Metabolic regulation of NK cell antiviral functions during cytomegalovirus infection

Abstract

Natural killer (NK) cells quickly mount cytotoxic responses, produce cytokines, and proliferate in response to infected or transformed cells. Moreover, they can develop memory, with enhanced effector responses following activation, in some cases with antigen-specificity. To optimally execute these functions, NK cells undergo metabolic reprogramming. Here, we discuss the interplay between metabolism and NK cell function in the context of viral infections. We review findings supporting metabolic regulation of NK cell effector functions, with a focus on NK cell antiviral infection in the context of cytomegalovirus in the mouse (MCMV) and human (HCMV).

Summary Sentence:

Here, we discuss the interplay between metabolism and NK cell function in the context of cytomegalovirus infection.

1. INTRODUCTION

Natural killer (NK) cells are innate lymphocytes important for the early response to viral infections and tumors through production of cytokines, in particular IFN-γ, and their cytotoxic capacity [1, 2]. NK cells recognize target cells through germline-encoded receptors that bind to pathogen-encoded or endogenous “stress” ligands, as well as activating Fc receptors for recognition of antibody-coated targets [2, 3]. Upon encountering target cells, NK cells have the capacity to rapidly secrete cytokines [1] and kill cells by inducing death receptor-mediated apoptosis or through release of their cytotoxic granules via directed exocytosis (reviewed in [4]).

While comparable in function to naïve CD8+ T cells (cytotoxic and IFN-γ producers), NK cells act in a shorter time span. In the context of cytomegalovirus (CMV) infection, NK cells carry out their cytotoxic function within minutes rather than the days required to mount an antigen-specific T cell response, and they are a primary source of IFN-γ prior to antigen-specific T cells responses [1, 5].

1.1. Antiviral functions of NK cells

The importance of NK cells is best highlighted by reports of patients with NK cell deficiencies, most frequently presenting with recurring herpesvirus infections with atypical and sometimes fatal consequences [6–9]. Herpesviruses establish persistent, life-long infection by entering a latency phase with virus reactivation typically during times of stress or acquired immunodeficiency [10]. Human cytomegalovirus (HCMV) is a herpesvirus that is highly prevalent worldwide (80–100% depending on the continent) [11], with one in three children becoming infected by age five, and one in two adults becoming infected by age 40 in the United States [12]. HCMV infection can have severe and sometimes deadly consequences in congenitally infected children and immunocompromised individuals [13, 14]. Although HCMV-infected heathy individuals are usually asymptomatic, viral shedding enables transmission to new hosts for months [15]. Moreover, HCMV infection can have a dramatic and long-lasting impact on the immune system of healthy individuals, outweighing heritable factors [16], and likely accelerating the progression to immunosenescent or aged immunophenotypes [17, 18]. Given the essential role NK cells play in preventing severe HCMV disease, as highlighted by individuals with NK cell deficiencies [6–9], extensive research has been conducted on NK cells in the context of cytomegalovirus infection.

Murine model of CMV infection

In the mouse model, murine CMV (MCMV) has been a valuable tool to interrogate the role of NK cells in viral control, including elucidation of the capacity of NK cells to display adaptive functions [19–21]. In the C57BL/6 mouse model, the timing and specific responses of NK cells have been elegantly elucidated by several groups, with early cytokine-driven NK cell activation and proliferation occurring during the first 24–48 hours of infection [22, 23], antigen-specific proliferation initiated between days 3–5 [24], and contraction of NK cells with long-lived adaptive NK responses between days 7–21 [20].

Shortly after MCMV infection, production of pro-inflammatory cytokines, including IL-12, IL-15 and IL-18, by antigen presenting cells (APCs) leads to activation of NK cells, which constitutively express receptors for these cytokines. In turn, NK cells produce IFN-γ, an immunomodulatory cytokine that further amplifies the immune response. IFN-γ is critical for MCMV control, with demonstrated contributions to dampening MCMV replication [25–29], clearing persistent viral replication [30, 31], and suppressing MCMV reactivation [30, 32]. In general, IFN-γ secretion can help recruit APCs to the infection site and promote effective antiviral immunity [33]. Therefore, early NK-derived IFN-γ is likely important for APC interaction with T cells, which later mount antigen-specific responses and form long-lived memory T cells capable of fast and efficient recall responses [34]. The first few days of MCMV infection are also characterized by high serum levels of other cytokines like IFN-α, IL-6, IL-12, IL-15, and IL-18, mainly produced by dendritic cells and stromal cells [35–38]. These cytokines can prime NK cell proliferation [39, 40], which, in the C57BL/6 model, is then specifically induced in NK cells bearing the activating receptor Ly49H, expressed on approximately 50% of NK cells [41]. BALB/c and 129/J mice do not express Ly49H, and NK cell memory has not been well-studied in mouse models beyond C57BL/6 [21]. Ly49H specifically recognizes the MCMV-encoded ligand m157 expressed on the surface of infected cells [19, 21], leading to killing of infected cells and preferential expansion of Ly49H+ NK cells 3–7 days post-infection, before the T cell antigen-specific response [24]. The absence of either Ly49H or m157 leaves mice highly susceptible to MCMV infection [19, 21, 42]. Because the process of acquiring antigen specificity is facilitated by somatic rearrangement of antigen receptor genes, a process which NK cells cannot undergo, it was long believed that NK cells lack antigen specificity and are not capable of developing classical immunologic memory. However, human and murine NK cells can develop memory to infection, with strong recall responses. In mice, most of the evidence is in the context of Ly49H+ NK cell responses to MCMV [20]. Contraction of the Ly49H+ NK cell population follows their expansion phase, forming a population of long-lived, self-renewing memory cells with enhanced recall response to MCMV and other stimuli [20, 43].

Virus-induced NK cell memory was first demonstrated in response to mouse CMV (MCMV) infection [20] (recently reviewed in [44]) and is best characterized in CMV infection models. In humans, HCMV infection has been associated with the expansion of a similar cell population of “adaptive” NK cells as the Ly49H+ in mice that can persist as memory cells and are marked by NKG2C and CD57 (see section 3.1 for additional details) [45, 46]. Elements associated with NK memory have also been reported in response to other viruses, including Friend virus [47], influenza [48–50], and vaccinia [51] in mice, and HCMV [46, 52, 53], Epstein-Barr virus [54–57], varicella-zoster virus [58] and hantavirus [59] in humans.

1.2. Immunometabolism

In recent years, there has been renewed interest in the contribution of metabolism to the regulation of immune cell function. Much progress has been made in uncovering the importance of metabolism for the development, differentiation, activation, and effector functions of lymphocytes, macrophages, and dendritic cells [60–63]. During these various processes, cells require energy as well as metabolites that can serve as building blocks and signaling molecules. For example, some metabolites such as acetyl-CoA exhibit non-canonical functions and help shape the epigenome by serving as co-factors or modulators of chromatin modifying enzymes [64].

To generate energy, cells mainly use glucose to fuel glycolysis and oxidative phosphorylation (OXPHOS). Glycolysis involves glucose conversion to pyruvate, high energy electrons carried by NADH, and 2 ATP molecules. Pyruvate can then be processed to lactate in the presence of oxygen, or it can be converted to acetyl-CoA for use in the TCA cycle and eventual production of high-energy electrons that feed into OXPHOS to produce large quantities of ATP. However, the journey from glucose to ATP production features several other pathways that branch from or feed into glycolysis or the TCA cycle, along with many entry and exit points for intermediates. Several metabolic regulators, including mTOR and c-Myc, can increase the rate of glycolysis by upregulating nutrient transporters and metabolic enzymes [65–67] or enhancing their activity [68]. Changes in metabolic pathways are thought to enable and support specialized functions of immune cell types. Many immunometabolism studies indicate glycolysis is critical for immune cell function, despite not being the most effective way to produce ATP. However, glycolysis facilitates rapid cell growth due to the resulting glycolytic intermediates feeding into pathways branching off glycolysis, such as the pentose phosphate pathway. Furthermore, several metabolic intermediates can have non-canonical roles that shape cell responses [69, 70]. For example, a prominent non-metabolic function of acetyl-CoA is to provide acetyl groups for post-translational modifications of proteins including histones [64, 71, 72]. Glycolytic enzyme GAPDH can act as an inhibitor of IFN-γ translation by binding to Ifng transcript in mouse CD4+ Th1 cells, and this block is released with increased rates of glycolysis triggered by CD4+ T cell activation [73].

Uncovering additional non-canonical functions of metabolic intermediates and deepening our understanding of the role metabolism can play in immune system function represent promising research avenues for better understanding and harnessing of the immune response.

2. NK CELL METABOLIC CHANGES IN RESPONSE TO ACTIVATION

NK cells, unlike their naïve CD8+ T cell counterparts that share many effector functions, require the ability to quickly respond to activation signals. NK cells rapidly produce inflammatory cytokines and cytotoxic machinery including perforin and granzymes, engage target cells in a directed manner for killing, and divide. These effector functions, which are critical to the antiviral response, require a coordinated program that includes changes to metabolic pathways in NK cells. In vitro studies have demonstrated pathways critical for these functional responses, while in vivo models have clarified metabolic pathways most relevant for antiviral functions. The majority of studies examining metabolic changes in NK cells has used mouse models and therefore constitute the bulk of the discussion here, followed by recent studies investigating human NK cell metabolism.

2.1. Lessons from in vitro studies

At steady state, NK cells have low levels of glycolysis and OXPHOS, yet this meets their energy demands during activation via short-term cytokine stimulation or activating receptor crosslinking [74–76]. In vitro, short-term (6 hours) activation via activating receptor crosslinking or stimulation with cytokines does not elicit significant changes in glycolysis or OXPHOS as measured by extracellular flux (Seahorse^™) assays. Despite the lack of changes, there is a requirement for glycolytic and OXPHOS flux for activating receptor-stimulated IFN-γ production by naïve, freshly enriched NK cells [76]. Activating murine NK cells through activating receptors NK1.1 and Ly49D, partnering with ITAM-bearing adaptors, results in impaired IFN-γ production if cells are deprived of glucose or if OXPHOS is inhibited with oligomycin (an inhibitor of ATP synthase) during stimulation [76]. Cytokine stimulation with IL-12 and IL-15 is partially dependent on OXPHOS, while stimulation with IL-12 and IL-18 does not have a similar metabolic requirement, suggesting there is a significant flexibility in the fuels used by NK cells to produce IFN-γ [76]. In contrast, effector CD8 T cells require glycolysis for IFN-γ production in response to both cytokine (IL-12+IL-18) and receptor stimulation [73, 77].

Several studies investigated the role of glycolysis, glycolytic intermediates, or glycolytic regulators in response to cytokine stimulation in vitro, making use of chemical inhibitors and genetic deletion models. A prominent metabolic regulator studied in this context is mTOR, a protein kinase that serves as the catalytic subunit of two protein complexes, mTORC1 and mTORC2. In NK cells, mTOR activity is upregulated downstream of IL-15 [75], and significantly increases metabolic rates, particularly glycolysis [75, 78]. Conditional models of Ncr1-specific deletion of Mtor, or a gene encoding a core protein from each complex, Rptor (mTORC1) or Rictor (mTORC2), demonstrated the importance of mTOR for NK cell maturation and homeostasis, as well as optimal responses to MCMV infection [75, 79, 80], discussed in section 2.2. Transcriptome analysis comparing Raptor- and Rictor-deficient NK cells revealed impaired OXPHOS pathway in Raptor-deficient NK cells, suggestive of subpar mitochondrial function [79]. This potentially altered metabolic state of Raptor-deficient cells coincided with diminished degranulation and cytokine production [79, 80]. mTORC1, rather than mTORC2, seems to be the main contributor to NK cell effector functions [79, 80], and is critical for cytokine-mediated priming of NK cells [75, 81]. In vitro, treatment with mTORC1 inhibitor rapamycin suggested that mTORC1 is required for IFN-γ production downstream of IL-2 and IL-12 [78] or activating receptor NKG2D, but not IL-12 and IL-18 or PMA and ionomycin [79]. Experiments with either mTOR- or Raptor-deficient NK cells confirmed the latter findings [75, 79, 80]. IL-2 and IL-12-stimulated expression of granzyme B in both mice and human NK cells also requires mTORC1 [75, 78].

2-deoxy-D-glucose (2-DG), a glucose analog which inhibits the first two enzymes of glycolysis, also impairs normal NK responses to IL-15 in vitro: proliferation, granzyme B production, and acquisition of enhanced cytotoxic functions [82]. Additionally, pharmacological activation of an isoform of the final enzyme in glycolysis, pyruvate kinase muscle 2 (PKM2), demonstrated the importance of PKM2-controled metabolism for in vitro IL-2+IL-12 cytokine-stimulated expression of granzyme B and production of cytokines IFN-γ, TNF-α, and IL-10 by murine NK cells [83].

Longer term cytokine stimulation (18 hours, IL-2+IL-12) increases glycolysis, OXPHOS (basal rate and maximal respiratory capacity), as well as mitochondrial mass [84]. The Finlay laboratory used isotope tracing to uncover that the citrate-malate shuttle, and not the full TCA cycle, is primarily used by these cytokine-stimulated NK cells to fuel OXPHOS and ATP synthesis [84]. This pathway involves mitochondrial uptake of glucose-derived pyruvate and conversion to citrate, utilizing part of the TCA cycle, with export to the cytosol in exchange for malate. Two key components of the citrate-malate shuttle (CMS), citrate-malate antiporter Slc25a1 and ATP-citrate lyase (ACLY), are sterol regulatory-element binding protein 1 (Srebp1) targets. In vitro pharmaceutical blockade and a genetic deletion model (Scap KO) that prevents Srebp activation demonstrated the importance of the CMS and Srebp signaling for sustaining the metabolic phenotype of cytokine-stimulated NK cells, as well as their proliferation [84].

In murine NK cells, cytokine (IL-2+IL-12) stimulation also upregulates c-Myc expression, which is important for NK cell metabolism and function [85]. In both T and NK cells, Myc and amino acid transporter Slc7a5 are needed for protein production [85, 86] and deleting either gene will phenocopy the deletion of the other. The Finlay laboratory demonstrated that cytokine-stimulated c-Myc- or Slc7a5-deficient NK cells have reduced metabolic rates, produce less IFN-γ and express less granzyme B compared to wild-type NK cells [85]. Furthermore, amino acid availability is necessary for sustained expression of c-Myc in activated lymphocytes, and IL-2+IL-12 stimulation increases glutamine transport into NK cells [85, 87]. The initial cytokine-stimulated upregulation of c-Myc requires mTORC1, but SLC7A5-mediated amino acid transport is critical for maintaining high levels of c-Myc [85]. A recent study uncovered that an obese state can lead to downregulation of c-Myc and mTORC1 signaling and diminished metabolic rates in response to IL-2 and IL-12 cytokine stimulation for 18 hours [88]. NK cells from obese mice or humans display impaired tumor killing in vitro, which was associated with failure to properly polarize their microtubule organizing center (MTOC) and lytic granules to the immune synapse [88].

In human NK cells, c-Myc binds to a distal upstream regulatory element of the KIR gene and promotes transcription, providing further evidence for c-Myc’s participation in directing NK cell function. c-Myc binding significantly increased with IL-15 stimulation and c-Myc overexpression in NK92 cells, which only express KIR2DL4, and led to de novo KIR acquisition [89].

Thus, in vitro studies suggest that several NK cell functions (cytokine production, polarization at the synapse) are energy-demanding processes. For some, like cytokine production, some metabolic flexibility exists. c-Myc, mTOR, and Srebp have emerged as important players in maintaining the metabolic phenotype and optimal function of activated NK cells. Conditions such as obesity that reduce the activity of these metabolic regulators also affect NK cell function, highlighting the need to understand how metabolic changes can affect function, and leverage that knowledge to maintain or enhance it.

Cytokines can change metabolic requirements for NK cell activation

Although freshly isolated NK cells require a second, metabolism-derived signal for IFN-γ production in response to activating receptor stimulation, this metabolic requirement can be almost completely abolished if the cells are cultured in high dose IL-15 (100 ng/ml) for 72 hours (priming) prior to stimulation. High-dose IL-15 induces proliferation, increased metabolic rates and enhanced cytotoxic capacity, but cells do not spontaneously produce IFN-γ [76]. These primed cells produce IFN-γ in response to receptor stimulation with OXPHOS inhibition and are less susceptible to glucose deprivation. Shorter culture times or a low (5–10 ng/ml) IL-15 dose did not produce similar results, suggesting this is a dose- and time-dependent effect. Proliferation does not seem to be required for metabolic independence of IFN-γ production, as both highly and poorly proliferated NK cells exhibited similar IFN-γ production with OXPHOS inhibition [76]. However, prolonged culture and expansion of NK cells with IL-15 can abrogate the metabolic flexibility observed with IL-15 priming, and even have negative consequences for NK cell metabolism and function. Donnelly et al. cultured NK cells for 7 days with an intermediate dose of IL-15 (50 ng/ml) prior to stimulation and reported impaired IFN-γ production in response to 18 hour IL-2+IL-12 in the presence of glycolytic or mTOR inhibitors [78]. Furthermore, Felices et al. showed that long term (9 days) exposure of human NK cells to high (10 ng/ml) doses of IL-15 in vitro results in reduced metabolic rates and impaired function [90]. Therefore, the effect of the exposure to high dose IL-15 is not only dependent on the duration of the exposure, but it can also become detrimental.

In summary, in vitro metabolic consequences of exposure to proinflammatory cytokines (IL-2, IL-12, IL-15) include upregulation of glycolysis, OXPHOS, mTOR activation [74, 75, 78–80], and Myc upregulation [85, 91]. Altogether, the findings suggest that the type, strength, or length of stimulation can alter IFN-γ production, including related metabolic requirements. Thus, it is important to note the status of the NK cells (primed vs. cytokine-expanded vs. naïve) when interpreting metabolic studies with NK cells.

2.2. In vivo metabolic requirements for NK cell antiviral function revealed by MCMV models

NK cell glycolytic requirements during MCMV

Acute viral infection with MCMV leads to increases in several metabolic parameters in splenic NK cells assessed in the first two days after infection, featuring increases in glycolysis and pentose phosphate pathway intermediates, as well as glycolysis rates [92]. Throughout the course of MCMV infection, NK cells also undergo mitochondrial changes, accumulating dysfunctional mitochondria and mitochondria-associated reactive oxygen species in the first seven days of infection [93]. As NK cells transition from an effector to a memory state, dysfunctional mitochondria are eliminated and mitochondria-associated reactive oxygen species gradually decrease [93].

The ability of NK cells to access metabolic pathways is critical for several aspects of NK antiviral function during MCMV infection. In vivo experiments administering 2-DG demonstrated the reliance of NK cells on glycolysis for optimal antiviral responses during MCMV infection, with the caveats that drug was administered systemically and that there is the potential for off-target effects [82]. Interestingly, 2-DG treatment did not alter most NK cell functions, including IFN-γ production, granzyme B expression, degranulation, or proliferation; however, mice had early mortality with MCMV infection in the timeframe of NK cell responses [82]. In vivo target clearance assays demonstrated that 2-DG administration impaired the ability of NK cells to kill m157-bearing targets, with in vitro correlates demonstrating altered adhesion of NK cells to targets at the immune synapse [82]. Pretreatment with an IL-15 superagonist rescued these defects, similar to the effect of IL-15 on metabolic requirements for IFN-γ in vitro. Furthermore, treatment with IL-15 of a post-hempatopoietic cell transplant patient experiencing recurring CMV viremia improved NK cell function and viral control [82].

Two genetic models investigated the role of glycolysis in NK cell response to MCMV. First, deletion of glycolytic enzyme PKM2 did not alter NK cell metabolism or antiviral function 4 days post-MCMV infection [83]. This is apparently due to compensatory activity of the other isoform of the enzyme, PKM1, which was increased in Pkm2^flox/flox Ncr1-Cre cells. PKM2-deficient NK cells exhibited metabolic activity similar to wild-type NK cells at baseline and in response to cytokine stimulation (18 hours, IL-2+IL-12): unaltered flux through glycolysis and OXPHOS; comparable levels of glycolytic intermediates, pentose phosphate pathway, and TCA cycle components; and similar pyruvate kinase activity measured by a direct biochemical enzymatic assay [83].

In a separate study, the Sun lab deleted the gene encoding lactate dehydrogenase A (LDHA) in NK cells to determine the dependence of NK antiviral responses on aerobic glycolysis. LDHA preferentially converts pyruvate to lactate and NADH to NAD+, helping maintain NAD+ levels, and sustaining glycolysis and the pentose phosphate pathway. The effects of NK-specific LDHA deficiency only become apparent in the context of a challenge such as a viral infection; the number and phenotype of NK cells at baseline was normal. Mice with constitutive Ldha deletion in NK cells (Ncr1^Cre X Ldha^flox/flox) succumb to MCMV infection more rapidly than WT mice, with reduced NK cell Ly49H expression at day 7 post-infection, impaired proliferation of Ly49H+ NK cells, and decreased activating receptor-stimulated IFN-γ production and m157-expresing target cell clearance. This underscores the necessity of intact aerobic glycolysis for mounting proper antiviral NK responses [92]. While this study suggests lactate is needed for optimal responses to MCMV infection, several other studies have reported negative effects of lactate/lactic acid on NK cell and effector CD8+ T cell functions such as IFN-γ production, and have linked this to impaired mitochondrial function [94–96]. Conflicting effects of lactate/lactic acid might be explained by timing and adaptations to varying lactate levels, or perhaps indirect effects due to acidification.

mTOR is now well-described to regulate NK cell metabolism and function, and activation of the mTOR complex 1 induced downstream of IL-15 signaling leads to metabolic reprogramming in NK cells, including a shift to glycolysis [75, 78]. Mice with constitutive deletion of mTOR in NK cells have significant impairment of NK cell differentiation with very low NK cell numbers, and subsequent impaired viral control during MCMV infection [75]. The few remaining mTOR-deficient NK cells have reduced degranulation and granzyme B expression 2 days post-MCMV infection, and reduced upregulation of nutrient receptors CD71 and CD98 in response to poly(I:C) injection, both cytokine-driven responses [75]. However, later activating receptor-dependent responses are less affected, suggesting the mTOR signal is not needed for Ly49H signaling [75]. IFN-γ secretion was unaffected by mTOR deficiency [75]. Most NK functional defects were comparable to responses seen in experiments using rapamycin [78, 81], and thus attributed to mTORC1. Inhibiting mTORC1 with rapamycin during MCMV infection leads to increased susceptibility to the infection [81, 82]. Interestingly, pre-treatment of mice with IL-15 restores viral control in an NK cell-dependent manner [82]. Therefore, the timing of mTOR signaling may be important: mTOR signaling and associated metabolic changes may be important during early MCMV infection, potentially for priming NK cell functions, but not for NK cell effector responses at later stages of infection. Fitting with this, rapamycin actually enhanced memory NK cell formation when administered after acute infection, potentially due to effects on autophagy [93]. Thus, the kinetics of when and how much mTORC1 signaling occurs are likely important for NK cell initial control of viral infection and generation of a pool of cells with adaptive features.

One way in which mTORC1 may exert its effects is through the endoplasmic reticulum stress sensor inositol-requiring enzyme 1 (IRE1) and its downstream substrate transcription factor X-box-binding protein 1 (XBP1) [91]. NK cells lacking either IRE1 or XBP1 display impaired proliferation in response to MCMV infection, and this effect was found to be mediated by c-Myc through experiments using genetic deletion and overexpression mouse models [91].

In addition to mTORC1 and c-Myc, transcription factor hypoxia-inducible factor-1α (HIF1α) also plays a role as a glycolysis regulator. HIF1α expression increases in response to MCMV and peaks at day 3 post-infection [97]. Absence of HIF1α in NK cells leads to impaired viral control via reduced NK cell numbers due to a predisposition to apoptosis, and reduced glucose uptake and lactate production. Interestingly, HIF1α-deficient NK cells had increased CD69 and IFN-γ expression at days 1.5 and 3 post-infection, suggesting HIF1α acts as a negative regulator of NK activation [97].

Therefore, in vivo studies underscore the need for intact aerobic glycolysis for optimal antiviral responses, although some functions like IFN-γ production are almost completely unaffected in vivo, in contrast to in vitro studies.

NK cell oxidative metabolism during MCMV infection

Compared to glycolysis, the role of OXPHOS in modulating NK cell function has been less studied in vivo. In vitro, cytokine stimulation of NK cells increases OXPHOS (basal rate and maximal respiratory capacity) and mitochondrial mass [84]. During MCMV infection, NK cells also undergo mitochondrial changes in vivo. The Sun laboratory demonstrated that adoptively transferred Ly49H+ NK cells have decreased mitochondrial membrane potential after MCMV-induced expansion, followed by an increase in mitochondrial membrane potential as NK cells contract and generate a memory pool (8–28 days post-infection) [93]. By contrast, a study using acute Friend virus infection showed increased NK cell mitochondrial mass and membrane potential at day 7 post-infection [47], highlighting potential differences in NK mitochondrial changes at different time points and in response to different viruses.

Our laboratory conditionally deleted Cox10, a gene encoding a component of electron transport chain complex IV COX10, in NK cells, and demonstrated that partial OXPHOS impairment does not significantly alter NK homeostasis, including homeostatic proliferation, maturation or in vitro cytotoxic capacity and cytokine production [98]. However, upon challenge of mice with MCMV virus, defects in NK cell antiviral function became apparent. Cox10-deficient NK cells had impaired expansion of Ly49H+ NK cells during MCMV infection. Coupled with the intact in vitro IL-15-driven proliferation and in vivo homeostatic proliferation in Rag2^−/−γ_c^−/−, this demonstrates that Cox10 and therefore OXPHOS is necessary for antigen-driven proliferation [98].

Collectively, in vivo studies demonstrate that while there is a degree of metabolic flexibility to compensate for aerobic glycolysis or OXPHOS defects for NK cell development, homeostasis, and some in vitro functions, there are specific metabolic requirements for antigen-driven NK cell responses to MCMV infection (Figure 1).

Figure 1. Metabolic requirements for antiviral NK cell functions in vivo.

A summary of the interplay between metabolic pathways and antiviral functions based on in vivo studies with MCMV demonstrating the consequences of disruption of different metabolic pathways. MCMV infection leads to mitochondrial changes in Ly49H+ NK cells, initially decreasing mitochondrial cell membrane potential (ψ) at day 7 post-infection. Metabolic pathway disruptions primarily affect proliferation, with impacts on viral control, host survival, and generation of a pool of memory NK cells, as indicated by the down arrows. Glycolysis is required for antigen-specific killing, while mTOR is needed for degranulation and granzyme B production. Checkmarks signify intact function despite pathway disruption. Figure created with Biorender.com.

3. HUMAN NK CELLS

Similar to the murine system, human NK cells arise in the bone marrow and are located predominantly in circulation and lymphoid organs. The majority of studies with human NK cells are conducted with blood samples, due to the availability of this compartment. Human NK cells in blood can be broadly divided into two subsets based on CD56 expression, with the vast majority (~90%) expressing CD56 at a low density (CD56^dim), while a minority of NK cells express CD56 at a high density (CD56^bright) and have relatively lower or absent CD16, an activating Fc receptor expressed by NK cells [99]. Notably, the proportions of CD56^dim and CD56^bright NK cells differ according to their location, with CD56^bright NK cells constituting a higher percentage of NK cells in tissues such as visceral adipose tissue, liver, uterus, and kidney [100]. These two NK cell populations are also functionally different, with CD56^bright cells generally more responsive to cytokines for IFN-γ production and producing more of this cytokine per cell [101]. Notably, CD56^bright cells also have higher expression of CD25 (IL-2Rα), which, when associated with CD122 (IL-2Rβ) and CD132 (common gamma chain, γ_c), forms a high affinity IL-2 receptor, which may influence their response to IL-2 in vitro, although in vivo, IL-15 is the more physiologically relevant ligand of the shared IL2Rβ/γ_c complex [102, 103]. CD56^dim cells are more responsive to activating receptor triggering, with enhanced cytotoxic capacity [100]. The CD56^bright subset also exhibits higher cytokine production and cytotoxicity after IL-15 priming [104].

These human NK subsets also appear to be metabolically distinct, which may support their functional differences. Keating et al. found that CD56^bright NK cells express relatively high levels of glucose transporter GLUT1 at baseline compared to CD56^dims, suggesting they are capable of faster glucose uptake in response to cytokines [74]. However, Surace et al. later reported higher levels of GLUT1 in CD56^dim NK cells compared to CD56^bright [105], potentially due to experimental differences, yet highlighting the need for additional studies. Keating et al. showed that cytokine stimulation (18 hours with either IL-2 or IL-12+IL-15) upregulates glycolysis and OXPHOS, as well as glucose (GLUT1) and nutrient (CD98, CD71) receptors in human NK cells, with a more prominent upregulation in CD56^bright NK cells [74]. Cytokine-stimulated mTORC1 activity indicated by pS6 increases in both subsets, but is highest in IL-2-stimulated CD56^brights, suggesting mTORC1 mediates glycolysis upregulation in CD56^brights after IL-2, but not IL-12+15 stimulation [74]. The presence of rapamycin during IL-2 and IL-12-stimulation reduced IFN-γ production and granzyme B expression, suggesting mTORC1 activity is important for these NK functions [78]. mTORC1 activity is also required for IFN-γ production, but not degranulation by NKG2D-stimulated IL-2 primed NK cells [106].

As in mice, amino acid uptake by human NK cells is important for intact function, particularly downstream of activating receptor activation. IL-2 priming (200 U/mL, 24 hours) upregulated the expression of amino acid transporters SLC1A5 and CD98, with the greatest increase observed in CD56^brights [106], consistent with their higher capacity for cytokine production. Inhibiting these transporters in IL-2 primed NK cells prevented NKG2D-stimulated IFN-γ production and degranulation [106].

Together, these studies suggest that human NK cells, similar to the mouse, increase glycolysis and mTORC1 with cytokine stimulation, and that mTORC1 signaling and amino acid uptake are important for optimal effector functions.

3.1. HCMV adaptive NK cells

Human NK cells acquire a memory, or ‘adaptive’, phenotype in individuals following HCMV infection [45, 46, 107]. HCMV-adaptive NK cells have been reported to have several distinguishing features: high expression of CD57 and activating receptor NKG2C (recognizing HLA-E), decreased expression of intracellular signaling molecules FcRγ, EAT-2/DAB2, SYK, and transcription factors PLZF (Zbtb16) and Helios (Ikzf2) [108–110]. Human NK cells do not have a receptor specific to a viral-encoded ligand, but NKG2C can recognize HCMV peptides presented in the context of HLA-E [111]. However, there are likely other mechanisms by which NK cells recognize HCMV infection, as HCMV-adaptive NK cells are also seen in individuals harboring a homozygous deletion of the gene encoding NKG2C [112], suggesting that NKG2C may not be necessary for adaptive NK cells responses and that there is heterogeneity within the adaptive population.

HCMV-adaptive NK cells have enhanced cytokine production and degranulation when stimulated with infected cells in the presence of virus-specific antibodies [108–110, 112–114]. However, NKG2C+ cells have diminished IFN-γ production in response to IL-12 and IL-18 stimulation alone, compared to conventional NK cells [109, 113], whereas engagement of NKG2C or CD16 by HLA-E-expressing or antibody-coated target cells leads to potent cytokine production, which is further amplified by IL-12 and IL-18 stimulation [113, 115]. HCMV-adaptive NK cells lacking FcRγ showed strong responses compared to FcRγ⁺ NK cells when cultured with HCMV-infected target cells in the presence of virus-specific antibodies. The addition of anti-HCMV antibodies was critical for the increased IFN-γ and TNF-α production [108, 110], in vitro proliferation [109, 110], and degranulation [108] observed. This enhanced effector function is associated with increased MAP kinase pathway and mTORC1 activity after CD16 crosslinking, and CD2 co-stimulation amplifies this signaling activity [112].

Cichocki et al. showed that adaptive (CD3⁻CD56^dimCD57⁺NKG2C⁺) NK cells have upregulated oxidative and glycolytic metabolic profiles relative to canonical NK cells. Adaptive NK cells have increased mitochondrial membrane potential and higher expression of multiple genes encoding components of the mitochondrial ATP synthase complex and electron transport chain (ETC) relative to canonical NK cells [116]. One of the significantly upregulated genes, ARID5B, is a chromatin-modifying transcriptional regulator. Through knockdown and overexpression studies using the NK-92 transformed NK cell line, they show that ARID5B can alter the mitochondrial membrane potential, OXPHOS, survival, and cytokine (IL-12+IL-18) stimulated IFN-γ production [116]. These data suggest that ARID5B is a key mediator of the metabolic changes observed in HCMV-adaptive NK cells.

Together, these studies suggest that despite the heterogeneity of HCMV-adaptive NK cells and the relative paucity of information regarding the role of metabolism in the formation, maintenance, and enhanced function of this NK cell population, current evidence suggests a positive correlation between adaptive NK cells and increased metabolism.

CONCLUDING REMARKS

NK cell activation elicits dynamic changes in their cellular metabolism, which, in turn, support rapid and proper execution of NK cell functions. Investigations into the mechanisms by which metabolism supports NK cell function are important for understanding how to modulate their function. Here, we focused on the metabolic regulation of NK cell responses during a viral infection. Disruption of metabolic pathways during MCMV infection revealed the important contribution of metabolism to viral control and reduced survival, with the highest impact on clonal expansion and antigen-specific killing. The metabolic requirements for contraction and memory phases of the antiviral response to MCMV remain largely uncertain (Figure 1), and a better understanding of this process may provide insight into pathways that could be manipulated in the context of human viral infections. In contrast to antigen-driven responses, cytokine-driven IFN-γ production exhibits substantial metabolic flexibility in vivo and raises the question of whether metabolic flexibility can be unlocked for other NK cell functions, as seen with IL-15 in vitro and in vivo. Finally, there is a need for additional human NK cell immunometabolism studies to determine potential differences between mouse and human NK cells for proper therapeutic targeting of metabolic pathways to modulate NK cell functions.

Abbreviations:

2-DG

2-deoxy-D-glucose

ACLY

ATP citrate lyase

APCs

antigen presenting cells

ATP

adenosine triphosphate

CMS

citrate malate shuttle

CMV

cytomegalovirus

ETC

electron transport chain

HIF1α

hypoxia-inducible factor 1-alpha

IFN-γ

interferon gamma

IL

interleukin

IRE1

inositol-requiring enzyme 1

ITAM

immunoreceptor tyrosine-based activation motif

LDHA

lactate dehydrogenase A

mTOR

mammalian target of rapamycin

mTORC1

mTOR complex 1

NAD

nicotinamide adenine dinucleotide

NK cell

natural killer cell

OXPHOS

oxidative phosphorylation

PKM1/2

pyruvate kinase M1/2

Srebp

sterol regulatory element-binding protein

TCA (cycle)

tricarboxylic acid

TNF

tumor necrosis factor

XBP1

X-box-binding protein 1