Polymorphic estrogen receptor binding site causes Cd2-dependent sex bias in the susceptibility to autoimmune diseases - Nature Communications
Results . We have set out to identify major genetic polymorphisms underlying the development of autoimmune arthritis, using animal models. As part of these efforts a major quantitative trait locus (QTL) was identified on chromosome 3 qF2.2, which was termed Cia21 9 . Cia21 was identified from an intercross between the collagen-induced arthritis (CIA)-susceptible C57BL/10.RIII (BR) and the CIA-resistant RIIIS/J (R3) mouse strains 11 . Cia21 contains several differentially expressed genes, including Cd2 and Ptpn22 9 . Both Cd2 and Ptpn22 play a key role in T cell activation and were proposed as strong candidate genes. The aim of the present study is to identify the polymorphisms underlying the Cia21 QTL. A minimal non-coding genetic interval proximal to Cd2 explains Cia21 . To dissect the Cia21 QTL, we bred heterozygous Cia21 mice and evaluated the resulting recombinant mice (shown in Fig.? 1a ) using CIA (Fig.? 1b–g ). Out of all the evaluated recombinants, only two, numbers 1 and 5, recapitulated the protective arthritis phenotype previously observed in Cia21 mice 9 . Thus, the Cia21 QTL results from individual contributions of these two sub-QTLs. Importantly, the phenotype driving recombinant regions 1 and 5 mapped to the previously proposed 9 candidate genes Cd2 and Ptpn22 , respectively. Recombinant fragment 1 (proximal to Cd2 ), however, was significantly smaller than fragment 5 providing better conditions for the positional identification of underlying polymorphisms. Therefore, we focused our efforts on the former. Fig. 1: Map of Cia21 QTL and critical D3KV1-MF31 interval. The Cia21 QTL resulted from an intercross between the CIA susceptible C57BL/10.RIII (BR) and the CIA-resistant RIIIS/J (R3) strains. Cia21 is present on chromosome 3 qF2.2 and is 3 Mbp in size. a Schematic representation of the Cia21 QTL and recombinant mice derived by intercrossing of Cia21 heterozygotes. Important genetic markers and genes are indicated on the left. The critical D3KV1-MF96 interval is highlighted in yellow. Uncertainty borders are dashed. b – g Collagen-induced arthritis in female recombinant mice from ( a ) compared to BR littermate controls. Incidence and total number of mice are indicated in parenthesis on the respective graphs. Data are summarised as mean (SEM). Statistical significance was evaluated using a two-tailed non-parametric Mann–Whitney U test. Data in each graph was pooled from two independent experiments. h Detailed view of D3KV1-MF31 (fragment 1, D3-31) and proximal genes. The critical D3KV1-MF96 interval is highlighted in yellow. Coordinates according to mouse NCBI37/mm9 build. n.s. not significant. Full size image Recombinant fragment 1 stretched from markers D3KV1 to MF31 (Fig.? 1a , ca. 0.2?Mbp), but could be further redefined to the significantly smaller D3KV1-MF96 interval (ca. 0.02?Mbp) through a recombination assisted breeding strategy. Although recombinant fragments 1, 2 and 3 overlapped significantly, only fragment 1 regulated arthritis. Thus, we concluded that the causative polymorphisms must be positioned between markers D3KV1 and MF96 (Fig.? 1a , highlighted yellow). D3KV1-MF96 is a non-coding 0.02?Mbp region proximal to Cd2 , located in-between the genes Atp1a1 and Igsf3 (Fig.? 1h ). We isolated the D3KV1-MF31 recombinant fragment (termed D3-31) in a congenic mouse line for further investigations. D3-31 congenic mice carry the parental R3 allele of D3-31 on an otherwise BR background. For simplicity, we hereon refer to the congenic line as D3-31 and to wild-type littermates as BR. Female D3-31 mice are protected from T-cell-dependent models of autoimmunity . In accordance with previous data on Cia21, the R3 allele of D3-31 protected congenic mice in T-cell-dependent 12 , 13 , 14 autoimmune inflammatory models, including collagen-induced arthritis (CIA), experimental autoimmune encephalomyelitis (EAE) and delayed type hypersensitivity (DTH) (Fig.? 2a–f ). We also investigated the T-cell-independent 15 collagen antibody-induced arthritis (CAIA) model, but observed no phenotypic differences (Supplementary fig.? 1 ). As the DTH model does not depend on B cells 12 , these results indicated a critical role for T cells. Interestingly, and as previously described for Cia21 9 , only female D3-31 mice were protected from T-cell-mediated autoimmunity (Fig.? 2a–f ). Thus, we concluded that D3-31 regulates T-cell-dependent autoimmune phenotypes, and likely T cells, in a sex-specific manner. Fig. 2: D3-31 mice are protected from T cell-dependent autoimmunity in a sex-specific manner. Collagen-induced arthritis (CIA) in a female and b male BR and D3-31 mice. Data was pooled from two independent experiments. Delayed-type hypersensitivity (DTH) reaction in c female and d male BR and D3-31 mice. Box upper and lower limits indicate interquartile range (25th/75th percentiles), the middle line indicates the median. Whiskers indicate 10th and 90th percentiles. Min. and max. values are plotted as individual dots. Data is representative of three independent experiments with similar results. MBP89-101-induced experimental autoimmune encephalomyelitis (EAE) in e female and f male BR and D3-31 mice. Data was pooled from two independent experiments. For all graphs, incidence and total number of mice used are indicated in parenthesis. Data are summarised as mean (SEM). Statistical significance was evaluated using a two-tailed non-parametric Mann–Whitney U test. Full size image Female sex hormones are required for the protective phenotype in D3-31 mice . To discriminate between influence of sex chromosomes versus hormones, we performed CIA and EAE experiments in castrated female mice (Fig.? 3a–c ). Castration of female mice depletes gonadal production of 17-β-estradiol (E2) 16 , which constitutes the major circulating oestrogenic compound in females. Castration reverted the protective effect of the D3-31 fragment both in CIA and EAE (Fig.? 3a–c ), which demonstrated the crucial contribution of female sex hormones, most likely E2, to the protective phenotype in female D3-31 mice. We next?set out to define the genetic mechanisms underlying this sexually dimorphic immune phenotype by sequencing the D3-31 fragment. Fig. 3: Female sex hormones are required for the protective phenotype in D3-31 mice. a CIA severity and incidence (in parenthesis) in ovariectomized D3-31 and BR mice. Data was pooled from two independent experiments. b Incidence of EAE in ovariectomized (OVX) and sham operated (SHAM) D3-31 and BR mice. c Table comparing incidence, maximal score and accumulated severity of?the?EAE experiment shown in b . Data was pooled from three independent experiments. Data are summarised as mean (SEM). Statistical significance was evaluated using a two-tailed non-parametric Mann–Whitney U test. P ?=?0.0486; ? P ?=?0.0419. Full size image Polymorphic ERBS in D3-31 interval affects E2-mediated transcriptional activity . DNA sequencing of the D3-31 BR and R3 alleles revealed four single-nucleotide polymorphisms (SNPs) in the critical D3KV1-MF96 interval (Fig.? 4a, b ). None of the variants affected the coding region of known genes, indicating distal (cis) regulation of gene expression, likely by interfering with regulatory elements. Given our previous observations, we speculated that the identified polymorphisms could be located within an ERBS, interfering with sex-dependent regulation of gene expression. Fig. 4: Polymorphism in D3-31 ERBS affects E2-mediated transcriptional activity. a Sequencing results showing genetic variants within critical D3KV1-MF96 interval. b Detailed schematic overview of polymorphisms (denoted by red lines) in the D3KV1-MF96 interval. SNP478 denotes an AC?>?GG substitution on chr3:101310478-79. c ChIP-seq data from mouse uterus (extracted from SRX129062 63 ) showing Erα binding intensity to polymorphic regions listed in a . Consensus ER binding motif (UN0308.1 64 ) and SNP478 are highlighted in blue and red squares, respectively, where double asterisk indicates the position of SNP478. Coordinates according to mouse NCBI37/mm9 build. d Rabbit anti-mouse Erα ChIP-qPCR data confirming binding of Erα to SNP478 in spleen cells. A gene dessert was used as negative control (-ctrl) and a known Erα binding site ( Csf2ra 30 ) as positive control (+ ctrl). Values are expressed as fold enrichment over rabbit IgG mock IP. Each dot represents an independent mouse biological replicate. The data shown is representative of two independent experiments with similar results. e Effect of SNP478 on the transcriptional activity of the D3KV1 Erα binding site shown in c . The candidate?D3KV1?Erα binding site (chr3:101310478?±?100?bp to each side) was cloned in its two variant forms (AC and GG) into luciferase reporter constructs. The constructs were transfected into MCF7 cells to evaluate transcriptional activity. The data shown is from a total n ?=?9 technical replicates pooled from two independent experiments. Data are summarised as mean (SEM). Statistical significance was evaluated using a two-sided non-parametric Mann–Whitney U test. Full size image Oestrogen receptors (ERα and ERβ) are nuclear hormone receptors that translate E2-mediated signalling. Both ERα and ERβ are expressed in immune cells 17 , and act as transcription factors regulating the expression of proximal and distant genes 18 , 19 . To test our hypothesis, we screened publicly available ChIP-seq data for Erα binding sites overlapping with one or more of the sequenced SNPs within D3KV1-MF96 interval. Indeed, one of the SNPs, AC?>?GG on chr3:101310478-479 (termed SNP478), clearly overlapped with an Erα binding site (Fig.? 4c ). In fact, bioinformatic analysis also revealed an oestrogen response element (i.e. an ER core binding motif) in close proximity to SNP478. We sought to verify this finding and could confirm binding of Erα to SNP478 in mouse spleen cells using ChIP-qPCR (Fig.? 4d ). Comparison of SNP478 between mouse inbred strains revealed that this SNP is in fact part of a highly polymorphic AC/GT simple repeat (Supplementary fig.? 2 , extracted from Kent et al. 20 ). To address whether SNP478 had functional consequences for E2-mediated transcriptional activity (i.e. interfered with the binding of Erα to the DNA), we cloned the candidate D3KV1 ERBS (±100?bp) in its two variant forms (AC and GG) into luciferase reporter constructs. Leveraging the fact that human and mouse ERα are highly conserved 21 , we assessed transcriptional activity of these constructs in the?ERα expressing?MCF-7 human?cell?line, treating them with increasing concentrations of E2 (Fig.? 4e ). In the context of the reporter construct, an increased occupancy of the candidate ERBS by ERα (as a function of increasing E2) resulted in dose-dependent suppression of transcriptional activity. Although counterintuitive, similar observations have been reported elsewhere 22 . Given the stronger transcriptional inhibition when using the BR derived construct, we concluded that ERα/Erα has a higher affinity for the BR allele than for the D3-31 allele. Importantly, these data demonstrate that SNP478 has functional consequences for E2-mediated transcriptional activity. Polymorphic ERBS in D3-31 interval leads to female-specific changes in Cd2 expression . Next, we sought to test the biological relevance of our findings by comparing the gene expression profile in lymph node cells from male and female D3-31 and BR mice. We observed female-specific changes in the expression of three genes adjacent to the polymorphic ERBS, namely Cd2 , Igsf3 and Mab21l3 (Fig.? 5a, b ). We also investigated the expression of Atp1a1 as well as more distal genes ( Cd101 and Slc22a15 ) previously implicated in the non-obese diabetic (NOD) mouse model of type 1 diabetes 23 , but found no changes in their expression level. Notably, the female-specific reduction of Cd2 expression in D3-31 mice was also evident at protein level (Fig.? 5c ), correlating with our gene expression results and those previously reported in Johanesson et al. 9 . Fig. 5: D3-31 mice show sex-specific differences in Cd2 expression. a , b Expression of genes surrounding the D3-31 congenic fragment in lymph nodes cells. Expression data for female mice is shown in a ?(red), and for male mice in b ?(blue). Dotted red?lines indicate congenic fragment borders. c Cd2 protein expression in lymph node Cd4 + T cells from female and male D3-31 and BR mice (flow cytometry). Data in a – c is representative of three independent experiments with similar results. d Expression of Cd2 and other surrounding genes in lymph node?cells from BR mice. e Cd2 protein expression in blood T cells, B cells and monocytes (flow cytometry). f Secretion of Il-17a and Ifn-? in T cells stimulated with soluble anti-Cd3 mAb only, or soluble anti-Cd3 and soluble anti-Cd2 mAb. g Cd2 expression in lymph node T cells after in vitro culturing with increasing concentrations of 17-β-estradiol (E2). Data in d – g is representative of two independent experiments with similar results. h Comparison of Cd2 expression in T cells from D3-31 and BR mice cultured in normal medium (ctrl), charcoal-stripped?medium devoid?of E2 (-E2), or -E2 medium supplemented with 10?nM E2. Data in h is pooled from two independent experiments. In all figures, each dot represents one independent mouse biological replicate. Data are summarised as mean (SEM). Statistical significance was evaluated using a two-tailed non-parametric Mann–Whitney U test. Sequential flow cytometry gating strategies for c and e are provided in Supplementary fig.? 9 . Full size image Out of the differentially expressed genes, Cd2 was the only gene predominantly expressed in lymphoid tissue (Fig.? 5d ), particularly in activated Cd4 + T cells (Fig.? 5e ). Igsf3 and Mab21l3 regulate neural 24 and ocular 25 development, whereas Cd2 has been involved in immune function 26 and associated with human autoimmune conditions 4 , 27 . Indeed, treatment of lymph node cells with anti-Cd2 mAb inhibited T cell activation as demonstrated by reduced secretion of pro-inflammatory cytokines (Fig.? 5f ). Considering these data and normal development of D3-31 mice, we concluded that Cd2 is driving the T-cell-dependent immune phenotype observed in D3-31 mice. Given the sex-specific differences in gene expression, we next investigated the relation between E2 and Cd2 expression. T cells cultured in the presence of E2 upregulated Cd2 in a dose-dependent manner (Fig.? 5g ). Conversely, use of E2 depleted medium (achieved by using charcoal-stripped serum) reduced the expression of Cd2, and, more importantly, neutralised the observed differences in Cd2 expression between BR and D3-31 mice. Additionally, differences in Cd2 expression could be re-established by reintroducing E2 to the medium (Fig.? 5h ). This not only demonstrates direct regulation of E2 on Cd2 expression, but also proves that the identified polymorphisms interfere with this regulation. Consequently, we speculated that E2-mediated regulation of Cd2 was contributing to sex-specific differences in the T cell responses. A sex-dependent reduction of Cd2 expression in female D3-31 mice could likely limit the T cell responses. E2-dependent regulation of Cd2 leads to sex-specific differences in T cell activation . To investigate the impact of sex hormone-dependent alterations in Cd2 expression on the T cell responses, we compared the activation of T cells between BR and D3-31 female mice. In a first set of in vitro experiments, we found an impaired response in D3-31 T cells to Tcr stimulation, as evidenced by reduced proliferation and Il-2 production (Fig.? 6a, b ). Importantly, the difference in T cell proliferation between BR and D3-31 mice could be enhanced in a dose-dependent manner by E2 (Fig.? 6c ), much like the E2-dependent expression differences observed for Cd2 (Fig.? 5h ). Fig. 6: Sex-specific differences in Cd2 expression limit the T cell responses in female D3-31 mice. a , b Proliferation ( a ) and Il-2 secretion ( b ) of Cd4 + lymph node T cells after stimulation with anti-Cd3/anti-Cd28 mAbs. c Proliferation of BR and D3-31 Cd4 + T cells as in a in the presence of increasing concentrations of E2 (10–100?nM). d Antigen recall assay showing pro-inflammatory cytokine secretion by lymph node cell cultures from CIA mice after recall with bovine collagen type II (bCII). Lymph nodes were harvested 10 days after immunisation with bCII (day 10). e , f Quantification of antigen experienced Cd40l + Cd4 + T cells in lymph nodes from CIA mice (day 10) ( e ), and expression of Cd2 in these cells ( f ). g , h Representative gating ( g ) and quantification ( h ) of Il-17a + Cd40l + lymph node T cells from CIA mice (day 10) after ex vivo restimulation with PMA in the presence or absence of soluble anti-Cd2 mAb. i , j Representative gating ( i ) and quantification ( j ) of Cd25 + Foxp3 + Tregs in lymph nodes from CIA mice (day 10). k Expression of Il-17a and Foxp3 in Cd4 + na?ve T cells stimulated with PMA in the absence or presence of anti-Cd2 mAb. The data in a – k are representative of two independent experiments with similar results. l Volcano plot comparing the proteomic profile of Cd4 + T cells stimulated with immobilised anti-Cd3 mAb in the presence and absence of immobilised anti-Cd2 mAb. This experiment was performed once with n ?=?8 independent mouse biological replicates per group. m , n Flow cytometry data showing Lag-3 expression in Cd4 + T cells after culture with anti-Cd2 mAb. Representative gating in m and quantification in n . This experiment was performed three independent times with similar results. In all figures, each dot represents one independent mouse biological replicate. Data are summarised as mean (SEM). Statistical significance was evaluated using a parametric two-tailed t -test in d , and non-parametric two-tailed Mann–Whitney U test in all other experiments. Sequential flow cytometry gating strategies for a , e – j , m , n , as well as cell purity for l are provided in Supplementary figs.? 10 – 15 . Full size image A diminished T cell response in D3-31 mice was also evident in vivo. Compared to BR mice, D3-31 mice showed a lower level of antigen-specific T cells responses 10 days after immunisation with CIA antigen bovine collagen type II, as demonstrated by reduced secretion of pro-inflammatory cytokines in lymph node cell?cultures recalled with antigen (Fig.? 6d ). Flow cytometry analysis of D3-31 draining lymph node?cells revealed lower numbers of antigen experienced Cd40l + Cd4 + T cells (Fig.? 6e ), which expressed reduced levels of Cd2 (Fig.? 6f ) and Il-17a after ex vivo restimulation with PMA (Fig.? 6g, h ). Differences in T cell activation status were also evident given lower numbers of induced regulatory T cells after immunisation (Fig.? 6i, j ). Importantly, the observed differences in T cell activation were strictly sex-specific (Fig.? 6e–j ), mirroring sex-specific differences in Cd2 expression. Treatment with anti-Cd2 strongly reduced the expression of Il-17a in autoreactive Cd4 + T cells (Fig.? 6h ), as well as expression of Il-17a and Foxp3 in na?ve T cells (Fig.? 6k ), demonstrating the importance of Cd2 costimulation for the differentiation of Th17 and Treg type cells. Consequently, we concluded that reduced Cd2 expression in female D3-31 mice limits T cell activation in a sex-specific manner. To further characterise how impaired Cd2 signalling affected T cells, we compared the proteomic landscape of anti-Cd3-stimulated Cd4 + T cells in the presence or absence of anti-Cd2 mAb. Blocking of Cd2 signalling by means of anti-Cd2 mAb resulted in the selective upregulation of the immune inhibitory marker Lag-3 (Fig.? 6l–n ). Thus, we concluded that Cd2 costimulation is required for T cell activation and that impaired Cd2 signalling results in the upregulation of the inhibitory marker Lag-3. CD2 is associated with rheumatoid arthritis and is regulated by E2 in humans . Our results in mice suggested a regulatory role for CD2 on T-cell-dependent autoimmunity, which is genetically determined in a sex-linked manner. We therefore went on to explore the relevance of our findings in humans, specifically in the context of rheumatoid arthritis (RA). In a genetic association study, we found a significant association between CD2 polymorphisms and RA (Fig.? 7a ?and Supplementary fig. 8 ). While this association was more often found in females than in males, this was likely due to higher prevalence of RA in females (female to male ratio 3:1). Interestingly, several of the SNPs associated with RA can enhance expression of CD2 (Fig.? 7b ), as we determined from the GTEx database 28 . Further analysis of available microarray datasets 29 revealed a mild yet significant correlation between CD2 expression in RA synovia and disease activity (Fig.? 7c ). Moreover, CD2 is strongly upregulated in the synovial tissue from RA patients when compared to osteoarthritis or healthy synovium (Fig.? 7d ). Thus, it is likely that CD2 is involved in joint inflammation, and that CD2 polymorphisms affecting its expression contribute to the development or perpetuation of joint autoimmunity. Fig. 7: E2-mediated regulation of CD2 is conserved in humans. a Genetic association data showing association between CD2 polymorphisms and rheumatoid arthritis (RA) in female (top) and male (bottom) patients (EIRA cohort, n ?=?1341 males and 3361 females). b Effect of indicated SNPs on expression of CD2 in human spleen as determined from GTEx database 65 . Number of samples is given in parenthesis below the graphs. Violin plots show Kernel density estimate KDE in green, interquartile ranges (25th/75th percentiles) in grey (squares), and the median in white (line). P values were obtained from GTEx, calculations are detailed in Oliva et al. 65 . c CD2 expression in synovia from RA patients plotted against disease activity (DAS28-CRP). Data was extracted from GEO Dataset GSE45867. R 2 and P were calculated using simple liner regression. d Expression of CD2 in synovial tissue from RA patients, osteoarthritis (OA) patients or healthy controls (GEO GDS5401-3). Females are shown in red and males in blue. e Expression of CD2 in PBMCs from healthy males and females (GEO GDS5363). f CD2 expression on antigen experienced CD45RO +? or na?ve CD45RA +? CD4 +? T cells from blood of a healthy donor. Data is representative of n ?=?3 independent human biological replicates. The experiment was done twice with similar results. g , h CD2 expression in CD45RO +? T cells after 24?h incubation with 10–100?nM E2. Representative flow cytometry histogram showing CD2 expression?in g and quantification in h . Data shown is pooled from two independent experiments. In all figures, each dot indicates one independent human biological replicate. Data are summarised as mean (SEM). In d , e , and g , statistical significance was evaluated using a two-tailed non-parametric Mann–Whitney U test. Sequential flow cytometry gating strategies for f – h are provided in Supplementary fig.? 16 . Full size image Importantly, women expressed higher levels of CD2 than men, both in RA synovium and healthy PBMCs (Fig.? 7 c, e , respectively), suggesting the E2-mediated regulation of CD2 observed in mice is conserved in humans as well. To corroborate our findings, we stimulated CD4 + ?T cells from healthy human donors with increasing amounts of E2. Firstly, we noticed a strong upregulation of CD2 in antigen experienced CD45RO + ?T cells compared to their na?ve CD45RA + ?counterparts (Fig.? 7f, g ). But more importantly, expression of CD2 could be enhanced in CD45RO + ?T cells by incubation with E2 in a concentration-dependent manner (Fig.? 7h ). Indeed, analysis of available ChIP-seq data 30 revealed that ERα robustly binds the human CD2 gene locus (Supplementary fig.? 8 ). Thus, these data demonstrate the evolutionary conserved nature of E2-mediated regulation of CD2. Anti-Cd2 mAb treatment affects T cells in female mice more than in male mice . We reasoned that hormonal regulation of CD2 expression could have implications for anti-CD2-mediated therapy, as previous research suggests that anti-CD2 (Alefacept) preferentially targets CD2 hi T cells 31 . To test this, we compared the in vivo effects of anti-Cd2 mAb administration on circulating T cells from male and female mice (Fig.? 8 ). Anti-Cd2 mAb treatment partially depleted circulating T cells and resulted in a relative expansion of effector Cd44 + T cells in the remaining T cell pool, skewing the na?ve Cd62L + /effector Cd44 + T cell ratio (Fig.? 8c–h ). This effect was significantly more pronounced in females, which, like in humans, expressed higher levels of Cd2 in circulating T cells. Taken together, these data demonstrate that sex-dependent differences in Cd2 expression determine the response to anti-Cd2 mAb. Fig. 8: Anti-Cd2 is more effective at skewing T cell phenotypes in females. a , b Initial titration experiment in mice showing in vivo depletion of T cells in dependency of administered anti-Cd2 mAb RM2-5. Representative flow cytometry plots of circulating blood T cells in a and total number of cells in b . Data shown is from one mouse. The experiment was performed three independent times with similar results. c – h Na?ve BR male and female mice were injected i.p. with 10??g anti-Cd2 mAb RM2-5. Circulating Cd4 + T cells were analysed before (day 0) and after (day 2) mAb injection. Representative flow cytometry plots are shown in b and d , quantification of the results including ratio of naive (Cd62l + ) to effector (Cd44 + ) Cd4 + T cells are shown in e – h . The experiment was performed two independent times with similar results. Individual dots represent independent mouse biological replicates. Data are summarised as mean (SEM). Statistical significance was evaluated using a two-tailed non-parametric Mann–Whitney U test. Sequential flow cytometry gating strategies for a – h are provided in Supplementary figs.? 17 – 18 . Full size image .