NK cells. C) Overall expansion of the YFP+ and YFP- Ly49H+ populations compared to all NK cells in the Bxd8 mouse. N=5-8 individuals/group, 3 separate experiments, both sexes represented. Error bars show SEM. Statistics by 2-way ANOVA, with Ncr1-Cox10 D/D vs. Cox10 D/D shown, ns = not significant (p≥0.05), *** p<0.001.

In this model, a population of “memory” Ly49H+ NK cells is present that persists for over a month (27). In contrast, Cox10-deficient “memory” NK cells never expanded and were difficult to detect in the spleen and blood after about 21 days post-infection (Fig. 3.6C). This experiment in

B6.Bxd8 mice shows that the production of “memory” in response to MCMV infection was hindered in Cox10-deficient NK cells due to their inability to expand their Ly49H+ population.

3.4 Discussion

We have shown that NK cells lacking Cox10, which encodes a necessary subunit for assembly of cytochrome c oxidase (Complex IV of the ETC), had normal development and energy generation at baseline. However, Cox10-deficient NK cells had a defect in expanding in response to Ly49H stimulation, which we found both in vivo during MCMV infection and in vitro using

Ba/F3-m157 cells. This defect appears to be cell-intrinsic and associated with decreased proliferation and decreased ROS production rather than increased rates of apoptosis. On the other hand, proliferation of Cox10-deficient NK cells in response to high-dose IL-15 was normal.

Defective Ly49H+ expansion led to increased susceptibility to MCMV in Ncr1-Cox10D/D mice and failure to generate “memory” NK cells. This is the first study to describe the role of oxidative metabolism in NK cell antiviral responses.

Given that NK cells have been shown to be primarily reliant on oxidative phosphorylation for energy generation at baseline (8), it was interesting to note that NK cell development and energy production at baseline was normal. It may be possible that COX turnover is slow relative to the lifespan of each NK cell, so that sufficient levels of Complex IV are present to produce ATP in resting cells. It might not be until cells are sufficiently stimulated, or induced to have their existing mitochondria turned over, that the lack of Cox10 affects cellular function and remaining levels of COX are no longer sufficient. Mitophagy, or mitochondrial turnover, has been shown to 92

be increased during MCMV infection (28), which may trigger defects in Cox10-deficient NK cells.

We did find defects in proliferative ability and production of mitochondrial ROS when Cox10- deficient NK cells were activated. Another possibility is that NK cells rely on alternate pathways such as aerobic glycolysis to fuel their ATP production and function. This would not explain the fact that we did not find a consistent decrease in oxygen consumption in our Cox10-deficient NK cells, as COX is the major oxygen-consuming enzyme complex in the cell.

We found that activated Cox10-deficient NK cells produce decreased levels of mitochondrial ROS compared to wildtype NK cells, which is a possible mechanism for how Cox10 deficiency causes decreased NK cell proliferation. In T cells, ROS are known to assist in signal transduction (7, 29), so it is possible that signaling downstream of activating stimuli are affected by Cox10 deletion. While T cell activation leads to production of ROS from several sources, the major mitochondrial sources are superoxide radicals from Complexes I and III. These radicals form H2O2, which can sulfenylate proteins and alter their function, e.g. altering phosphorylation cascades (7). Although H2O2 can cross membranes, it is not entirely clear how much mitochondrial

ROS affects signaling at the plasma membrane. In fact, there is evidence suggesting that T-cell proliferation does not require intracellular ROS, though proliferation can be blocked by exogenous

ROS produced by phagocytes & neutrophils (7). Little is known about the role of ROS in NK cell signaling. In vitro, neither exogenous nor cell-intrinsic H2O2 affects IFN-g production during 6 hour stimulation (8). Its role during MCMV infection is also unclear. NK cells increase their mitochondrial ROS production by day 7 post-infection, but mitophagy is required for contraction of Ly49H+ NK cells into a “memory” pool, which suggests that high ROS levels are detrimental for maintenance of “memory” NK cells (28). In NK cells, we have evidence that COX10 deficiency leads to decreased production of mitochondrial ROS in activated NK cells, independent

of the activating stimulus. Further study is required to discern whether or not changes in mitochondrial ROS production affect signaling cascades downstream of Ly49H or other activating stimuli and lead to decreased proliferation.

It appears that Cox10 deficiency has different effects in T cells and in NK cells. Both

Cox10-deficient T cells and NK cells proliferated more slowly with receptor stimulation (CD3 +

CD28 or Ba/F3-m157 co-culture, respectively) (17). However, Cox10-deficient T cells had higher rates of apoptosis after activation (17), which was not noticed in Cox10-deficient NK cells. Thus, the involvement of mitochondrial signaling in apoptosis after activation appears to be different in

NK cells and T cells. In T cells, activation-induced cell death has been linked to mitochondrial

ROS production (5); this may not be the case in NK cells. Also, cytochrome c is a key regulator of apoptosis. Since COX plays a role in cytochrome c control, Cox10 deficiency might alter cytochrome c regulation and lead to induction of apoptosis. If this is occurring in T cells but not

NK cells, NK cells may have a way to avoid triggering this pathway.

Multiple studies have found that IL-15 fundamentally alters NK cell metabolism; here, we show that IL-15-treated NK cells overcome Cox10 deficiency. Cox10-deficient NK cells stimulated with high-dose IL-15 do not have a defect in proliferation compared to co-culture with

Ba/F3-m157 cells. This may be due to differences in signaling pathways downstream of cytokines compared to activating receptors, as we have previously shown these two activation pathways are differentially-regulated by OXPHOS in resting NK cells (8). However, high-dose IL-15 is also known to upregulate the rate of glycolysis and OXPHOS and change metabolic requirements for activation (8, 10, 29). The IL-15àPI3KàAKTàmTOR signaling pathway has been shown to be required for normal NK cell function, including proliferation during MCMV infection (29, 30).

Given the role of mTOR in upregulating glycolysis and anabolic cell metabolism, pro-glycolytic

signals downstream of IL-15 may allow cells to use aerobic glycolysis or other non-oxidative pathways to compensate for decreased COX function.

We show here that NK cell expression of Cox10 is required for optimal NK cell antiviral responses. MCMV infection can be controlled by cytokine production and NK cell killing before

Ly49H+ expansion begins (18), which we have yet to examine in Ncr1-Cox10D/D mice. However, our data suggest that at higher doses, deficient expansion of the Ly49H+ population is associated with increased viral copy number and increased susceptibility to infection. Additionally, we used adoptive transfer into Ly49H-deficeint B6.Bxd8 mice to show that loss of Cox10 led to failure to expand the Ly49H+ NK cell population when competing with wildtype Ly49H+ NK cells. This competition assay controls for possible effects of viral titer, killing, and cytokine production, indicating that the defect in expansion is cell-intrinsic. This assay also showed that loss of Cox10 led to failure to produce a MCMV-adaptive “memory” NK cell pool. While memory T cell generation is associated with increased mitochondrial mass and oxidative capacity (1), NK cell memory generation is associated with clearance of dysfunctional mitochondria (28), suggesting that the metabolisms underlying NK and T cell memory are very different. This model did not allow us to examine the role of Cox10 and oxidative metabolism in the contraction to form a memory pool and the maintenance of enhanced survival and function associated with memory; work using inducible Cre models will be necessary to ascertain whether memory NK cells are as highly oxidative as memory T cells are. However, our data suggest that OXPHOS is required for initial proliferation to generate memory in NK cells.

In conclusion, we have found that deficiency in the mitochondrial OXPHOS-related gene

Cox10 leads to deficient expansion of the Ly49H+ NK cell population during MCMV infection, increased susceptibility to infection, and failure to generate “memory” NK cells. At baseline,

Cox10-deficient NK cells appear normal, but have in deficient proliferation in vitro with Ly49H stimulation, though not with IL-15 stimulation. Cox10-deficient NK have decreased production of mitochondrial ROS, which we speculate may lead to altered signaling or mitochondrial function to cause decreased proliferation. These studies describe an initial foray into using genetic models of oxidative metabolism deficiency to investigate the metabolic regulation of NK cell functions, including antiviral responses, in vivo. We show that oxidative metabolism is required for normal

NK cell responses to MCMV infection.

3.5 Materials and Methods

Mice and MCMV infection

All mice were maintained in specific pathogen-free conditions, bred in-house, used between 8-12 weeks of age, and euthanized in accordance with institutional guidelines. Cox10 floxed mice (14) and YFP reporter mice, where YFP is driven by the Rosa26 promoter (31), were purchased from Jackson Laboratories. They were bred to Ncr1-Cre mice (32) to produce Ncr1-

Cox10D/D mice. All mice were on the C57BL/6 background. 8-12 week old mice were injected i.p. with 0.5 x 104 - 2 x 105 PFU of Smith strain salivary gland MCMV (ATCC, prepared in BALB/c mice), or media as a control. To assess proliferation during MCMV infection, mice were injected with 2mg BrdU (BD Biosciences) i.p. 3 hours before harvest. For memory experiments, with 20-

30 x 106 splenocytes were adoptively transferred into Ly49H-deficient Bxd8 mice, which were then infected with 2 x 103 PFU MCMV 18-24 hours later. Ly49H+ expansion was tracked through bleeding from the cheek vein.

NK purification and sorting

Murine NK cells were enriched from spleens using negative selection beads, using EasySep

Mouse NK Cell Isolation Kits (StemCell Technologies) according to manufacturer’s directions.

Sorting was accomplished on a FACSAria Fusion sorter (BD) using FACSDiva software (BD).

In vitro cell assays

Primary splenocytes from mice were cultured in RPMI 1640 (Mediatech) supplemented with 10% FBS (Sigma-Aldrich), L-glutamine to a final concentration of 4mM (Sigma), 50µM 2-

ME (Sigma), and antibiotics (penicillin/streptomycin). Ba/F3-m157 (from W. Yokoyama) were cultured in R10 supplemented with puromycin for selection of m157 expression and 5% IL-3- containing supernatant from X63 cells. NK cells were enriched from spleens using negative- selection magnetic bead kits (StemCell Technologies).

For short-term stimulation assays, NK cells were stimulated with 10 ng/ml murine IL-12

(PeproTech) and 50ng/mL murine IL-15 (MBL), or 10ng/mL IL-15 and 100 ng/ml IL-15. Ab stimulation was performed by culturing splenocytes or enriched NK cells in plates coated with 20

μg/ml purified anti-NK1.1 (PK136; BioXcell), 40ug/mL purified anti-Ly49H (3D10, BioXcell), or 40ug/mL IgG2A control (BioXcell). Brefeldin A was added to cells after 1 h of culture to inhibit secretion of IFN-γ. Intracellular IFN-γ production was assayed using BD Cytofix/Cytoperm (BD

Biosciences), as recommended by the manufacturer.

For proliferation assays, cells were stained with 2.5-5µM CellTrace Violet before culture.

Splenocytes were cultured in 10ng/mL murine IL-15 (PeproTech) for two days, and then Ba/F3- m157 cells were added at a 5:1 to 10:1 splenocyte:Ba/F3-m157 ratio for three additional days. For purified NK cells, cells were cultured in 20ng/mL IL-15 for two days, then Ba/F3-m157 cells were added at approximately a 1:1 ratio for three additional days. As controls, cells were cultured in 10-

20ng/mL or 100ng/mL IL-15 for five days.

Flow cytometry

Cell viability was measured by cell counts on a hemacytometer with trypan blue staining, live/dead stain (Zombie Yellow, BioLegend), or incubation with 7-aminoactinomycin D (7AAD,

Calbiochem) for 15-60 minutes. Cells were blocked with anti-FcγRIII (2.4G2) prior to surface staining. Cells were fixed with 1% paraformaldehyde for surface analysis, or fixed with

Cytofix/Cytoperm (BD Bioscience) for intracellular staining. For intranuclear staining, cells were fixed in Cytofix/Cytoperm and Cytoperm Plus buffer (BD) and DNAse treated. Data was acquired on a Cytek-modified FACScan (BD and Cytek) or LSRFortessa (BD); data was analyzed using

FlowJo 7.6.5. Geometric mean fluorescence intensity and % proliferated were calculated by

FlowJo. The following fluorochrome-conjugated antibodies were used: Annexin V (Apoptosis

Detection Kit I, BD), BrdU (BrdU kit, BD), CD3ε (145-2C11, BD/BioLegend), CD11b (M1/70,

BD), CD19 (eBio1D3, eBiosciences), CD49b (DX5, BioLegend), CD122 (TM-Beta1, BD),

DNAM-1 (10E5, BioLegend), anti-human Gzmb (GB12, Invitrogen/Life Technologies), IFN-γ

(XMG1.2, BioLegend), KLRG1 (2F1, Sirigen/BioLegend), Ly49C/I (BD), Ly49D (4E5,

BioLegend), Ly49H (3D10, BD/BioLegend), NK1.1 (PK136, BD), NKG2A/C/E (20d5, BD),

NKG2D (CX5, Biolegend) and NKp46 (29A1.4, BD).

Transmission Electron Microscopy

For ultrastructural analyasis, sorted NK cell pellets were fixed in fresh 2% paraformaldehyde/2.5% glutaraldehyde fixative (Polysciences) in 100mM sodium cacodylate buffer, pH 7.2 for 1 hour at room temperature. Samples were washed in sodium cacodylate buffer and postfixed in 1% osmium tetroxide (Polysciences) for 1 hour. Samples were then rinsed extensively in dH2O prior to en bloc staining with 1% aqueous uranyl acetate (Ted Pella) for 1 hr. Following several rinses in dH20, samples were dehydrated in a graded series of ethanol and

embedded in Eponate 12 resin (Ted Pella). Sections of 95 nm were cut with a Ultracut UCT ultramicrotome (Leica Microsystems), stained with uranyl acetate and lead citrate, and viewed on a JEOL 1200 EX transmission electron microscope (JEOL USA) equipped with an AMT 8 megapixel digital camera and AMT Image Capture Engine V602 software (Advanced Microscopy

Techniques).

ATP assay

Purified NK cells were plated at 5 x 104 to 1 x 105 cells/well in a 96-well Optilux plate and

ATP production was assessed using the ATPLite Luminescence Assay Sytem (Perkin Elmer) according to the manufacturer’s protocol. Briefly, cells were lysed and exposed to luciferin, then dark-adapted. Luminescence was read on a Synergy 2 Multi-Dection Microplate Reader (BioTek).

Values were obtained by comparison to a standard curve and normalized to pMol ATP per million cells.

Extracellular flux

Extracellular flux assays were performed using a XF96 Analyzer (Seahorse Bioscience).

Cell were washed and plated on poly-L-lysine-coated plates in at least duplicate for extracellular flux analysis of oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) as previously described (10, 76). Mitochondrial stress tests were performed in Seahorse XF media with glucose, L-glutamine, sodium pyruvate, and 1% FBS (pH 7.4). Drug concentrations during the assay were: 10µM oligomycin, 10µM FCCP, 10µM rotenone, and 10µM antimycin A.

Cox10 quantitative PCR

Enriched NK cells were sorted on YFP expression and lysed in TRIzol (Thermo Fischer), for serial extraction of RNA, DNA, and protein according to the manufacturer’s directions. cDNA from RNA was reverse transcribed using random hexamers (Promega). Cox10 primer/probe set

was from IDT, and b-actin primer/probe set was from Applied Biosystems. Fold change was calculated using the DDCt method.

MCMV quantitative PCR

5mm3 tissue pieces were minced and placed in Gentra Puregene Cell Lysis Buffer

(Qiagen), and DNA extracted following the manufacturer’s recommended protocol and diluted to

MCMV, and all values below 100 were rounded up. Graphs represent [(copies MCMV ie1)/(copies

Actb)] x 1000. Statistics were performed on log-transformed data.

Statistics

Statistics calculated in Prism 7 include mean, standard deviation, standard error of measurement (SEM), student’s t-test, Log-rank Mantel-Cox test, ordinary one-way or two-way

ANOVA. All treatment groups used in an experiment were included in the ANOVA analysis, including controls; in non-survival experiments, mice that appeared uninfected were excluded from analysis. Tukey’s test and Sidak’s test were used to correct for multiple comparisons, and adjusted p values are shown. P values represent 2-tailed probability, a=0.05. Data are presented with data averaged from each mouse or experiment as a separate point, or as mean±SEM.

Animal study approval

All animal studies were performed at Washington University using an Animal Studies

Committee (ASC) approved protocol, and conducted in accordance with ASC regulations.

100

3.6 Supplementary Figures

A) B o n e M a rr o w L iv e r L u n g L y m p h N o d e S p le e n B lo o d

n s

2 .5 n s 1 5 n s 1 5 n s 1 .0 n s 4 1 5 n s

)

s e t 2 .0 0 .8

y 3 c

o 1 0 1 0 1 0

h 1 .5 0 .6 p

y l

1 .0 0 .4

5 5 5

% (

K 0 .5 0 .2 N

0 .0 0 0 0 .0 0 0 )

s 1 .5 n s 1 0 n s 1 5 0 .8 n s 2 .5 n s 1 0 n s

e n s

y c

o 8 2 .0 8

h 0 .6

p 1 .0 1 0

m 6 1 .5 6

y l

0 .4 %

( 4 1 .0 4

0 .5 5 K

N 0 .2

2 0 .5 2

P F

Y 0 .0 0 0 0 .0 0 .0 0

) 0 .5 n s 0 .2 5 0 .6 n s 0 .0 1 5 n s 2 .5 n s 6 0 0 n s

6 n s

x 0 .4 0 .2 0 2 .0

(

o l

s 0 .4 0 .0 1 0 4 0 0

b l

0 .3 0 .1 5 1 .5

K /

0 .2 0 .1 0 1 .0

N K

0 .2 0 .0 0 5 2 0 0

N P

0 .1 0 .0 5 0 .5 #

# 0 .0 0 .0 0 0 .0 0 .0 0 0 0 .0 0

/D /D /D /D /D /D T D T D T D T D T D T D W 0 W 0 W 0 W 0 W 0 W 0 1 1 1 1 1 1 x x x x x x o o o o o o C C C C C C B) Bone Marrow Liver Lung Lymph Node Spleen Blood

Ncr1-YFP (WT)

Ncr1-Cox10D/D CD27 CD11b C) C D 4 9 b (D X 5 ) C D 1 2 2 D N A M 1 L y 4 9 C I

n s n s

4 0 0 8 0 0 2 0 0 0 n s 8 0 n s

s l

3 0 0 6 0 0 1 5 0 0 l 6 0

I I

F F

2 0 0 4 0 0 1 0 0 0 K 4 0

M M

2 0 0 1 0 0 5 0 0 % 2 0

0 0 0 0 r r g n d g N n d r g n d r g n d M e N M e n e M e N M e N n L e o B v L o n L e o n L e o B iv u e o i u le lo B iv u e o B iv u e o L l l L L L l l L l l L p B p B L p B L p B S S S S

L y 4 9 D L y 4 9 H N K G 2 A /C /E N K G 2 D

8 0 n s 7 0 n s 7 0 n s 1 5 0 0 n s

s s s

l l l

l l

l 6 0 6 0 6 0

e e

e 1 0 0 0

c c c

K K

K 4 0 5 0 5 0

N N N

5 0 0

% % % 2 0 4 0 4 0

0 3 0 3 0 0

r g n r g n r g n r M e N d M e N d M e N d M e g N n d n L e o n L e o n L e o n L e o B iv u e o B iv u e o B iv u e o B iv u e o L L l l L L l l L L l l L l l p B p B p B L p B S S S S Supplementary Figure 3.1. Normal development of Cox10-deficient NK cells. Organs were collected from 8-12 week old mice of both sexes. Bone marrow (2 femur, 2 tibia) n=7 individuals/group, 3 separate experiments. Liver n=7, 3 exp. Lung n=8-9, 3 exp. Lymph node (average of axillary & inguinal) n=8-9, 3 exp. Spleen, n=10-11, 3 exp. Blood n=10, 4 exp. Black circles = WT, orange squares = Cox10D/D. Statistics by unpaired t-test, error bars show SEM. ns=not significant (p≥0.05). A) % NK cells and % YFP+ of all NK cells were assessed by flow cytometry. # of YFP+ NK cells was calculated based on cell counts. B) Representative flow showing maturity by CD11b x CD27 staining in YFP+ NK cells from various organs. C) Summary graphs showing expression of NK receptors. 101

3.7 References

1. Buck, M. D., D. O'Sullivan, and E. L. Pearce. 2015. T cell metabolism drives immunity. J Exp Med 212: 1345-1360. 2. O'Neill, L. A., and E. J. Pearce. 2016. Immunometabolism governs dendritic cell and macrophage function. J Exp Med 213: 15-23. 3. Pearce, E. J., and B. Everts. 2015. Dendritic cell metabolism. Nat Rev Immunol 15: 18- 29. 4. Fox, C. J., P. S. Hammerman, and C. B. Thompson. 2005. Fuel feeds function: energy metabolism and the T-cell response. Nat Rev Immunol 5: 844-852. 5. Kaminski, M., M. Kiessling, D. Suss, P. H. Krammer, and K. Gulow. 2007. Novel role for mitochondria: protein kinase Ctheta-dependent oxidative signaling organelles in activation-induced T-cell death. Mol Cell Biol 27: 3625-3639. 6. Starkov, A. A. 2008. The role of mitochondria in reactive oxygen species metabolism and signaling. Ann N Y Acad Sci 1147: 37-52. 7. Belikov, A. V., B. Schraven, and L. Simeoni. 2015. T cells and reactive oxygen species. J Biomed Sci 22: 85. 8. Keppel, M. P., N. Saucier, A. Y. Mah, T. P. Vogel, and M. A. Cooper. 2015. Activation- specific metabolic requirements for NK Cell IFN-gamma production. J Immunol 194: 1954-1962. 9. Keating, S. E., V. Zaiatz-Bittencourt, R. M. Loftus, C. Keane, K. Brennan, D. K. Finlay, and C. M. Gardiner. 2016. Metabolic Reprogramming Supports IFN-γ Production by CD56bright NK Cells. J Immunol 196: 2552-2560. 10. Mah, A. Y., A. Rashidi, M. P. Keppel, N. Saucier, E. K. Moore, J. B. Alinger, S. K. Tripathy, S. K. Agarwal, E. K. Jeng, H. C. Wong, J. S. Miller, T. A. Fehniger, E. M. Mace, A. R. French, and M. A. Cooper. 2017. Glycolytic requirement for NK cell cytotoxicity and cytomegalovirus control. JCI Insight 2. 11. Diaz, F. 2010. Cytochrome c oxidase deficiency: patients and animal models. Biochim Biophys Acta 1802: 100-110. 12. Valnot, I., J. C. von Kleist-Retzow, A. Barrientos, M. Gorbatyuk, J. W. Taanman, B. Mehaye, P. Rustin, A. Tzagoloff, A. Munnich, and A. Rotig. 2000. A mutation in the human heme A:farnesyltransferase gene (COX10 ) causes cytochrome c oxidase deficiency. Hum Mol Genet 9: 1245-1249. 13. Antonicka, H., S. C. Leary, G. H. Guercin, J. N. Agar, R. Horvath, N. G. Kennaway, C. O. Harding, M. Jaksch, and E. A. Shoubridge. 2003. Mutations in COX10 result in a defect in mitochondrial heme A biosynthesis and account for multiple, early-onset clinical phenotypes associated with isolated COX deficiency. Hum Mol Genet 12: 2693- 2702. 14. Diaz, F., C. K. Thomas, S. Garcia, D. Hernandez, and C. T. Moraes. 2005. Mice lacking COX10 in skeletal muscle recapitulate the phenotype of progressive mitochondrial myopathies associated with cytochrome c oxidase deficiency. Hum Mol Genet 14: 2737- 2748. 15. Fukui, H., F. Diaz, S. Garcia, and C. T. Moraes. 2007. Cytochrome c oxidase deficiency in neurons decreases both oxidative stress and amyloid formation in a mouse model of Alzheimer's disease. Proc Natl Acad Sci U S A 104: 14163-14168.

102

16. Diaz, F., S. Garcia, D. Hernandez, A. Regev, A. Rebelo, J. Oca-Cossio, and C. T. Moraes. 2008. Pathophysiology and fate of hepatocytes in a mouse model of mitochondrial hepatopathies. Gut 57: 232-242. 17. Tarasenko, T. N., S. E. Pacheco, M. K. Koenig, J. Gomez-Rodriguez, S. M. Kapnick, F. Diaz, P. M. Zerfas, E. Barca, J. Sudderth, R. J. DeBerardinis, R. Covian, R. S. Balaban, S. DiMauro, and P. J. McGuire. 2017. Cytochrome c Oxidase Activity Is a Metabolic Checkpoint that Regulates Cell Fate Decisions During T Cell Activation and Differentiation. Cell Metab 25: 1254-1268 e1257. 18. Orange, J. S., and C. A. Biron. 1996. Characterization of Early IL-12, IFN-alphabeta, and TNF Effects on Antiviral State and NK Cell Responses During Murine Cytomegalovirus Infection. J Immunol 156: 4746-4756. 19. Dokun, A. O., S. Kim, H. R. Smith, H. S. Kang, D. T. Chu, and W. M. Yokoyama. 2001. Specific and nonspecific NK cell activation during virus infection. Nat Immunol 2: 951- 956 20. Daniels, K. A., G. Devora, W. C. Lai, C. L. O'Donnell, M. Bennett, and R. M. Welsh. 2001. Murine Cytomegalovirus Is Regulated by a Discrete Subset of Natural Killer Cells Reactive with Monoclonal Antibody to Ly49h. J Exp Med 194: 29-44. 21. Smith, H. R., J. W. Heusel, I. K. Mehta, S. Kim, B. G. Dorner, O. V. Naidenko, K. Iizuka, H. Furukawa, D. L. Beckman, J. T. Pingel, A. A. Scalzo, D. H. Fremont, and W. M. Yokoyama. 2002. Recognition of a virus-encoded ligand by a natural killer cell activation receptor. Proc Natl Acad Sci U S A 99: 8826-8831. 22. Sun, J. C., J. N. Beilke, and L. L. Lanier. 2009. Adaptive immune features of natural killer cells. Nature 457: 557-561. 23. Chiossone, L., J. Chaix, N. Fuseri, C. Roth, E. Vivier, and T. Walzer. 2009. Maturation of mouse NK cells is a 4-stage developmental program. Blood 113: 5488-5496. 24. Hackenbrock, C. R. 1966. Ultrastructural bases for metabolically linked mechanical activity in mitochondria. I. Reversible ultrastructural changes with change in metabolic steady state in isolated liver mitochondria. J Cell Biol 30: 269-297. 25. French, A. R., H. Sjolin, S. Kim, R. Koka, L. Yang, D. A. Young, C. Cerboni, E. Tomasello, A. Ma, E. Vivier, K. Karre, and W. M. Yokoyama. 2006. DAP12 signaling directly augments proproliferative cytokine stimulation of NK cells during viral infections. J Immunol 177: 4981-4990. 26. Geurs, T. L., E. B. Hill, D. M. Lippold, and A. R. French. 2012. Sex differences in murine susceptibility to systemic viral infections. Journal of autoimmunity 38: J245-253. 27. Franchina, D. G., C. Dostert, and D. Brenner. 2018. Reactive Oxygen Species: Involvement in T Cell Signaling and Metabolism. Trends Immunol 39: 489-502. 28. O'Sullivan, T. E., L. R. Johnson, H. H. Kang, and J. C. Sun. 2015. BNIP3- and BNIP3L- Mediated Mitophagy Promotes the Generation of Natural Killer Cell Memory. Immunity 43: 331-342. 29. Marcais, A., J. Cherfils-Vicini, C. Viant, S. Degouve, S. Viel, A. Fenis, J. Rabilloud, K. Mayol, A. Tavares, J. Bienvenu, Y. G. Gangloff, E. Gilson, E. Vivier, and T. Walzer. 2014. The metabolic checkpoint kinase mTOR is essential for IL-15 signaling during the development and activation of NK cells. Nat Immunol 15: 749-757. 30. Nandagopal, N., A. K. Ali, A. K. Komal, and S. H. Lee. 2014. The Critical Role of IL- 15-PI3K-mTOR Pathway in Natural Killer Cell Effector Functions. Front Immunol 5: 187. 103

31. Srinivas, S., T. Watanabe, C.-S. Lin, C. M. William, Y. Tanabe, T. M. Jessell, and F. Costantini. 2001. Cre reporter strains produced by targeted insertion of EYFP and ECFP into the ROSA26 locus. BMC Developmental Biology 1: 4-4. 32. Eckelhart, E., W. Warsch, E. Zebedin, O. Simma, D. Stoiber, T. Kolbe, T. Rülicke, M. Mueller, E. Casanova, and V. Sexl. 2011. A novel Ncr1-Cre mouse reveals the essential role of STAT5 for NK-cell survival and development. Blood 117: 1565-1573.

104

Chapter 4

Conclusions and Future Directions

105

In this dissertation, we examined the role that metabolic pathways—specifically glycolysis and oxidative phosphorylation—play in regulating NK cell function during viral infection. In the following section, we review data from the two models we used to manipulate metabolism during infection with murine cytomegalovirus (MCMV): injection of a glycolytic inhibitor drug, and NK cell-specific genetic deletion of Cox10, a subunit of mitochondrial electron transport chain

Complex IV. Then, we discuss how these findings relate to the greater body of work in the field of immunometabolism and NK cell regulation. Finally, we suggest new areas of study based on our findings.

4.1 Glycolytic Metabolism in NK Cell Antiviral Function

Previous work examining the role of glycolysis in NK cell function shows that while naïve murine NK cells are dependent on glucose for survival, glycolysis is required for short-term IFN- g production downstream of NK activating receptor (NKR) activation, but not cytokine activation

(1). Human NK cells also depend on glycolysis for IFN-g production in response to certain stimuli

(2). Most of this work has been done in vitro, so studies in vivo are necessary to understand how glycolysis regulates other NK cell functions in physiological contexts such as during viral infection. While mTOR has been shown to be required for NK cell function in response to MCMV infection (3, 4), it is not known whether mTOR is required because of its upregulation of glycolysis, or because of its other functions. We hypothesized that NK cells require upregulation of glycolysis in order to fuel their antiviral function during MCMV.

In order to examine the role of glycolysis in NK cell antiviral function, we used injection of the glucose metabolism inhibitor 2-deoxy-ᴅ-glucose (2DG) to non-specifically inhibit glucose uptake and metabolism during MCMV infection. Despite administration of 2DG to the entire

106

mouse, we found that NK cell production of IFN-g at 36 hours post-infection and proliferation of

Ly49H+ NK cells at 4 days post-infection was normal, indicating 2DG-treated NK cells were responsive to inflammatory cytokines and proliferated normally. However, we found that clearance of m157-expressing targets in vivo was inhibited by 2DG treatment. Decreased NK cell killing in 2DG-treated mice was associated with poor control of viral infection and led to death in the majority of mice around day 6 of infection. Using in vitro assays to explore potential mechanisms, we found that killing defects were present when murine and human NK cells were treated with 2DG during priming with IL-15. This was associated in vitro with altered actin accumulation at the immunological synapse and decreased Gzmb production in murine 2DG- treated NK cells, providing two potential mechanisms for deficient killing.

Treatment with IL-15 has been shown to alter NK cell function, including their metabolic requirements for function (1). We found that pre-treatment of mice with the recombinant IL-15 superagonist ALT-803 was able to maintain high levels of NK cell killing and protect mice from

2DG-induced susceptibility to MCMV in an NK-dependent manner. Additionally, ALT-803- treated mice were also resistant to the effects of mTOR inhibitor rapamycin, which lies downstream of IL-15 and upregulates glycolytic metabolism. Combining this data with work by others examining the role of mTOR in IL-15 signaling (4, 5), we speculate that mTOR signaling and upregulation of glycolysis are required for IL-15 priming, but after priming has occurred, NK cell function is independent of both mTOR and glycolysis.

In conclusion, 2DG treatment in vitro inhibited IL-15 priming of multiple NK cell functions, while 2DG treatment in vivo led to selective loss of NK cell cytotoxicity. Decreased killing was associated with altered actin dynamics at the immunological synapse, and was associated with increased susceptibility to infection. Pre-treatment with an IL-15 agonist before

107

2DG treatment restored normal cytotoxicity and resistance to infection. These findings indicate a key role for glycolysis in NK cell antiviral functions, with NK cell killing especially dependent on glucose metabolism. Interestingly, NK cell proliferation in vivo did not appear to require a transition to aerobic glycolysis as has been described in other immune cell types. This may indicate that NK cell proliferation is regulated much differently than in other immune cells, or that additional signals in vivo compensate for glycolytic inhibition. Finally, our results showing ALT-

803 enhances NK cell function and aids in clearance of CMV viremia suggests that treatment with

ALT-803 may help NK cells clear other refractory viral infections in patients.

4.2 Oxidative Metabolism in NK Cell Antiviral Function

Previous work examining the role of oxidative and mitochondrial metabolism in NK cells has been more limited than for glycolysis. Blockade of ATP synthesis using oligomycin was shown to inhibit IFN-g production in murine NK cells activated through NKRs but not cytokine receptors

(1), as well as in human NK cells activated by IL-12 + IL-15 (2). Also, activation of NK cells during MCMV was associated with increased mitochondrial reactive oxygen species (ROS) production, while successful formation of the Ly49H+ memory population required mitophagy (6).

As discussed above, we found that glycolysis was required for NK cell killing, but not cytokine production or proliferation, during MCMV infection. Based on these findings, we hypothesized that oxidative metabolism is required for NK cell antiviral functions that were not dependent on glycolysis, such as proliferation, as well as formation of “memory” NK cells.

To investigate the role of oxidative metabolism during MCMV infection, we used genetic deletion of the Complex IV assembly factor COX10 to inhibit oxidative metabolism in an NK- specific manner (Ncr1-Cox10D/D). Cox10-deficient NK cells had normal development, energy 108

production, and IFN-g production in response to short-term stimulation. However, we found that

Cox10-deficient NK cells had poor expansion of Ly49H+ cells that respond specifically to the m157 ligand during MCMV infection. An in vitro model showed Cox10-deficient NK cells had poor Ly49H-driven proliferation but normal IL-15-driven proliferation. The role of Cox10 in NK cell IFN-g production and killing during MCMV infection remains to be examined. However, a co-transfer experiment showed that the expansion defect in Cox10-deficient NK cells was cell- intrinsic. Ncr1-Cox10D/D mice appear to be more susceptible to infection than wildtype, although these effects may be dependent on the dose of virus and/or the sex of the mouse. Additionally, production the Ly49H+ “memory” NK cell pool, which persists with higher function for weeks to months after infection, was severely inhibited in Cox10-deficient NK cells.

How Cox10 deficiency leads to decreased Ly49H+ expansion is not yet clear. Preliminary data from both in vitro and in vivo experiments suggest that the expansion defect in Ly49H+

Cox10-deficient NK cells as not associated with increased rates of apoptosis after activation, which occurs in Cox10-deficient T cells. One possible mechanism is that decreased NK cell proliferation in Cox10-deficient NK cells is due to decreased production of ROS, which are known to contribute to signaling in T cells.

In conclusion, NK cell-specific deletion of Cox10 did not affect NK cell development, but did impair the ability of the cells to expand the Ly49H+ pool in response to both MCMV infection and in vitro co-culture with Ly49H-stimulating target cells. This expansion defect in Cox10- deficient NK cells was cell-intrinsic and not due to increased apoptosis of activated cells, but may be related to their decreased production of ROS. Finally, Cox10 deficiency in NK cells led to increased susceptibility to infection and poor formation of “memory” NK cells. These findings support our hypothesis that oxidative metabolism is required for normal NK cell antiviral function.

109

They also imply that NK cells may have flexible fuel requirements for development and short- term function, and suggest that ROS may play a role in regulating NK cell function, which has not been shown before.

4.3 NK Cells in the Immunometabolism Paradigm

NK cells are the closest, developmentally and functionally, to CD8+ T cells. My original hypothesis was that NK cells would be similar to CD8+ T cells in their metabolic regulation, especially for the periods of extended activity required for antiviral responses. CD8+ T cells have been well-described to require aerobic glycolysis (but not oxidative phosphorylation) for their activation and proliferation through their T cell receptor, and to require decreased glycolysis and increased oxidative metabolism during their transition to memory (7-10). To my surprise, we were able to inhibit NK cell proliferation in vivo by impeding oxidative metabolism but not glucose metabolism. In fact, it was NK cell cytotoxicity that was especially sensitive to glycolytic regulation, which was not expected as NK cell killing can occur without the extended activation

(≥18h) required for upregulating glycolysis in murine NK cells. While the discrepancy between results in T cells and NK cells as differences may reflect the differences in in vitro and in vivo inhibition (as we were able to block NK cell proliferation using 2DG in vitro), our findings nevertheless suggest that NK cells are regulated significantly differently at a metabolic level compared to CD8+ T cells. Our work examining the role of metabolic regulation of cytotoxicity in

NK cells also provides a basis for much-needed work examining metabolic regulation of CD8+ T cell cytotoxicity.

Another of my initial hypotheses was that both T and NK cell memory would be associated with upregulation of oxidative metabolism and mitochondrial content. In memory T cells, 110

enhanced longevity and increased functional capacity upon reactivation has been associated with increased mitochondrial mass and spare respiratory capacity. We found that Cox10-deficient NK cells form almost no memory population in a co-transfer model, although this is likely due to their inability to proliferate and may not reflect the changes undergone by wildtype NK cells to form memory. However, another group has shown that production of “memory” NK cells requires mitophagy, or the recycling of mitochondria that are over-activated and producing too many ROS

(6). They show that “memory” NK cells have less mitochondrial mass than naïve or MCMV- activated NK cells. If this is true—that NK cells require high oxidative metabolism to proliferate and decreased mitochondrial activity to form memory—then memory is fundamentally different in NK and T cells at a metabolic level. These differences could be explained by the differences in functional capacity of memory T cells and NK cells. While memory T cells last years, we are not sure that NK cell “memory” lasts beyond months (11). If this is the case, then memory T cells may rely more heavily on mitochondria for their longevity while the shorter-lived “memory” NK cells do not prioritize long-term survival.

It is also interesting to note that Cox10 deficiencies defects cause apoptosis in T cells but not NK cells (12). In Cox10-deficient T cells, increased apoptosis may be associated with decreased energy availability and/or destabilization of COX leading to cytochrome c-driven activation-induced cell death (13). It appears that NK cells may be avoiding these apoptotic signals through an unknown mechanism.

Altogether, we have found multiple ways in which NK cells seem to violate the typical immunometabolism paradigm, of which T cells are the prototype; that aerobic glycolysis is associated with proliferation, while oxidative metabolism is associated with memory formation.

111

4.4 Future Directions

Novel mechanisms for metabolic regulation

Our data suggest two potential mechanisms by which metabolism regulates NK cell function that may be novel to the field of immunometabolism. First, I hypothesize that glucose blockade inhibits NK cell killing by altering the stability of the immune synapse. Based on our findings that actin accumulation at the synapse between an NK cell and its target is more variable in 2DG-treated NK cells compared to untreated NK cells, I think that it is likely that some element of glucose metabolism is required for normal actin accumulation. For example, a specific metabolite may play a role in signaling at the immune synapse, or sufficient ATP must be generated to provide fuel and substrate for actin polymerization. Loss of actin motility can lead to failure to kill target cells, as granules may be unable traverse the network to the surface for degranulation. Another possibility is that changes in actin dynamics interfere with release of dying target cells by the NK cell, limiting serial killing behavior, as degranulation was normal but overall killing was not.

Second, I hypothesize that a mitochondrial signal is required for signaling through NKRs, but that this mechanism is redundant in cells that have been primed by IL-15. In vitro, we have previously shown that blockade of ATP production by oligomycin inhibits NK cell activation through NKRs (1), and here we show that COX10 deficiency affects proliferation in response to

Ly49H, but not IL-15. NKRs have different signaling pathways downstream than cytokine receptors, one of which may be metabolically regulated. Our current best guess for metabolites that may be involved in this signaling are ROS, which are required for signaling in T cells. Cox10- deficient NK cells have decreased production of mitochondrial ROS, but whether this observation is causative or symptomatic still remains to be seen. For example, decreased ROS might disrupt 112

signaling, leading to slower proliferation. This would possibly be the first example of ROS affecting signaling in NK cells. Alternately, decreased ROS might be a byproduct of decreased activity of the ETC, associated with but not causing decreased ATP production or interference with other pathways that are the true cause of the proliferation defect. Our work shows that proliferation is deficient in Cox10-deficient NK cells downstream of Ly49H, but I hypothesize that

NKRs signaling through the same signaling chains as Ly49H will also be affected.

Additionally, I support a previously stated hypothesis on how NK cell “memory” is regulated by metabolism. O’Sullivan et al. hypothesize that while mitochondrial function is upregulated during MCMV infection, increased production of ROS and high mitochondrial function is dangerous to NK cells, and mitophagy of these activated mitochondria is required for longevity of the “memory” NK cell population (6). As discussed before, this may define some of the functional differences between memory T cells and memory NK cells. An inducible Cre model is needed to test the hypothesis that NK cells will form a more stable pool of “memory” cells if they are induced to lose Cox10 function after Ly49H+ expansion has already occurred.

In conclusion, investigating the mechanisms of how inhibiting metabolic pathways cause defects in NK cell cytotoxicity and proliferation will be the next step toward understanding the complex interplay between metabolism and canonical immune activation pathways in NK cells, as well as unearth novel forms of metabolic regulation that might be relevant in other cell types.

The role of IL-15 in altering metabolism

I am very interested in how IL-15 is able to so comprehensively change the metabolism and metabolic requirements of NK cells. We found that IL-15 priming of cytotoxicity and IL-15- driven proliferation in vitro can be blocked by 2DG, indicating that glycolysis plays a crucial but unknown role in producing the results of IL-15 priming. However, after IL-15 priming has

113

occurred, NK cells are much more resilient to metabolic inhibition. We previously observed in vitro that IL-15 treatment for three days allowed NK cells to produce IFN-g in culture in glucose- free media or the ATP synthetase inhibitor oligomycin (1). In this dissertation, we find that ALT-

803 pretreatment in vivo maintained high levels of NK cell cytotoxicity despite treatment with

2DG. I hypothesize that IL-15 priming alters the epigenetic availability of metabolic regulators and metabolic enzymes to allow for greater metabolic flexibility. I also suspect that in order for

IL-15 signaling to alter metabolism, glucose metabolism is required, possibly for histone acetylation. Epigenetic changes downstream of IL-15 might allow multiple competing metabolic pathways to function at once, as well as upregulate nutrient uptake, allowing for metabolic flexibility. This would explain the observed increase of both aerobic glycolysis and OXPHOS in vitro and in vivo, as well as provide the resistance to single inhibitors that we see in IL-15-primed

NK cells.

Therapeutic applications

Two models that we use, MCMV and Cox10 deficiency, are especially relevant to human disease. MCMV and human CMV (HCMV) share many features, including production of a

“memory” subset with increased number and enhanced function. HCMV is a major problem in utero and in immunocompromised individuals, and a better understanding of MCMV antiviral responses may help us combat persistent HCMV infection. Cox10 deficiency is known to cause mitochondrial disease in humans. Although immunodeficiency is not a defining feature of mitochondrial disease, it has been noted that such patients do have compromised immune responses (12). The potential role of NK cell deficiency in patients with mitochondrial disease has not yet been investigated.

114

Several of our findings have the potential for therapeutic application in humans. First and foremost, the IL-15 superagonist ALT-803 has already been shown to help a CMV reactivation patient clear their viremia, and other groups are investigating its role in boosting NK cell antitumor immunity. Understanding how IL-15 and ALT-803 function to make NK cells more metabolically flexible and resistant to inhibition may also be useful for immunotherapy, as cancerous and infectious microenvironments are often nutrient-poor compared to conditions in circulation.

Especially in tumors, which compete with infiltrating lymphocytes for glucose, allowing NK cells to be more flexible with their fuel usage might lead to enhanced NK antitumor function. ALT-803 could be given in concert with drugs that affect metabolism in order to boost its function, ensuring

NK cells have the fuel necessary for IL-15 priming to occur. Second, understanding how Cox10- deficient NK cells are metabolically flexible to allow short-term function, and how their other functions are limited, might also help us understand and boost NK cell function in oxygen-limited environments such as within tumors.

We hope that by understanding the basic regulation of NK cell function by glycolysis and oxidative metabolism, we can contribute to both a better understanding of how diverse immune cells are regulated by metabolism, and find new ways to boost NK cells for therapy.

115

4.5 References

1. Keppel MP, Saucier N, Mah AY, Vogel TP, and Cooper MA. Activation-specific metabolic requirements for NK Cell IFN-gamma production. J Immunol. 2015;194(4):1954-1962. 2. Keating SE, Zaiatz-Bittencourt V, Loftus RM, et al. Metabolic Reprogramming Supports IFN-γ Production by CD56bright NK Cells. J Immunol. 2016;196(6):2552-2560. 3. Marcais A, Cherfils-Vicini J, Viant C, et al. The metabolic checkpoint kinase mTOR is essential for IL-15 signaling during the development and activation of NK cells. Nat Immunol. 2014;15(8):749-757. 4. Nandagopal N, Ali AK, Komal AK, and Lee SH. The Critical Role of IL-15-PI3K- mTOR Pathway in Natural Killer Cell Effector Functions. Front Immunol. 2014;5(187. 5. Donnelly RP, Loftus RM, Keating SE, et al. mTORC1-dependent metabolic reprogramming is a prerequisite for NK cell effector function. J Immunol. 2014;193(9):4477-4484. 6. O'Sullivan TE, Johnson LR, Kang HH, and Sun JC. BNIP3- and BNIP3L-Mediated Mitophagy Promotes the Generation of Natural Killer Cell Memory. Immunity. 2015;43(2):331-342. 7. Chang CH, Curtis JD, Maggi LB, Jr., et al. Posttranscriptional control of T cell effector function by aerobic glycolysis. Cell. 2013;153(6):1239-1251. 8. Sukumar M, Liu J, Ji Y, et al. Inhibiting glycolytic metabolism enhances CD8+ T cell memory and antitumor function. J Clin Invest. 2013;123(10):4479-4488. 9. Pearce EL, Walsh MC, Cejas PJ, et al. Enhancing CD8 T-cell memory by modulating fatty acid metabolism. Nature. 2009;460(7251):103-107. 10. Buck MD, O'Sullivan D, and Pearce EL. T cell metabolism drives immunity. J Exp Med. 2015;212(9):1345-1360. 11. Sun JC, Beilke JN, and Lanier LL. Adaptive immune features of natural killer cells. Nature. 2009;457(7229):557-561. 12. Tarasenko TN, Pacheco SE, Koenig MK, et al. Cytochrome c Oxidase Activity Is a Metabolic Checkpoint that Regulates Cell Fate Decisions During T Cell Activation and Differentiation. Cell Metab. 2017;25(6):1254-1268 e1257. 13. Huttemann M, Helling S, Sanderson TH, et al. Regulation of mitochondrial respiration and apoptosis through cell signaling: cytochrome c oxidase and cytochrome c in ischemia/reperfusion injury and inflammation. Biochim Biophys Acta. 2012;1817(4):598- 609.

116