No Time to Die: Epigenetic Regulation of Natural Killer Cell Survival

Summary

NK cells are short-lived innate lymphocytes that can mediate antigen-independent responses to microbial infection and cancer. However, studies from the past two decades have shown that NK cells can acquire transcriptional and epigenetic modifications during inflammation that result in increased survival and lifespan. These findings blur the lines between the innate and adaptive arms of the immune system and suggest that the homeostatic mechanisms that govern the persistence of innate immune cells are malleable. Indeed, recent studies have shown that NK cells undergo continuous and strictly regulated adaptations controlling their survival during development, tissue residency, and following inflammation. In this review, we summarize our current understanding of the critical factors regulating NK cell survival throughout their lifespan, with a specific emphasis on the epigenetic modifications that regulate the survival of NK cells in various contexts. A precise understanding of the molecular mechanisms that govern NK cell survival will be important to enhance therapies for cancer and infectious diseases.

1. Introduction

Natural killer (NK) cells are innate immune cells known for their cytotoxic activity against virally infected and malignant cells in addition to their ability to rapidly produce large amounts of the anti-viral cytokine IFN-γ¹. NK cells express germline-encoded activating and inhibitory receptors that interact with ligands expressed on target cells, after which the balance between both signals will determine whether an NK cell will become activated or inhibited². Activated NK cells can kill target cells via the release of cytotoxic granules containing granzymes and perforin, or by engaging death receptors on target cells³. NK cells also shape the innate and adaptive immune responses through the secretion of a wide variety of cytokines, including IFN-γ, TNF-α, IL-10, and GM-CSF⁴.

In addition to the classical functions of NK cells such as cytotoxicity and cytokine production, NK cells have also been shown to recognize pathogens in an antigen-specific manner, even though germline-encoded NK cell receptors are unable to undergo somatic recombination. Since NK cells are known to play a critically important role in controlling the early stages of herpesvirus infections, both in humans as well as in mice^5–8, the first described and most studied subject in this regard is the specific interaction of NK cell receptors (NKG2C in human and Ly49H in mice) with proteins encoded by the beta-herpesvirus cytomegalovirus (CMV) (UL40 in HCMV and m157 in MCMV)⁹. These studies led to the understanding that NK cells have several functions previously solely attributed to cells of the adaptive immune system (i.e. clonal expansion, contraction, longevity of primed cells, and enhanced secondary effector responses), thereby challenging the existing dogma of NK cells functioning solely as innate killer cells (reviewed in^10,11). Although the plurality of studies on NK cell memory have been conducted in the context of CMV infections, NK cell memory was first observed in response to stimulation with haptens¹² and has also been noted in response to other viral infections^13–16 and cytokines¹⁷. Thus, NK cells can enhance their lifespan through a myriad of mechanisms. Given the fact that naïve NK cells are unable to persist for more than 14 days, and the transcriptome of memory NK cells is closely related to naïve NK cells¹⁸, epigenetic processes were proposed as critical regulators of NK cell memory formation and survival. Diversity, functionality, and survival of peripheral blood NK cells are also largely affected by epigenetic modifications. In addition, epigenetic mechanisms are also thought to have a large impact on the phenotype and functions of NK cells residing in specific tissue microenvironments, such as lung, liver, and decidua, and the tumor microenvironment (TME)¹⁹. However, the epigenetic signatures of tissue-resident NK (trNK) cells and the impact of epigenetic modifications on NK cells in the tumor microenvironment remain largely unexplored. Epigenetic modifications thus critically influence NK cell survival during their development, homeostasis, memory formation, and adaptation to specific tissue sites. Therefore, the scope of this review is to provide an overview of the molecular processes that occur during NK cell survival through the lifespan and tissue adaption of NK cells during homeostasis or following inflammation (Figure 1). These insights allow us to speculate on potential novel mechanistic processes that influence NK cell biology and determine critical knowledge gaps to enhance the therapeutic applications of NK cells.

Figure 1: Overview.

The epigenetic landscape of NK cells can be modified by DNA methylation, histone modifications, and non-coding RNA molecules. Development, homeostatic processes, metabolism, tissue-residency, and biological sex can critically influence the epigenetic state of NK cells. Demethylation of H3K27 and methylation of H3K4 are critical for the proper development of NK cells and lineage-defining transcription factors (LDTFs) influence the epigenetic landscape and urge NK cell development. Homeostatic levels of IL-15 lead to enhanced expression of pro-survival proteins belonging to the Bcl-2 family, keeping the levels of these molecules above the needed threshold for survival. Upon activation, metabolic requirements of NK cells change dramatically. Changes in metabolism are interwoven with epigenetic modifications that allow for prolonged NK cell survival. Mitophagy plays a crucial role in this process and increases NK cell survival during memory formation. The lncRNA Rroid and the TF Hobit can induce epigenetic modifications in certain group 1 ILC subpopulations depending on their tissue environment and have been shown to exert tissue-specific functions. Biological sex differences also have an impact on the epigenetic landscape of NK cells. This is regulated by the differential expression of UTX, which impacts NK cell effector functions and numbers between males and females.

2. Epigenetic modifications

The central dogma of molecular biology states that DNA replicates itself, and will be transcribed to produce RNA, which in turn will be translated into a protein. Each of these steps can be influenced by regulating factors, resulting in an increased or decreased presence of a certain protein. As such, cells can develop and differentiate from each other although they share the same DNA sequence. Alterations in the transcription of a certain gene that are not caused by a modification of the DNA sequence, are referred to as epigenetic modifications, which are often heritable to daughter cells^20,21. During the last decade, ample studies have demonstrated that epigenetic processes impact several immune cell functions, including lineage commitment and maturation, effector functions, memory formation, and exhaustion^22–29. These epigenetic alterations are typically classified into three categories, namely DNA methylation, histone modifications, and RNA-based epigenetic regulation²¹, which will be briefly introduced below.

2.1. DNA methylation

The covalent binding of a methyl group to the C-5 position of cytosine is called DNA methylation and is mediated by the family of DNA methyltransferases (DNMTs). Whereas DNMT3A and DNMT3B are responsible for de novo DNA methylations during development, DNMT1 ensures the maintenance of DNA methylations during replication and is thus crucial for the inheritance of the modification^21,30–32. In somatic cells, DNA methylation is most frequently observed in CpG sites (regions where cytosine is immediately followed by guanine in the DNA sequence) and methylation in so-called CpG islands (genomic regions with a high CpG frequency which are often associated with gene promoter sites) typically result in stable gene silencing^33,34. Methylation in non-CpG sites is more commonly found in specific tissues such as neurons and embryonic stem cells, however, their functional role is not yet fully understood³⁵. Although DNA methylation is a stable modification, it can be reversed both actively and passively. For example, lack of maintenance of the modification by DNMT1 results in a progressive decrease in methylation status after several replication rounds, a process called passive DNA demethylation. While the precise mechanisms behind active DNA demethylation were heavily debated, it is now generally accepted that ten-eleven translocation (TET) proteins can actively modify methylated cytosines (5mC) to an oxidated state, which can then be passively removed during replication or actively removed by thymine DNA glycosylase^36–40. Demethylation of CpG sites within the NCR1 gene locus was shown to define NK cells and differentiate them from other leukocytes. Moreover, this modification predicted NK cell numbers in blood and tissues⁴¹. Upon activation, hypomethylation of effector function genes, such as the Ifng locus, have been observed in NK cells, epigenetically priming them for enhanced IFN-γ production⁴².

2.2. Histone modifications

The DNA of a eukaryotic cell winds around histones to create nucleosomes that protect DNA from knotting and damage. The tails of histones are extensively post-translationally modified, resulting in changes in chromatin structure that lead to differences in gene expression⁴³. Different histone modifications exist and the exact function of many of them is not yet understood, however, the most studied histone modifications are histone acetylation and histone methylation⁴⁴. Histone acetylation is a highly dynamic modification mediated through the activity of histone acetyltransferases (HATs) and histone deacetylases (HDACs) and is usually associated with transcriptional activation. Methylation of histones can activate or repress transcription, depending on the amino acid that is modified and on the number of methyl groups added. Histone modifications are for example important for inducing trained immunity in innate immune cells. This is mediated by marking active promoters with H3K4me3, H3K4me1, and H3K27ac and reducing the repressive H3K9me3 and H3K27me3⁴⁵. In NK cells, increased methylation of H3K4 was observed in response to in vitro stimulation with IFN-α and IL-12/IL-18, and upon infection with MCMV in vivo⁴⁶. Cytokine stimulation of NK cells can also induce enhancer-specific histone modifications such as H3K27ac and H3K4me1, thereby allowing a rapid and dynamic NK cell response during inflammation⁴⁷.

2.3. Non-coding RNA

Functional RNA molecules that are not translated into proteins, such as microRNA (miRNA) and long noncoding RNA (lncRNA), are called non-coding RNA (ncRNA). They are mostly involved in regulating other RNA molecules and thereby indirectly impact the transcription of proteins. Although ncRNAs have been shown to impact the development and function of NK cells, a detailed description of these processes is beyond the scope of this review but was discussed elsewhere⁴⁸.

3. NK cell-intrinsic epigenetic regulation of survival

3.1. Development

NK cells develop from hematopoietic stem cells (HSCs) present in the bone marrow through a stepwise process called hematopoiesis, regulated by the sequential expression of different transcription factors (TFs) and cytokine receptors⁴⁹. HSCs develop into common lymphoid progenitor cells (CLPs), after which NK cells will start developing separately from T and B cells under the influence of lineage-defining transcription factors (LDTF) such as Nfil3, ETS1, and Id2^50–53. Nfil3 and Id2 were shown to strongly increase the sensitivity of NK cells to IL-15 signaling, thereby enhancing their survival^51,54–56. ETS1 indirectly enhances the response of NK cells to IL-15 by inducing the expression of Nfil3, but also directly enhances NK cell survival by upregulating anti-apoptotic genes and down-regulating the pro-apoptotic gene PMAIP-1⁵⁰. RUNX2 has also been described as a LDTF for NK cells and functions by inducing the expression of the IL-2Rβ on NK cells, which is essential for their homeostasis⁵⁷.

After initial lineage commitment, NK cells mature under the influence of a different set of TFs including T-bet, Eomes, and ZEB2^49,58,59. NK cell maturation is highly controlled by epigenetic mechanisms through the acquisition of lineage-specific activity of regulatory elements that are established in early progenitors^60,61. The Granzyme A and Ncr1 enhancers were shown to be established de novo during the differentiation process by histone methylation⁶⁰. Moreover, it was shown that many of the enhancers that gained activity during NK cell development display motifs for LDTFs, thereby further enhancing their effects^60,61. Epigenetic rearrangements also occur in mature NK cells at genomic sites where the maturation-related TFs Eomes and T-bet bind, suggesting that both T-box family TFs cooperate to shape the epigenetic landscape of these cells during their maturation⁶². In human and mice, Eomes is required to promote the survival of less mature NK cells, whereas T-bet is important for their survival during final maturation^62–64. It has been suggested that the optimal response to IL-15, a critical factor for NK cell survival, depends on a balanced expression of both TFs⁶². For example, in mice, loss of Eomes led to fewer CD11b⁺CD27⁺ NK cells due to increased apoptosis, altered cell cycling, and impaired maturation from CD27 single-positive cells⁶⁴. The importance of both T-box family TF was also shown in humans by a study that overexpressed T-bet and Eomes in hematopoietic progenitor cells (HPCs) derived from umbilical cord blood. This treatment altered the chromatin landscape of the HPCs, resulting in the induction of an NK cell-specific transcriptome⁶³. In another study, knock-out of T-bet and Eomes in primary human NK cells resulted in reduced proliferation and persistence of these cells in vivo after engraftment in NSG mice and impaired their long-term cytotoxic capacity. Eomes and T-bet are thus needed to maintain mature NK cell numbers and functions in mice and humans⁶⁵. T-box family TFs thus critically contribute to the survival of NK cells throughout their maturation process and simultaneously influence their epigenetic landscape, which in turn facilitates the binding of TFs to regulatory elements of genes critically involved in shaping the development and maturation of NK cells.

DNA methylation of Killer Immunoglobulin-like Receptor (KIR) genes has also been shown to suppress their expression, leading to a specific KIR expression profile depending on the degree of DNA methylation^66,67. Inhibition of DNA methylation in NK cells increased the expression levels of inhibitory KIRs, leading to a strong decrease in their cytolytic activity. This suggests that aberrant methylation of KIR genes during NK cell development impacts their function⁶⁷. Similarly, CD56^dim NK cells that underwent epigenetic remodeling of the Ifng promoter were able to produce IFN-γ more efficiently upon receptor stimulation compared to their immature CD56^bright counterparts. This remodeling was caused by successive demethylation of the Ifng transcriptional start site during terminal differentiation, accompanied by the addition of the permissive histone mark H3K4me3, which enables transcriptional activation⁶⁸. IFN-γ production of NK cells is also regulated by the H3K4me3 demethylase Kdm5a, which suppresses H3K4me3 at the promoter site of SOCS1, a suppressor of cytokine signaling, ensuring the repressive chromatin configuration of this promoter⁶⁹. Another histone modification that was shown to be important during NK cell development is H3K27me3. It was shown that inhibition of the histone methyltransferase Ezh2 decreased H3K27me3, resulting in an enhanced NK cell generation from HSCs and increased cytotoxicity in mature NK cells caused by the increased expression of the activating receptor NKG2D⁷⁰. Furthermore, inhibition of Ezh2 led to increased survival of NK progenitors, probably due to the induction of higher expression levels of IL2RA and IL7R. Ezh2 deficiency also increased survival in immature NK cells. Interestingly, the enhanced survival observed throughout NK cell maturation was found to be dependent on the expression of NKG2D, although the precise mechanisms remain to be determined⁷⁰. In line with this, inhibition of H3K27 demethylases resulted in decreased production of pro-inflammatory cytokines by NK cells upon activation⁷¹. Both studies show that H3K27 demethylation is critical for the proper development and functional activity of NK cells and their survival. Taken together, epigenetic modifications of NK cells initiate from the start of their development, continue during their maturation, and result in long-lasting effects on the function and survival of mature NK cells. These findings will need to be considered when developing NK-cell therapies that require ex vivo activation.

3.2. The Homeostatic Survival Clock

Homeostatic survival and proliferation are essential to maintain a functional level of mature NK cells in peripheral tissues and the circulation^72,73. IL-15 is a critical factor in NK cell homeostatic proliferation and survival during homeostasis^72,74–77. The pro-survival effect of IL-15 relies on the trans-presentation of the IL-15/IL-15Rα complex by parenchymal and myeloid cells to NK cells^78–81. Upon receptor ligation, the IL-15 signal is transmitted via the activation of 2 separate pathways: the JAK1/3 pathway and the AKT pathway (Figure 2). Upon activation of the JAK1/3 pathway, the TF STAT5 is recruited and activated. In contrast, the activation of the lipid kinase phosphatidylinositol-3-OH kinase (PI3K) leads to the production of phosphatidylinositol triphosphate (PIP3) which is bound by the kinase AKT and finally results in the phosphorylation of the kinase mammalian target of rapamycin (mTOR)^82–85. IL-15-mediated activation of mTOR leads to increased NK cell proliferation and increased expression of NKG2A⁷³, whereas STAT5 activation results in the induction of the anti-apoptotic protein Bcl-2 in resting NK cells⁸⁶. However, apoptotic NK cells show only a limited reduction of Bcl-2 expression, indicating that Bcl-2 alone is not sufficient to maintain NK cell viability⁸⁷. In contrast, Mcl-1 was shown to be necessary and sufficient for NK cell survival and IL-15 can promote Mcl-1 expression. Furthermore, IL-15 antagonizes NK cell apoptosis by limiting the expression of Bim, which binds and withdraws pro-survival proteins such as Mcl-1 and Bcl-2 thereby making NK cells sensitive to metabolic stress-induced apoptosis⁸⁸.

Figure 2: Homeostasis.

IL-15 trans-presentation by myeloid cells to NK cells initiates multiple pro-survival pathways. Via the JAK1/3 pathway, STAT5 is recruited to the nucleus and leads to increased transcription of pro-survival molecules such as Bcl-2 and Mcl-1. High levels of Bcl-2 and Mcl-1 counteract the pro-apoptotic signal provided by BIM, resulting in increased survival of NK cells. Via PI3K, AKT is phosphorylated and activates mTORC1, leading to increased CD94/NKG2A expression on the NK cell surface and increased NK cell proliferation through enhanced glycolysis. Downstream of AKT, XBP1s is stabilized by PIM-2 and recruits T-bet to the promoter region of IFN-γ, resulting in an increased NK cell effector function.

In general, homeostasis can be looked at as the continuous suppression of cell death by pro-survival signals. Recently, a threshold-based model for the lifespan of T and B cells has been proposed⁸⁹. Three pro-survival proteins of the Bcl-2 family, namely Bcl-2, Mcl-1, and Bcl-xL, were shown to be critical drivers of survival and cells would undergo apoptosis when these fall below a certain threshold, allowing the authors to predict the moment of cell death of individual cells through a mathematical model⁸⁹. Presumably, the same model could be applied to predict the lifespan of NK cells, however, it might be necessary to add additional components to the analysis and address the critical importance of Mcl-1 for NK cell survival to achieve the same precision. In this regard, it is important to take epigenetic modifications into account since they can largely influence the critical balance between pro- and anti-apoptotic molecules present in NK cells. For example, the transcriptional regulator haematopoietically expressed homeobox (Hhex) binds to Bcl2l11 (encoding Bim), thereby repressing its expression and promoting NK cell survival⁹⁰. Likewise, the transcription factor interferon regulatory factor 4 (IRF4) was found to prevent NK cell apoptosis by decreasing Bcl2l11 transcripts and by increasing iron and amino acid uptake⁹¹.

Downstream of IL-15-mediated Akt signaling, the transcriptional activator spliced X-box binding protein 1 (XBP1s) is also important for homeostatic NK cell survival and effector functions^92,93. Recently, it was shown that XBP1s induced transcription of the anti-apoptotic gene proviral integrations of moloney virus 2 (PIM-2) by binding to its promoter region. PIM-2 was essential for NK cell development and survival during homeostasis since knock-out of PIM-2 resulted in reduced NK cell numbers in the spleen and bone marrow of PIM-2 knock-out mice compared to wild-type mice. In turn, PIM-2 phosphorylates XBP1s thereby protecting it from rapid proteasomal degradation. This allows XBP1s to recruit T-bet to the promoter region of IFN-γ, resulting in an enhanced IFN-γ production⁹⁴. Apart from its effects on survival and proliferation, IL-15 signaling has thus also been shown to result in the upregulation of genes that control effector functions of NK cells. In addition, IL-15 has been shown to induce metabolic reprogramming towards the glycolytic pathway^82,85,95. Given the critical role of IL-15 in NK cell homeostasis and function, it has been proposed that IL-15 levels in the environment might control the degree of proliferation of NK cells to regulate the size of the NK cell pool, like the mechanism of quorum sensing in bacteria⁹⁶. Of course, this positive proliferative role of IL-15 must be counteracted to keep control over NK cell numbers and activity. Therefore, senescent NK cells, characterized by the expression of KLRG1, are less sensitive to IL-15, which coincides with their less functional phenotype and decreased survival^97–99.

3.3. Metabolism

Upon activation, metabolic requirements for NK cells change dramatically. Whereas resting NK cells only need a limited amount of energy, which is provided through glucose-driven oxidative phosphorylation (OXPHOS), the energy requirements for activated NK cells greatly increase and depend on the effector function and the type and duration of the stimulus¹⁰⁰. Both in humans and mice, IFN-γ production of NK cells generally relies on both glycolysis and OXPHOS, whereas NK cell cytotoxicity is mostly dependent on glycolysis^100–102. The duration of the stimulus was also shown to influence the NK cell metabolic reprogramming. In mice for example, short-term cytokine activation of NK cells showed no changes in OXPHOS or glycolysis metabolism, indicating that NK cells have a flexible metabolic system, which guarantees quick production of IFN-γ when needed¹⁰⁰. Longer-term stimulation led to metabolic reprogramming and resulted in increased glycolysis via mTORC1 and elevated amino acid uptake⁸⁵. Indeed, amino acids and glutamine are important to sustain NK cell effector functions in stimulated cells¹⁰³.

In humans, CD56^bright NK cells are thought to be more dependent on glycolysis compared to their CD56^dim counterparts since they express higher levels of glucose-transporter 1 (GLUT-1), allowing for increased glucose uptake. This is in line with their increased capacity to produce IFN-γ since glycolysis was shown to be important for IFN-γ production^104,105. In CD56^dim NK cells, mitochondrial fusion driven by OXPHOS was shown to be critical for their survival¹⁰⁶. The ability to reprogram their cellular metabolism has also been proven to be indispensable for the anti-viral activity of NK cells. For example, it was shown that mouse NK cell responses during viral infection required increased glycolysis and mitochondrial metabolism, but also increased amino acid transport and sufficient iron availability¹⁰⁷. Altogether, these results show that NK cell effector functions are largely influenced by NK cell metabolism, which may vary depending on the specific NK cell subset or activation status, and that NK cells can react to stimuli by changing their metabolism to optimize their response.

The survival of memory NK cells also critically depends on metabolic regulation. Accumulation of damaged mitochondria in cells leads to oxidative stress, resulting in cell death through the production of reactive oxygen species (ROS)¹⁰⁸. It has been shown that in response to MCMV infection, late effector Ly49H⁺ NK cells accumulate mitochondrial-associated ROS and exhibit oxidative stress¹⁰⁹. However, the mitochondrial-associated proteins BNIP3 and BNIP3L mediated the removal of ROS during the contraction phase through the induction of mitophagy (i.e. the autophagic lysosomal degradation of mitochondria), thereby leading to the enhanced survival of these memory NK cells. Moreover, induction of autophagy at the peak of NK cell expansion resulted in a larger number of MCMV-specific memory NK cells, showing that mitophagy promotes cellular homeostasis and survival of memory NK cells upon MCMV infection¹⁰⁹. In line with this, it was shown that the NK-cell specific deletion of cox10, critical for cytochrome c oxidase (COX) assembly and thus an essential element in the mitochondrial electron transport chain, resulted in an impaired receptor-mediated activation and an impeded expansion of Ly49H⁺ NK cells, whereas differentiation, homeostasis and IL-15-driven proliferation of NK cells were unaffected¹¹⁰. Another study showed that lactate dehydrogenase A (LDHA) was essential for the clonal expansion and effector functions of NK cells upon MCMV infection in mice¹¹¹. In HCMV-positive humans, the adaptive pool of NKG2C⁺ NK cells showed increased glycolysis and OXPHOS and it was shown that this enhanced cellular metabolism was needed to support the expansion and survival of these NK cells¹¹². Interestingly, these effects were shown to be mediated by the chromatin-modifying transcriptional regulator AT-rich interaction domain 5B (ARID5B), and the authors also observed a selective hypomethylation of the 5’ region and the transcriptional start site upstream of ARID5B transcript variant 2 in adaptive NK cells¹¹². These studies show that both OXPHOS and glycolysis are crucial for the establishment and survival of mouse and human memory NK cells and that NK cell metabolism is influenced by epigenetic modifications. NK cell effector functions and memory formation are thus critically influenced by metabolic processes, however, the molecular mechanisms regulating these processes are not yet completely understood. Therefore, it will be important to further research the biology of NK cell metabolism in the future to efficiently harness NK cells for clinical applications¹⁰¹.

3.4. Trafficking and Tissue-residency

Growing evidence has suggested that NK cells are but one of a growing number of group 1 innate lymphoid cells (ILCs) classified on their developmental dependence on Tbet, IFN-γ production, and natural cytotoxicity. While heavily debated previously, fate-mapping experiments in addition to single-cell RNA sequencing datasets have shown that mammalian group 1 ILCs consist of distinct and stable lineages of circulating NK cells (cNK), tissue-resident NK cells (trNK), and ILC1^113–116. While group 1 ILCs may have similar host-protective functions, they differ based on their localization in non-lymphoid tissues and lifespan, which allow for non-redundant spatiotemporal differences in the regulation of the immune response to pathogens¹¹⁷. For instance, ILC1s represent long-term tissue-resident group 1 ILCs that are the first lymphocytes to produce IFN-γ in response to IL-12 and IL-18 in initially infected tissues of naïve mice¹¹³. As ILC1 have been recovered from all mouse peripheral organs analyzed, in addition to certain human tissues, they likely represent the first critical line of non-specific lymphocyte defense in response to myeloid-derived inflammation during infection and tissue injury^{113,114,118–120}. Current evidence from mouse studies in the salivary gland and uterus supports the hypothesis that trNK cells are derived from circulating NK cells and are likely epigenetically imprinted based on tissue-specific factors¹²¹. Because the developmental and phenotypic differences between group 1 ILCs have been reviewed previously¹¹⁷, we will focus our discussion on studies detailing the impact of epigenetic modifications on the survival of distinct subsets of group 1 ILCs.

The development and survival of ILCs require a coordinated interplay between the epithelial and stromal cells of the tissue and the tissue-resident immune cells. These interactions influence their epigenetic landscape, allowing for tissue-specific adaptations^19,122. For example, it has been demonstrated that the gut microbiome influences the epigenetic landscape of ILCs in the small intestine, resulting in phenotypical and functional differences¹²³. In support of this hypothesis, Nussbaum et al. showed that the effector functions of type 3 ILCs were largely influenced by the tissue of residence and not solely by the tissue of origin¹²⁴. Similarly, the lncRNA RNA-demarcated Regulatory region of Id2 (Rroid) promotes the expression of Id2, a transcriptional repressor essential for the lineage commitment of ILCs, in ILC1s residing in the spleen, liver, and lung, thereby controlling their homeostasis¹²⁵. Although Rroid was not found to be required for the survival of ILC1s in the salivary gland or the small intestine, it was needed for the maturation (but not the early development), the homeostasis, and the function of ILC1s, but not ILC2s or ILC3s. Mechanistically, this was shown to depend on epigenetic modifications of the Id2 promoter region; in the absence of Rroid, the histone marks for active promoters (H3K27Ac) and for transcription (H3K36me3) dramatically decreased, leading to decreased chromatin accessibility in the Id2 promoter region in these cells, which inhibited the binding of STAT5. Altogether, this means that the lncRNA Rroid is needed for the IL-15-mediated epigenetic regulation of Id2 expression, thereby determining the survival of ILC1 populations in specific tissues¹²⁵. A similar tissue-dependent role in ILC1 biology was described for the transcriptional repressor Hobit (encoded by ZNF683)¹¹⁵. While Hobit was critical for the survival of ILC1 in the liver and the salivary glands, it was dispensable for ILC1s in the small intestine. However, Hobit was also shown to be an important regulator of ILC1 differentiation after initial lineage commitment. It was needed for CD127⁺IL18R⁺TCF1⁺ helper-like ILC1s to develop into CD127⁻ IL18R⁺TCF1⁻ cytotoxic ILC1s, but not for the commitment to the ILC1 lineage and thus also shows subset-dependent effects¹¹⁶. Altogether, these data show again that ILC1s can adapt to their local tissue environment, thereby changing their response to certain stimuli, however, the tissue-specific signals that can regulate ILC1 maturation are still poorly understood and will need to be explored in future studies.

NK cells can also be impacted by discrete tissue microenvironments during their differentiation to trNK cells. Besides ILC1s and cNK cells, the salivary gland also harbors a distinct and specific tissue-resident population. In contrast to cNK cells, these salivary gland trNK cells do not need Nfil3 for their development, phenotype, or function^126,127. The salivary gland-specific development of trNK cells was shown to be regulated by TGF-β through the suppression of Eomes¹²⁸. Loss of TGF-β signaling reduced the numbers of salivary gland trNK cells and led to the loss of certain markers associated with tissue residency, such as CD49a. Interestingly, the TGF-β-mediated regulation of salivary gland trNK cells happened simultaneously with the development of the salivary gland itself, highlighting the importance of temporal regulation of tissue-specific signals influencing their tissue residency¹²⁸. This synchronized development of tissue and trNK cells might also imply that resident lymphocytes are also important in shaping the development of the tissue they reside in¹²⁹. Indeed, following MCMV infection in mice cNK cells strongly increased in the salivary gland in a CX3CR1-dependent manner, resulting in a relative decrease of other group 1 ILCs^121,130. The recruited NK cells displayed an activated phenotype (CD49a⁺CD11b⁺TRAIL⁺), but were CD200R⁻ and formed a pool of tissue-resident memory-like NK cells that prevented auto-immunity through the killing of autoreactive CD4⁺ T cells¹²¹.

The impact that tissue-resident lymphocytes can have on their environment can also be illustrated by decidual trNK cells^131,132. During early pregnancy in humans, the number of trNK cells increases dramatically, resulting in more than 50% of the decidual lymphocytes being NK cells at that time point^133,134. These cells are, in contrast to their counterparts in the blood, poorly cytotoxic¹³⁵ and are thought to exert constructive functions, assisting in fetal development¹³⁶. Decidual trNK cells, both in humans and mice, were shown to be important for immune tolerance against the fetus, extravillous trophoblast cell invasion, and remodeling of the uterine spiral artery, thereby facilitating successful implantation^137–139. Interestingly, human decidual trNK cells were found to be highly plastic and were able to quickly upregulate cytotoxic effector functions when stimulated with cytokines or upon NKp46 engagement^140–142. It was shown that reversely, the decidual microenvironment also impacts the residing decidual trNK cells. A phenotypically and functionally unique subset of decidual trNK cells was found in multigravid women, characterized by increased accessibility at the Ifng and Vegfa loci and increased production of IFN-γ and VEGF-A upon stimulation. This study shows that the unique environmental stimuli generated during pregnancy resulted in the generation of a unique NK cell subset through the induction of epigenetic modifications in decidual trNK cells¹⁴³. The fact that the decidual microenvironment can change the epigenetic landscape of trNK cells was also described in another study, showing that these cells have a distinct pattern of CpG island methylation compared to NK cells found in breast or lymph nodes¹⁴⁴. As described above, DNA hypomethylation leads to increased inhibitory KIR expression and decreased cytotoxicity, but in addition it was shown that demethylating agents caused NK cells to promote invasion of human trophoblast cell lines and increase the production of VEGF¹⁴⁵.

The liver has also been identified as a unique immunological environment for group 1 ILCs. Interest in liver trNK cells, and trNK cells in general, arose from the discovery that ILC1s in the liver could mount an antigen-specific contact hypersensitivity response¹². Later, it was discovered that the NLRP3 inflammasome plays a non-redundant role in the induction of these hapten-dependent memory ILC1s in the liver¹⁴⁶. Although the epigenetic processes involved in hapten-induced memory are currently unknown, it has been shown that epigenetic modifications modulate the expression, stability, and activation of the NLRP3 inflammasome¹⁴⁷. Therefore, it is probable that epigenetic reprogramming of liver ILC1s is crucial to mount hapten-dependent memory responses. Liver ILC1s in mice and liver trNK cells in humans are phenotypically characterized by the expression of the chemokine receptor CXCR6, although other surface markers vary substantially between both species^148–150. In mice, it has been shown that adoptively transferred hepatic memory ILC1s can protect recipient mice after a lethal challenge with the sensitizing virus and that this effect was dependent on CXCR6¹⁵¹. In humans, antigen-specific NK cell-mediated memory responses were described and the cells mediating this response were shown to possess a tissue-resident phenotype (such as the expression of CXCR6), however, it was impossible to determine whether these cells originated from the liver or another tissue¹⁵. Recently, a CXCR6-expressing CD49a⁺CD16⁻ ILC1-like population in the human liver showed more accessible chromatin at certain promoter regions compared to CD49a⁻CD16⁺ NK cells, which predisposes the underlying genes for activation, leading to an increased potential for antigen-specific memory responses¹⁵². Altogether, these studies show that specific environmental stimuli, even though they might be temporary, can influence the epigenetic landscape of locally residing group 1 ILCs, resulting in cell states that are well adapted to exert the peculiar functions required by that specific environment to maintain tissue homeostasis and function. Of course, further research will be necessary to fully understand the mechanisms behind the regulation of these tissue-specific phenotypes on the epigenetic state of group 1 ILCs.

3.5. Sex Differences

Sex differences in mammals can be driven by gonadal hormones and sex chromosome gene dosage¹⁵³. Expression of a subset of X-linked genes is higher in females (XX) than males (XY) due to random X-chromosome inactivation (XCI) to maintain similar levels of X-linked protein expression between sexes¹⁵⁴. However, XCI is incomplete, with 3–7% of X-chromosome genes escaping inactivation in mice and 20–30% escaping inactivation in humans^154–156. A recent study identified the X-linked escapee KDM6A (UTX) as an epigenetic regulator that is increased in female NK cells compared to males (mice and humans). Heterozygous loss of UTX in female NK cells phenocopied male NK cell phenotype with an increase in NK cell fitness and numbers but decreased cytotoxicity and cytokine production. UTX controlled chromatin accessibility at several loci associated with NK cell survival (Bcl2, Casp3) and effector function (Ifng, Prf1, Csf2), demonstrating that X-linked gene dosage creates sexually dimorphic gene regulatory networks in NK cells. While UTX can poise chromatin for active gene expression through its function as an H3K27me3 demethylase¹⁵⁷, UTX can also form complexes with other chromatin modifiers, such as p300, MML4, and SWI/SNF to regulate chromatin accessibility in a demethylase-independent manner^157,158. In NK and CD8⁺ T cells, the role of UTX is independent of its demethylase function, whereas it is required in iNKT cells^159–161. These results suggest that the mechanisms of UTX-mediated epigenetic regulation are cell lineage-specific. This observation could be partially explained by the cell type-specific expression of UTX interaction partners, which motif analyses of UTX-bound peaks suggest consisted of Runx and KLF family members in addition to Eomes in mouse NK cells¹⁶⁰. Furthermore, because UTX enzymatic function is dependent on the byproducts of cellular metabolism such as iron, oxygen, and α-ketoglutarate, differences in tissue localization and metabolic dependencies could also dictate whether the demethylase function of UTX is essential for immune functions. Thus, further research will be necessary to identify the cell type-specific interaction partners and metabolic regulators of UTX to promote sexual dimorphism in the immune system.

The question remains as to why sexual dimorphism in NK cell fitness and function has been selected through evolution in mice and humans. It is tempting to speculate that conferring a hormone-independent enhancement in NK cell-intrinsic function in females would increase protection against infection of the mother and trophoblast during pregnancy-induced systemic immunosuppression associated with pregnancy hormones and increased numbers of regulatory T cells^162,163. While several studies have implicated decidual NK cells in the control of microbial and viral infections in trophoblasts^164–166, whether increased UTX expression is required for protective NK cell responses in females during pregnancy is unknown. On the other hand, decidual-resident NK cells are critical for fetal growth and development by promoting vascularization and remodeling of uterine spiral arteries through IFN-γ and VEGF-α production, and trophoblast invasion through GM-CSF secretion^137–139. Because decidual NK cells are functionally inhibited in the presence of trophoblasts derived from early pregnancy in an HLA-G-dependent manner¹⁶⁷, enhanced epigenetic poising of the Ifng and Csf2 (GM-CSF) locus in females may confer decidual NK cells with a lower activation threshold to produce key cytokines during critical activation windows in fetal development. Thus, evolutionary pressure centered on pregnancy outcomes may underlie sexually dimorphic epigenetic regulation by UTX in NK cells, although more evidence will be necessary to support this hypothesis. Because of selection for sexual dimorphism in NK cell effector function, compensation in NK cell fitness through boosting NK cell numbers in males may reflect an evolutionary trade-off to enhance viral surveillance at the cost of optimal effector function. Given that NK cells are critical for protection against viral infection, this trade-off could predispose males with heightened susceptibility to severe viral infections as observed during the COVID-19 pandemic with severe disease associated with hypofunctional NK cell responses¹⁶⁸.

4. Epigenetic regulation of NK cell survival during inflammation

Apart from the antigen-specific memory responses that NK cells exhibit (which are discussed below), they are also able to develop an antigen-independent form of enhanced persistence and effector function. This type of immunity is called trained immunity and has also been observed in monocytes, macrophages, neutrophils, ILC, and even non-immune cells^169,170. Trained immunity leads to long-lasting epigenetic reprogramming of transcriptional and metabolic pathways of the activated cell, which results in an altered response upon reactivation¹⁷¹. Interestingly, this does not only comprise an increased reaction towards a stimulus upon re-exposure; immune cells can be “trained” to tolerate certain signals and thereby become hyporesponsive in the presence of microorganisms. This dampened inflammatory response, referred to as ‘tolerance’, prevents excessive tissue damage and is needed in certain tissues to maintain homeostasis. In addition, a particular characteristic of trained immunity is that this type of memory can show a heterologous recall capacity. The signal eliciting the recall response can thus be unrelated to the stimulus that evoked the primary activation¹⁷¹.

In 2009, Cooper et al. showed that murine NK cells stimulated with IL-12 and IL-18 exhibited an increased production of IFN-γ upon restimulation with IL-12 or upon stimulation of activating NK cell receptors via plate-bound antibodies up to three weeks post-stimulation¹⁷. In contrast, the cytotoxic activity of these cytokine-induced memory-like (CIML) NK cells was not elevated. This effect was visible in NK cells activated for only 5 hours in the presence or absence of the pro-survival cytokine IL-15. Although the exact mechanism of the prolonged persistence of these activated cells was not investigated, it was thus not solely dependent on the effect of the pro-survival cytokine IL-15. Interestingly, they also showed that the memory-like capacity of the CIML NK cells was passed on to daughter cells, implying an underlying epigenetic mechanism. Importantly, these results could be repeated in vivo, where adoptively transferred CIML NK cells proliferated to maintain homeostasis and retained the capacity to produce increased amounts of IFN-γ for at least one month in NK-cell deficient recipient mice¹⁷². Human NK cells were shown to form CIML NK cells with similar characteristics upon cytokine priming¹⁷³. Both CD56^bright and CD56^dim NK cells were able to mount CIML responses and NKG2A, CD94, and NKp46 were surface receptors associated with CIML NK cells¹⁷³. The increased IFN-γ production of CIML NK cells is indeed shown to be epigenetically imprinted through specific demethylation in the Ifng locus, which has been linked to increased Ifng transcription levels¹⁷⁴. In subsequent mice experiments, CIML NK cells showed rapid proliferation in vivo under the influence of IL-2 produced by local CD4⁺ T cells and reduced tumor growth of established tumors only if the transfer of CIML NK cells was combined with total body radiation therapy¹⁷⁵. CIML NK cells can thus be homeostatically maintained in vivo and show an enhanced anti-tumor capacity under certain circumstances. Therefore, these cells have been researched as a potential novel immunotherapeutic strategy. Indeed, human CIML NK cells showed enhanced IFN-γ production and increased cytotoxicity against leukemia cell lines and primary acute myeloid leukemia (AML) blasts in vitro¹⁷⁶. Moreover, xenografting these cells into NSG mice with induced leukemia decreased the AML burden and increased survival. Interestingly, this study was followed by an in-human study which showed that allogeneic human CIML NK cells were able to proliferate in patients with AML and improved clinical outcomes¹⁷⁶. These promising results have led to several subsequent CIML NK cell clinical trials in AML patients (extensively reviewed in^177,178), which have shown promising results and led to more innovative strategies such as the use of CIML NK cells in combination with chimeric antigen receptor (CAR) technology. So, if we manage to better understand the epigenetic and metabolic reprogramming necessary for NK cell survival and effector activity in a pro-inflammatory environment, we can further develop pioneering off-the-shelf NK cell-based therapies in both prophylactic as well as in therapeutic settings and turn fundamental research into applicable solutions for patients.

Trained immunity in NK cells has also been studied in the context of systemic inflammation induced by LPS and upon priming with activating cytokines. NK cells primed under these conditions are mostly referred to as memory-like NK cells to distinguish them from antigen-specific memory NK cells. Endotoxemia in mice is established through intraperitoneal injection of LPS and is used as a model to study systemic inflammation. NK cells isolated from mice that underwent such treatment, showed increased production of IFN-γ and GM-CSF when restimulated with LPS in vitro¹⁷⁹. Moreover, upon adoptively transferring naive NK cells into mice primed with LPS two weeks in advance, the endogenous NK cells of LPS-primed mice showed a higher capacity to produce IFN-γ compared to the naïve NK cells that were adoptively transferred, proving that NK cells can indeed develop this memory-like phenotype after endotoxemia in vivo. Similar results were obtained if an adoptive transfer was performed in mice primed with LPS 9 weeks in advance. These endotoxemia-induced memory-like NK cells thus persisted for at least 9 weeks. Moreover, the authors showed that LPS-trained memory-like NK cells contributed to protection from E. Coli infection. Mechanistically, the memory-like phenotype of LPS-primed NK cells was shown to rely on an H3K4me1 histone mark at an enhancer site close to the Ifng transcription start-site, and inhibition of this methylation completely abrogated the observed effects¹⁷⁹. However, the precise mechanisms that lead to trained memory in this context remain elusive.

5. Epigenetic regulation of NK cell survival during viral infection

Group 1 ILCs undergo significant changes in their epigenetic state in response to viral infection^18,180,181. The coordinated epigenetic reprogramming of NK cells is temporally regulated and distinct during three main phases: activation, clonal proliferation, and persistence to form memory cells. The transcriptional and epigenetic changes induced during the activation and proliferation phases have been extensively reviewed and are initiated by direct sensing of viral glycoproteins by germline-encoded activating receptors, co-stimulation of activating receptors, and proinflammatory cytokines^10,24,182. While epigenetic modifications have been described in mouse and human NK cells during viral infections, we will focus our discussion on studies using mouse cytomegalovirus (MCMV) infection due to their mechanistic interrogation of the epigenetic mechanisms that impact NK cell survival.

During MCMV infection, NK cells receive direct multiple activating signals from the environment (IL-12, IL-15, type I Interferons, IL-18, IL-33) and from virally infected cells (CD155, CD112, m157) that lead to their activation and epigenetic remodeling (Figure 3)^10,182. Chromatin modifications at gene promoters are triggered by type I IFN-STAT1 signaling, involving the deposition of H3K4me3⁴⁶. These epigenetic modifications are necessary to protect rapidly proliferating Ly49H⁺ NK cells from being killed by other activated NK cells (called fratricide) through STAT1/STAT2–IRF9 mediated increase in inhibitory MHC-I expression¹⁸³. IL-12–STAT4 signaling primarily remodels enhancers by acting as a pioneering factor, priming de novo enhancers through the induction of chromatin accessibility and recruiting p300/CBP for H3K27 acetylation^46,47. The resulting IL-12–STAT4 induced regulome is required for the increased expression of Runx3 to support the survival of MCMV-specific Ly49H⁺ NK cells for optimal memory generation¹⁸⁴, although the precise mechanisms of how Runx3 controls activated NK cell survival are not yet understood. IRF4, induced by a combination of cytokines and activating receptors, forms a transcriptional complex with Batf to repress Bcl2l11 (BIM) expression (Figure 3)¹⁸². Collectively, these results suggest that while coordinated signals from activating receptors and cytokines are necessary to support the enhanced lifespan of virus-specific NK cells, activating receptor stimulation alone is not sufficient. In support of this hypothesis, NFAT-mediated epigenetic modifications downstream of activating receptor signaling enhance the chromatin accessibility of STAT4 binding sites, creating a permissive epigenetic state for the IL-12-STAT4 regulome¹⁸⁵. While these findings were used to argue that activating receptor activation precedes cytokine stimulation in virus-specific NK cells in vivo, these data could also reflect epigenetic priming due to a gradient of Ly49H avidity signaling, in which Ly49H^hi NK cells have been shown to preferentially survive to form memory NK cells¹⁸⁶. Irrespective of this point, it is likely that individual NK cell clones experience a stochastic mixture of activating signals determined by spatial location in infected tissue, creating a heterogeneous mixture of epigenetic states with disparate survival potential. Further profiling of these NK cell states by single-cell ATAC and spatial sequencing will be necessary to understand the critical gene regulatory networks that impact NK cell survival during viral infection.

Figure 3: Molecular factors influencing NK cell memory formation upon MCMV infection.

Upon infection with MCMV, NK cells become activated through the interaction between their activating receptors (Ly49H and DNAM-1) and ligands on the infected cells (m157 and CD155 respectively) and through cytokines secreted by activated myeloid cells (phase 1: activation and expansion phase). The induced signaling pathways modify the epigenetic landscape of the activated Ly49H+ cells, resulting in enhanced proliferation and survival of these cells until D7 post-infection. Around D7–10 post-infection, survival of late effector cells is challenged by a reduced availability of IL-15 (phase 2: late effector phase). This will result in a large reduction of NK cell numbers after D10 (phase 3: contraction phase). However, Ly6C- cells overcome cell death by increasing their expression of Bcl-2, are enriched in Fli-1 and Ets-1 binding sites, and may represent a memory-precursor (MP) NK cell population. The presence of the transcriptional repressors Zfp740 and Fli-1 in these MP NK cells, which are induced during early activation, leads to decreased levels of pro-survival genes and increased BIM levels, limiting their lifespan. This results in a memory phase that is characterized by a continuous loss of memory NK cells (phase 4).

Following activation and clonal proliferation, the expanded pool of late effector Ly49H⁺ NK cells face several barriers to persist and form memory NK cells. The first is extrinsic, IL-15 is reduced to homeostatic levels following viral clearance, unable to support the survival of the increased number of late effector NK cells through continuous STAT5 signaling and resulting in the death of most of these cells through the contraction phase⁴⁶. Other obstacles are cell-intrinsic, with Bcl2 levels dramatically reduced in late effector compared to naïve NK cells, leading to an increased sensitivity to BIM-mediated apoptosis¹⁸⁷. Furthermore, late effector NK cells acquire dysfunctional mitochondria following rapid clonal proliferation, which can cause decreased persistence through increased intracellular reactive oxygen species¹⁰⁹. How then do any late effector NK cells survive to form memory NK cells? Although only a few studies have attempted to address this question directly, recent evidence demonstrated that late effector NK cells are not a homogeneous population. Single-cell RNA sequencing and lineage tracing experiments were used to support the hypothesis that late effector NK cells consist of two phenotypically distinct Ly6C⁻ and Ly6C⁺ NK cells. When co-transferred into naïve or previously infected mice, Ly6C⁻ late effector NK cells displayed enhanced persistence compared to Ly6C⁺ late effector NK cells and selectively gave rise to Ly6C⁺ memory NK cells¹⁸⁷. Comparison of late effector NK cell populations with memory precursor (MP) CD8⁺ T cells revealed an overlapping epigenetic signature of Ly6C⁻ late effector NK cells and MP CD8⁺ T cells enriched in Fli1 and Ets1 binding sites, suggesting that this population may represent a MP NK cell state (Figure 3). MP NK cells displayed higher levels of Bcl2, which is also required for MP CD8⁺ T cell survival and memory formation^188,189. However, MP NK cells did not display epigenetic or transcriptional signatures associated with stem-like progenitor CD8⁺ T cells¹⁹⁰, which could explain the continuous loss of memory NK cells observed following viral clearance. While these results suggest that the epigenetic state of MP NK cells fundamentally alters the NK cell homeostatic survival clock through control of Bcl2 levels, future studies will need to determine the mechanisms by which this occurs.

Given that the half-life of mammalian circulating NK cells during homeostasis is ~7–14 days^64,191, epigenetic modifications acquired by virus-specific NK cells to enhance their persistence following viral infection likely create competitive gene regulatory networks with activation-induced negative feedback loops that function to limit activated NK cell survival. In support of this hypothesis, recent studies have observed that the transcriptional repressors Zfp740 and Fli1 are induced downstream of Ly49H and IL-15 signaling respectively^187,192. Genetic deletion of either Zfp740 or Fli1 increased late effector and memory NK cell formation through Zfp740-mediated repression of putative pro-survival genes, or Fli1-mediated enhancement of BIM levels. While Fli1 preferentially reduced the survival of MP NK cells compared to Ly6C⁺ late effector cells, Zfp740 deletion did not, suggesting that distinct activation-induced negative transcriptional feedback loops regulate nonoverlapping mechanisms of NK survival following viral infection (Figure 3). However, the mechanism of how Fli1 regulates BIM levels, and the gene targets of Fli1-mediated repression in NK cells are unknown. Because IL-12 stimulation is also sufficient to increase BIM levels in NK cells ex vivo¹⁸⁵, it will be of interest for future studies to determine how specific activation signals and their combinations can modify NK cell heterogeneity and the epigenome to impact the survival of NK cells during viral infection.

6. Epigenetic regulation of NK cell survival in the tumor microenvironment

NK cells were named based on their natural killing of tumor cells in vitro^193,194. Subsequent clinical studies have shown that increased NK cell infiltration correlates with better prognosis in multiple solid tumor types^195–200. NK cells function not only to recognize and kill tumor cells directly but also activate and recruit conventional type 1 dendritic cells (cDC1s) that perform neoantigen cross-presentation to T cells and are essential for immune checkpoint blockade (ICB) efficacy²⁰¹. However, current evidence suggests that NK cells derived from many human solid tumor types are phenotypically immature and dysfunctional when compared to peripheral blood NK cells from the same patient, suggesting local imprinting of NK cell dysfunction upon infiltration into the tumor^202–205. This hypothesis was recently supported by in vivo photolabeling and fate-mapping of tumor-infiltrating mouse NK cells²⁰⁶, suggesting that TME-induced transcriptional changes occur rapidly following infiltration into solid tumors. While the potential mechanisms underlying NK cell dysfunction in the tumor microenvironment (TME) have been reviewed previously^207,208, we will briefly focus on studies identifying distinct epigenetic modifications in NK cells that may regulate NK cell survival in the TME.

NK cells, in comparison to other immune populations in common human solid tumors, are rare²⁰⁹. Although some studies have linked reduced tumor infiltration of NK cells to decreases in or altered chemokine expression, solid tumors with low NK cell infiltration can still express abundant amounts of chemokines that should cause NK cells to be recruited²¹⁰. These findings suggest that other factors, such as NK cell survival or homeostasis, may also be inhibited in the TME. While studies have shown that infiltrating NK cells can localize to the tumor tissue, tumor microarray studies using immunofluorescence markers for NK cells in renal cell carcinoma and periampullary adenocarcinomas demonstrate that NK cells are found excluded from tumor regions in regions rich in stromal cells^211–213. Stromal cell interactions in liver cancer can cause NK cells to become quiescent through CXCL12-CXCR4 signaling²¹⁴, likely decreasing the lifespan of NK cells in tumors, although the epigenetic mechanisms by which this occurs remain elusive. In the presence of PGE₂ derived from stroma-rich “cold tumors”, cDC1s become dysfunctional through decreased IRF8 expression and produce less IL-12²¹⁵, which can enhance NK cell persistence through epigenetic remodeling. cDC1 and NK cells are often located near each other in tumors, likely because of high NK cell-mediated production of the cDC1-specific chemokine Xcl1 in tissue^119,201,216. However, PGE₂ can also directly decrease NK cell abundance in tumors by either suppressing NK cell viability or indirectly by limiting interactions with IL-15-producing cDC1 due to lower Xcl1 production²⁰¹, thus limiting NK cell survival. Thus, a combination of PGE₂ and CXCL12 in addition to proximity with activated or suppressed cDC1 in tumors likely influences the overall epigenetic state of NK cells to dictate their survival in tumors, although further experiments will be necessary to confirm this hypothesis.

While typically excluded from the tumor core, NK cells can undergo epigenetic remodeling in response to tumor cell contact. αV-integrin-mediated TGF-β activation following NK cell contact with glioblastoma stem cells can cause NK cell dysfunction²¹⁷, likely due to p300-mediated histone modifications downstream of SMAD signaling²¹⁸. Similarly, co-culture of NK cells with breast cancer organoids induced a hyporesponsive NK cell state associated with increased expression of TIGIT, LAG3, and KLRG1, reduced expression in genes associated with NK cell activation and proliferation, and decreased control of metastatic tumor cells in vivo²¹⁹. This effect was found to be reversed by treatment of co-cultured NK cells with a DNA methyltransferase inhibitor. Similarly, inhibition of Ezh2 function during NK cell development enhances antitumor function²²⁰, suggesting that altering histone methylation status can be used to counteract inhibition of NK cells in the TME. Recent studies have suggested that the expression of critical activating and inhibitory receptors on NK cells, such as NKG2A and NKG2D, are also regulated by histone methylation by histone demethylases UTX and JMJD3, and the histone methyltransferases Ezh2^71,221–223. While NKG2D signaling does not impact NK cell survival, trogocytosis of NKG2D ligands by tumor-interacting NK cells can render them susceptible to fratricide by other NK cells in a Rae-1-dependent manner²²⁴. Interestingly, NK cell-mediated fratricide can be suppressed in CAR-NK cells through the expression of inhibitory receptor domains fused to the CS1 co-receptor expressed in NK cells to greatly enhance their antitumor function through enhanced persistence and effector function²²⁵. As Fas-mediated fratricide can increase in response to continuous stimulation with IL-2 and IL-15 in vitro^226,227, and NKG2D ligand expression increases in cytokine-activated NK cells^183,228, the cytokine-induced epigenetic mechanisms that regulate Fas and NKG2D ligand expression on endogenous NK cells will require further study to enhance the survival of NK cells in the TME by decreasing their susceptibility to fratricide.

7. Conclusions and Future Outlook

NK cells have a broad functional repertoire that is influenced by the epigenetic landscape created throughout development, inflammation, and unique tissue microenvironments. Since NK cells are innate immune cells that intrinsically have a short lifespan, they must rely on external stimuli for prolonged survival, which allows them to bridge the gap between the innate and adaptive immune response. Future research will need to focus on the heterogenous epigenetic cell states at the single cell level that can be induced by activating signals and their combinations ex vivo, increasing our understanding of the mechanisms that support the enhanced lifespan of NK cells following activation. However, caution must be taken to not only study mature NK cell responses since the shaping of the epigenetic landscape starts early during NK cell development and continues to influence the function and survival of mature NK cells. Likewise, the metabolism of activated NK cells requires the presence of the appropriate building blocks in the local environment to fuel effector functions and epigenetic modifications. We are only beginning to understand the metabolic requirements for activated mouse and human NK cells, and future metabolomics and lipidomics experiments will be necessary to understand the interplay between cellular energetics and epigenetic modifications. In addition, the precise signals that govern the distinct epigenetic states observed in trNK populations will require further study. Advances in spatial sequencing and single-cell ATAC sequencing will aid our understanding of the cell-cell interactions in intact tissues that govern the epigenetic state of trNK cells. While studies from the past two decades have uncovered many of the molecular players that regulate NK cell survival, a clearer understanding of the transcription factor networks and epigenetic modifications that influence the lifespan of NK cells will be critical in developing more efficacious NK cell immunotherapies for cancer and severe viral infections.