Preferential Binding to Elk-1 by SLE-Associated Risk Allele Upregulates Expression

Summary

Immunoregulatory cytokine interleukin-10 (IL-10) is elevated in sera from patients with systemic lupus erythematosus (SLE) correlating with disease activity. The established association of IL10 with SLE and other autoimmune diseases led us to fine map causal variant(s) and to explore underlying mechanisms. We assessed 19 tag SNPs, covering the IL10 gene cluster including IL19, IL20 and IL24, for association with SLE in 15,533 case and control subjects from four ancestries. The previously reported IL10 variant, rs3024505 located at 1 kb downstream of IL10, exhibited the strongest association signal and was confirmed for association with SLE in European American (EA) (P = 2.7×10⁻⁸, OR = 1.30), but not in non-EA ancestries. SNP imputation conducted in EA dataset identified three additional SLE-associated SNPs tagged by rs3024505 (rs3122605, rs3024493 and rs3024495 located at 9.2 kb upstream, intron 3 and 4 of IL10, respectively), and SLE-risk alleles of these SNPs were dose-dependently associated with elevated levels of IL10 mRNA in PBMCs and circulating IL-10 protein in SLE patients and controls. Using nuclear extracts of peripheral blood cells from SLE patients for electrophoretic mobility shift assays, we identified specific binding of transcription factor Elk-1 to oligodeoxynucleotides containing the risk (G) allele of rs3122605, suggesting rs3122605 as the most likely causal variant regulating IL10 expression. Elk-1 is known to be activated by phosphorylation and nuclear localization to induce transcription. Of interest, phosphorylated Elk-1 (p-Elk-1) detected only in nuclear extracts of SLE PBMCs appeared to increase with disease activity. Co-expression levels of p-Elk-1 and IL-10 were elevated in SLE T, B cells and monocytes, associated with increased disease activity in SLE B cells, and were best downregulated by ERK inhibitor. Taken together, our data suggest that preferential binding of activated Elk-1 to the IL10 rs3122605-G allele upregulates IL10 expression and confers increased risk for SLE in European Americans.

Introduction

The gene cluster that includes interleukin 10 (IL10), IL19, IL20 and IL24 is located on chromosome 1q31-32, a genomic region that is linked with susceptibility to systemic lupus erythematosus (SLE, OMIM 152700) [1], [2]. Recent genome-wide association (GWA) and follow-up replication studies in European ancestry have identified an association between the minor allele of rs3024505, a SNP located at 1 kb downstream of IL10, and increased risk for SLE [3], inflammatory bowel disease (IBD) including both Crohn's disease (CD) [4] and ulcerative colitis (UC) [5], [6], and decreased risk for type 1 diabetes [7], [8]. In addition, SNPs in the second intron (rs1518111) and the promoter region (rs1800871) of IL10 have been reported to be associated with Behçet's disease (BD) in GWAS of Turks [9] and Japanese [10]. These findings indicate IL10 as a common susceptibility locus shared by SLE and several other autoimmune diseases.

Dysregulation of IL-10 family cytokines contributes to autoimmune disease and tissue damage (reviewed in [11]). IL-10 is an important immunoregulatory cytokine with a wide variety of functions in T cells, B cells, natural killer cells, dendritic cells and macrophages [12]. The observations of elevated serum IL-10 levels in SLE patients correlating with increased disease activity [13], [14], and promising findings of anti-IL-10 monoclonal antibody treatment in patients with SLE [15] support a pivotal role for IL-10 in the pathogenesis of SLE. Of interest, elevated IL-10 levels were also reported in first-degree relatives of SLE patients [16], [17], suggesting that levels of IL10 expression may be determined genetically.

In this study, we fine mapped the IL10 gene cluster for genetic association with SLE in 15,533 case and control subjects from four diverse ancestries, identified a causal variant rs3122605 at IL10 5′ upstream using both genetic and functional assays, and explored the underlying molecular mechanism in explaining the elevated IL-10 levels in patients with SLE associated with increased disease activity.

Results

Association of four IL10 SNPs with SLE susceptibility in European Americans

To fine map the IL10 gene cluster, we genotyped 19 tag SNPs in 15,533 case and control subjects from four ancestries, including European American (EA, 3,820 cases vs. 3,412 controls), African American (AA, 1,670 vs. 1,904), Asian (AS, 1,252 vs. 1,249) and Amerindian/Hispanic (HS, 1,445 vs. 781). Each SNP was assessed for the association with SLE susceptibility under a logistic regression model adjusted for gender and global ancestry (Figure 1A). In the largest EA dataset, rs3024505 located at 1 kb downstream of IL10 exhibited the strongest association with SLE (minor allele frequency of 18.2% in cases vs. 14.8% in controls, P = 2.7×10⁻⁸, OR [95%CI] = 1.30 [1.19–1.43]), which exceeded the GWAS significance level of P<5×10⁻⁸ (Table S1). To identify additional SLE-associated SNPs, we performed SNP imputation using 1000 Genomes Project data as a reference. In EA, a total of 109 well-imputed SNPs spanning 154 kb from IL10 downstream to FAIM3 were assessed for association with SLE. Of them, three imputed SNPs (rs3122605, rs3024493 and rs3024495 located at 9.2 kb upstream, intron 3 and 4 of IL10, respectively), which were in tight linkage disequilibrium (LD, r²>0.9) with the genotyped SNP rs3024505, were strongly associated with SLE and remained significant after Bonferroni correction (rs3122605: P = 1.3×10⁻⁸, OR = 1.34 [1.21–1.48]; rs3024493: P = 5.0×10⁻⁸, OR = 1.29 [1.18–1.42]); rs3024495: P = 1.0×10⁻⁷, OR = 1.29 [1.17–1.41]) (Figure 1B and Table S1). None of the genotyped or imputed SNPs was significantly associated with SLE in three non-European datasets including AS, AA and HS after Bonferroni correction (Table S2). These data confirmed that IL10 is a risk locus for SLE in EA, and thus we subsequently focused on EA only to identify the causal variant(s).

IL10 promoter SNPs rs1800872 (also named as −592T/G), rs1800871 (−819G/A) and rs1800896 (−1082T/C), which were identified to be associated with elevated IL-10 production and SLE susceptibility in some, but not all of previous studies (reviewed in [18], [19]), showed nominal association with SLE in our EA dataset (Table S1). The BD-associated SNP rs1518111 showing effect on decreased IL-10 levels [9] was not associated with SLE in EA (Table S1). We performed haplotypic analysis to investigate relationships between these four previously reported SNPs and rs3024505, rs3024495, rs3024493 and rs3122605. Only the haplotype H4 carrying risk alleles of rs3024505, rs3024495, rs3024493 and rs3122605 was strongly associated with SLE (frequency of 16.8% in cases vs. 13.6% in controls, P = 1.3×10⁻⁷) (Figure 1D).

Due to strong LD among rs3122605, rs3024493, rs3024495 and rs3024505 in EA ancestry, their associations with SLE were highly correlated and could not be distinguished from each other using the conditional haplotype-based association test (Table S1). Conditioning on rs3122605, rs3024493, rs3024495 and rs3024505, respectively, association signals (P<0.05) of all other SNPs within the IL10 gene cluster were completely eliminated (Table S1), suggesting that these four SNPs within tight LD could capture all associations of the IL10 cluster region with SLE in EA. Of note, searching +/−200 kb of IL10 based on the 1000 Genomes Project data, we found that rs61815643 located at 10.3 kb upstream of IL10 was also in strong LD with rs3122605 (r² = 0.9) in European subjects, which suggested that this SNP might account for association signals detected within the IL10 gene cluster. However, because the imputation quality of rs61815643 did not reach the threshold of information score >0.9, it was not included for association test in this study.

Taken together, our data provide evidence supporting IL10 as a risk locus for SLE in EA and the underlying causal variant(s) might be or tagged by rs3122605, rs3024505, rs3024495 and rs3024493.

Dose-dependent association between SLE-risk allele and elevated IL10 expression levels

To explore potential functional consequences of the SLE-associated SNPs (rs3122605, rs3024493, rs3024495 and rs3024505), we assessed their genetic effects on influencing IL10 expression. IL10 mRNA levels in peripheral blood mononuclear cells (PBMC) and IL-10 protein levels in plasma from EA subjects were measured by quantitative real-time PCR and ELISA, respectively. Using rs3122605 as a surrogate of the other three SNPs, we compared IL10 expression levels among subjects carrying different genotypes of rs3122605. In control subjects, the SLE-risk (G) allele of rs3122605 was dose-dependently associated with elevated IL10 expression at both mRNA (n = 55; P = 0.001, R² = 0.18 in linear regression) and protein (n = 116; P = 3.3×10⁻⁷, R² = 0.21) levels (Figure 2). Consistently, dose-dependent association of the risk allele with elevated IL10 expression was also observed in patients with SLE at both mRNA (n = 58; P = 6.4×10⁻⁶, R² = 0.31) and protein (n = 132; P = 1.4×10⁻⁶, R² = 0.16) levels (Figure 2).

Compared to healthy controls, higher IL10 expression was observed in patients with SLE carrying the same genotype at both mRNA (genotype AA: P = 1.3×10⁻⁴, AG: P = 4.3×10⁻⁴, GG: P = 3.9×10⁻⁵, cases vs. controls in t test) and protein (AA: P = 4.6×10⁻⁴, AG: P = 4.9×10⁻⁶, GG: P = 0.047) levels (Figure 2), probably due to the activated immune status of SLE patients.

The risk allele of rs3122605 creates a novel binding site to transcription factor Elk-1 at 5′upstream of IL10

We hypothesized the presence of transcription factors activated in SLE patients upregulating IL10 expression and prepared nuclear extracts of peripheral blood lymphocytes from active SLE patients (defined as SLEDAI score≥4) [20], [21] to perform electrophoretic mobility shift assays (EMSA) for testing allelic differences in transcription factor binding conferred by the SLE-associated SNPs rs3122605, rs3024505, rs3024495 and rs3024493. Because we could not exclude the possibility that rs61815643 is a SLE-risk SNP affecting IL10 expression, it was also tested by EMSA.

Upon incubation with nuclear extracts, specific mobility-shift bands were only detected using the oligodeoxynucleotide probe containing the risk, but not the non-risk allele of rs3122605 (Figure 3A). Of interest, no specific binding of nuclear proteins was observed with the oligodeoxynucleotide probes containing either the risk or the non-risk allele of rs3024505, rs3024493, rs3024495 and rs61815643 (Figure S1). In silico analysis using the program TFSEARCH indicated that the risk allele of rs3122605 might create a novel binding site of the transcription factor Elk-1 (ETS-like transcription factor 1). We validated this prediction by showing that the addition of polyclonal rabbit IgG anti-ElK-1 antibody produced a super shift band only to the probe containing the risk but not the non-risk allele of rs3122605 in EMSA (Figure 3A).

Taken together, these data showed that rs3122605, the risk allele of which binds to Elk-1 in peripheral blood lymphocytes from SLE patients with active disease, is more likely to be the causal variant upregulating IL10 expression than the other four candidate SNPs.

Aberrant activation of Elk-1 in the nuclei of PBMCs from patients with SLE

Elk-1 is activated through phosphorylation and the phosphorylated Elk-1 (p-Elk-1) translocates into the nucleus to induce gene transcription [22]. To investigate the role of Elk-1 in SLE, we compared the distribution and activation of Elk-1 in PBMCs between SLE cases (n = 4) and healthy controls (n = 4) using Western blot. The amount of total Elk-1 was higher in nuclear (Figure 3B) but lower in cytoplasmic extracts (Figure 3C) of cases than controls. Of interest, p-Elk-1 was detected only in nuclear extracts of cases but not in controls (Figure 3B) and not in cytoplasmic extracts of either cases or controls (Figure 3C). These data suggest that Elk-1 is aberrantly activated and accumulates in nuclei of SLE PBMCs. Furthermore, the amount of total Elk-1 and p-Elk-1 appeared to be increased in SLE patients with higher SLEDAI scores (Figure 3B).

Co-expression of IL-10 and p-Elk-1 increases with SLE disease activity in B cells and could be best down-regulated by ERK inhibitor

Using flow cytometry, we quantified the co-expression of IL-10 and p-Elk-1 in specific cell subsets of PBMCs including CD3⁺ T cells, CD19⁺ B cells and CD14⁺ monocytes from healthy controls and SLE patients and showed representative data plotted in two-dimensional dot plots in Figure 4A. Compared to healthy controls (n = 3), percentages of IL-10⁺p-Elk-1⁺ cells were significantly increased in B (n = 11), T cells (n = 12) and monocytes (n = 12) from SLE patients (Student's t-test: P = 0.013, 0.012 and 0.012, respectively, Figure 4B). Active SLE patients had significantly elevated IL-10⁺p-Elk-1⁺ double positive B cells compared to non-active SLE patients (n = 7 vs. 4, P = 0.038). Similar trends for association with SLE disease activity were detected in IL-10⁺p-Elk-1⁺ T cells and monocytes, but the difference was not statistically significant (P = 0.74 and 0.57, respectively).

Elk-1 is known to be activated by mitogen-activated protein kinases (MAPK), including ERK (extracellular-signal-regulated kinases), JNK (c-Jun N-terminal kinases) and p38. We wondered which MAPK inhibitor could best down-regulate the elevated co-expression of IL-10 and p-Elk-1 in PBMCs. Because of the low expression of IL-10 in freshly isolated control PBMCs (Figure S2), we stimulated control PBMCs with IFN-α, a pivotal cytokine upregulated in most SLE patients [23], to mimic SLE PBMCs and incubated them with the MAPK inhibitors specific to ERK (PD 98059), JNK (SP 600125) or p38 (SB 203580), respectively. IFN-α-stimulation could significantly increase the percentage of IL-10⁺p-Elk-1⁺ cells in B, T cells and monocytes (n = 5, P = 0.0002, 0.0018 and 0.0079, respectively, Figure 4C), and such increase could be best down-regulated by the addition of the ERK inhibitor (P = 2.8×10⁻⁵, 0.0046 and 0.0079 in B, T cells and monocytes, respectively, Figure 4C). To a lesser extent, the p38 inhibitor also significantly suppressed IFN-α-induced double positive cells in all three cell subsets (P = 0.0016, 0.01 and 0.0039 in B, T cells and monocytes, respectively). Under our experimental conditions, the JNK inhibitor significantly inhibited the percentage of IL-10⁺p-Elk-1⁺ cells in B cells and monocytes (P = 0.0048 and 0.009, respectively), but not in T cells (P = 0.10).

Discussion

Our data provide strong evidence for a dose-dependent association between SLE-predisposing IL10 genotypes and corresponding mRNA and protein levels of IL-10, and identify one underlying molecular mechanism to explain previous findings of elevated IL-10 serum levels in SLE patients that positively correlated with increased disease activity. In addition to confirming the previously reported association with SLE at the IL10 3′ downstream SNP (rs3024505) in European Americans, we identified a SLE-associated risk haplotype, defined by the minor alleles of four SNPs in tight LD, rs3024505, rs3024495, rs3024493 and rs3122605, which could best explain the association with SLE and capture underlying causal variant(s) within the IL10 gene cluster in EA ancestry. The minor allele of rs3122605, which tags the IL10 SLE-risk haplotype, exhibited a dose-dependent association with elevated IL10 expression at both mRNA levels in PBMCs and protein levels in plasma samples from SLE patients and healthy controls, suggesting that these four SLE-associated SNPs may act by influencing IL10 regulation. Further functional studies showed that only rs3122605 was experimentally validated to confer preferential allele-binding to the transcription factor Elk-1 present in SLE PBMCs, hence the most likely functional variant present on the IL10 risk haplotype. Compared to normal PBMCs, nuclear localization of activated p-Elk-1 was observed only in SLE PBMCs. Co-expression of p-Elk-1 and IL-10 was significantly increased in all SLE PBMC subsets compared to normal PBMC subsets. Of interest, SLE patients with active disease had higher double positive (p-Elk-1⁺IL-10⁺) B cells than those with inactive disease. These data suggested that nuclear accumulation of activated Elk-1 in SLE peripheral lymphocytes contributes to overproduction of IL-10 in SLE patients associated with disease activity.

An abnormally high production of IL-10 in patients with SLE has been consistently demonstrated in many studies (reviewed in [18]), but the underlying molecular mechanism remains less well-characterized. The observation that healthy relatives of SLE patients also exhibit increased levels of IL-10 [16], [24], [25] suggests a possibility of genetically regulated IL-10 production. A number of genetic polymorphisms in the IL10 promoter region have been reported [26]–[30], in particular, three SNPs shown in Figure 1D located at −1082 (rs1800896, C/T), −819 (rs1800871, G/A) and −592 (rs1800872, G/T) have been inconsistently associated with IL-10 production levels and risk of SLE (reviewed in [18], [19]). Identification of SNPs in the −1.3 to −4 kb region of the IL10 promoter associated with both IL-10 production phenotypes and SLE susceptibility [31] suggested that further evaluating the contribution of SNPs in the more distal promoter of IL10 might be warranted. Consistently, the SNP (rs3122605) we identified that tags the SLE-risk haplotype in EA ancestry and confers genetic effect on IL10 expression is located at 9.2 kb upstream of IL10. According to the ENCODE Project, rs3122605, rs3024505, rs3024493 and rs61815643 were located within DNasel hypersensitive and transcription factor binding sites in at least one cell type (as shown in UCSC genome browser), suggesting that each of them may affect gene expression through interaction with regulatory elements. Given that gene regulation by genetic variants often occurs within the specific cell types most relevant to the disease phenotype [32], we used nuclear extracts from PBMCs of active SLE patients to perform EMSA assays and found only the SLE-risk allele of rs3122605 preferentially binds to the transcription factor Elk-1. Therefore, our data support rs3122605 as the most likely causal variant on the SLE-associated haplotype and implicate a potential importance for Elk-1-mediated upregulation of IL10 expression in SLE patients, particularly in those patients carrying the risk allele of rs3122605.

Elk-1 is a member of the Ets oncogene family of transcription factors characterized by a conserved DNA-binding domain and a C-terminal activation domain containing multiple phosphorylation sites targeted by three major MAP kinase pathways [33]. Different phosphorylation patterns of Elk-1 mediated through activation of MAPK signaling cascades by distinct external stimuli are important for Elk-1 to execute its physiologic functions [34]–[37]. We used an antibody to measure phosphorylation of Elk-1 at S383, and detected the presence of Elk-1 pS383 in the nuclei of SLE but not normal PBMCs. Quantification of IL-10 and p-Elk-1 co-expression confirmed a higher proportion of IL-10⁺p-Elk-1⁺ cells in SLE than normal PBMC subsets. These findings imply that Elk-1 in nuclei of SLE PBMCs has been biologically activated likely due to a higher baseline immune activation status, which may enhance ability of Elk-1 to regulate IL10 transcription. In the absence of microenvironmental activation in control PBMCs, Elk-1 remains within the cytoplasm and less translocated into nuclei, limiting its regulation effects. This hypothesis may in part explain the observation of higher IL-10 expression in SLE patients than healthy controls even if they carry the same risk genotype of rs3122605. In support of this possibility, previous studies revealed an increased expression of activation markers on peripheral lymphocytes of SLE patients, including phosphorylated ERK, JNK and p38 which are prerequisite for subsequent activation of Elk-1 [38], [39]. In addition, our data showed a significantly increased proportion of IL-10⁺p-Elk-1⁺ cells in normal PBMC subsets (Figure 4C) when exposed to IFN-α which may induce a partial activation phenotype in lymphocytes mimicking that of SLE.

We used healthy control, rather than SLE, PBMCs for testing effects of MAPK inhibitors on co-expression of IL-10 and p-Elk-1, because disease activity and medications of SLE patients might confound the results and leucopenia of SLE patients could limit the amount of PBMCs available for our experiments. In all three cell subsets of PBMCs, inhibition of ERK could best suppress IFN-α induced increase in IL-10 and p-Elk-1 co-expression, highlighting the importance for phosphorylated Elk-1 via ERK signaling in regulation of IL10 transcription. The ERK-dependent activation of Elk-1 has been clearly demonstrated in neuronal cells (reviewed in [40]) in which phosphorylation of Elk-1 at S383/389 by ERK is tightly linked to its activation and nuclear translocation and inhibition of phosphorylation results in cytoplasmic Elk-1 retention, limiting its transcriptional properties [22].

Elk-1, like all members of Ets-domain containing transcription factors, can bind genomic regions similarly as well as uniquely to regulate distinct classes of target genes [41]. Another member of the Ets family, Ets-1, has a dual function in regulating IL10 gene expression acting as both a transcriptional activator with the binding partner Sp-1 in HIV-1Tat-induced IL10 transcription in THP-1 cells [42], and as a repressor interacting with histone deacetylase 1 (HDAC1) in Th1 cells [43]. Increasing evidence indicates the involvement of Ets-1 in the pathogenesis of SLE: (1) ETS1 has been identified as a risk locus for SLE in GWAS [44], [45] and the risk allele is associated with decreased levels of ETS1 transcripts in healthy control PBMCs [45]; (2) Ets-1 is critical in maintaining B cell identity and its absence drives terminal differentiation of B cells into immunoglobulin-secreting plasma cells [46], [47]; (3) Ets-1 functions as a cofactor for T-bet essential for Th1 effector function and differentiation [48]; (4) Ets-1 negatively regulates Th17 cell differentiation and Ets-1 deficiency results in elevated production of IL-17 and IL-17-related cytokines by Th17 cells [49]. These emerging findings support an important role of Ets family transcription factors in the development of SLE manifestations.

Previous findings indicated that increased production of IL-10 by SLE PBMCs was mainly derived from B cells [16], [50]–[53]. One explanation might be due to elevated expression of toll-like receptor 9 (TLR9) on B cells of SLE patients with active disease, as the study showed that TLR9-CpG interaction could enhance the production of anti-dsDNA antibody and IL-10 [54]. B cell receptor (BCR) stimulation or BCR-TLR9 costimulation have been shown to activate the Erk pathway in B cells of NZB×NZW F1 mice [55]. A plausible explanation of elevated co-expression of IL-10 and p-Elk-1 in B cells from SLE patients, especially during active disease, might be attributable to activated ERK and Elk-1 involved in the BCR-dependent IL-10 production in SLE B cells.

Overexpression of B-cell-derived IL-10 contributes to the pathogenesis of SLE likely dependent on its ability to promote B cell proliferation, differentiation and autoantibody production [56], [57]. Recently identified IL-10 producing regulatory B cells (Bregs) may exert immunosuppressive effects to modulate murine lupus [58]–[63]. The phenotypic markers of human Breg cells have not reached consensus; they may be enriched in CD19⁺CD24^hiCD38^hi and CD24^hiCD27⁺ peripheral blood cells [64], [65]. Interestingly, CD19⁺CD24^hiCD38^hi B cells from SLE patients produce less IL-10 upon stimulation and are functionally impaired in suppressive capacity [64]. Thus, it seems unlikely that Bregs are the major producer of elevated IL-10 we observed in SLE patients.

The minor allele of rs3024505 showed consistently higher frequency in SLE patients than controls in all four ancestries, but only reached statistical significance for association with SLE in EA. Upon considering different genetic models, the additive model yielded the best genotypic association in EA (P = 2.7×10⁻⁸, OR [95%CI] = 1.30 [1.19–1.43]). Given the lack of evidence for genetic heterogeneity across EA, AS, AA and HS (P = 0.66 for Q statistic), the lack of significant association between rs3024505 and risk of SLE in AS, AA and HS might be due to low minor allele frequency and small sample size. Under the assumption that the minor allele of rs3024505 confers genetic risk with an odds ratio of 1.3 (determined in EA), the power to detect a significant association (P<0.05) for EA samples reaches 100%, whereas it is only 31% in AS, 55% in AA and 58% in HS datasets. According to the 1000 Genome Project data, the LD strength between rs3024505 and our proposed causal SNP rs3122605 is similarly strong in Asians (r² = 1.0) and Americans/Hispanics (r² = 0.8) as in Europeans (r² = 0.85), but not in Africans (r² = 0.1), suggesting that rs3122605 can be tagged by rs3024505 in non-Europeans except for the African-derived population.

It is possible that other SLE-associated variant(s) specific for AS, AA or HS failed to be captured by SNPs used in this study due to different LD pattern in each ancestry. Among non-European populations, SLE GWAS conducted in AA and HS are not in the currently available literature. To our knowledge, there have been four published SLE GWA studies conducted in AS, including two Chinese [44], [45], one Japanese [66] and one Korean GWAS [67], and a meta-analysis study based on Chinese GWAS [68]. In these studies, the Illumina Human 610 BeadChip is the commonly used genotyping platform, which contains nine IL10 SNPs (including rs3024505) and can capture 23 of the 38 common SNPs (MAF>1%) within +/−5 kb of IL10 with r²>0.9 in Asians (according to the 1000 Genome Project data). Because none of these studies have reported association signals of IL10 SNPs with SLE, current data are consistent with our findings that the IL10 locus is not a strong genetic risk factor for SLE in AS. Taken together, the association of IL10 with SLE in non-EA ancestries awaits further investigation.

We have searched four publically available cis-eQTL datasets conducted in immune cells from healthy Europeans [69]–[72] but found no convincing evidence to support the presence of another SNP that can better capture the association signal of IL10 expression trait in these datasets than our data of SLE-associated rs3024505/rs3122605. In addition, there was no convincing evidence to support the association of rs3024505 with differential expression of other genes within +/−1 Mb flanking region of IL10.

Given that the SNP rs3024505 confers increased risk for both SLE and IBD and could be tagged by rs3122605, it is possible that patients with IBD carrying the risk allele of rs3024505 may exhibit high serum levels of IL-10. There is evidence supporting elevated circulating IL-10 levels in both CD and UC patients that positively associated with disease activity [73]–[75], similarly to previous reports in SLE patients. However, genetically engineered mice exhibiting low to no IL-10 signaling in the intestinal tract develop severe IBD manifestations [76]–[78], supporting a pivotal role of IL-10 in down-regulation of inflammation. Increased levels of circulating IL-10 may be elicited by chronic inflammation in IBD, but may not be sufficiently strong to dampen intestinal inflammation [75], raising the possibility of defective IL-10 signaling at sites of organ damage in patients with SLE.

In conclusion, by characterizing genetic variations within the IL10 gene cluster region, we have identified the IL10 upstream SNP rs3122605 as the best likely causal variant responsible for association with SLE in European Americans. The SLE-associated rs3122605 G-allele preferentially binds to the activated Elk-1 conferring elevated IL10 expression. The observation that SLE patients, particular those with increased disease activity, showed enhanced activation of Elk-1 in nuclei and elevated co-expression of IL-10 and phospho-Elk-1 in peripheral lymphocytes highlights the involvement of aberrant Elk-1 signaling in development of SLE and suggests potential targeting therapy for disease amelioration.

Materials and Methods

Ethics statement

This study was approved by the Institutional Review Boards (IRBs) or the ethnic committees at the institutions where subjects were recruited. All subjects were enrolled after informed consent had been obtained. The overall study was approved by the IRB of the Oklahoma Medical Research Foundation (OMRF).

Subjects

To test the association of IL10 family genes with SLE, we used a large collection of case-control subjects from the collaborative Large Lupus Association Study 2 (LLAS2) [79], including European American (EA: 4,248 cases vs. 3,818 controls), African American (AA: 1,724 vs. 2,024), Asian (AS: 1,328 vs. 1,348) and Hispanic enriched for the Amerindian-European admixture (HS: 1,622 cases vs. 887 controls). African Americans included 286 Gullahs (155 cases vs. 131 controls), who are subjects with African ancestry. Asians were composed primarily of Koreans (906 cases and 1,012 controls) but also included Chinese, Japanese, Taiwanese and Singaporeans. Cases were defined by meeting at least four of the 1997 American College of Rheumatology (ACR) revised criteria for the classification of SLE [80].

To test functional consequences of SLE-associated variants, SLE patients and healthy controls of European ancestry were recruited at the University of California, Los Angeles and through the Lupus Family Registry and Repository (LFRR, lupus.omrf.org) for blood donations.

Genotyping and quality control

DNA samples were processed at the Lupus Genetics Studies Unit of OMRF. SNP genotyping was performed using an Illumina custom bead array on the iSCAN instrument for 19 tag SNPs covering over 135 kb of IL10-IL24 region and 347 admixture informative markers (AIMs). SNPs meeting the following criteria were included for subsequent genetic association tests: well-defined cluster scatter plots, SNP call rate >90%, minor allele frequency >1%, total proportion missing <5%, P>0.05 for differential missing rate between cases and controls, and Hardy-Weinberg proportion (HWP) test with a P>0.01 in controls and P>0.0001 in cases.

Subjects with individual genotyping missing rate >10% (due to low quality), shared identity by descent >0.4 or showing mismatch between the reported and estimated gender were removed. The global ancestry of each subject was estimated based on genotype of AIMs, using principal components analysis (PCA) [81] and ADMIXMAP [82]–[84], as described in another LLAS2 study [85], and then genetic outliers were removed.

Finally, a total of unrelated 15,533 subjects including EA (3,820 cases vs. 3,412 controls), AA (1,670 vs. 1,904; composed of 92.5% African Americans and 7.5% Gullahs), AS (1,252 vs. 1,249; composed of 74.6% of Koreans, 16.1% of Chinese and subjects from Japan and Singapore) and HS (1,445 vs. 781) were analyzed for 19 SNPs.

SNP imputation

Imputation was performed using IMPUTE 2.1.2 [86], with SNP genotypes of 379 Europeans (CEU, TSI, GBR, FIN and IBS), 246 Africans (YRI, ASW and LWK), 286 Asians (CHB, JPT and CHS) and 181 Americans (MXL, PUR and CLM) from the 1000 Genomes Project (version 3 of the phase 1 integrated data, March 2012 release) as references in imputation for our EA, AA, AS and HS subjects, respectively. Imputed genotypes had to meet the threshold of information score >0.9, as well as the quality control criteria as described above. After imputation, we obtained an additional 109 SNPs for EA, 45 for AA, 80 for AS and 64 for HS (the number varied due to different LD structure) for further analysis.

Real-time quantitative PCR

Total RNA was purified with TRIzol reagent (Life Technologies) from PBMCs of EA individuals (58 SLE cases and 55 healthy controls) and reverse-transcribed into cDNA with SuperScript II Reverse Transcriptase kit (Life Technologies). Messenger RNA levels of IL10 and a housekeeping gene RPLP0 were measured by quantitative real-time PCR using TaqMan assays (IL10 probe: Hs00961622_m1; RPLP0 probe: Hs99999902_m1, Applied Biosystems). All samples were run in triplicate. Relative IL10 mRNA levels were normalized to that of RPLP0, calculated by the 2^−ΔΔCt method and Log10 transformed.

Enzyme-linked immunosorbent assay (ELISA)

Plasma IL-10 levels from 132 SLE patients and 116 healthy controls of EA ancestry were measured by ELISA (R&D systems).

Cell cultures

To examine whether inhibition of MAPK pathway may affect co-expression of IL-10 and p-Elk-1, control PBMCs (1×10⁶) were cultured in growth medium with or without interferon alpha (IFNα) (1000 U/ml; PBL Biomedical Laboratories) in the presence or absence of one MAPK inhibitor (EMD Millipore), PD 98059 (20 µM; ERK/MEK inhibitor), SP 600125 (20 µM; JNK inhibitor) or SB 203580 (10 µM; p38-MAPK pathway inhibitor), respectively. Addition of Brefeldin A (eBiosciense) to cells in culture blocks intracellular protein (IL-10) transport processes.

Flow cytometry

The patients with SLE recruited in this part of study were evaluated for disease activity by the SLE Disease Activity Index (SLEDAI) 2000 [87] at the time of blood draw, and SLEDAI≥4 was considered as active disease [20], [21]. Freshly isolated or cultured PBMCs were incubated with mouse reference serum to block nonspecific binding to Fcγ receptors and then incubated with peridinin chlorophyll protein (PerCP)-conjugated anti-human CD3, allophycocyanin (APC)-conjugated anti-human CD19 and phycoerythrin (PE)-conjugated anti-human CD14 (eBiosciense) to identify T cell, B cell and monocyte subpopulations, respectively. For intracellular staining of IL-10 and p-Elk-1, cells were fixed with IC Fixation Buffer (eBiosciense), washed with Permeabilization Buffer (eBiosciense), and stained with fluorescein isothiocyanate (FITC)-conjugated anti-human IL-10 (eBiosciense) and PE- or Alexa Fluor647-conjugated anti-phospho-Elk-1 antibody (BD Biosciences). Background fluorescence was assessed using appropriate isotype- and fluorochrome-matched control antibodies. Cells were collected and analyzed by FACSCalibur flow cytometer equipped with the manufacturer's software (CellQuest; BD Biosciences).

Bioinformatic prediction of transcription factors, electrophoretic mobility shift assay (EMSA) and supershift assay

Bioinformatic analysis using the program TFSEARCH (conducted on 01/18/2011) showed predicted binding to Elk-1 and STRE in DNA sequence containing the minor but not the major allele of rs3122605 (Figure S3 and S4). In addition, HSF, ADR1 and MZF1 were predicted to bind with sequence containing either allele of rs3122605. Given that we were interested in identifying transcription factors that preferentially bind to the minor allele of rs3122605, and that the STRE (stress response element) binding factor includes two yeast transcription factors, Msn2p and Msn4p, we prioritized to test Elk-1 using EMSA.

EMSA and supershift assays were performed as previously described [88]. Nuclear extracts were prepared from peripheral blood lymphocytes of SLE patients with active disease using NE-PER Nuclear Extraction Reagent (Thermo scientific) and incubated with biotin-labeled oligodeoxynucleotides (synthesized by Integrated DNA Technologies, depicted in Table S3). EMSAs were performed with the LightShift Chemiluminescent EMSA kit (Thermo scientific). The antibody used in the supershift reactions was polyclonal rabbit anti-human Elk-1 (Santa Cruz Biotechnology).

Immunoblot analyses

Cytoplasmic or nuclear proteins from PBMCs were prepared using NE-PER Nuclear Extraction Reagent (Thermo scientific). Following SDS/PAGE, proteins were transferred onto Immobilon-P membrane (Millipore). After blocking with membrane blocking solution (Invitrogen, Life Technologies), the membrane was successively incubated with the anti-Elk-1 (Santa Cruz Biotechnology) or anti-phospho-Elk-1 (Cell Signaling Technology) primary antibody and the horseradish peroxidase (HRP)-conjugated secondary antibody (Santa Cruz Biotechnology). Blots were developed using the ECLPlus Western Blotting Detection System (GE Healthcare), visualized with ChemiDoc XRS imager and analyzed by Quantity One software (BIO-RAD). β-actin or Lamin B was used as internal control.

Statistical analysis

Allelic association tests in each ancestral group and conditional haplotype-based association tests in EA ancestry were performed by PLINK v1.07 software [89] under a logistic regression model adjusted for gender and the first three principal components estimated using AIMs. The Bonferroni corrected P-value threshold was adjusted to P<3.9×10⁻⁴ ( = 0.05/128 SNPs in EA). Pairwised LD values between SNPs and haplotypic association with SLE were evaluated using Haploview 4.2 [90]. The linear regression test was used to evaluate the association of SNP genotypes with IL-10 mRNA or protein levels. The Student's t-test was used to compare the mean values between two groups. A P value<0.05 was considered to be statistically significant.

Supporting Information

Attachment 1

Attachment 2

Attachment 3

Attachment 4

Attachment 5

Attachment 6

Attachment 7