LILRB1 Intron 1 Has a Polymorphic Regulatory Region That Enhances Transcription in NK Cells and Recruits YY1

This article is featured in In This Issue, p.2865

The human leukocyte Ig-like receptors (LILRs) family has 11 multifunctional regulatory receptors involved in immune tolerance, inflammation, hematopoietic differentiation, and neural functions. LILRs are widely expressed within the immune system and interact with diverse ligands ranging from MHC class I (MHC-I) molecules, CNS-derived molecules, and host immune-modulatory proteins to pathogen-derived molecules and intact pathogens (1). The LILR family contains stimulatory receptors that transmit signals through associated signaling subunits and inhibitory receptors that repress signaling through ITIMs within their cytoplasmic tails. The family is characterized by substantial polymorphism and copy number variation, and the genetic diversity has been linked to disease susceptibility (2–4). Each family member displays a unique expression pattern on immune cells and can be constitutively expressed or induced under particular conditions (5).

One of the most highly studied LILR is LILRB1, an inhibitory receptor for which genetic variation has been linked to various autoimmune conditions, including rheumatoid arthritis, systemic lupus erythematosus, and atherosclerosis (6–13). LILRB1 is expressed on monocytes, dendritic cells, B cells, and subsets of T and NK cells (2, 14, 15). LILRB1 regulates cellular responses through binding classical and nonclassical MHC-I molecules (4, 16) as well as S100A9 (17). LILRB1 is exploited by several bacteria, viruses, and parasites that interact directly with LILRB1, such as dengue virus, malaria, and human CMV (HCMV) (14, 18, 19). For example, HCMV expresses an MHC-I mimic named UL18 that binds tightly to LILRB1 (14, 20). The presumed main function of UL18 is to inhibit NK cell responses. LILRB1-negative NK cells control virus dissemination in vitro better than LILRB1⁺ NK cells (21–23). LILRB1 is the only LILR expressed on NK cells, and its expression is variegated. Moreover, the frequency of LILRB1⁺ NK cells differs among individuals. HCMV infection is also associated with the expansion of LILRB1⁺ NK cells, and LILRB1 expression on other lymphoid cells can fluctuate with time, particularly in response to HCMV replication (21, 24–26). We recently described that LILRB1 polymorphisms correlated with LILRB1 expression patterns on NK cells are also correlated with the control of HCMV in transplant patients (27).

Although the precise mechanism underlying variegated LILRB1 expression in NK cells and the variation in the LILRB1⁺ NK cell frequency are not known, we and others have linked particular haplotypes with the frequency of LILRB1⁺ NK cells (28, 29), a feature reminiscent of the expression patterns of the highly related and syntenic killer-cell Ig-like receptors (KIRs). For example, KIRs have allele-specific expression patterns (30), and KIR expression is acquired at a late stage of NK differentiation as is LILRB1 (31). The KIR expression patterns are correlated with DNA methylation at the promoter region (32–34), and there is evidence that DNA methylation is involved in keeping LILR genes quiescent (35). During NK development, a probabilistic switch controls expression of KIR through a bidirectional promoter for which the relative strength of the forward activity over the reverse activity correlates with the frequency of a particular of KIR gene progressing to permanent expression (36, 37). Polymorphisms within the bidirectional promoter are associated with allele-specific frequencies of expression for a given KIR gene (30, 38). Although multiple KIR genes can be expressed in a single NK cell, typically only one allele for each KIR gene is expressed, and a piwi-like system arising from the antisense transcript may silence the other allele of a KIR gene that has initiated reverse transcription (39). The LILRB1 gene has 16 exons, and transcription is initiated at two distinct promoters (15). The first (5′) promoter is used by lymphoid cells, and the resulting transcript has all 16 exons (15). We previously defined the core region of the distal promoter with a functional JunD site that is required for expression in NK cells (40). The second promoter is positioned just upstream of exon 2 and contains functional PU.1 and Sp1 sites in line with its function in myeloid lineage cells (35).

In the current study, we investigated the transcriptional regulation of LILRB1 in NK cells with a view to uncovering how lineage-specific patterns of expression are generated and how polymorphisms selectively control the expression patterns in NK cells. We used the features of KIR transcriptional regulation and publicly available epigenomic data as a guide. We show that individual NK cells can express one or both alleles at the same time and that there is a correlation between high cell surface LILRB1 and expressing both alleles. Although we did not find strong evidence of a bidirectional promoter, we identified an additional polymorphic enhancer within the first intron. We provide evidence that the deletion of the region reduces LILRB1 expression in an NK cell line, that the transcription factor Yin Yang 1 (YY1) associates with the region as well as the promotors, and that the enhancer and promotors are in proximity, implicating YYI as a scaffold. The identification of this regulatory region provides new (to our knowledge) insight into the regulation of LILRB1 expression in NK and likely other lymphoid cells.

The human NK cell line NKL was cultured in Iscove's medium supplemented with 10% characterized FBS (HyClone), 1 mM l-glutamine, and 200 U/ml recombinant human IL-2 (Tecin; Biological Resources Branch, National Cancer Institute at Frederick). The human B lymphoma cell line 721.221 was maintained in Iscove's medium supplemented with 10% characterized FBS (Life Technologies) and 1 mM l-glutamine. PE-Cy5 mouse anti-human CD85j (clone: GHI/75) and PE-Cy5 mouse IgG2b κ isotype control were purchased from BD Biosciences (San Jose, CA). Allophycocyanin-conjugated mouse anti-human CD85j (clone: HPF1) and allophycocyanin-conjugated mouse IgG1κ isotype control were purchased from eBioscience (Waltham, MA). Chromatin immunoprecipitation (ChIP) grade anti-human YY1 (clone: D5D9Z) was obtained from Cell Signaling Technology, and the normal rabbit IgG control was included in the SimpleChIP Enzymatic Chromatin IP Kit (no. 9003; Cell Signaling Technology, Danvers, MA).

Blood samples were collected from healthy donors with written informed consent, and the experiments were performed as approved by the Health Research Ethics Board at the University of Alberta (Edmonton, AB, Canada). PBMCs were isolated from the whole blood using Lympholyte-H Cell Separation Media (Cedarlane, Burlington, ON, Canada). NK cells were then isolated from the PBMCs using the EasySep Human NK Cell Isolation kit (STEMCELL Technologies, Vancouver, BC, Canada) following the manufacturer’s protocol. NK cell purity was determined by flow cytometry, and at least 90% purity was accepted. Purified NK cells were plated at 3, 1, and 0.3 cells per well with 10⁴ irradiated 721.221 feeder cells per well to generate single clones using limiting dilution and cultured in Iscove's medium with 10% human serum and 1 mM l-glutamine, supplemented with 0.5 μg/ml PHA-P (Sigma-Aldrich) and 100 U/ml recombinant human IL-2, as previously described (41). Clones were derived from plates with growth in <30 wells after 30 d.

Ex vivo NK cell clones were first stained by anti-human CD85j (clone: HPF1) for surface LILRB1 expression using flow cytometry. Approximately 8 × 10⁴ cells from each LILRB1⁺ clone were lysed using the SingleShot Cell Lysis Kit (Bio-Rad, Hercules, CA) following the manufacturer’s protocol and immediately subjected to reverse transcription using iScript Advanced cDNA Synthesis Kit (Bio-Rad). A custom TaqMan genotyping assay (identifier: ANGZE69) was used to differentiate the expression of alleles specific for rs1061079 (C/T) that was designed using the online tool (https://www.thermofisher.com/order/custom-genomic-products/tools/genotyping/). The sequences of primers and probes are listed in Supplemental Table I. Each Droplet Digital PCR (ddPCR) reaction was prepared by mixing the cDNA template derived from ∼1000 cells of every single clone, TaqMan primers probes mix, and ddPCR Supermix for Probes (no dUTP; Bio-Rad) and then subjected to droplet generation using QX200 Droplet Generator (Bio-Rad) following the instruction manual. The ddPCRs were amplified in a thermal cycler according to the kit instructions, and the fluorescence signals were detected and recorded using the QX200 Droplet Reader (Bio-Rad). The copy number per sample of the two alleles for the single-nucleotide polymorphism (SNP) rs1061079 (C/T) was determined using QuantaSoft Software (Bio-Rad).

Genomic DNA was isolated from ex vivo primary NK cells using the Illustra tissue and cells genomicPrep Mini Spin Kit (GE Healthcare, Chicago, IL). Bisulfite conversion was done using the EZ DNA Methylation kit (Zymo Research) following the manufacturer’s protocol. Each conversion reaction was done with 500 ng of genomic DNA, and half of each reaction was used in the subsequent PCR. Primers were designed to amplify the DNA regions of interest avoiding the inclusion of CpG sites in the primer binding regions. The regions were amplified using the bisulfite primers listed in Supplemental Table I. The PCR was done with AccuStart Taq DNA Polymerase HiFi (QuantaBio, Beverly, MA). PCR products were run on an agarose gel and extracted according to the manufacturer’s protocol using the QIAquick Gel Extraction Kit (Qiagen, Hilden, Germany). An A-overhang reaction was performed on purified PCR products before performing TA cloning into the pCRII vector following the TOPO TA Cloning Kit (Life Technologies, Carlsbad, CA). The cloned PCR products in pCRII were transformed into TOP10 Escherichia coli, as per the manufacturer’s protocol. Ten to 20 clones for each PCR product were selected and cultured in Luria–Bertani broth with ampicillin for 16 h, and the plasmid DNA was isolated using the QIAprep Spin Miniprep Kit (Qiagen) and subsequently sent for sequencing at McLab (San Francisco, CA). Each CpG site was analyzed in all clones to determine the percentage of methylation for each CpG cytosine.

The putative enhancer region was amplified using iProof Hi-Fi DNA Polymerase (Bio-Rad) with the primers listed in Supplemental Table I. CRISPR guide RNAs were designed using the online software (http://chopchop.cbu.uib.no/). The boundaries of the 3-kb region encompassing the entire putative enhancer were targeted by two single guide RNAs (5′-AATCAGTACTAAAAATCTTC-3′ and 5′-TGATGAGCATAGTATTGGTG-3′) using an Alt-R CRISPR-Cas9 System (Integrated DNA Technologies, Coralville, IA) combined with a 4D-Nucleofector System (Lonza, Basel, Switzerland). Briefly, the guide RNA and tracrRNA (conjugated with ATTO 550) were preincubated to form duplex, followed by incubating with Cas9 nuclease to form ribonucleoprotein complex. The ribonucleoprotein complex was then electroporated into NKL cells within the nucleofector system using the reagents from SF Cell Line 4D-NucleofectorTM X Kit S and the CM150 program. Nucleofection efficiency was determined by flow cytometry after 24 h, and the NKL cells were further incubated for 2 d and assessed for LILRB1 surface expression with Ab staining (clone: GHI/75) by flow cytometry. The population with decreased LILRB1 expression was sorted and followed with single-cell sorting to generate clones using a BD FACSAria III sorter (BD Biosciences). The genomic DNA of the clones was isolated using QuickExtract DNA Extraction Solution (Lucigen, Middleton, WI), and the knockout of the putative enhancer was validated using PCR and Sanger sequencing.

The chromosome conformation capture (3C)–quantitative PCR (qPCR) assay was performed following the protocol as described (42) with some minor modifications. NKL cells were cross-linked by 1% formaldehyde for 10 min at room temperature and quenched by 0.125 M glycine. The cells were then washed twice with ice-cold PBS and lysed with lysis buffer containing 10 mM Tris-HCl (pH = 8), 10 mM NaCl, and 0.2% Nonidet P-40 with protease inhibitor mixture (Cell Signaling Technology) for 45 min on ice. Nuclei were collected and resuspended in 1.2× restriction enzyme buffer supplemented with 0.3% SDS and then incubated at 37°C with shaking for 1 h. A total of 2% Triton X-100 was added to the nuclei to sequester the SDS and incubated at 37°C with shaking for 1 h. Cross-linked genomic DNA within the nuclei was digested by EcoRI overnight at 37°C with shaking. A total of 1.6% SDS was then added to the sample, and the restriction enzyme was inactivated by incubating at 65°C for 25 min. The digested DNA was diluted into 1.15× ligation buffer supplemented with 1% Triton X-100 and incubated for 1 h at 37°C with gentle shaking. T4 DNA ligase (Thermo Fisher Scientific, Waltham, MA) was added to ligate the DNA fragments, and the reaction was incubated at 16°C for 4 h, followed with 30 min at room temperature. The cross-linking was reversed by adding proteinase K and incubating at 65°C overnight. After treatment with RNase for 45 min at 37°C, the sample was purified by phenol-chloroform extraction and ethanol precipitation. The concentration of the 3C library was determined using Qubit Fluorometric Quantitation System (Thermo Fisher Scientific). Specific ligation products were detected by TaqMan qPCR with the primers and probe listed in Supplemental Table I. To generate a reference control library, we purchased the bacterial artificial chromosome (BAC) clone (WI2-1436-K15), which contains the whole LILRB1 gene, from BACPAC Resources (Oakland, CA) and performed parallel digestion, ligation, and purification steps as described above. The values presenting the relative cross-linking frequency were calculated using the formula of 10^{(Ct − b)/a} (b = intercept; a = slope) where the parameters were from the standard curves generated using the BAC reference control library.

ChIP assay was performed using SimpleChIP Enzymatic Chromatin IP Kit (Cell Signaling Technology) following the manufacturer’s protocol. Briefly, the chromatin of NKL cells was cross-linked as described for the 3C assay. Nuclei were isolated and incubated with micrococcal nuclease to digest the chromatin to fragments of between 150 and 900 bp. The nuclei were sonicated using the Sonic Dismembrator Model 100 (Thermo Fisher Scientific) to break down the nuclear membrane. The digested chromatin was collected and then incubated with anti-human YY1 (1:50 dilution) or normal rabbit IgG overnight at 4°C with rotation. Protein G magnetic beads were added to each immunoprecipitated sample and incubated at 4°C for 2 h with rotation. The beads were washed, and the chromatin was eluted at 65°C for 30 min with vortexing in a thermomixer (Thermo Fisher Scientific). The cross-linking was reversed by incubating with proteinase K overnight at 65°C. Predicted YY1 binding loci were assessed by standard PCR using the purified ChIP DNA and then analyzed in agarose gel. The primers used for ChIP and ChIP-loop were designed based on the assay for transposase-accessible chromatin using sequencing results (Supplemental Fig. 2A) and listed in Supplemental Table I.

The ChIP-loop assay was performed as described previously (43) combined with the SimpleChIP Enzymatic Chromatin IP Kit. The digested and ligated chromatin was generated as described above before proceeding to the immunoprecipitation using the YY1 or normal rabbit IgG Ab. The purified ChIP-loop library was subjected to PCR analysis.

Our previous work showed that both alleles of LILRB1 are expressed in a mixed population of ex vivo NK cells (29). To examine the relative expression of each allele within single NK cells, we selected three individuals with varying LILRB1⁺ NK cell profiles as shown in Fig. 1 (left panels) known to be heterozygous at SNP rs1061679. We initially performed a single-cell analysis of ex vivo NK cells to determine the relative expression of each allele for rs1061679, but the sensitivity was not sufficient to draw conclusions (data not shown). Therefore, we grew out clones from purified NK cells and analyzed relative rs1061679 expression in at least eight clones from each donor. The clones had varying surface staining of LILRB1 (Supplemental Fig. 1), and we included a clone without detectable surface expression for each donor in the analysis. As expected, detection of LILRB1 transcript was correlated with surface expression; however, in each donor, we found a different pattern of expression for the two alleles. For D183, two clones expressed only rs1061679-C and three clones expressed both alleles, whereas three clones had no detectable transcript. For D185 and D500, we detected both alleles in 5 of 12 and 8 of 11 clones, respectively. In contrast to D183, we detected the rs1061679-T variant alone in several clones for both D185 and D500. For those clones in which both alleles were detected, the ratio was variable with either allele being dominant (Fig. 1, right panels).

The trend that there is a more frequent expression of the rs1061679-T allele in two out of three donors fits with our earlier observation of a correlation between the frequency of LILRB1⁺ NK cells and LILRB1 gene polymorphisms in the regulatory regions of the gene (29). The SNP rs1061679 is in the coding domain but has strong linkage disequilibrium with the SNPs within the distal promoter (r² = 0.92) and proximal promoter (r² = 0.91), which we showed to correlate with the frequency of LILRB1 expression on NK cells (27). Therefore, it is fitting that the rs1061679-T allele was detected more frequently and on its own in two donors. Although D183 showed the opposite trend, it also had the fewest clones analyzed and perhaps points to additional polymorphic regulatory regions.

As shown earlier, there are a series of SNPs in the distal and proximal promoter that are in near-perfect linkage disequilibrium with each other (27), forming haplotypes that correlate with the frequency of LILRB1⁺ NK cells. The two typical haplotypes are denoted as the major GGTG-AGG and minor CGTA-GAA, and there is a very minor G—G-GAA variant, which is illustrated in Fig. 2A. We previously reported there is not a difference in the activity of the allelic variants of the distal promoter using reporter assays (27), which implies that other regulatory mechanisms other than differential recruitment of transcription factors to the promoters are involved. Although there are no obvious CpG islands in the region of the distal promoter, methylation of individual CpG sites can alter transcription factor binding to regulate expression (44), and CpG methylation is an epigenetic modification that could limit transcription in an allele-specific manner.

To determine if CpG methylation near the promoter was correlated with promoter activity in general, we performed bisulfite sequencing analysis of the regions around the LILRB1 distal promoter from ex vivo purified NK cells, B cells, and monocytes in four heterozygous donors and one homozygous donor for the AGG haplotype (Fig. 2B). We found that the extent of methylation was very low for the CpG site at −14,052 in NK and B cells, the only CpG in the core distal promoter. The CpG site at −14,052 is quite prominently methylated in monocytes, suggesting its methylation might be linked to preventing transcription from the distal promoter in monocytes and fit with our observations that THP-1 cells can transcribe from the distal promoter when introduced as a plasmid (45). In all three cell types, the degree of methylation increases progressively for sites further upstream, and such a pattern is consistent with that location of the core promoter of the LILRB1 gene, as previously reported (40). Curiously, the two CpG sites just upstream of −14,205 have higher methylation in NK cells relative to B cells or monocytes, perhaps indicative of NK-specific regulation through DNA modification.

To test if there are differences between two alleles in terms of the methylation status of the distal promoter, we examined the DNA methylation patterns of the distal promoters in isolated NK cells from the four heterozygous donors (Fig. 2C). The bisulfite sequencing technique allowed us to segregate methylation status according to the alleles. The CpG site located at −14,290 from the translational start site is the only site to be substantively differentially methylated between the two haplotypes (p = 0.02). However, the allele with the higher methylation at −14,290 is the haplotype associated with lower expression in NK cells (Fig. 2D). Although it is possible a negative regulator occupies the undermethylated site or a positive regulator requires the methylation, the data do not fit well with an epigenetic silencing mechanism similar to KIR arising from an antisense transcript.

Given that enhancers are elements typically controlling the cell type–specific gene expression that could mediate allele-specific effects, we looked for evidence of additional regulatory elements in the LILRB1 gene using data from the Roadmap Epigenomics project (46). The dataset includes patterns of histone modifications and DNase sensitivity for populations of cells denoted as CD56, CD3, CD19, and CD14 markers for NK, T, B, and monocyte cells. First, we examined the patterns of DNA accessibility and histone modification at the known promoters in the various cell populations (Fig. 3A). Active promoters are marked by DNase hypersensitivity (green) and H3K4me3 (blue) and H3K27ac (orange). The region upstream of the distal promoter showed both H3K4me3 and H3K27ac in CD56, CD3, and CD19 cells but not CD14 cells (Fig. 3). This pattern of histone modification is consistent with the fact that only lymphocytes use the distal promoter to transcribe LILRB1. A reciprocal situation is found at the proximal promoter, which is marked as an active promoter only for the CD14 cells.

In the three lymphoid cell types, there is a 3-kb region within intron 1 with high sensitivity to DNase and three peaks densely marked by H3K4me1 (red) and H3K27ac (region boxed in Fig. 3A), a pattern of modifications associated with active enhancers (47). The intensity of the signal is greatest for CD19, followed by CD56 and then CD3, which corresponds to the expected proportion of cells that would express LILRB1 for each lineage, and the region does not exhibit the modifications in CD14 cells. Published assay for transposase-accessible chromatin using sequencing data from decidual NK cells (48) also showed the 3-kb region has high accessibility, further supporting a role for this region in regulating transcription (Supplemental Fig. 2A). In addition, analysis of available ChIP sequencing (ChIP-Seq) data from the Encyclopedia of DNA Elements or Gene Expression Omnibus (GEO) datasets (49, 50) also contains histone modifications indicative of an active enhancer at the same region in three lymphoblastoid cell lines and Jurkat cells, further supporting the 3-kb region as a regulatory site in lymphoid cells (Supplemental Fig. 2B, 2C). The region starts at 55,135,000 and ends at 55,137,700 (human hg19) and, as expected for an intron, has a high degree of variability with many SNPs recorded in the SNP and 1000 genomes databases (Fig. 3B).

To characterize the putative enhancer region and the relationship to the promoter haplotypes, we amplified the predicted enhancer region from NKL cells that have both promoter haplotypes as well as two donors, D230 and D258, which are homozygous for the promoter haplotypes (27). The alignment of a portion of the sequence is shown in Fig. 3C. Despite being heterozygous for the promoter haplotypes, only one sequence was derived from NKL for the entire 3-kb region. Only one sequence was derived from D258, as well; however, it differs from NKL’s in a number of places, four of which are shown in Fig. 3C. We obtained two alleles from D230 and although the sequence is more similar to that of NKL for the region shown in the left panel of Fig. 3C, it differs in other regions as shown in the right panel of Fig. 3C. As will be discussed in more detail below, algorithms to predict transcription factor binding sites indicate a large number of potential sites for factors known to be expressed by lymphoid cells, such as STAT4, Pax5, and c-Ets-1 (Table I). The sequence variation in this putative enhancer region suggests alleles could differentially recruit transcription factors. In addition, the region has predicted YY1 sites, and YY1 is a factor that has been implicated in enhancer function.

The physical interaction of enhancers with promoters can be mediated by certain cofactors, including cohesion and DNA-binding proteins such as YY1, that facilitate and stabilize the DNA looping structure by forming dimers (51–54). YY1 is also a factor that can promote or prevent transcription, and several YY1 binding sites are predicted in the putative enhancer region as well as in the distal and proximal promoters. There is also evidence YY1 binds to the putative enhancer region and the two promoters in published ChIP-Seq data from the Encyclopedia of DNA Elements and GEO datasets (Supplemental Fig. 2B, 2C) (49, 50). These observations suggest YY1 could mediate physical interaction between the enhancer and the two promoters. To test if YY1 also binds to the putative enhancer and/or the two promoter elements of the LILRB1 gene in NK-type cells, we applied YY1-ChIP in NKL cells. We analyzed 13 YY1 sites within the putative enhancer region and two promoter regions and detected YY1 binding at 10 sites (Fig. 4). Two of the sites were excluded because of high background for the negative control (E1 and D5), and all the remaining 11 sites shown except D3 have YY1 association (Fig. 4). The D1 and P3 sites consistently provided the most intense signal, suggesting the highest occupancy with YY1 for these sites. The ChIP results indicate YY1 is present at the promoter and enhancer regions and support the possibility YY1 scaffolds these regulatory elements together.

Given genome-wide high throughput 3C by sequencing assay is a powerful tool used to identify chromatin interactions across the whole genome (55), we first examined the chromatin interacting pattern at the LILRB1 gene locus using high throughput 3C by sequencing data in the 4D Nucleome Data Portal (56) available for several lymphoblastoid cell lines. We observed signals indicating the interaction between the putative enhancer and both of the LILRB1 promoters in most cell lines, but the data were derived from the HindIII-digested chromatin, which did not provide the best resolution for the fragments we are interested in.

Therefore, we went on to test if the putative enhancer is in direct physical contact with the LILRB1 gene promoters in NK cells by performing 3C-qPCR on NKL cells. We used EcoRI to generate suitable fragments to resolve the putative enhancer from the LILRB1 promoters and specify their interaction (Fig. 5A). The probe and the anchoring primer were designed within the enhancer fragment close to the EcoRI site at -7667 from the translational start codon, and all the test primers for the fragments were designed close to the upstream restriction end (Fig. 5A). Among the three distal fragments examined, the −16,483 fragment encompassing the distal promoter has a significantly stronger interaction with the anchor in the enhancer than fragments on either side (Fig. 5A). There were two additional major and minor interactions observed with −1972 and −1474 in the proximal region, respectively, where the proximal promoter and the YY1 site denoted as P3 in Fig. 4 are included in the latter fragment (Fig. 5A). The stronger signal for the −1972 fragment suggests the region upstream of the previously defined “core promoter” is involved.

To investigate if YY1 is part of the complexes that contain the enhancer and promoter in the 3C assay, we used ChIP-loop, a technique that combines ChIP and 3C. In brief, the chromatin-capture procedure was initiated, and then the YY1-associated complexes were immunoprecipitated and subsequently analyzed by PCR for the ligation produces. As shown in Fig. 5B, specific ligation products with the three fragments denoted as −16,483, −1972, and −1474 were detected in the anti–YYI-immunoprecipitated samples relative to an IgG control and the negative control region at −17,925. Collectively, the 3C and ChIP-loop data are consistent with the YY1-ChIP placing YYI at the site of the enhancer and promoters in complex with each other in NKL cells.

We tested the ability of the 3.2-kb region to enhance the distal promoter using the pGL3-basic luciferase system. However, the region actually repressed transcription, and its size made it difficult to pursue this approach. To more directly test the role of the putative enhancer in LILRB1 gene expression in the context of chromatin, we applied CRISPR-Cas9 technology to delete the 3.2-kb region in NKL cells as illustrated in Fig. 6A. We sorted single-cell clones with lower LILRB1 expression by FACS and analyzed by PCR to more readily ascribe the relative LILRB1 expression to the deletion and assess the variability of LILRB1 expression in unmanipulated but cloned NKL cells. We selected clones with decreased surface LILRB1 expression compared with the NKL control. In these six clones, we detected bands corresponding to the intact locus at 5286 bp from NKL control and the expected deletion at around 1500–2000 bp and sequenced the product to ensure the deletion was of the correct locus. Among the six clones, KO-3, KO-4, KO-5, and KO-6 with two alleles knocked out showed lower surface LILRB1 expression than that of KO-1 and KO-2 with one allele knocked out (Fig. 6B, 6C). To investigate whether the decreased surface LILRB1 expression was due to decreased LILRB1 transcription, we isolated total RNA from the six knockout clones and did quantitative real-time PCR to detect the change of LILRB1 transcript. Consistent with the FACS data shown in Fig. 6B, the LILRB1 mRNA levels of the six knockout clones were all significantly decreased compared with the NKL control (Fig. 6D). Importantly, the LILRB1 mRNA levels of those six clones were well matched with the LILRB1 mean fluorescence intensities detected by FACS (Fig. 6B, 6D), which suggested the knockout of the putative enhancer inhibited the LILRB1 gene transcription.

To ensure the lower LILRB1 transcript was not an artifact of subcloning the NKL line, we also isolated 19 NKL clones from the parental NKL cells. Although these NKL subclones have slightly different LILRB1 surface expression, they were all close to the parental level (Fig. 6E). The mRNA was also analyzed for six of these clones, showing minimal changes in the transcript level (Fig. 6F). These results indicate the intronic putative enhancer plays a positive role in regulating LILRB1 gene transcription.

In the current study, we wanted to characterize the NK-specific regulation of LILRB1 transcription with a view to understanding the influence of allelic variation on LILRB1 expression. To this end, we investigated the allelic expression of LILRB1 in NK clones from heterozygous individuals. We detected the major allele at a higher frequency as it is detectable in more clones from two of the three individuals examined. This correlates well with our previous results that associate the major allele with higher overall expression in NK cells (29) and is reminiscent of the allele-specific expression patterns of KIRs. However, the expression of LILRB1 alleles differs from KIR in important ways. First, a substantial proportion of NK clones express both LILRB1 alleles, whereas, for most KIRs, a minority express both alleles (33). Second, assuming there is concordance between the signal obtained in the ddPCR and the total transcript per cell, there is significant variability in the amount of transcript among the NK clones, which correlates imperfectly with the surface expression. Curiously, for the majority of clones that express both alleles, the amount of transcript for the rare allele is actually the higher one. Previously we showed that the distal promoter variants provide similar levels of transcription when tested ectopically in an NK cell line, indicating polymorphisms in the core promoter do not directly explain the expression patterns (27). Together, the results suggest the mechanism that leads to differential expression between alleles is not the probability of initiating transcription-limiting expression to one allele during a tight developmental window but rather other aspects that vary at a clonal level and determine the set point for transcription of each allele. In addition, the inverted frequency of the alleles for D183 suggests the element that is involved is distinct from the polymorphisms in the promoters we previously identified.

We have previously shown that the pattern of LILRB1 expression on peripheral NK cells remains quite stable over time for healthy individuals (41). LILRB1 is first expressed during later stages of NK development but signals that initiate LILRB1 in NK cells are not known. We were unable to identify any additional promoters for the LILRB1 gene from the epigenomic datasets. However, the proximal promoter region does display some elements suggesting a transcript could arise in reverse (TATA box and poly-A site), and using reporter constructs, we previously found modest reverse activity of the proximal promoter in THP-1 cells and very weak reverse activity in NKL cells (45). We also could not detect any reverse activity from the distal promoter. Therefore, it remains a possibility that the proximal promoter or another region has reverse activity during some point in NK cell development and mechanism of locus activation/repression akin to that of KIRs might occur.

The LILRB1 distal promoter region lacks obvious CpG islands; nonetheless, our DNA methylation analysis suggests one CpG site in the distal promoter needs to be unmethylated for expression. A second CpG site upstream of the distal promoter has differential methylation between the two haplotypes. Although most methyl-CpG binding proteins are thought to play a repressive role in gene transcription by recruiting transcriptional corepressor protein, there are certain methyl-CpG binding proteins that can activate transcription, which indicates DNA methylation may not always be a repressing marker (57–61). Therefore, if this CpG site is playing a role in the difference between the two haplotypes, it is likely by promoting methylation-dependent binding of an activating transcription factor or preventing the association of a negative regulatory factor. There are several transcription factors predicted to bind to this site, including STAT4 and SPI1, and, as this site is also 83 bp from one of the YY1 sites, it may warrant further investigation.

In this study, we identified a new regulatory region within intron 1 through its enhancer-specific histone modification pattern in CD56⁺ cells. Of note, we also observed similar profiles in T cells and B cells but not monocytes, indicating the putative enhancer is lymphoid-specific. The region possesses many sites predicted to recruit transcription factors known to be expressed by lymphoid cells, including some that are more T and NK (e.g., STAT5) or B lineage specific (e.g., Pax5) (Table I).

The presence of the regulatory region in the intron may explain why the LILRB1 gene has maintained a 13-kb-long intron 1 when the leukocyte receptor complex genes tend toward very compact structures. Analysis with RepeatMasker annotations (62) reveals that the putative enhancer region is flanked by a long interspersed nuclear element (LINE) indicative of its formation by a transposable element (Supplemental Fig. 3A) (63–66). The distal promoter and the enhancer flanking sequences also have LINE, long terminal repeats, and short interspersed nuclear elements (67) (Supplemental Fig. 3A). The presence of the LINE, short interspersed nuclear elements, and long terminal repeats suggests the distal promoter and first exon were acquired through insertion events, and this may explain how the locus evolved to differentially regulate LILRB1 in myeloid and lymphoid cells. Multiple alignment of the LILRB1 gene in different species reveals differences between primates and nonprimates not only in sequence conservation but also in retaining all or partial sequences upstream of the proximal promoter, although some primates such as gorilla and rhesus lost exon1 (68) (Supplemental Fig. 3B).

The evidence indicating the 3-kb region in intron 1 is a positive regulator of the distal promoter is that the deletion of the region in NKL cells reduces LILRB1 transcript and protein expression. However, we were unable to show that the 3-kb region or 1-kb fragments can enhance the distal promoter in typical luciferase reporter assays (data not shown), suggesting the native chromatin structure is required for function. It is tempting to speculate that the region has silencing activity for the proximal promoter as the 3C data shows the enhancer in physical complexes with the proximal promoter as well, but the biological relevance of the interaction with the enhancer and the proximal promoter requires further investigation. Despite the ability of NKL cells to support transcription from the proximal promoter from a plasmid, the use of the proximal promoter to transcribe LILRB1 is negligible in ex vivo NK cells, although it can be increased under cytokine stimulation (15, 40, 41). It is possible the enhancer region identified in this study may positively regulate the LILRB1 expression through recruiting trans-activating transcription factors to the distal promoter in NK cells and at the same time block the formation of the transcription preinitiation complex at the proximal promoter.

The 3-kb region contains three peaks of enhancer marks with two valleys of unmarked DNA in between, suggesting the region may have multiple distinct elements. We used a transcription factor prediction tool (ALGGEN-PROMO program) with the sequence of the putative enhancer and found many binding sites for Ccaat-enhancer-binding proteins (C/EBPs) and two binding sites (−6581 and −6520) of CREB, which could recruit another marker of potential enhancers named CREB-binding protein (CBP) (69–72). Among the many transcription factors predicted to associate with the enhancer element, we focused on YY1 because YYI was shown recently to be a structural regulator mediating enhancer/promoter interactions (54). Our results demonstrate YY1 is bound to the enhancer region as well as both promoters and can pull down the ligated products with each promoter formed by 3C, suggesting YY1 dimers bridge the elements together. Coincidently, the proximal promoters of KIR genes also bind YY1, and a polymorphic mutation at the YY1 site increases promoter activity and the ratio between the reverse and forward transcripts (30, 38, 73). In the cell lines and blood donors we analyzed, the YY1 sites are conserved and therefore unlikely to directly control allele-specific expression patterns. Nonetheless, polymorphisms do alter the prediction of other transcription factor binding sites (Table I), which could lead to differential levels of transcription between alleles. It is also worth noting that there are many predicted Pax5 sites, and Pax5 is essential for B cell commitment and is expressed specifically during B cell development (74). Although this study has illustrated an additional polymorphic regulatory sequence within the LILRB1 gene, additional layers of regulation might also be involved in the steady-state expression of the transcript. Along these lines, a very recent report showed that NF90 associated with the LILRB1 transcript in THP-1 cells and inhibits LILRB1 protein expression (75).

The current and further characterization of the regulatory mechanisms of LILRB1 gene expression provides new perspectives to understand the heterogeneity of human NK cells. Beyond the role of LILRB1 in NK cell control of HCMV and in decidual NK cells, there many associations of LILRB1 in other infections, such as dengue and malaria, and in autoimmune diseases, such as systemic lupus erythematosus, rheumatoid arthritis, pemphigus foliaceus, and autoimmune thyroiditis (6–13). Mounting evidence also supports the immune checkpoint function of LILRB1 in viral chronic infection and cancer (76–78). LILRB1 is highly expressed on tumor-associated macrophages, and the LILRB1–MHC-I axis modulates phagocytic function (79). How LILRB1 is regulated in each of these contexts will require further characterization of LILRB1 transcriptional elements in relation to developmental programs and differentiation into effector states.

We thank Heather Eaton and Bara’ah Azaizeh for technical assistance and Drs. James Smiley and Robert Ingham for helpful discussions. We also thank Drs. Lynee Postivit, Tom Hobman, Maya Shmulevitz, and Matthias Gotte for use of laboratory equipment. We acknowledge the support from the staff of the Flow Cytometry Facility, Faculty of Medicine and Dentistry, University of Alberta.

The authors have no financial conflicts of interest.