Cross-Phenotype GWAS Supports Shared Genetic Susceptibility to Systemic Sclerosis and Primary Biliary Cholangitis

Objective An increased risk of primary biliary cholangitis (PBC) has been reported in patients with systemic sclerosis (SSc). Our study aims to investigate the shared genetic susceptibility between the two disorders and to define candidate causal genes using cross-phenotype GWAS meta-analysis. Methods We performed cross-phenotype GWAS meta-analysis and colocalization analysis for SSc and PBC. We performed both genome-wide and locus-based analysis, including tissue and pathway enrichment analyses, fine-mapping, colocalization analyses with expression quantitative trait loci (eQTL) and protein quantitative trait loci (pQTL) datasets, and phenome-wide association studies (PheWAS). Finally, we used an integrative approach to prioritize candidate causal genes from the novel loci. Results We detected a strong genetic correlation between SSc and PBC (rg = 0.84, p = 1.7 × 10−6). In the cross-phenotype GWAS meta-analysis, we identified 44 non-HLA loci that reached genome-wide significance (p < 5 × 10−8). Evidence of shared causal variants between SSc and PBC was found for nine loci, five of which were novel. Integrating multiple sources of evidence, we prioritized CD40, ERAP1, PLD4, SPPL3, and CCDC113 as novel candidate causal genes. The CD40 risk locus colocalized with trans-pQTLs of multiple plasma proteins involved in B cell function. Conclusion Our study supports a strong shared genetic susceptibility between SSc and PBC. Through cross-phenotype analyses, we have prioritized several novel candidate causal genes and pathways for these disorders.


Introduction
Systemic sclerosis (SSc) is a multi-system autoimmune disease characterized by a complex interplay of fibrosis, vasculopathy, and inflammation.Unlike in other systemic autoimmune rheumatic diseases, the therapeutic response to immunosuppressive medications in SSc is organ-dependent.Certain organ involvement, including gastrointestinal tract fibrosis, has not been found to be responsive to immunosuppressive therapy. 1 Primary biliary cholangitis (PBC) is an autoimmune liver disease characterized by inflammation of the intrahepatic bile ducts leading to liver fibrosis.Similar to the gastrointestinal involvement of SSc, the efficacy of immunosuppressive therapies in PBC has not yet been established. 2The prevalence of PBC in patients with SSc is 2-2.5%, substantially higher than its prevalence of 0.4% in the general population. 3,4us, there is likely an overlap of etiopathogenesis between SSc and PBC.
Genome-wide association studies (GWAS) have been conducted in both SSc and PBC, identifying numerous genomic loci associated with these two disorders. 5,6Cross-phenotype GWAS analytic approaches that leverage existing GWAS summary statistics have emerged as a powerful new strategy for identifying shared mechanisms and novel risk loci. 7We used this approach to systematically assess overlapping susceptibility and identify novel candidate causal genes that contribute to the common etiopathogenesis of the two disorders.

Study Design
An overview of the study design is shown in Figure 1.

GWAS Summary Statistics
for use under a CC0 license.This article is a US Government work.It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted July 3, 2024.; https://doi.org/10.1101/2024.07.01.24309721 doi: medRxiv preprint We obtained summary statistics of SSc and PBC from their recent GWAS meta-analyses.The GWAS for SSc was comprised of 26,679 individuals (9,095 cases and 17,584 controls), 5 while the GWAS for PBC was comprised of 24,510 individuals (8,021 cases and 16,489 controls). 6For comparison, we also obtained GWAS summary statistics for rheumatoid arthritis (RA) 8 and systemic lupus erythematosus (SLE) 8,9 since both are prevalent in patients with SSc.1][12][13] The GWAS summary statistics were harmonized using reference data from the 1000 Genome Project (phase 3) and underwent quality control with MungeSumstats. 14,15The included GWAS datasets are summarized in Table 1.

Genetic Correlation Analysis
To quantify the degree of shared genetic susceptibility, we used linkage disequilibrium score regression (LDSC) to estimate the global genetic correlation (r g ), excluding the human leukocyte antigens (HLA) region, between each phenotype pair (SSc, PBC, RA, and SLE). 16,17A Bonferroni-corrected p-value of 8.3 x 10 -3 was used as the significance threshold.

Cross-Phenotype GWAS Meta-Analysis
We performed a cross-phenotype GWAS meta-analysis to identify pleiotropic loci shared between SSc and PBC.We combined the summary statistics of SSc and PBC using the fixed-effect model with effect size estimates and standard errors using METAL. 18Genomic control correction was applied to the summary statistics of each phenotype before the meta-analysis. 19After the meta-analysis, we excluded SNPs in the HLA region or with evidence of heterogeneity (heterogeneity p < 0.05).We used wANNOVAR to annotate the lead SNPs of the significant loci. 20We defined novel loci as those that were significant in for use under a CC0 license.This article is a US Government work.It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted July 3, 2024.; https://doi.org/10.1101/2024.07.01.24309721 doi: medRxiv preprint the cross-phenotype meta-analysis but not significant in either SSc or PBC input GWAS.The loci were named based on the annotation of the lead SNPs from wANNOVAR, which relies on the distance to nearby genes. 20e fixed-effect model is limited in examining SNPs with heterogeneity of effects.Therefore, we performed a sensitivity analysis using the PLEIO (Pleiotropic Locus Exploration and Interpretation using Optimal test) method. 21PLEIO is designed for cross-phenotype meta-analysis and can account for heritability, genetic correlation, and sample overlap.There is no established method for extrapolating Zscores from PLEIO statistics.Therefore, we used the fixed-effect model statistics for subsequent analyses.

Fine-Mapping and Credible Set Analyses
We prioritized the most likely causal SNPs from the cross-phenotype meta-analysis statistics by calculating the 99% credible sets using CARMA (CAusal Robust Mapping method with Annotations). 22ese sets represent the smallest sets of SNPs with the probability of including the causal variant exceeding 99%.CARMA is a novel Bayesian model for fine-mapping that can better account for the uneven measurement of SNPs in each GWAS study of a meta-analysis, as well as the discrepancies between summary statistics and LD from external reference panels.We incorporated functional annotation into CARMA using the prior causal probabilities based on the meta-analysis of 15 UK Biobank traits from PolyFun (POLYgenic FUNctionally-informed fine-mapping). 23

Tissue and Pathway Enrichment Analyses
5][26] MAGMA and DEPICT perform enrichment analyses at pathway and tissue levels but use different approaches to associate loci with genes.MAGMA annotates SNPs based on their locations relative to genic regions (transcription start and stop sites +/-10 kb for use under a CC0 license.window).DEPICT prioritizes genes in a locus if genes in different loci have similar predicted functions.We reported the Bonferroni-corrected p values.

Colocalization between SSc and PBC
We performed colocalization analyses between SSc and PBC in loci that are significant in the crossphenotype meta-analysis with the fixed-effect model.Colocalization analyses infer the probability that a single genetic variant is causal to both traits of interest -SSc and PBC.We performed colocalization using the Wakefield's method from the R package "coloc". 27For loci with independent signals in the conditional analysis (p < 5 x 10 -8 ), we performed additional colocalization analyses using statistics from the conditional analyses.Evidence of colocalization was defined as at least one signal with a colocalization probability (PP4) above 70%.

Colocalization between Meta-Analysis Statistics and eQTL/pQTL Datasets
We prioritized loci that met the following criteria for additional colocalization analyses: (1) lead SNPs without evidence of heterogeneity (P het ≥ 0.05); (2) significant in the fixed-effect cross-phenotype metaanalysis; and (3) colocalized between SSc and PBC.We performed colocalization analyses between the SSc-PBC meta-analysis statistics of these loci and cis-eQTL and cis-pQTL datasets to predict transcripts or proteins associated with the genomic signal.For the eQTL colocalization analyses, we included relevant tissues and cells, including blood, skin, lung, liver, and immune cells.We first screened genes using SNPs in that locus's 99% credible set to query the eQTL databases.Next, we selected genes with significant eQTL signals for the colocalization analyses.For loci that colocalized with a cis-pQTL signal, we performed additional colocalization analyses between the meta-analysis statistics and trans-pQTLs measured at that locus to investigate the downstream effects of the candidate causal gene.A colocalization probability (PP4) above 70% was used as the significance threshold, and 50% as the suggestive threshold.
for use under a CC0 license.This article is a US Government work.It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Phenome-wide association studies (PheWAS)
We performed PheWAS for the lead SNP for each novel locus that colocalized between SSc and PBC.The PheWAS was performed within three biobanks: the Electronic Medical Records and Genomics III (eMERGE-III), All of Us, and the UK Biobank.[33] To further explore pleiotropic associations, we performed meta-PheWAS analyses on the polygenic risk scores (PRS) of the cross-phenotype meta-analysis (PRS-meta-PheWAS).We used PRS-CS (continuous shrinkage), a method based on high-dimensional Bayesian regression, to generate the weights for PRS. 34e HLA region was excluded.We set the Bonferroni-corrected statistical significance threshold for phenome-wide significance at 2.75x10 -5 (0.05/1,817 phecodes tested).Lastly, we manually queried the top SNPs using PheWAS results from the Open Targets Genetics webpage 35,36 , which includes data from the GWAS Catalogue, UK Biobank, and FinnGen.We designated the effect allele as the GWAS risk allele in SSc-PBC cross-phenotype analysis.

Integrative prioritization of novel candidate causal genes
For each novel locus, we prioritized candidate causal genes using an integrative approach.Each candidate causal gene was scored with the following criteria and we then calculated the number of the satisfied criteria ("priority score"): (1) genes most proximal to the lead SNP at the locus; (2) genes colocalizing with the locus in the examined cis-eQTL datasets; (3) genes colocalizing with the locus in the examined cis-pQTL datasets; (4) genes with a nonsynonymous coding variant in the credible set; (5)   Genes prioritized by MAGMA (false discovery rate [FDR] q-value < 0.05 ) 24 ; (6) genes prioritized by DEPICT (FDR q-value < 0.05) 26 ; (7) genes receiving the top score from the Variant-to-Gene (V2G) pipeline for use under a CC0 license.
on the Open Targets Genetics webpage 35,36 , using the fine-mapped SNPs (PIP > 10%); (8) genes whose predicted regulatory elements from the ENCODE-rE2G model 37 intersected with the fine-mapped SNPs (PIP > 10%); (9) genes prioritized by the large language model (LLM) GPT-4, which was recently suggested as a systematic way to mine literature for candidate causal gene prioritization (input described in the Supplemental Note). 38We prioritized the genes with the highest priority scores within each locus.
Patients or the public were not involved in the design, or conduct, or reporting, or dissemination plans of our research.

Global Genetic Correlation
There was a strong global genetic correlation between SSc and PBC (rg = 0.84, p = 1.7 x 10 -5 ), in which the effect estimate is comparable to the genetic correlation between SSc and SLE (rg = 0.84, p = 1.6 x 10 - 15 ).The pairwise comparison of global genetic correlation in SSc, PBC, RA and SLE is shown in Figure 2.

Cross-Phenotype GWAS Meta-Analysis
We performed a cross-phenotype meta-analysis for SSc and PBC using the fixed-effect model.The Manhattan plot is shown in Figure 3.There were 44 non-HLA loci that reached genome-wide significance (p < 5 x 10 -8 , Supplemental Table 1).The genomic inflation factor (λ) was 1.065 and the LDSC intercept was 1.010 (standard error 0.012).In 16 out of the 44 significant loci (36%), there was evidence of heterogeneity in the lead SNPs (Phet < 0.05).However, these loci remained significant after removing SNPs with heterogeneity.
Given the high proportion of loci with evidence of heterogeneity, we performed another crossphenotype meta-analysis between SSc and PBC using PELIO as a sensitivity analysis.There were 58 non-HLA loci that reached genome-wide significance (p < 5 x 10 -8 , Supplemental Table 2).The genomic for use under a CC0 license.inflation factor (λ) was 1.091.Forty-one of the 44 loci (93%) that were significant in the fixed-effect model were also significant in PLEIO.Regarding the novel loci, five out of the seven (71%) identified in the fixed-effect model were also significant in PLEIO.The two novel loci that were only significant in the fixed-effect model, CD40 and AHNAK2, had p-values of 7.37 x 10 -8 and 3.37 x 10 -7 in PLEIO.Two novel loci, NDFIP1 and PPHLN1, were significant only in PLEIO but not in the fixed-effect model.However, the evidence for the association of these loci with SSc was insufficient (lead SNP p values 3.93 x 10 -2 and 1.33 x 10 -2 , respectively).Thus, no further analyses were performed for these loci.

Fine-Mapping and Credible Set Analyses
for use under a CC0 license.This article is a US Government work.It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.To infer the causal SNPs underlying the association signals, we performed fine-mapping and generated 99% credible sets.There were 53 predicted causal signals across the 44 significant loci.Eleven (21%) of the credible sets contained only one SNP.Twenty-six (49%) credible sets contained fewer than five SNPs.
Thirty-nine (74%) credible sets contained fewer than ten SNPs.In 30 (68%) loci, the maximum posterior inclusion probability (PIP) in their credible sets contained the lead SNPs of that loci.There were no nonsynonymous coding variants in any of the credible sets.The credible SNP sets in each locus are summarized in Supplemental Table 5.

Colocalization between SSc and PBC
Due to LD, significant SNPs within genomic loci may not necessarily be causal for the associated trait.
Therefore, we conducted colocalization analyses to determine whether there was at least one shared causal variant between SSc and PBC in loci significant in the cross-phenotype meta-analysis.We identified 9 loci that colocalized between SSc and PBC (PP4 > 70%) and did not have evidence of heterogeneity (P het ≥ 0.05), as detailed in the Table 2 and Supplemental Figure 1.Notably, five were novel: CSNK2A2/CCDC113, SPPL3, CAST/ERAP1, AHNAK2, and CD40.

Colocalization with Tissue-and Immune Cell-Specific cis-eQTL
In the nine loci that were significant in cross-phenotype meta-analysis without evidence of heterogeneity and colocalized with SSc and PBC, we further performed colocalization between SSc-PBC meta-analysis statistics and eQTL statistics in blood, skin, lung, liver and immune cells.In seven of the nine loci, the SSc-PBC meta-analysis statistics colocalized with at least one transcript in the examined eQTL statistics (PP4 > 50%), prioritizing candidate causal genes at these loci (Figure 5).The eQTL colocalized transcripts were IRF5, TNPO3, ANP32B, IL12RB1, ERAP1, ERAP2, SPPL3, AKT1, PLD4, LINC00638 and CD40.

Colocalization with plasma cis-and trans-pQTL
for use under a CC0 license.

PheWAS
We performed a meta-PheWAS for the five novel loci that colocalized between SSc and PBC.Three novel loci had significant associations with at least one phecode (Supplemental Figure 3).Rs10083496-G (ANNAK2 locus) was associated with "Systemic lupus erythematosus" (OR = 1.12, p = 2.05 x 10 -7 ).

Integrative prioritization of novel candidate causal genes
We prioritized five candidate causal genes from the novel loci using a scoring approach that integrates nine in silico annotation methods (Figure 6. MAGMA, DEPICT and ENCODE-rE2G results in the Supplemental Table 9-11.GPT-4 output described in the Supplemental Note).CD40 received the highest priority score of 8, meeting all criteria except for credible sets containing a non-synonymous coding variant.ERAP1 received a priority score of 6, followed by PLD4 with a score of 5. SPPL3 and CCDC113 received priority scores of 4 and 3, respectively.

Discussion
Our study demonstrates a strong genetic correlation between SSc and PBC, with the correlation effect estimate comparable to that between SSc and SLE.The prevalence of PBC in SSc was 2-2.5%, lower than the 8.4-14.7%3][44][45][46] Consequently, compared to the general population, the for use under a CC0 license.
relative risk (RR) of PBC and SLE in SSc likely mirrors their genetic correlation.Hence, our genetic correlation results, corroborated by the extent of phenotypic overlaps, support the existence of shared genetic susceptibilities and biological mechanisms between SSc and PBC.
We identified 44 significant non-HLA genomic loci in the fixed-model cross-phenotype GWAS metaanalysis.The robustness of our meta-analysis was supported by the PRS-meta-PheWAS analysis.In this analysis, using independent external datasets, SSc and PBC ranked among the top hits associated with the PRS, derived from our SSc-PBC meta-analysis statistics.Moreover, the SSc-PBC PRS demonstrated associations with a broad spectrum of autoimmune disorders, highlighting that the shared genetic susceptibilities between SSc and PBC captured by our cross-phenotype meta-analysis represent pleiotropic genomic regions.
We identified five novel loci that were significant in the cross-phenotype GWAS meta-analysis and colocalized between SSc and PBC: CSNK2A2/CCDC113, SPPL3, CAST/ERAP1, AHNAK2, and CD40.Three of the five loci were independently confirmed in a GWAS in another population: CD40 and AHANK2 loci were significant in a recent SSc GWAS meta-analysis that included the Japanese population, 47 and the CSNK2A2/CCDC113 locus was significant in a PBC GWAS from the Chinese population. 48Moreover, we found that the lead SNP in the CD40, AHNAK2, and CAST/ERAP1 loci was associated with other autoimmune disorders in PheWAS.The pleiotropic effects observed at these loci underscore their potential role in promoting autoimmunity.We subsequently prioritized five novel candidate causal genes for SSc and PBC based on integrating nine analytic approaches: CD40, ERAP1, PLD4, SPPL3, and CCDC113.
At the CD40 locus, the SSc-PBC meta-analysis statistics colocalized with not only the reduced transcript but also reduced plasma protein levels of CD40.Such associations have also been observed in other autoimmune diseases, including inflammatory bowel disease and multiple sclerosis, as well as malignancy in our PheWAS analysis. 49,50This seems paradoxical given CD40's established role in for use under a CC0 license.
promoting autoimmunity. 51However, CD40 deficiency, a rare monogenic disorder caused by bi-allelic loss-of-function variants in CD40, is characterized not only by humoral immunodeficiency but also by malignancy and autoimmunity, including sclerosing cholangitis and colitis. 52Thus, a causal relationship between reduced CD40 expression due to polymorphisms and an increased risk of autoimmunity and malignancy is biologically plausible, mimicking the phenotypic manifestations of its monogenic disease counterpart.Interestingly, at the CD40 locus, the SSc-PBC meta-analysis statistics also colocalized with increased levels of multiple plasma proteins involved in B cell functions, including CD40L, FCER2, CD22, TRAF2, and TCL1A.This suggests a potential compensatory response.Soluble CD40L (sCD40L), the circulating form measured in the proteomics assay, is the ligand of CD40, which also binds to other receptors on endothelial cells and promotes vascular pathology. 53Elevated soluble CD40L levels have been found in patients with SSc and are associated with its vascular manifestations. 54Overall, the complex B cell dysregulation mediating genetically determined reduced CD40 expression in SSc warrants further investigation.
Our study also suggests that SSc and PBC may be associated with reduced major histocompatibility complex (MHC)-I-mediated immune response, potentially affecting cancer immunosurveillance.The SSc-PBC meta-analysis statistics in CAST/ERAP1 and SPPL3 loci colocalized with reduced transcripts of endoplasmic reticulum aminopeptidase 1 (ERAP1), increased transcripts of endoplasmic reticulum aminopeptidase 2 (ERAP2) and reduced transcripts of signal peptide peptidase-like 3 (SPPL3).These genes are implicated in the MHC-I-mediated immune response.ERAP1 and ERAP2 process and trim antigen peptides prior to their binding to MHC-I.6][57] In contrast, both SSc and PBC are associated with MHC-II and the risk alleles in CAST/ERAP1 identified in our study are in the opposite direction of ERAP1 and ERAP2 expression compared to those found in the MHC-Iopathy entities. 58,59Recent research has shown that reduced SPPL3 activity can dampen the MHC-I-for use under a CC0 license.This article is a US Government work.It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
Our study has several limitations.First, we used summary statistics from published GWAS meta-analyses and were unable to perform standardized quality control with individual-level genotype data.
Nevertheless, the GWAS studies included in our study were recent, comprised of large sample sizes from multiple cohorts, and led by international experts.Second, our study was performed in European ancestries; thus, our results may not be generalizable to other populations.Third, the sample size of SSc and PBC in our external datasets was limited, which precluded us from performing a replication GWAS to validate the newly discovered loci.Instead, we conducted a PRS-meta-PheWAS analysis to confirm that the PRS derived from our SSc-PBC meta-analysis captures the overall genetic risk of SSc and PBC in independent EHR-based datasets.Fourth, the sample size of GTEx and CEDAR was smaller than that of eQTLGen, which could limit statistical power in our eQTL colocalization analyses for the relevant tissues and cells.Finally, for our meta-PheWAS analyses, the diagnoses in the EHR-based datasets rely on administrative codes, which may have non-random missingness and low sensitivity for phenotype detection. 65 conclusion, our study revealed a strong genetic correlation between SSc and PBC and provided insights into their shared genetic susceptibility.We prioritized seven novel genes that were potentially involved in the common causal mechanisms between SSc and PBC.These discoveries prioritize therapeutic targets for both SSc and PBC.Moreover, our study advocates for heightened awareness among rheumatologists about the possibility of concurrent PBC in patients with SSc.
for use under a CC0 license.This article is a US Government work.It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.We performed multiple cross-phenotype GWAS analyses to identify the shared genetic susceptibility between SSc and PBC.Additionally, we also performed a single-center retrospective chart review to evaluate the prevalence of PBC, including potentially undiagnosed cases, in patients with SSc.SSc: systemic sclerosis.PBC: primary biliary cholangitis.GWAS: genome-wide association studies.PLEIO: Pleiotropic Locus Exploration and Interpretation using Optimal test.eQTL: expression quantitative trait loci.PheWAS: phenome-wide association studies.We performed genetic correlation analyses using linkage disequilibrium score regression pairwise among SSc, PBC, RA and SLE.The effect estimates of genetic correlation between SSc and PBC, and between SSc and SLE, were both 0.84, higher than other pairwise estimates.
Figure 3 Manhattan plot of the cross-phenotype GWAS meta-analysis in SSc and PBC using the fixedeffect model.We performed a cross-phenotype GWAS meta-analysis in SSc and PBC using the fixed-effect model.We identified 44 significant genomic loci (p < 5 x 10 -8 ).We found 9 loci that colocalized between SSc and PBC and did not show evidence of heterogeneity (indicated in red), among which five were novel (indicated with an asterisk).SNPs with evidence of heterogeneity (P het < 0.05) were excluded.a Tissue enrichment analysis using MAGMA prioritized tissues related to the immune system (spleen, whole blood and EBV-transformed lymphocytes), respiratory system (lung) and digestive system for use under a CC0 license.This article is a US Government work.It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted July 3, 2024.; https://doi.org/10.1101/2024.07.01.24309721 doi: medRxiv preprint (terminal ileum).b Tissue enrichment analysis using DEPICT prioritized multiple tissues and cells related to the immune system, respiratory system and musculoskeletal system.c Significant enrichment of multiple immune-related pathways associated with SSc and PBC using MAGMA.
Figure 5 Colocalization analysis with expression quantitative trait loci (eQTL) in blood, skin, lung, liver and immune cells.
Out of the nine loci that were significant in the fixed-model cross-phenotype meta-analysis, showed no evidence of heterogeneity, and colocalized between SSc and PBC, seven colocalized with expressed genes in at least one of the examined tissues.

Figure 1
Figure 1 Flowchart for the overview of the study.

Figure 2
Figure 2 Pairwise genetic correlation in SSc, PBC, RA and SLE.

Figure 4
Figure 4 Tissue and pathway enrichment analyses.

Figure 6
Figure 6 Integrative prioritization of novel candidate causal genes

Table 1
GWAS summary statistics included in this study.: systemic sclerosis.PBC: primary biliary cholangitis.RA: Rheumatoid arthritis.SLE: systemic lupus erythematosus.eQTL: expression quantitative trait loci.pQTL: protein quantitative trait loci.GTEx: Genotype-Tissue Expression.CEDAR: Correlated Expression and Disease This article is a US Government work.It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. SSc

Table 2
Genomic loci significant in the cross-phenotype GWAS, colocalized between SSc and PBC and do not have evidence of heterogeneityThis article is a US Government work.It is not subject to copyright under 17 USC 105 and is also made available (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.