Lineage-Specific Transition of Histone Signatures in the Killer Cell Ig-Like Receptor Locus from Hematopoietic Progenitor to NK Cells¹

Killer cell Ig-like receptors (KIR)³ are a polymorphic gene family that is expressed by NK cells as well as subsets of T cells and comprise inhibitory, as well as stimulatory, receptors (1, 2). Most, if not all, inhibitory KIR are specific for allotypes of the polymorphic HLA class I family and enable NK cells to recognize aberrant levels of HLA class I molecules, which are frequently found on virus-infected or malignant cells (1). In contrast, the function and specificity of stimulatory KIR is still elusive, albeit genetic evidence is accumulating that they are involved in recognition of viral structures (3).

An important feature of KIR genes is their clonally-distributed expression, which leads to a repertoire of structurally and functionally distinct NK cells. Individual KIR expression patterns are established during differentiation from early NK cell progenitors and are stably maintained upon clonal expansion of mature NK cells (4, 5). A comparable expression mode is also found in rodent Ly-49 genes, a structurally unrelated but functionally equivalent family of NK cell receptors (6). An important physiological role of the stochastic receptor expression mode probably is the creation of diversity, which similar to T and B cell diversity, leads to a broad spectrum of NK cell specificities. Consequently, lack of NK cell diversity might lead to immunodeficiency (7). In contrast, an important issue of randomly distributed repertoires in the immune system is maintenance of tolerance (8, 9). The rules of NK cell tolerance induction and the implications for KIR expression are just starting to become unraveled: whereas the majority of NK cells express at least one “fitting” inhibitory receptor and thus is tolerant to self (4), a recently described subset of NK cells does not express any inhibitory receptor for self HLA class I (10, 11, 12). Notably, those cells are hyporesponsive and thus are not part of the functional NK cell repertoire. It should be noted that KIR expression is not entirely random – genetic polymorphism of HLA class I-encoded KIR ligands was recently described to shape NK cell repertoires by increasing frequencies of cognate inhibitory KIR (13).

The expression of KIR genes is, at least partly, under epigenetic control (14). The transcriptional activity of KIR genes is tightly regulated by the degree of DNA methylation of a CpG cluster overlapping the promoter region (15, 16, 17). Whereas in active KIR genes, the CpG cluster is largely unmethylated, silent KIR genes exhibit dense DNA methylation. All KIR genes except the recently discovered primordial KIR3DL0 are tightly clustered in head-to-tail configuration as part of the leukocyte receptor complex on chromosome 19q13.4 (18, 19). This leads to a clone-specific patchwork pattern of DNA methylation on the KIR locus. Experimentally, demethylation of KIR genes by 5-aza-2′-deoxycytidine leads to rapid induction of KIR expression, whereas in vitro DNA methylation of the CpG cluster leads to inhibition of KIR promoter activity (15).

It is so far unknown how epigenetic silencing of KIR genes by DNA methylation is controlled. It was previously reported (20) that the regulation of DNA methylation is associated with changes in chromatin accessibility and specific covalent histone modifications. Some of the best-studied modifications are acetylation of lysines, e.g., at lysine 9 of histone H3 (H3K9ac) or lysine 8 of histone H4 (H4K8ac), which is associated with euchromatin and gene activation. Lysine methylation either is associated with active genes, e.g., dimethylation of lysine 4 at histone H3 (H3K4dime) or heterochromatin, e.g., dimethylation of lysine 9 at histone H3 (H3K9dime) (21). Other markers such as trimethylation of lysine 27 of histone H3 (H3K27trime) are context-dependent and can be found in heterochromatic regions, but also in regions which are destined for later expression, for example in lineage-specific genes of embryonic stem cells when the modification is found in combination with the H3K4dime mark (22, 23, 24).

Different histone marks are read out by different chromatin-associated proteins. The H4K8ac mark provides a docking site for the bromodomain-containing subunits of the SWI/SNF chromatin-remodeling complex that facilitates access of the basic transcription machinery to the promoter (25). In contrast, the repressive H3K9dime mark represents a recognition site for chromodomain-containing proteins, such as heterochromatin protein 1, which are involved in induction and maintenance of heterochromatin (26). The combination of histone marks, referred to as histone signature, is believed to carry information not only about the current activity state of a given gene but also about its potential fate in the course of differentiation and development (27).

The KIR gene cluster provides the opportunity to compare the epigenetic features of members of a highly homologous gene family, some of which are active and others silent in the same cell. The purpose of the present study was to find out how the different layers of epigenetic regulation, namely DNA methylation, chromatin packaging, and histone modification, change during differentiation starting from hemopoietic stem cells to the mature lymphoid lineages of NK, T, and B cells. The comparative analysis shows that the KIR locus is already DNA-methylated in early hemopoietic progenitor cells (HPC) and remains so in all lymphocytes that do not express KIR. Whereas DNA demethylation was found to be gene-specific and tightly associated with KIR expression in NK cells, the acquisition of active histone modifications was lineage-specific and apparently precedes the DNA demethylation step.

Samples from peripheral blood and G-CSF-mobilized peripheral blood were taken from healthy donors after informed consent. Placental cord blood was collected as previously described with the informed consent of the mothers (28). All samples were typed for KIR gene polymorphisms as described previously and samples with genotype 1 (homozygous for group A haplotype) were chosen for all experiments unless indicated otherwise (29). Lymphocyte-enriched fractions were isolated from the different blood sources by density gradient separation (Biocoll separating solution; Biochrom). NK cells were enriched using the CD56 MultiSort kit (Miltenyi Biotec) according to the supplier’s instructions. When appropriate, NK cells were short term cultured in RPMI 1640 (Lonza) supplemented with 10% FCS, 5% human serum, 1% penicillin/streptomycin and 1000 U/ml IL-2 (proleukin; Chiron). CD34⁺ HPC were enriched from G-CSF-mobilized peripheral blood using the Direct CD34 Progenitor Cell Isolation kit (Miltenyi Biotec). B cells, CD4⁺, and CD8⁺ T cells were enriched from cord blood by staining with anti-CD19, -CD4, and -CD8 mAbs, respectively and subsequent magnetic cell sorting using anti-PE MicroBeads as secondary Ab (Miltenyi Biotec) to >98% purity. CD34⁺ cells and subsets of NK cells were further purified by flow cytometric cell sorting to a purity of >98% using the following mAbs: anti-CD34 for HPC enrichment and a combination of anti-CD3, -CD56, and either KIR-NKAT2 (anti-KIR2DL2/2DL3/2DS2) or NKB1 (anti-KIR3DL1) (the latter two from BD Biosciences) for enrichment of NK cell subsets divergent for KIR expression. Cell sorting was performed on a FACStar^Plus (BD Biosciences) equipped with a 2-W argon ion laser (Innova 70; Coherent) operating at 488 nm and 140 mW. All mAbs including isotype-matched control mAbs for IgG1 and IgG2a were purchased from Beckman Coulter unless otherwise noted. NK3.3 cells (KIR2DL4⁺ and KIR2DL3⁺) were provided by J. Kornbluth (St. Louis University, St. Louis, MO). The human bone marrow stromal cell line L88/5 was provided by K. Thalmeier (GSF, Munich, Germany), (30).

DNA was extracted using the QIAamp DNA Blood Mini kit (Qiagen). Genomic sequencing of bisulfite-converted DNA was performed as described (15). In short, bisulfite conversion was performed using the MethylEasy kit (Human Genetic Signatures). PCR primers for specific amplification of KIR promoters, promoter-associated Alu elements and exon 4 of various KIR genes are listed in Table I. Following digestion with EcoRI and BamHI, the amplification products were cloned into pBluescript SK⁺ (Stratagene). Sequence evaluation was performed with the BigDye Terminator Cycle Sequencing kit (Applied Biosystems) on a DNA analyzer (3700; Applied Biosystems) using the T7 primer (5′-TAATACGACTCACTATAGGG-3′).

To characterize chromatin accessibility MIRECAL (Micrococcus nuclease/real-time PCR chromatin accessibility assay with locus specificity) assays were performed as described previously (31) with some modifications. In short, nuclei were prepared using the Nuclei Ez Prep Nuclei Isolation kit (Sigma-Aldrich) and subsequently digested for different periods of time with Mnase (Sigma-Aldrich). Each sample contained nuclei of 2 × 10⁶ cells and was incubated with 0.5 U Mnase and 0.25 μmol CaCl₂ at 28°C. The reaction was stopped by adding 0.5 μmol EGTA on ice. The DNA concentration in each sample was adjusted to 15–20 ng/μl depending on the experiment. Primer sequences for amplification of KIR and GAPDH promoters are listed in Table I. Specificity of PCR conditions was verified by direct DNA sequencing of PCR products. Real-time PCR was performed with 40 ng template DNA using the QuantiTect SYBR Green PCR kit (Qiagen). After initial denaturation for 13 min at 95°C (for activation of HotStar TaqDNA polymerase as recommended by the manufacturer), 40 cycles were performed, consisting of 1 min at 95°C, 35 s at 57°C and 25 s at 72°C for the KIR promoters and 65 s at 95°C, 25 s at 63°C and 25 s at 72°C for the GAPDH promoter. Each PCR was run in triplicate to control for PCR variation. Data quantification for MIRECAL was performed as follows: C_T values of all samples were converted into template amount (in nanograms) via a calibration curve. The ratio of KIR promoter templates to GAPDH promoter templates was determined by dividing the template content of each sample by the corresponding GAPDH sample. Finally, these ratios were normalized to the ratio of KIR promoter templates to GAPDH promoter templates of the corresponding undigested sample resulting in the relative accessibility index (RAI):

ChIPs were performed using the ChIP Assay kit according to the manufacturer’s recommendations (Upstate Biotechnology). For shearing of genomic DNA to 200–800 bp fragments, the ultrasonic processor Vibracell 75022 (Novodirect) with a 2-mm probe was used with the following settings: amplitude, 30; time, 1 min; pulser, 6 × 10 s. From 2 ml of sonicated sample, 100 μl were taken as input control and 1900 μl for immunoprecipitation. Abs used for immunoprecipitations were anti-H4K8ac, anti-H3K4dime, anti-H3K9ac (Upstate Biotechnology), anti-H3K9dime, which was a gift from Thomas Jenuwein (IMP Vienna, Ref. 32), and rabbit anti-human Ig fraction as isotype control (3 μg/sample; DakoCytomation). Subsequently, input control as well as immunoprecipitates were amplified with specific primers using either conventional PCR (31 cycles) or real-time PCR (40 cycles), respectively. Real-time PCR was performed using the QuantiTect SYBR Green PCR kit (Qiagen). PCR were always conducted in triplicate using the following conditions: initial denaturation by 95°C for 14 min, 95°C for 50 s and T_M for 30 s and 72°C for 20 s. For relative quantitation of gene expression the ΔΔC_T method was applied as outlined in the Applied Biosystems protocol (Applied Biosystems). Real-time PCR assays to be compared were verified to have similar amplification efficiencies. First, a ΔC_T value was calculated for each sample by subtracting the mean value of triplicate C_T values of the normalizing input DNA sample from the mean value of triplicate C_T values of the precipitated output DNA sample. Second, a ΔΔC_T value was calculated by subtracting the ΔC_T for the “no antibody” control from each sample ΔC_T. Finally, the amount of precipitated sample DNA, normalized to the input reference and relative to the “no antibody” calibrator is given by 2^−ΔΔCT (relative fold enrichment). SD bars were determined by evaluating the value 2^−ΔCT with ΔΔC_T +s and ΔΔC_T–s, where “s” is the SD of the ΔΔC_T value. Data calculation leading to “fold differences” was conducted as described (33). In this case, sample values were normalized by subtraction of the nonspecific signal derived from the rabbit anti-human Ig fraction instead of calibration to the “no antibody” control.

We have previously shown (15) that clonally distributed expression of KIR genes is tightly coupled to the DNA methylation status of KIR promoters in NK cells. As it is so far unknown at which stage of NK cell development DNA methylation patterns are established on KIR genes, the DNA methylation status was determined in early HPC and the main lymphocyte lineages. As shown in Fig. 1, the KIR2DL4 promoter, which is constitutively active in NK cells, was unmethylated in primary NK cells but densely methylated in HPC (Fig. 1,A). In HPC, similar dense DNA methylation was observed in KIR2DL3 (Fig. 1,B) and other KIR genes that are expressed in a clonally-distributed fashion in NK cells, for example KIR3DL1 (data not shown). Moreover, dense DNA methylation patterns were also observed in the KIR2DL3 promoter of cord blood-derived (>99% KIR⁻) primary T cells, as well as B cells (Fig. 1,B). Notably, analyses of a CpG cluster in the coding region (exon 4) of various KIR (Fig. 1,C) as well as of an Alu element upstream of the KIR3DL1 promoter (969–703 bp upstream of start codon) (Fig. 1 D) showed dense DNA methylation not only in silent but also in expressed KIR genes. Thus, demethylation was confined to the promoter area of expressed KIR. The data suggest that KIR promoters are already DNA-methylated at the hemopoietic stem cell stage and remain so in all progenitor and differentiated hemopoietic cell lineages that do not express KIR. Moreover, the present results show that removal of DNA methylation in active KIR genes of NK cells is restricted to the promoter area.

We next analyzed whether the DNA methylation status would correlate with chromatin compaction of the KIR promoter region. For this purpose, we used a quantitative chromatin accessibility assay, which is based on DNA digestion by Micrococcus nuclease and subsequent locus-specific PCR amplification and normalization against a housekeeping gene. As shown in Fig. 2, A and B, in primary NK cells KIR promoters were more sensitive to Mnase digestion than in HPC thus reflecting higher accessibility of chromatin in NK cells. To differentiate between active and silent KIR genes, which were both represented in the polyclonal NK cell sample, NK cell subpopulations were enriched by flow-cytometric cell sorting using the NKB1 mAb, which is specific for KIR3DL1 but does not bind the closely related KIR3DS1 (34). KIR3DL1⁺ NK cells exhibited a much higher accessibility of the respective KIR locus than the KIR3DL1⁻ fraction (Fig. 2,C). In fact, the low chromatin accessibility of the KIR3DL1 promoter in KIR3DL1⁻ NK cells was comparable to that in HPC (data not shown). Similar low accessibility of KIR promoters was found in the KIR-negative human bone marrow stromal cell line L88/5 (data not shown). In general, the state of chromatin packaging tightly correlated with the degree of DNA methylation of the respective KIR promoter (Figs. 1 and 2 and Ref. 15).

To assess the contribution of specific histone modifications on clonal KIR expression patterns, ChIP assays were performed. The panel of histone-specific Abs for ChIP analysis was selected to cover modifications, which were reported to be associated with expressed genes, namely acetylation of lysine 8 of histone H4 (H4K8ac) and dimethylation of lysine 4 of histone H3 (H3K4dime), as well as a marker for repressive chromatin, which was dimethylation of lysine 9 of histone H3 (H3K9dime). Analysis of the NK cell line NK3.3 revealed high levels of the H4K8ac mark in KIR2DL3 and KIR2DL4 promoters, which are expressed, but also in KIR3DL1 and KIR3DL2, two genes that are not expressed in NK3.3 cells. The other active histone mark H3K4dime was restricted to the two expressed genes and virtually absent from the silent ones. In contrast, the KIR-negative human bone marrow stromal cell line L88/5 exhibited a histone signature characteristic of repressive chromatin with low levels of H4K8ac and H3K4dime and high levels of H3K9dime in all four KIR promoters (Fig. 3).

Next, the analysis of histone modifications was extended to polyclonal NK cells from peripheral blood, which were sorted by flow cytometry into subsets that were divergent for expression of selected KIR genes. As seen in Fig. 4,A, comparative analysis of KIR2DL3⁺ and KIR2DL3⁻ NK cells did not reveal significant differences in any of four selected histone modifications on the KIR2DL3 promoter. The promoters consistently exhibited high levels of H3K9ac (albeit slightly reduced in KIR⁻ NK cells), H4K8ac, and H3K4dime, which are indicative of gene activation and no heterochromatin-associated H3K9dime mark. Notably, the difference in the H3K4dime mark previously seen in the NK3.3 line (Fig. 3) was not observed in ex vivo-isolated NK cells, which suggests that the observed down-regulation of H3K4dime in NK3.3 is a cell culture or cell line-specific effect. The combination of these histone modifications constitutes a euchromatic histone signature, which can similarly be found in promoters of constitutively active genes such as GAPDH (Fig. 4,C). Similar results were seen for the KIR3DL1 promoter in KIR3DL1⁺ and KIR3DL1⁻ NK cell subpopulations (Fig. 4 B). Thus, active as well as silent KIR genes of NK cells possess highly similar histone signatures, which are typically found in expressed genes.

The analyses of DNA methylation patterns and chromatin accessibility did not so far give any clues as to why KIR gene expression is confined to NK cells and subpopulations of T cells (Figs. 1 and 2). In fact, KIR genes that were not expressed were generally densely methylated and inaccessible in all cell types analyzed so far. In contrast, the analysis of histone modifications revealed marked differences between different hemopoietic cell types (Fig. 5). Analysis of the KIR2DL3 promoter showed that the H4K8ac modification, a consistent marker of active genes, was strongly present on KIR promoters of NK cells but also on CD8⁺ T cells from cord blood that do not express KIR (Fig. 5,A). In contrast, low levels were seen on promoters of KIR-negative CD4⁺ T cells, B cells, and HPC. Analyses of other active marks like H3K4dime, showed similar cell type-specific differences (data not shown). In contrast, the heterochromatic marker H3K9dime exhibited low levels in NK and CD8⁺ T cells and increasingly higher amounts in CD4⁺ T cells, B cells, and HPC, respectively (Fig. 5 B).

Of note, direct comparison of ChIP values between different cell types is inherently difficult due to cell type-specific variability of the immunoprecipitation procedure. By taking the ratio of the two opposing histone modifications, H4K8ac as active and H3K9dime as repressive mark, we tried to create a more reliable parameter, which should in principal be independent of the cell type. For KIR2DL3, and even more pronounced for KIR2DL4 promoters, the H4K8ac/H3K9dime ratio was consistently highest in NK cells and CD8⁺ T cells and lowest in B cells and HPC (Fig. 5, C–D). Vice versa, the reciprocal value H3K9dime/H4K8ac ratio was highest in the latter two cell types. The CD4⁺ T cell lineage exhibited intermediate values that were clearly different from NK and CD8⁺ T cells, but also distinct from B cells and HPC.

To find out whether lineage-specific differences in histone modifications are confined to the promoter area the analysis was extended to the coding region of KIR genes. In NK cells, examination of H4K8ac and H3K9dime marks in exons 3, 4, and 8 of KIR2DL3 encoding the two extracellular and the cytoplasmic domains, respectively, revealed a euchromatic signature that was highly similar to the respective promoter region (Fig. 6, A and B). Similarly, in HPC a repressive histone signature extended from the promoter to the end of the coding region (Fig. 6 C). Thus, whereas changes in DNA methylation of the KIR locus are strictly confined to the promoter area of expressed KIR genes, changes in histone patterns extend across all parts of the gene. Moreover, because histone signatures seem to be independent of clonal expression patterns, our data suggest that lineage-specific changes in histone marks are propagated over the whole KIR locus.

The clonally-distributed expression of KIR is tightly correlated with the DNA methylation status of CpG islands that are located in the promoter region of KIR genes (15, 16). In vertebrates, the majority of promoter-associated CpG islands remain unmethylated at all stages of development. Only a small proportion of all CpG islands become methylated during development, usually at early embryonic stages (35). Our study suggests that KIR promoter-associated CpG islands (although not for all KIR adhering to published definitions for CpG islands (36)) are indeed DNA-methylated at an early developmental stage because dense DNA methylation was already found in CD34⁺ HPC. Furthermore, these observations implicate that KIR genes are DNA-demethylated in the course of NK cell differentiation. DNA demethylation is a well-known but enigmatic phenomenon that occurs during early embryonic development and germ cell differentiation and was also described for immune response genes in lymphocytes (37, 38). Of note, it is currently unknown how DNA demethylation is achieved, although some recent reports (39, 40) suggest a role for components of the DNA repair machinery in this process. The demethylating step appears to be specific for actively-transcribed KIR genes because silent KIR generally remain DNA-methylated in NK cells as previously shown as well as in other lymphocyte populations such as primary B cells and KIR⁻ CD4⁺ and CD8⁺ T cells. Furthermore, analyses of other CpG-dense regions either in the Alu elements upstream of the promoter region or in exon 4 in the coding region show that the demethylating activity is restricted to the promoter region. Although it is likely that the demethylated state, once it is established, is stably maintained in activated KIR genes the data do not exclude that KIR are demethylated in an unspecific way at an earlier stage of NK cell maturation and that clonal expression patterns are generated subsequently through inactivation of selected KIR genes by DNA methylation. Of note, the CpG cluster in the Alu repeat that was densely methylated in HPC and expressed as well as silent KIR genes of NK cells was recently described as part of a novel distal KIR promoter (41). If this promoter is active in hemopoietic progenitor or NK cells its activity seems to be regulated independently of changes in DNA methylation.

DNA methylation frequently is associated with chromatin condensation (20) and both effects might contribute to KIR gene silencing through recruitment of specific repressors and reduction of accessibility to the transcription machinery, respectively. Our analysis of chromatin accessibility of the KIR promoter region indeed reveals a close association between both epigenetic parameters. Whenever KIR promoters were found to be DNA-methylated, DNA accessibility was low. This was not only the case in cell types that generally do not express KIR such as HPC or L88/5 stroma cells but also for KIR genes in NK cells that were silenced in a clone-specific way. The other way round, DNA accessibility was high in expressed (and thus DNA-demethylated) KIR genes. The data support previous observations that accessibility of a restriction site in the KIR promoter of NK cells correlates with KIR expression (17). Thus, it can be concluded that both condensed and open chromatin domains are found next to each other in the KIR locus of NK cells and that this domain structure coincides with the clonal patterns of DNA methylation and expression.

Whereas DNA methylation and chromatin accessibility both exhibited a close correlation with the expression status of KIR genes, this was clearly not the case for N-terminal histone modifications. In NK cells, KIR genes generally showed active histone modifications that were highly similar between expressed and silent genes. These observations are in agreement with a recent study, which found only moderate differences between active histone marks in NK cells that were divergent for KIR expression (17). Furthermore, in the present study, silent KIR genes exhibited lineage-specific differences between KIR-expressing and nonexpressing cell types e.g., a silent KIR3DL1 gene showed active histone signatures in NK cells but repressive ones in B cells or HPC. The fact that histone signatures of KIR genes were largely uncoupled from DNA methylation patterns suggests that demethylation of DNA is not required for acquisition of active or removal of repressive histone modifications. Comparative analyses of lymphocyte lineages rather suggest that active histone modifications are applied to the KIR locus before DNA demethylation takes place. This notion is supported by the observation that cord blood-derived CD8⁺ T cells, which generally do not express KIR (>99% KIR⁻), have already acquired active histone signatures but still exhibit dense DNA methylation. This, in turn, suggests that in KIR-expressing CD8⁺ T cells, which represent a variable, but often substantial fraction of CD8⁺ T cells in peripheral blood (42), DNA demethylation occurs subsequent to the acquisition of active histone marks.

What is the biological significance of lineage-specific differences in histone signatures of the KIR locus? Active histone signatures, characterized by high levels of H4K8ac and low levels of H3K9dime, were restricted to NK and CD8⁺ T cells, which are the main lymphocyte populations that express KIR. CD4⁺ T cells that under nonpathological conditions exhibit a very low frequency of KIR expression (0.2–1%, which is ∼1/10 of the frequency found in CD8⁺ T cells (43, 44)) have a more repressive histone signature characterized by low levels of H4K8ac. Nonetheless, naive CD4⁺ T cells are fully competent for KIR expression as shown by Xu et al. (45) using a KIR promoter reporter construct. Even more pronounced repressive histone signatures, characterized by low levels of H4K8ac and high levels of the H3K9dime mark are found in early hemopoietic progenitors and cell types that are not competent for KIR expression, such as B cells and nonhemopoietic cell lines. Based on these observations, we suggest a model of epigenetic KIR gene regulation in which the lineage-specific acquisition of active histone signatures puts the KIR locus into a state of readiness, which is required for subsequent rapid demethylation and expression. In cell types that are not “supposed” to express KIR, repressive histone signatures would deny access of factors that could initiate DNA demethylation and in this way prohibit aberrant expression.

It is not known how lineage-specific transitions from repressive to active histone marks in the KIR locus are achieved in the first place. One possibility is that KIR-specific transcription factors recruit epigenetic modifiers like histone acetylases to the histones. Suitable candidates would be members of the family of runt-related proteins (RUNX), transcription factors that have a conserved binding site in the KIR promoter (46). A mutation of this site, that abolishes binding of RUNX to the KIR promoter is selectively found in KIR variants that are transcriptionally silent (47, 48). As RUNX1 is known to interact with various histone acetylases (49, 50), the inability to recruit these epigenetic modifiers via RUNX1 could inhibit the establishment of active histone acetylation marks in these silent KIR variants. Similarly, the repressive H3K9dime mark could be removed by recently described H3K9-specific histone demethylases via binding to KIR-specific transcription factors although it is not known so far if they interact with RUNX proteins (51, 52, 53).

In summary, the present study suggests that N-terminal histone modifications constitute an additional level of epigenetic regulation for KIR genes that is largely uncoupled from DNA methylation and chromatin accessibility. The acquisition of active histone signatures correlates with the general ability of cell types to express KIR and might even define lineage-specificity for KIR expression. However, the fact that histone marks were lineage-specific and transcended clone-specific expression patterns makes it unlikely that histone marks are directly involved in establishment of clonal expression pattern. They rather provide an activated environment, which might be required for recruitment of factors that initiate stochastic demethylation of the locus. Indeed, the tight association of KIR expression with demethylation of KIR promoters and the reactivation of silent KIR genes by chemical demethylation points toward DNA methylation as crucial factor in this process. In contrast, N-terminal histone modifications might be regarded as gene-specific metadata that not only describe the current status of transcriptional activity but provide additional information, for example, on the “readiness” for expression of a given gene.

We thank G. Kögler for supply of cord blood samples, J. Fischer for supply of G-CSF-mobilized peripheral blood, B. Giebel and S. Weinhold for helpful discussions, and T. Jenuwein (Research Institute of Molecular Pathology, Vienna) for the gift of anti-H3K9dime Abs.

The authors have no financial conflict of interest.