Limb Bud and Heart Development Gene Transcription is Regulated by the Interplay of an Enhancer Risk Allele and DNA Methylation in Rheumatoid Arthritis

Abstract

Objective

To identify non-obvious therapeutic targets for rheumatoid arthritis, we performed an integrative analysis incorporating multiple ‘omics data and the ENCODE database for potential regulatory regions. This analysis identified the Limb Bud and Heart development (LBH) gene, which has risk alleles associated with RA/celiac disease and lupus, and can regulate cell proliferation in RA. We identified a novel LBH transcriptional enhancer with an RA-risk allele (rs906868 G (risk, Ref) /T) 6kb upstream of the LBH gene with a differentially methylated locus. The confluence of three regulatory elements, rs906868, an RA differentially methylated locus and a putative enhancer, led us to investigate their effect on LBH regulation in fibroblast-like synoviocytes (FLS).

Methods

We cloned the 1.4kb putative enhancer with either the rs906868 Ref allele or SNP variant into reporter constructs. The constructs were methylated in vitro and transfected into cultured FLS by nucleofection.

Results

We found that both variants increased transcription, thereby confirming the region’s enhancer function. Unexpectedly, the transcriptional activity of the Ref risk allele was significantly lower than the SNP variant and is consistent with low LBH levels as a risk factor for aggressive FLS behavior. Using RA FLS lines with homozygous Ref or SNP allele, we confirmed that homozygous Ref lines expressed lower LBH mRNA levels than the SNP lines. Methylation significantly reduced enhancer activity for both alleles, indicating that enhancer function is dependent on its methylation status.

Conclusion

This study shows how the interplay between genetics and epigenetics can affect expression of LBH in rheumatoid arthritis.

Rheumatoid arthritis (RA) is a chronic, immune-mediated disease characterized by synovial inflammation and joint damage(1-3). Fibroblast-like synoviocytes (FLS) are key players in rheumatoid synovial pathology and cartilage destruction through the elaboration of cytokines, small molecules, and proteases(4). These cells form the synovial intimal lining and assume a unique tumor-like phenotype, including the ability to invade into cartilage explants and to migrate between sites in a SCID mouse model(5, 6). However, therapies that target FLS are not currently available. With the advent of new unbiased genome-wide platforms, RA FLS are now amenable to genomic and epigenomic analyses that can potentially identify non-obvious therapeutic targets.

To understand the aggressive behavior of RA FLS, we previously analyzed DNA methylation patterns and identified an RA-associated epigenetic signature that implicated genes involved with cell migration, cell recruitment and matrix regulation(7). We then performed an integrative analysis of three genome-wide datasets, namely, RA risk-associated genetic variants (GWAS), changes in DNA methylation at gene promoters, and changes in gene expression in FLS(8). This analysis identified a limited subset of genes, where an overlap was observed in the three datasets(9). One of these genes, LBH (Limb Bud and Heart Development gene), is a transcription cofactor that regulates cell differentiation during development and is linked to cardiac and skeletal abnormalities in the partial trisomy 2p syndrome(10). We subsequently showed that LBH regulates synoviocyte proliferation by arresting cell growth at the G1 checkpoint(11, 12). Reduced LBH expression has been associated with inflammation and autoimmunity(12, 13). We hypothesized that low LBH expression could, therefore, be an RA risk factor by permitting the G1-S phase transition. Several LBH polymorphisms have been associated with autoimmune diseases including RA, celiac disease and systemic lupus erythematosus (SLE), which suggests a key role in regulating immune function(14, 15).

In this study, we expanded our DNA methylation analysis beyond promoters to consider distal regulatory loci by including genome-wide DNaseI hypersensitive sites (DHS). These open chromatin regions often function as regulatory elements such as transcriptional enhancers that interact with promoters through DNA looping(16). We identified a new putative LBH enhancer 6kb upstream of the transcription start site that contains a differentially methylated locus (DML) in RA FLS, but unexpectedly, also contains an adjacent RA-associated variant. This striking confluence of an RA risk-associated variant, an RA DML and an open DNA regulatory element suggests that LBH and this regulatory region might be important in RA pathogenesis. Here, we show that the open chromatin region is, in fact, an enhancer and that the RA-associated SNP decreases LBH gene transcription. The data show how a functional SNP and DNA methylation interact to alter expression of a gene that regulates the aggressive behavior of pathogenic cells in RA.

Materials and Methods

Human fibroblast-like synoviocytes and culture conditions

This study was approved by the Institutional Review Board of University of California, San Diego School of Medicine, and informed consent was obtained from all participants. Synovial tissue was obtained from patients with osteoarthritis (OA) and RA at the time of total joint replacement or synovectomy, as previously described. The diagnosis of RA conformed to American College of Rheumatology 1987 revised criteria(17). The synovium was minced and incubated with 0.5mg/ml collagenase type VIII (Sigma-Aldrich, St. Louis, MO) in serum-free RPMI 1640 (Life Technologies, Grand Island, NY) for 1h at 37°C, filtered, extensively washed, and cultured in DMEM (Life Technologies, Grand Island, NY) supplemented with 10% heat-inactivated fetal bovine serum (Gemini Bio Products, Calabasas, CA), and supplements (penicillin, streptomycin, gentamicin, and glutamine) in a humidified 5% CO2 atmosphere. Cells were allowed to adhere overnight, non-adherent cells were removed. Adherent FLS were split at 1:3 when 70–80% confluent and used from passages 4 through 7(18).

Quantitative Real time PCR

RA FLS were serum starved for 24h in DMEM containing 0.1% FCS and supplements. Total RNA was isolated from FLS using RNASTAT-60 and reverse transcribed (Applied Biosystems, Foster City, CA). The cDNA served as template for amplification by qPCR using LBH TaqMan Gene Expression assays (Hs00368853_m1; StepOnePlus Instruments, Applied Biosystems, Foster City, CA). The Ct values were normalized to GAPDH (Hs02758991_g1)(19).

Processing omics datasets and pathway analysis

Data processing for integrative analysis was performed as described previously(9). The putative enhancer regions were based on the ENCODE database of 1,281,988 DNaseI-seq peaks from 125 cell-types/condition combinations, including fibroblast cell lines. These were filtered to remove any peaks that overlap promoters (−2500bps to +500bps from a gene transcription start site (TSS)), resulting in a set of 1,055,570 putative enhancers. Promoter regions were identified from GENCODE gene models and represent 177,999 unique TSS. Putative enhancers containing DMLs were identified and then linked to potential target regulatory genes by identifying the closest TSS. KEGG pathway enrichment analysis of the double evidence gene sets was performed as described in our previous studies(8).

Chromatin Immunoprecipitation (ChIP)-qPCR

ChIP assay was performed using Zymo-Spin ChIP kit according to the manufacturer’s protocol (Zymo Research, Irvine, CA). Briefly, FLS were serum-starved for 24h with DMEM/0.1%FCS with penicillin, streptomycin, gentamicin and glutamine). The cells were fixed in 1% formaldehyde (ThermoScientific, Waltham, MA) for 8 minutes at RT. After sonication, the chromatin was immunoprecipitated with specific antibodies overnight at 4°C. The antibodies used were anti-H3K4me1 (ab8895, Abcam, Cambridge, MA), anti-RNA polymerase II (ab26721) or rabbit IgG (2729, Cell Signaling Technology, Danvers, MA). The complexes were then immunoprecipitated with protein A magnetic beads for 1h at 4°C, and reverse-crosslinked at 65°C overnight. The eluted DNA was purified and used as template in the qPCR reaction. A 10% input sample was used as control. The enhancer primers (EpiTect ChIP qPCR primer GPH1007377(-)07A) were purchased from Qiagen (Valencia, CA). Control GAPDH promoter primers were purchased from Clontech (Mountain View, CA).

Identifying RA FLS lines with homozygous Ref and SNP alleles

Purified genomic DNA from RA FLS was subjected to PCR and agarose gel analysis. Tetra-arms PCR primers were designed using BatchPrimer3 program(20). The outer primers were forward 5’-CACTTCAACAGCCAGAAAAGAGAGGTTA-3’ and reverse 5’-GCAAGTGCTAGCTCAGGTTACAGTTAGG-3’. The inner primers were forward 5’-ACGTCTCCCACTATTTTACCCACGAC-3’ and reverse 5’-ATGATGGAGTTTCCCTGGTCCAAAT-3’. Pyrosequencing of Ref and SNP FLS at chr2:30448529 (cg 20495819) was performed by EpigenDx (Hopkinton, MA).

Constructs, transfections and luciferase reporter assays

A 1311bp enhancer fragment was amplified from homozygous Ref or SNP genomic DNA with specific primers (Forward – 5’-CTGAGCTCGCTAGCCTCGAGCAGGCTGATCTCGAACTC-3’ and Reverse - 5’-ATACCCTCTAGTGTCTAAGCTTAGGCCTCAGGGTCTAGTAAG-3’). The enhancer fragment was cloned into the pGL4.23 vector with a minimal promoter and the empty vector was used as control. The constructs were validated by sequencing. Control, ref and SNP plasmids were methylated by incubating 50ug of plasmid with 20U/ul of CpG methyltransferase (M.SssI) and 32 mM S-adenosyl methionine, overnight at 37°C, according to manufacturer’s instructions (New England Biolabs, MA). Complete methylation was verified by plasmid bisulfite modification and pyrosequencing (EpigenDx, Hopkinton, MA). For luciferase assays, 2ug of control, homozygous Ref or SNP plasmids were co-transfected with 50ng of renilla plasmid into 5×10⁵ FLS by nucleofection (Lonza, Walkersville, MD). The cells were serum starved for 24h prior to harvesting. The cells were lysed and assayed for luciferase activity using the dual luciferase reporter assay kit (Promega, Madison, WI). The firefly luciferase activity was normalized to renilla and expressed relative to empty control vector.

Results

Expanding RA FLS integrative analysis to include putative transcription enhancers

Our previous integrative analysis of RA FLS focused on promoter DMLs and identified several candidate genes of interest in RA therapy, namely ACOXL, AIRE, CASP8, CSF2, ELMO1, ETS1, HLA-DQA1 and LBH(9). To increase the coverage, we expanded the number of DMLs to include non-promoter regulatory regions. Putative enhancers containing DMLs were identified as described in Material and Methods and then linked to potential target regulatory genes by identifying the closest transcription start site. Integrative analysis of these newly identified differentially methylated genes (DMG), differentially expressed genes (DEGs), and RA GWAS variants, yielded 672 genes that were significantly different between RA and non-RA in at least 2 of the 3 datasets. The 672 genes represent 614, 28, and 18 double evidence genes and 12 triple evidence genes (Fig 1A, Table S1). The 12 genes included the 8 genes found previously and the new genes, IKZF3, IL2RB, PRKCH, PTPN11, and LBH. Interestingly, LBH was identified twice, once for the promoter DML and the second, for a putative enhancer DML.

Fig 1. Expanded integrative analysis of omics data and pathway analysis.

(A) A Venn diagram showing the overlap of three datasets: differentially methylated genes (DMGs) associated with differentially methylated loci at promoters and putative enhancers, differentially expressed genes (DEGs) and genome-wide association studies data (GWAS). (B) KEGG pathway enrichment analysis was performed using genes that were present in 2 or 3 of the datasets.

$$\begin{array}{cc}\hfill EnrichmentFactor& \hfill \\ \hfill & =(numberoflocifromgeneAbeneathcut-off\hfill \\ \hfill & ∕totallocifromgeneA)∕(totalnumberoflocibeneathcut-off\hfill \\ \hfill & ∕totalnumberofloci)\hfill \end{array}$$

P-values were calculated using the hypergeometric distribution and were corrected for multiple testing to produce q-values. Pathways with q-values <0.05 are shown.

Pathway analysis of the expanded multi-evidence gene set

The number of double evidence genes (i.e. identified in at least two of the three databases) and triple evidence genes increased from 357 genes found in our previous study(9) to 672 genes when the enhancer-related genes were included (multi-evidence genes (MEGs); Fig 1B; Table S2). KEGG pathway analysis was performed, and the ‘Immune’ and ‘Signaling’ groups were enriched with at least four pathways relevant to RA. The ‘Cell adhesion molecules’ pathway was enriched 3.9-fold (P=2.71E-07, q=3.75E-05) with 19 out of 129 genes in the MEGs. The ‘Cytokine-cytokine receptor interaction’ pathway was enriched 2.88-fold (P=4.8E-07, q=3.75E-05) with 27 out of 248 MEGs and the ‘JAK-STAT signaling pathway’ was enriched 2.85-fold (P=2.1E0-4, q=5E-03) with 15 out of 139 MEGs. The KEGG ‘Antigen processing and presentation’ pathway was enriched 3.55-fold (P= 8.4E-04, q=1.1E-02) with 9 out of 67 genes. Interestingly, the fold enrichment of these four pathways with 672 MEGs was similar to our previous data using 357 MEGs. Therefore, the expanded integrative analysis confirms and extends the association of immune-related pathways and abnormally methylated genes in RA FLS.

Identification of a novel LBH regulatory region in RA FLS

We were particularly struck by the fact that the analysis of regulatory regions identified LBH, which was also discovered in our original analysis due to promoter DMLs. The LBH DML in the present analysis was found in an intergenic regulatory region 6kb upstream of the LBH transcription start site (Fig 2A). It contains a non-CpG RA/celiac-associated risk variant (rs906868, G [RA-risk allele, Ref] → T [protective, SNP]), and a DML that we identified, located 185bp downstream of the variant (Fig S1). We first confirmed hypomethylation of this DML by pyrosequencing 6 OA and 9 RA FLS lines (Fig 2B). We then performed chromatin immunoprecipitation on RA FLS nuclear extracts with anti-H3K4me1 antibody and qPCR with primers specific to the region of interest (hg19- chr2:30447672-30448982). The region is significantly enriched in H3K4me1 mark in FLS (1.35±0.3 percent of input; t-test, n = 4 RA FLS lines, p=0.005, Fig 2C), which supports its potential role as an enhancer(21). Myt1 primer, which binds repressed chromatin histone mark H3K27me3, was used as a negative control and was enriched 0.17±0.03 percent of input in the enhancer region.

Fig 2. Identification of a novel LBH regulatory region in RA FLS.

(A) The putative enhancer is in a DNaseI hypersensitive region located approximately 6kb upstream of the TSS. This region contains a non-CpG risk-variant (rs906868) and an RA hypomethylated locus (cg20495819) 185bp apart. (B) Pyrosequencing analysis confirmed significant hypomethylation of cg20495819 in RA compared with OA FLS. (C) FLS lysates were immunoprecipitated with anti-H3K4me1 antibody and the DNA was subjected to qPCR with enhancer-specific or negative control Myt1 primer. The enhancer region was significantly enriched for the H3K4me1 mark.

LBH enhancer region regulates gene transcription in FLS

The LBH enhancer region containing either the Ref (i.e., risk) or SNP (i.e., protective) allele was cloned into pGL4.23 luciferase reporter construct, transfected into RA FLS and analyzed for luciferase activity. As shown in Fig 3, both alleles increased luciferase activity compared with control, which confirms the enhancer’s regulatory function. Of interest, SNP allele-driven transcription was significantly higher than the Ref allele (SNP: 2.9±0.6; Ref: 1.6±0.3, n=12 RA lines; paired t-test, p<0.0001). These data confirm that the region is an enhancer and that SNP and Ref alleles differentially regulate LBH enhancer function, with lower transcriptional activity associated with the Ref (risk) allele.

Fig 3. LBH enhancer region regulates gene transcription in FLS.

The enhancer sequence with Ref or SNP allele was cloned into luciferase constructs, which were then transfected into RA FLS. Luciferase assays show significantly higher activity in SNP plasmids compared with Ref.

RA risk allele regulates LBH expression in FLS.

The differential effects of the Ref and SNP on transcription was then confirmed by determining if the risk allele for RA (G/G) is associated with lower LBH gene transcription in FLS. To identify RA FLS lines with homozygous Ref (G) or SNP (T) alleles, we genotyped 58 cell lines. Seven RA lines (12%) were homozygous for the Ref allele, 19 (32%) were homozygous for the SNP allele and 32 (55%) lines were heterozygous. The allele frequencies of G and T were 34.5% and 65.5%, respectively. LBH expression in the two homozygous genotypes was then determined. RA FLS lines with SNP alleles expressed significantly higher levels of LBH than the lines with Ref alleles, consistent with the enhancer functional assay (Fig 4; Ref: 0.65 ± 0.08, n=7; SNP 1.2 ± 0.13, n=9; t-test, p=0.01). These data indicate that the Ref allele decreases LBH enhancer activity, contributing to low LBH gene expression in RA FLS.

Fig 4. RA risk allele regulates LBH expression in FLS.

RA FLS lines with Ref- or SNP-homozygous alleles were identified and LBH gene expression was measured by qPCR. LBH expression was higher in SNP- than the Ref-homozygous FLS lines.

Effect of DNA methylation on LBH enhancer function

Because a DML is located 185bp from the SNP, we determined whether the presence of the Ref or SNP allele is responsible for the DML methylation, such as by recruiting DNA methyltransferases to specific loci nearby. Pyrosequencing of 7 Ref homozygous and 14 SNP homozygous RA lines showed that the methylation at the RA DML was similar in both groups, indicating that the Ref and SNP alleles are likely not responsible for hypomethylation at the DML (median % methylation: Ref = 9.4, SNP = 9.0; Mann-Whitney test, p=0.62, Fig 5A). We then evaluated whether enhancer methylation alters its transcriptional activity by methylating all CpG loci in the Ref, SNP or control plasmids in vitro (Fig S2). We transfected methylated Ref, SNP or control plasmids and compared enhancer activity to their respective unmethylated plasmids. Figure 5B shows that enhancer methylation decreased transcription in both the SNP and Ref alleles (Inhibition: Ref 36±5% and SNP 30±9%, n=7 different RA lines/group, paired t-test; Fig 5B). These data suggest that the enhancer function is determined by a combination of its sequence (SNP vs. Ref) and its methylation state.

Fig 5. Regulation and function of DNA methylation in LBH enhancer.

(A) Pyrosequencing was performed in RA FLS lines identified as Ref- or SNP-homozygous. Methylation of the DML was similar in both groups, indicating that the presence of the variant did not alter methylation status. (B) Fully methylated or unmethylated enhancer plasmids with Ref or SNP alleles were transfected into RA FLS and luciferase activity was measured. Methylation significantly decreased luciferase activity compared with unmethylated plasmids.

Transcription factor binding to the enhancer region in Ref and SNP RA FLS

To determine if altered transcription factor binding to enhancer with Ref or SNP allele might contribute to differences in LBH expression, we screened the variant sequences using HaploReg(22), ENCODE ChIP-seq database (Genome Browser), TOMTOM(23) and LASAGNA –Search2.0(24). LASAGNA-Search2.0 and TOMTOM indicated a sequence specific binding of the transcriptional repressor RP58 and GATA6 while the HaploReg and ENCODE ChIP-seq database indicated that RNA polymerase II (pol II) and CTCF bind the enhancer region. RP58 and GATA6 mRNA expression was confirmed in RA FLS (data not shown). However, ChIP-qPCR analysis did not detect RP58 or GATA6 binding to the enhancer. In contrast, RNA pol II binding was detected in the enhancer but was unaffected by the presence of the variant (Fig 6). CTCF binding was very low or undetectable in both groups (data not shown). Together, these data show that RNA pol II binds the enhancer region to increase LBH expression in both SNP and Ref lines.

Fig 6. Detection of transcription factor binding to the enhancer region.

Chromatin immunoprecipitation was performed using anti-RNA pol II antibody in Ref- or SNP-homozygous RA FLS lines and qPCR was performed with enhancer-specific primers. RNA pol II binding was detectable in the enhancer region with the SNP or Ref alleles.

Discussion

Over 100 non-HLA GWAS SNPs have been identified as RA-risk loci, at least 19 of which are categorized as missense variants(25). The best characterized are functional variants in PTPN22 and IL-6 receptor genes, which alter protein function and therefore, increase susceptibility to multiple autoimmune diseases(26-28). However, the majority of GWAS RA-risk alleles are intra- or intergenic non-coding variants without obvious phenotype or effect on protein function(29, 30). Non-coding GWAS variants are frequently located in DNaseI hypersensitivity sites and can potentially regulate gene expression in a cell- or stimulus-specific way(29). Identifying how these non-coding variants regulate gene expression, function, and disease susceptibility can increase our understanding of disease pathogenesis.

We previously identified candidate genes by a systematic integrative analysis of three genome-wide FLS data sets, namely promoter DMLs, mRNA expression and GWAS SNPs(9). Of the original candidate genes, we focused on LBH, a highly conserved transcriptional coactivator, since very little was known about its function in RA. Subsequent functional studies showed that LBH expression in FLS is regulated by PDGF, which is a major growth factor that contributes to synovial lining hyperplasia in RA(11). Low LBH expression, which has been previously documented for RA FLS(31), allows cells to progress from G1 to S phase and can contribute to hyperproliferative state of RA FLS. Understanding why low LBH expression could participate in RA pathogenesis can provide insights into disease mechanisms and possibly novel therapeutic targets.

In the present study, we expanded our original promoter DML analysis to include DMLs in DNaseI hypersensitive sites, which are likely transcriptional enhancers with RA specific activity. The full integrative analysis identified a set of 12 genes, in which LBH appeared twice: once because of a DML in its promoter and second, in a novel regulatory region. The intriguing co-localization of an RA-associated variant and an RA DML in an enhancer led us to dissect how this enhancer might regulate a pathogenic gene in RA.

After confirming that the enhancer regulates transcription, we demonstrated that the RA-risk (Ref) allele is associated with reduced enhancer activity compared with the protective SNP allele (rs906868). Furthermore, LBH expression was lower in RA FLS that were homozygous for the Ref allele compared with the SNP allele. This observation suggested that the polymorphism could contribute to increased FLS proliferation in RA. LBH gene expression is also reduced in SLE peripheral blood mononuclear cells (PBMC) compared with control, suggesting that LBH could play a role in multiple autoimmune diseases and that various LBH risk alleles might participate(15). LBH expression was modestly reduced in Ref compared with SNP PBMCs in SLE, although the mechanisms were not defined(13). In other studies, LBH over-expression increased phosphorylation of NF-κB inhibitor, IκB, while loss of LBH enhanced NF- κB activity, indicating that lower LBH expression promotes inflammation(12).

Although the focus of our study was rs906868, several LBH reference variants outside the enhancer region have been identified and are associated with RA(14). These variants have high linkage disequilibrium (D’>0.95) with rs906868 (Fig S3). Recent studies showed four LBH SNPs (rs7579944, rs1355208, rs1396838, rs906868) are modestly protective in RA but confer risk in celiac disease; while the role of LBH vary depending on the disease, the data suggest that the gene plays a key role in many inflammatory states(14). These results are consistent with a meta-analysis study on Asian and European RA patients demonstrating that rs7579944 confers significant risk(32). The individual and cumulative effects of these variants on enhancer function can ultimately be evaluated using recently developed high-throughput strategies(33, 34).

We were particularly interested in the interactions between epigenetic marks, RA-associated SNPs and enhancer function. The fact that a differentially methylated locus was identified in such close proximity to a functional SNP led us to explore how methylation might affect the enhancer function. The complex relationship between non-CpG GWAS variants, the methylation status of neighboring CpG loci and their effect on gene expression is not well understood. In response to temporal or environmental cues, methylation can be associated with increased or decreased gene expression in enhancers, introns, and gene bodies, but the mechanisms are unclear(35). Recent genome-wide studies have shown that disrupting transcription factor binding motifs causes changes in DNA methylation(36, 37) and that SNPs can regulate (i) gene expression through methylation, (ii) methylation through expression or (iii) methylation and gene expression independently(38).

Our data show that altering LBH enhancer methylation decreased its activity, independent of the Ref or SNP variant. The CpG locus near the RA-associated variant is hypomethylated and is associated with lower LBH expression might appear counter-intuitive. However, we cannot assume that low methylation levels at this particular site is a primary event, as it could also be a specific regulatory mechanism to increase LBH expression and suppress FLS proliferation. The mechanism of differential methylation in vivo is an area of active investigation to address these questions. For example, non-coding RNAs that direct DNA methyltransferases(39) or steric hindrance from transcription factors could play a role, but are not yet defined in LBH. A direct effect of the SNP itself is unlikely because the SNP and Ref did not influence the methylation status of the enhancer. Future studies to evaluate the contribution of other CpGs in the enhancer and the mechanisms responsible for altered DNA methylation are needed to define the interplay between epigenetics and LBH expression in RA.

We also explored whether differential binding of transcription factors could account for lower transcription in Ref. Several candidate factors that could potentially bind at or near the SNP site were identified. However, ChIP-qPCR did not demonstrate a consistent difference between the two alleles, including RNA pol II. The SNP might affect enhancer activity by disrupting the DNA-binding site of an uncharacterized epigenomic regulator in FLS. Exploring this possibility would require the creation of genome-wide chromatin maps for RA FLS, which is an area of active investigation(30). Another possibility is that the SNP alters the sequence of enhancer RNAs that are transcribed from active enhancers(40) and regulate gene expression(41, 42) by forming complex with transcription factors and stabilizing their interaction with DNA(43).

In conclusion, an expanded integrative analysis of regulatory regions in RA FLS identified a novel LBH enhancer with an RA-associated polymorphism that affects enhancer function. The polymorphism is closely associated with a differentially methylated CpG locus, and methylation status of the region can dramatically affect enhancer function and gene expression. In RA, the combination of genetic and epigenetic marks can lead to the dysregulation of a pathogenic gene and subsequently, might contribute to synovial hyperplasia by allowing cells to progress through the G1 checkpoint and initiating the cell cycle.

Supplementary Material

Supp Info

Fig S1. Methylation status of CpGs neighboring cg20495819 from Illumina 450K chip in RA, OA and normal FLS

Fig S2. Confirmation of enhancer plasmid methylation pyrosequencing

Fig S3. LBH variants in high linkage disequilibrium with rs906868 that are associated with RA

Supp Table S1

Table S1. Overlap of DMG, DEG and RA GWAS variants in two or three data sets.

Supp Table S2

Table S2. KEGG pathway analysis