Ascorbic Acid Promotes KIR Demethylation during Early NK Cell Differentiation

This article is featured in In This Issue, p.1477

Natural killer cells are cytotoxic lymphocytes that are important for the control of viruses and malignancies. NK cell recognition of abnormal tissue occurs through a process known as “missing-self” whereby absent or reduced expression of self-MHC molecules on the surface of aberrant cells triggers NK cell cytotoxicity (1). There are two sets of genes encoding NK cell receptors that bind MHC molecules. One set is the NK complex (NKC) of lectin-like genes located on chromosome 12p13.1. Among the genes within the NKC are KLRD1 (encoding CD94) and KLRC1 (encoding NKG2A) (2). CD94 forms heterodimers with NKG2A on the cell surface and binds the nonclassical HLA-E. Triggering of CD94/NKG2A induces an inhibitory signal consistent with the presence of two ITIMs within the cytoplasmic domain of NKG2A that recruit the tyrosine phosphatases SHP-1 and SHP-2 (3–7). The other gene set consists of the leukocyte receptor complex (LRC) of Ig-like genes on chromosome 19q13.4. Within the LRC is the killer Ig-like receptor (KIR) locus (2). The KIR genes are arranged ∼2 kbp apart in a head-to-tail orientation from centromere to telomere, with just one unique section of sequence 14 kbp upstream of KIR2DL4. KIR gene sequences are over 90% identical, and this gene family likely evolved through extensive duplication/deletion and intergenic sequence exchange (8). With the exception of KIR2DL4 (which binds HLA-G), the ligands for KIR are classical MHC class I molecules (HLA-A, -B, and -C) (9–11). Functionally, there are two types of KIR: stimulatory and inhibitory. Stimulatory KIR have short cytoplasmic tails and associate with DAP12 to transduce strong activating signals (12, 13). Inhibitory KIR have long ITIM-containing cytoplasmic tails that recruit tyrosine phosphatases to dampen activation (14).

Interactions between KIR and HLA have been exploited therapeutically in donor selection for allogeneic hematopoietic cell transplantation and adoptive NK cell transfer to treat patients with advanced cancer. NK cell alloreactivity driven by KIR ligand mismatch has been shown to induce durable remissions in acute myelogenous leukemia patients transplanted with T cell–depleted allogeneic hematopoietic grafts (15), and donor KIR haplotypes have been shown to impact relapse-free survival after T cell replete–unrelated hematopoietic cell transplantation (16–18). Induction of complete hematologic remission in poor-prognosis acute myelogenous leukemia patients has been achieved with adoptive transfer of haploidentical NK cells (19), and response rates were improved when combined with T regulatory cell depletion (20).

With the exception of KIR2DL4, which is expressed on all peripheral blood CD3⁻CD56⁺ NK cells, KIR are expressed predominantly within the CD56^dim NK cell population in a variegated pattern (21, 22). This pattern is maintained at the allelic level in CD56^dim NK cells by DNA methylation at CpG sites within the promoter regions proximal to KIR transcriptional start sites (23, 24). How the methylation patterns within the KIR locus are established in NK cells is still unknown. In CD34⁺ hematopoietic progenitor cells, KIR genes are densely methylated, inaccessible to nuclease digest and exhibit a repressive histone signature (25). Thus, there is a presumptive stage during NK cell maturation in which epigenetic remodeling and DNA demethylation occurs within the KIR locus to allow for gene expression.

Previous analyses of the expression pattern of KIR2DL5 provide potential insight into the mechanism of KIR promoter demethylation. The KIR2DL5 promoter is remarkably diverse, as there are three unique promoter sequences that differ by 1.61–2.46% in their nucleotide sequence. One sequence is shared by the KIR2DL5B*002 and *004 alleles, and the other two promoter sequences are found within the KIR2DL5A*001 and KIR2DL5B*003 alleles. This promoter diversity likely explains the expression pattern of KIR2DL5. Whereas NK cells and CD8⁺ T cells transcribe the KIR2DL5A*001 and KIR2DL5B*003 alleles in a variegated manner, the mRNA of the KIR2DL5B*002 and *004 alleles are undetectable by RT-PCR (26). The expression profile of KIR2DL5 alleles correlates with a single nucleotide polymorphism in an Runx binding site ∼100 bp upstream of the translation start codon. This binding site recruits Runx3, but not Runx1 or Runx2. Runx3 binding likely plays a key role in promoter demethylation, as the promoters of silent KIR2DL5 alleles can drive transcription of reporter genes when transfected into NK cell lines, and pharmacological DNA demethylation results in de novo transcription from silent KIR2DL5 alleles (27).

Over the past several decades, different models have been put forth to explain the mechanisms underlying DNA demethylation. The prevailing model had been that an absence or reduction in the levels of DNA methyltransferase (DNMT) enzymes in cells leads to a gradual and passive removal of DNA methylation in cells as they divide (28, 29). Other studies described demethylation of DNA through a more rapid and active mechanism independent of cell division (30–32). However, a mammalian enzyme capable of cleaving methyl groups remained elusive. A flurry of recent studies has shown that ten-eleven translocation (TET) enzymes catalyze the conversion of 5-methylcytosine (5mC) to 5-hydroxymethylcytosine (5hmC), which can then be deaminated and replaced with cytosine through DNA glycosylase-mediated excision repair (33–36) or passively diluted during DNA replication (37, 38). The enzymatic activity of TET proteins is markedly enhanced by ascorbic acid (vitamin C), and TET enzymes are enriched near the transcriptional start sites of genes affected by ascorbic acid treatment (39–41). Ascorbic acid can interact with the C-terminal catalytic domain of TET enzymes, which may promote protein folding and/or recycling of the TET cofactor Fe(2+) (40).

In light of the recent literature showing a role for TET proteins in active DNA demethylation, we sought to determine whether TET enzyme activity contributes to demethylation of KIR genes in human NK cells and at which stage of NK cell development this may occur. In this study, we show that the promoters of stochastically expressed KIR genes are densely methylated in CD56^bright NK cells and progressively demethylated in CD56^dimCD94^high and CD56^dimCD94^low NK cells. Ex vivo culture of sorted CD56^bright, but not CD56^dim NK cells with increasing concentrations of ascorbic acid resulted in increased frequencies of KIR expression and KIR promoter demethylation. CD56^bright NK cells cultured with ascorbic acid exhibited significant enrichments of TET2, TET3, and Runx3 binding within the proximal promoters of stochastically expressed KIR genes. Furthermore, overexpression of TET3 in CD56^bright NK cells combined with ascorbic acid supplementation resulted in elevated frequencies of KIR expression after ex vivo culture. Double transduction of TET3 and Runx3 in combination with ascorbic acid strongly induced KIR expression in the NK-92 cell line. Together, our results show that the process of KIR demethylation is facilitated by ascorbic acid at an early stage of NK cell maturation and that TET enzymes and Runx3 contribute to this process.

Peripheral blood products obtained by healthy donor apheresis were purchased from Memorial Blood Bank (Minneapolis, MN). All studies were approved by the institutional review committee at the University of Minnesota. All studies involving human subjects were conducted in accordance with the guidelines of the World Medical Association’s Declaration of Helsinki. PBMCs were isolated by Ficoll-Paque (GE Healthcare, Chicago, IL) density gradient centrifugation. To isolate NK cells, T and B cells were first depleted using anti-CD3 and anti-CD19 magnetic beads (STEMCELL Technologies, Vancouver, BC, Canada) according to the manufacturer’s protocol. The remaining cells were then stained with anti-CD3 (OKT3), anti-CD56 (HCD56), and anti-CD94 (DX22) fluorochrome-conjugated Abs (BioLegend, San Diego, CA) at a concentration of 1 μl per 1 × 10⁶ cells. NK cell subsets were sorted using an FACS Aria II instrument (BD Biosciences, San Diego, CA). Sorted cells were cultured at a concentration of 0.5 × 10⁶ cells per milliliter in B0 media (DMEM plus Ham’s F-12 media, 2:1, supplemented with 10% heat-inactivated human AB sera, 1% penicillin-streptomycin, 25 μM 2-ME, and 0.5 μg/ml sodium selenite). Recombinant human IL-15 (10 ng/ml) (National Cancer Institute, Bethesda, MD) was added at the beginning of culture. Ascorbic acid (Sigma-Aldrich, St. Louis, MO) was added at various concentrations depending on the experimental design. In some experiments the sodium-dependent vitamin C transporter (SVCT) inhibitor sulfinpyrazone (Sigma-Aldrich) was added at a concentration of 0.2 mM. NK cells were cocultured with irradiated EL08-1D2 murine embryonic liver cells (provided by Dr. Elaine Dzierzak, Erasmus University) to promote ex vivo proliferation (42). NK-92 cells (American Type Culture Collection) were cultured in DMEM-α media containing 12.5% FCS, 12.5% horse serum (HyClone Laboratories, Logan, UT), 1% penicillin-streptomycin, 0.2 mM inostitol, 0.1 mM μM 2-ME, 0.02 mM folic acid, and 500 U/ml recombinant human IL-2 (Chiron, Emeryville, CA).

Cultured primary NK cells and NK-92 cells were stained with the LIVE/DEAD Fixable Dead Cell Staining Kit (Thermo Fisher Scientific, Waltham, MA) and anti-CD56 (HCD56), anti-CD3 (OKT3), anti-KIR2DL1 (HP-MA4), anti-KIR2DL2/L3 (DX27), anti-KIR3DL1 (DX9), and anti-CD94 (DX22) flurochrome-conjugated Abs (all from BioLegend) and analyzed on an LSR II Flow Cytometer (BD Biosciences). Data were analyzed using FlowJo (v10.6.0) software (BD Biosciences). For all flow cytometry experiments, lymphocytes were first gated as determined by size in forward-/side-scatter plots. NK cells were then gated as CD3⁻CD56⁺. For transduction experiments, cells were further gated as GFP⁺, mCherry⁺, or blue⁺.

Freshly isolated or cultured NK cell subsets were harvested, and DNA was isolated using the DNeasy Blood and Tissue Kit (Qiagen, Hilden, Germany). Bisulfite conversions of DNA were performed using the TrueMethyl Whole-Genome Kit (Cambridge Epigenetix, Cambridge, U.K.). For selection of specific regions spanning the KIR2DL1 and KIR3DL1 proximal promoters, forward and reverse primers were designed using MethPrimer (43). Primer sequences are listed in Supplemental Table I. The following PCR components were combined in a total volume of 25 μl: 5 μl of Expand High Fidelity Plus buffer without MgCl₂, 1 μl of 10 mM dNTPs, 2.5 μl of 25 mM MgCl₂, 11.5 μl of nuclease-free H₂O, 1 μl of 10 mM each forward and reverse primers, and 2 μl of DNA template. All PCR reagents were from Invitrogen (Carlsbad, CA). Thermal cycling was performed as follows: 35 cycles of denaturation at 94°C for 2 min, annealing at 60°C or 53°C for 1 min, and extension at 72°C for 30 s followed by 10 min at 72°C. The PCR products were purified and cloned into the pCR4-TOPO vector using the TOPO TA Cloning Kit (Invitrogen). Independent clones were sequenced by Sanger sequencing. Sequences were analyzed online with Bisulfite Sequencing DNA Methylation Analysis software using the FASTA format (44).

Quantitative real-time PCR was used to analyze mRNA transcript levels for KIR2DL4, KIR2DL2, KIR2DS3, KIR2DS4, KIR2DL1, KIR2DL2, KIR2DL3, and KIR2DL5, and was performed using custom TaqMan primers and probes (Thermo Fisher Scientific). Reactions were run on a 7500 Real-Time PCR System (Applied Biosystems, Foster City, CA) using TaqMan Gex Master Mix (Thermo Fisher Scientific). All data were normalized to ACTB (Thermo Fisher Scientific). Primer and probe sequences for amplifying individual KIR genes have been published previously (45).

Chromatin immunoprecipitation (ChIP) assays were performed as previously described (46). Briefly, aliquots of 2.5 × 10⁶ formaldehyde-fixed cells were resuspended in 50 μl of nuclei isolation buffer (Abcam, Cambridge, U.K.), and chromatin was digested with 15 U MNase (Thermo Fisher Scientific) for 5 min at 37°C. EDTA was added to stop the reaction. After enzymatic digestion, Abs against Runx1 (ab23980; Abcam), Runx2 (AF2006; R&D Systems, Minneapolis, MN), Runx3 (353604; BioLegend), TET2 (PA5-35847; Thermo Fisher Scientific), and TET3 (PA5-34431; Thermo Fisher Scientific) were added to precipitate the sheared chromatin. An IgG isotype Ab (MAB004; R&D Systems) was also added for a background control. After washing steps, protein and DNA complexes were eluted and cross-links were reversed. Purified DNA samples were analyzed by quantitative real-time PCR with primers listed in Supplemental Table I.

To construct the TET3-mCherry and Runx3-blue expression plasmids, the open reading frames of TET3 and RUNX3 were amplified by PCR using the primers listed in Supplemental Table I. The PCR products were cloned into the pMSCV-IRES-GFP (86537; addgene, Cambridge, MA), pMSCV-IRES-mCherry (80139; addgene), and pMSCV-IRES-Blue FP (52115; addgene) vectors using the NEBuilder HiFi DNA Assembly Kit (New England Biolabs, Ipswich, MA).

For virus production, the Platinum-E Retroviral Packaging Cell Line (Cell Biolabs, San Diego, CA) was transfected with 25 μg of control, TET3, or RUNX3 plasmid along with 7.5 μg pCMV-VSV-G and 7.5 μg pUMVC plasmids (addgene) using 1 μg/μl polyethylenimine (Sigma-Aldrich). Viral supernatants were collected after 48 h and used for transduction. A total of 1 × 10⁴ sorted CD56^bright NK cells were plated per well in 96-well plates containing confluent layers of irradiated EL08-1D2 cells. One hundred microliters of viral supernatant was added to each well, and spin transductions were performed for 2 h at 360 × g.

For RNA sequencing (RNA-seq), RNA was isolated from cultured cells using the RNeasy Mini Kit (Qiagen). RNA quality was analyzed using High Sensitivity RNA ScreenTape Analysis (Agilent Technologies, Santa Clara, CA). RNA was then multiplexed and sequenced on a NextSeq 550 Sequencing system with 150-paired end output (Illumina, San Diego, CA). Quality of data in fastq files was assessed using FastQC. Low quality bases and adapter sequences were removed using Trimmomatic. Reads were aligned using Hisat2. Fragments per kilobase of transcript per million mapped reads expression values were generated using Cuffquant and Cufnorm from the Cufflinks package, and Raw read counts were generated using featureCounts from the Subread R package. For whole-genome DNA methylation analyses, DNA was isolated from cultured cells using the DNeasy Blood and Tissue Kit (Qiagen). DNA was quantified, bisulfite-converted, and analyzed on Illumina Infinium Methylation EPIC BeadChips (Illumina) at the University of Minnesota Genomics Center according to the manufacturer’s instructions. Data were analyzed using GenomeStudio software (Illumina). Raw and processed data have been submitted to the Gene Expression Omnibus database curated by the National Center for Biotechnology Information (GSE152571, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE152571).

Data in all figures are expressed as the mean ± SEM. One-way ANOVA with multiple comparisons was used to determine statistical significance. Statistical analyses were performed using the GraphPad software in Prism 8 (v.8.3.1).

Mature human NK cell subsets can be distinguished phenotypically on the basis of their relative density of CD56 surface expression. Peripheral blood NK cells are composed of CD56^bright (∼5–10%) and CD56^dim (∼90–95%) subsets. CD56^bright NK cells express high levels of CD94/NKG2A but very low levels of surface receptors associated with NK cell maturation, including KIR, CD16, and CD57 (47–49). Although there is experimental evidence suggesting that NK cell development proceeds from a CD56^bright to CD56^dim phenotype (50–52), definitive in vivo proof in humans has been lacking. Nevertheless, one plausible model for late stage human NK cell development supported by phenotypic and functional analyses of subsets is that CD56^bright NK cells transition into CD56^dimCD94^high intermediate stage and then further mature into a CD56^dimCD94^low phenotype (53). We sought to determine whether proximal KIR promoters were progressively demethylated in accordance with this model of development. To this end, we isolated PBMCs from two healthy donors and analyzed KIR expression by FACS. In agreement with previous work (53), we found that KIR2DL1 and KIR3DL1 were expressed at low frequencies on CD56^bright NK cells, exhibited intermediate frequencies on CD56^dimCD94^high NK cells and were expressed at the highest frequencies on CD56^dimCD94^low NK cells (Fig. 1A). We then sorted CD56^bright, CD56^dimCD94^high, and CD56^dimCD94^low NK cells from these two donors and performed bisulfite sequencing using primers specific for the KIR2DL1 and KIR3DL1 proximal promoters. CD56^bright NK cells from both donors exhibited dense methylation within the proximal promoters of KIR2DL1 and KIR3DL1. Demethylation of both promoters was evident in CD56^dimCD94^high NK cells, coinciding with receptor expression. Interestingly, we observed similar levels of KIR2DL1 and KIR3DL1 promoter demethylation in CD56^dimCD94^low cells despite higher frequencies of surface receptor (Fig. 1B). Our results suggest that KIR proximal promoter demethylation primarily occurs between the CD56^bright and CD56^dimCD94^high stages of NK cell development.

Because ascorbic acid is a known catalyst of TET activity and DNA demethylation (39–41), we hypothesized that addition of ascorbic acid to NK cells cultured ex vivo could facilitate KIR acquisition. To test this hypothesis, we sorted CD56^bright NK cells, CD56^dimCD94^high NK cells and CD56^dimCD94^low NK cells and cultured all three populations on irradiated EL08-1D2 feeder cells with 10 ng/ml IL-15 and either DMSO or increasing concentrations of ascorbic acid for 7 d. Cells were then harvested and analyzed by FACS for expression of KIR2DL1, KIR2DL2/3, and KIR3DL1. We observed a dose-dependent increase in the frequency of KIR expression for CD56^bright NK cells cultured with ascorbic acid, whereas no effect was observed for CD56^dimCD94^high NK cells or CD56^dimCD94^low NK cells (Fig. 2A–C). The maximal effect of ascorbic acid on KIR expression was at 50 ng/ml. Because no decreases in viability or cell counts postculture were observed at 50 ng/ml, we performed all subsequent experiments with ascorbic acid at this concentration.

To confirm that the increases in KIR expression frequencies with ascorbic acid occurred at the transcriptional level, we performed quantitative real-time RT-PCR with primers specific for KIR2DL4, KIR2DS2, KIR2DS3, KIR2DS4, KIR2DL1, KIR2DL2, KIR2DL3, and KIR2DL5 using mRNA from CD56^bright NK cells cultured in either DMSO or ascorbic acid. With the exception of KIR2DL4, which contains a unique promoter region and is ubiquitously expressed on NK cells (8, 21, 22), we observed elevated levels of all KIR transcripts in CD56^bright NK cells cultured with ascorbic acid (Fig. 2D). To verify that ascorbic acid treatment impacted KIR promoter demethylation, we took DNA from CD56^bright NK cells cultured in either DMSO or 50 ng/ml ascorbic acid and performed bisulfite sequencing to analyze the KIR2DL1 and KIR3DL1 proximal promoters. For CD56^bright NK cells cultured with DMSO, both KIR promoters remained densely methylated. Importantly, significant demethylation was observed for both KIR promoters in CD56^bright NK cells cultured with ascorbic acid (Fig. 3).

Two isoforms of ascorbic acid cotransporters have been identified and are named SVCT1 and SVCT2. SVCT1 is located mostly in epithelial tissues, whereas SVCT2 has a wider distribution (54). To determine whether the effect of ascorbic acid on KIR expression could be blocked through inhibition of ascorbic acid transport, we sorted CD56^bright NK cells and cultured them with ascorbic acid in the presence or absence of the SVCT2 inhibitor sulfinpyrazone (55). We observed a partial inhibition of KIR induction with sulfinpyrazone treatment, suggesting that ascorbic acid transport into CD56^bright NK cells is at least in part dependent on SVCT2 (Supplemental Fig. 1). Together, our data show that ascorbic acid facilitates demethylation of proximal KIR promoters in CD56^bright NK cells, leading to increased frequencies of KIR expression.

Having demonstrated the effect of ascorbic acid on KIR promoter demethylation, we next sought to determine whether we could detect binding of TET family proteins within KIR proximal promoters. We were also interested in whether TET proteins were located in the same vicinity as Runx transcription factors given previous work demonstrating the importance of the Runx binding site to expression of KIR2DL5 (26, 27). To this end, we sorted CD56^bright NK cells and cultured them for 7 d with either DMSO or ascorbic acid. Cells were then harvested for ChIP experiments using Abs against Runx1, Runx2, Runx3, TET2, and TET3 and primers specific for the proximal promoters of KIR2DL1 and KIR3DL1. We found that Runx3, TET2, and TET3 were enriched within the promoters of both KIR genes in CD56^bright NK cells cultured with ascorbic acid. No statistically significant enrichment of Runx1 or Runx2 was observed (Fig. 4).

To substantiate the role of TET proteins in driving KIR expression, we sorted CD56^bright NK cells and retrovirally transduced them with a TET3 overexpression vector containing a GFP reporter or a control vector containing GFP only. Cells were cultured for 5 d in media containing either DMSO or ascorbic acid and analyzed by FACS to determine the frequencies of KIR expression on transduced cells. We observed significantly higher KIR frequencies on CD56^bright NK cells overexpressing TET3 relative to GFP⁺ control cells in DMSO cultures. Ascorbic acid increased KIR expression on GFP- and TET3-overexpressing cells, with the highest frequency of KIR expression observed on TET-overexpressing cells cultured with ascorbic acid (Fig. 5A).

We performed similar experiments using the NK cell line NK-92, which expresses low-to-absent levels of KIR. NK-92 cells were transduced with either a control vector containing an mCherry reporter only or a TET3 overexpression vector with an mCherry reporter. Transduced cells were sorted and expanded in vitro. We then took a fraction of NK cells transduced with the TET3 overexpression vector and transduced these cells with an Runx3 overexpression vector containing a blue reporter. Cells were then cultured for 5 d in media containing either DMSO or ascorbic acid and analyzed by FACS to determine the frequencies of KIR3DL1, KIR2DL2/3, and KIR2DL1 expression. For each KIR, we found that expression was induced to varying degrees only on cells that were transduced with both the TET3 and Runx3 overexpression vectors and cultured in media containing ascorbic acid (Fig. 5B). Together, these data demonstrate direct roles for TET3 and Runx3 in promoting KIR expression.

Motivated by our finding that ascorbic acid facilitated KIR promoter demethylation, we were interested in determining whether there were broad transcriptional or epigenetic changes induced by ascorbic acid in cultured CD56^bright NK cells. To this end, we sorted CD56^bright NK cells and cultured them for 7 d with either DMSO or ascorbic acid. RNA expression levels from these cells were analyzed by RNA-seq, and whole-genome DNA methylation was analyzed using Illumina Infinium MethylationEPIC BeadChips. Surprisingly, the transcriptional profiles of CD56^bright NK cells cultured with DMSO were nearly identical to those of CD56^bright NK cells cultured with ascorbic acid. This was evident in principal component analysis (Fig. 6A) and hierarchical clustering (Fig. 6B) of the RNA-seq data, where the only major differences in transcriptome profiles were due to intraindividual variability. The only transcript that was significantly upregulated by ascorbic acid in this analysis was KIR3DL1 (4.6-fold, p = 0.003). Similarly, in our analyses of whole-genome DNA methylation profiles, only a small fraction of CpG sites exhibited >2-fold differences in methylation levels between the DMSO and ascorbic acid culture conditions (Fig. 6C), and these differences were not consistent between donors. Furthermore, whole-genome DNA methylation profiles grouped by donor rather than NK cell treatment in hierarchical clustering analyses (Fig. 6D). We conclude that promoter DNA demethylation and elevated transcription in CD56^bright NK cells in response to ascorbic acid treatment is largely restricted to the KIR locus.

In recent years, it has become clear that TET enzymes play a key role in DNA demethylation within tissue-specific gene regulatory elements. Studies in mice have shown that TET enzymes augment DNA demethylation of conserved noncoding sequence elements within the Foxp3 gene in regulatory T cells and the Aicda superenhancer that controls expression of activation-induced cytidine deaminase (AID) in B cells (55–57). The addition of ascorbic acid to cultured regulatory T cell stabilizes Foxp3 expression and facilitates conserved noncoding sequence element demethylation (55, 56). TET enzymes are also involved in lineage-specification of B cells and invariant NKT cells (58, 59). In this study, we show that KIR proximal promoters exhibit active demethylation between the CD56^bright and CD56^dimCD94^high stages of human NK cell development, and the addition of ascorbic acid to cultured CD56^bright NK cells resulted in an increased frequency of KIR promoter DNA demethylation and surface KIR expression. It is interesting to note that increases in KIR frequencies were not observed for CD56^dimCD94^high or CD56^dimCD94^low NK cells that were cultured with ascorbic acid. This result is in line with previous studies showing that during lymphoid differentiation, the most striking consequences of TET loss of function are observed in cells undergoing rapid proliferation (58, 59). The most likely explanation for this phenomenon is that methyl groups, such as 5hmC, on DNA are stable and are primarily lost as a consequence of passive dilution during DNA replication (60, 61). CD56^bright NK cells represent the most highly proliferative NK cell subset in peripheral blood (62), and these high rates of proliferation are likely necessary for passive dilution and complete removal of methyl groups within proximal KIR promoters.

Importantly, we also found that TET2, TET3, and Runx3, but not Runx1 or Runx2, were enriched within KIR proximal promoters in CD56^bright NK cells that were cultured with ascorbic acid. TET3 overexpression in CD56^bright NK cells and NK-92 cells also resulted in an increased frequency of KIR expression. Because of the association between Runx3 binding and expression of KIR2DL5 (26, 27) it is tempting to speculate that Runx3 functions to promote KIR expression by recruiting TET proteins to the proximal promoters of KIR genes during NK cell differentiation. TET2 haploinsufficiency has been associated with increased DNA methylation in regions surrounding RUNX binding sites (63). Runx1 has been shown to be physically associated with TET2 and TET3 (64). Runx3 is the major Runx family member expressed by NK cells and cytotoxic T cells (65). Support for Runx3 as a transcription factor involved in DNA demethylation comes from a study in which 293T cells were transduced with lentivirus vectors containing overexpression constructs for several different transcription factors, including Runx3, and analyzed by DNA methylation arrays. In this system, 114 promoter-associated regions were demethylated in response to Runx3 overexpression (66). Our overexpression studies support a role for both TET3 and Runx3 in promoting KIR expression. Additional experiments testing TET2 and TET3 knockout alone and in combination will be necessary to definitively determine the relative contributions of these proteins in driving KIR promoter demethylation and gene expression.

Our updated model for regulation of KIR gene expression is that transcription from the distal promoter ∼1 kb upstream of the transcriptional start site is initiated in NK cell precursors and is associated with activation of the proximal promoter during NK cell development (67, 68). During the CD56^bright to CD56^dim transition, Runx3, TET2, and TET3 cooperate to induce DNA demethylation of the proximal promoter, which is bidirectional and can initiate transcription in either the sense or antisense direction in a probabilistic fashion (69). If antisense transcripts are produced, they can pair with transcripts from the distal promoter and be processed into a 28-base piwi RNA that silences KIR expression. If sense transcripts are produced, the activated state of the proximal promoter is maintained (70). Transcription from an intermediate KIR promoter located in between the proximal and distal promoters also contributes to KIR expression in mature NK cells (71).

In our transcriptome and whole-genome DNA methylation analyses of CD56^bright NK cells cultured with and without ascorbic acid, we were surprised at how few changes in gene expression and DNA methylation levels were observed. Similar findings were reported by Sasidharan et al. who found that ascorbic acid strongly promoted DNA demethylation of conserved noncoding sequence element 2 within Foxp3, but only 36 genes were upregulated more than 2-fold by microarray analysis (50). An interesting question that remains to be answered is how the TET protein-dependent activity of ascorbic acid can influence DNA demethylation of regulatory elements within tissue-specific genes such as KIR and FOXP3 with such apparent specificity. This could reflect instances in which there is competition between active demethylation by TET2/3 and antisense RNA-mediated silencing. High expression or activation of TET enzymes could tip the balance in favor of demethylation.

We thank the University of Minnesota Genomics Center, the University of Minnesota Informatics Institute, and the University of Minnesota Flow Cytometry Core for their services.

F.C. consults for Fate Therapeutics and has received research funds from this relationship. J.S.M. serves on the Scientific Advisory Board and consults for GT BioPharma and Fate Therapeutics. He has received research funds from these relationships. J.S.M. also serves on the Scientific Advisory Board for CytoSen and Onkimmune. None of these companies had a role in funding this research. All conflicts are managed according to institutional policies. The other authors have no financial conflicts of interest.