Identification of new risk loci shared across systemic vasculitides points towards potential target genes for drug repurposing

Abstract

Objectives

The number of susceptibility loci currently associated with vasculitis is lower than in other immune-mediated diseases due in part to small cohort sizes, a consequence of the low prevalence of vasculitides. This study aimed to identify new genetic risk loci for the main systemic vasculitides through a comprehensive analysis of their genetic overlap.

Methods

Genome-wide data from 8467 patients with any of the main forms of vasculitis and 29 795 healthy controls were meta-analysed using ASSET. Pleiotropic variants were functionally annotated and linked to their target genes. Prioritised genes were queried in DrugBank to identify potentially repositionable drugs for the treatment of vasculitis.

Results

Sixteen variants were independently associated with two or more vasculitides, 15 of them representing new shared risk loci. Two of these pleiotropic signals, located close to CTLA4 and CPLX1, emerged as novel genetic risk loci in vasculitis. Most of these polymorphisms appeared to affect vasculitis by regulating gene expression. In this regard, for some of these common signals, potential causal genes were prioritised based on functional annotation, including CTLA4, RNF145, IL12B, IL5, IRF1, IFNGR1, PTK2B, TRIM35, EGR2 and ETS2, each of which has key roles in inflammation. In addition, drug repositioning analysis showed that several drugs, including abatacept and ustekinumab, could be potentially repurposed in the management of the analysed vasculitides.

Conclusions

We identified new shared risk loci with functional impact in vasculitis and pinpointed potential causal genes, some of which could represent promising targets for the treatment of vasculitis.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• The genetic component of systemic vasculitides is still largely unknown, mainly due to the low prevalence of these disorders.

• Combination of genome-wide data from different traits, through cross-phenotype association studies, is a powerful tool to identify new genetic risk loci shared across related diseases.

WHAT THIS STUDY ADDS

• Meta-analysis of large-scale summary statistics of the main forms of vasculitis allowed identification of 16 risk loci with a pleiotropic role in these disorders.

• Functional annotation of the identified cross-phenotype associations led to the prioritisation of causal genes with a potential effect in vasculitis.

• Many of the pleiotropic variants affect genes encoding proteins that are therapeutic targets of drugs indicated for other pathologies; these could potentially be repurposed for the treatment of the analysed vasculitides.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• A better understanding of the key molecular pathways shared across different forms of systemic vasculitis may help to identify common therapeutic mechanisms that lead to improved clinical management of these disorders.

Introduction

Systemic vasculitides comprise a heterogeneous group of immune-mediated disorders characterised by blood vessel inflammation. According to the Chapel Hill Consensus Conference, the most common vasculitides can be classified into three main categories based on the distribution of the predominant type of vessel involvement.¹ Large-vessel vasculitis, including giant cell arteritis (GCA) and Takayasu’s arteritis (TAK), is characterised by the inflammation of the aorta and its main branches; medium vessel vasculitis, which comprises Kawasaki disease (KD) and polyarteritis nodosa, mainly affects the major visceral arteries and their branches; and, finally, small vessel vasculitis, which affects arterioles, small arteries, capillaries and venules, can be subdivided into ANCA-associated vasculitis (AAV) that includes eosinophilic granulomatosis with polyangiitis (EGPA), granulomatosis with polyangiitis (GPA), and microscopic polyangiitis (MPA), and immune complex vasculitis, including IgA vasculitis (IgAV). In addition, other vasculitides such as Behçet’s disease (BD) involve vessels of any size. Since inflammation can affect vessels of various sizes and locations, these diseases show a wide spectrum of clinical manifestations, many of which are shared among them.^2–4

The aetiology of the vasculitides is still unclear; however, it is well known that they are complex disorders caused by the interaction between multiple genetic and environmental factors. Although several large-scale genetic scans have been performed in vasculitis,⁵ only a few loci have been consistently associated with these disorders, including the human leucocyte antigen (HLA) region, which represents the main genetic risk locus for all of them.⁵ This limited knowledge of their genetic component, as compared with other immune-mediated disorders, is likely due to the low prevalence of vasculitides, which greatly limits the statistical power of genetic studies.

In the last years, combined analysis of the results of large-scale genetic studies of different diseases, through cross-phenotype association studies, has made it possible to overcome the obstacle of sample size, leading to the identification of genetic risk loci shared among related disorders. This approach has already been used to explore the genetic overlap across vasculitides, specifically between KD and IgAV, using genome-wide association study (GWAS) data,⁶ and among GCA, TAK, AAV and IgAV, using Immunochip data.^{7 8}

Nevertheless, these studies did not include some of the main forms of vasculitis and/or lacked the greater genomic coverage offered by the GWAS platform compared with the Immunochip. The purpose of the current study was, therefore, to comprehensively assess the genetic overlap among the major phenotypes of systemic vasculitis by combining GWAS data of GCA, TAK, KD, IgAV, BD and AAV as a single phenotype.

Methods

Study cohort

A total of 8467 patients with vasculitis, including 2134 GCA, 1091 TAK, 405 KD, 215 IgAV, 3197 BD, 914 AAV (MPA and GPA; 268 MPO-positive and 478 PR3-positive), 159 ANCA-positive EGPA and 352 ANCA-negative EGPA, and 29 795 healthy individuals were included in the study. A mixed population was analysed, including mainly Europeans (77.8%), but also Turkish (15.8%), Asians (5.8%) and Middle Easterners (0.64%). All the analysed datasets came from previously published GWAS,^{6 9–12} except for BD for which we included two case/control sets, one from a previous GWAS (BD_GWAS) (obtained from dbGAP, accession number: phs000272.v1.p1) and one from a recent Immunochip study (BD_Ichip).^{13 14} To avoid overlap between cases of BD, we selected only those SNPs from the GWAS that were not included in the Immunochip study. In addition, several control sets, specifically those from Italy, Spain and UK, overlapped among diseases, which was taken into account for the later meta-analysis, as described below. A summary of the cohorts included in the study is provided in online supplemental table 1.

Patient and public involvement

Patients or the public were not involved in any of the stages of our research.

Quality control and imputation of the GCA, BD_GWAS and AAV datasets

For TAK, KD, IgAV, BD_Ichip and EGPA, summary statistics from the previous studies6 10 12 13 were used for the cross-disease meta-analysis. However, in the case of GCA, AAV and BD_GWAS, data from previous GWAS^{9 11 14} were not imputed or imputed with small reference panels and, therefore, genotypic data were newly imputed using a larger reference panel to increase the number of single-nucleotide polymorphisms (SNPs) to analyse.

First, stringent quality controls were applied to each GCA cohort and the BD_GWAS. Specifically, we individually analysed ten independent case/control sets for GCA and one for BD. GWAS data were filtered prior to imputation using PLINK V.1.9 software (www.cog-genomics.org/plink/1.9/).¹⁵ SNPs with low allele frequency (minor allele frequency <1%), low genotyping rate (<98%) and deviating from Hardy-Weinberg equilibrium (p<0.001) were removed. In addition, duplicates and first-degree relatives as well as individuals with successful call rates <95% were discarded. Finally, sex chromosomes were removed from further analyses.

SNP genotype imputation of GCA and BD_GWAS datasets was performed using the Michigan Imputation Server V.1.0.3¹⁶ and the TOPMed Imputation Server (https://imputation.biodatacatalyst.nhlbi.nih.gov/), with the HRC reference panel (HRC V.r1.1 2016) and the TOPMed Imputation Reference panel, respectively. A probability threshold for merging genotypes of r²=0.9 was applied. After imputation, principal component (PC) analysis was performed with PLINK 1.9 and the GCTA64 (Genome-wide Complex Trait Analysis) software¹⁷ and R-base under GNU Public licence V.2. Around 100 000 independent SNPs were selected and used to calculate the first ten PCs per individual. Individuals showing >4 SD from the cluster centroid were excluded by considering them as outliers.

In the case of the AAV cohort, prephasing and genome-wide SNP imputation were performed using Eagle2 and Minimac3, respectively, on the Michigan Imputation Server V.1.0.3 (HRC V.r1.1 2016). Postimputation, SNPs with MAF <0.01 or r² <0.3 were removed from the dataset using BCFtools V.1.2.

The total number of SNPs and individuals that remained in the final filtered GCA, BD_GWAS and AAV datasets is shown in online supplemental table 2.

Statistical analysis of the GCA, BD_GWAS and AAV datasets

First, we performed independent association analyses for GCA, BD_GWAS and AAV. In the case of GCA and BD_GWAS, each case/control set was analysed by logistic regression including the first 10 PCs as covariates in PLINK 1.9, whereas case–control association testing of the AAV dataset was performed using a linear mixed model with BOLT-LMM software V.2.3.2.¹⁸ Then, for GCA, for which 10 independent case/control sets were available, we combined the different cohorts using an inverse variance weighted meta-analysis in METASOFT.¹⁹.

Cross-disease meta-analysis

Summary statistics of the eight phenotypes of vasculitis were used to identify those SNPs showing shared associations with at least two of the analysed disorders. We used ASSET²⁰ to perform a subset-based meta-analysis, which allows identifying the best subset containing those diseases contributing to the overall association signal. Specifically, ASSET explores all possible subsets of phenotypes to identify the strongest association signal and then evaluates the significance of the signal while accounting for multiple tests required by the subset search. Since part of the control individuals included in the analysis were shared among different datasets (online supplemental table 1), correlation matrices were used to adjust for this overlapping. We tested all the SNPs shared by at least two summary statistics. Considering the complexity of the HLA region and that the associations within this locus with the different forms of vasculitis are well established, this region (Chr6: 20–40 MB) was excluded from further analysis. After cross-disease meta-analysis, those SNPs showing p values lower than 5E-08 were considered as statistically significant and, of these, we only focused on those signals for which more than one disease contributed to the association. In the case of SNPs for which opposite effects across diseases were found, they were considered as significant when both positively and negatively associated subsets showed p values lower than 0.05. When several SNPs were statistically significant within the same loci, those including the higher number of phenotypes in the best subset were considered as lead SNPs. In addition, for each pleiotropic locus, signals reaching genome-wide significance and showing r² <0.1 were considered as independent.

Moreover, taking into account that MPO-positive and PR3-positive AAV patients can be considered as two genetically different entities, ASSET meta-analysis was also performed including both subgroups of AAV as two independent phenotypes.

Functional annotation of genetic variants and gene prioritisation

Next, to identify potential target genes of the identified pleiotropic variants, we used the SNP2GENE function of FUMA GWAS (Functional Mapping and Annotation of GWAS).²¹ Functional annotation of the predefined independent lead SNPs, and their proxies (r² >0.6), were obtained to prioritise potential causal genes. Specifically, FUMA maps SNPs to genes based on three strategies, positional mapping, expression quantitative trait loci (eQTL) mapping and chromatin interaction mapping. In addition, using epigenomic data from the Roadmap Epigenomic Project, enhancers and promoters were also annotated to significantly interaction regions. Additionally, we also assessed the overlap of the cross-phenotype signals with splicing QTL obtained from the GTEx project (V.8). For the functional annotation, we only focused on relevant tissues in vasculitis (whole blood, immune cells and arterial tissue).

Regulatory element enrichment analysis

We used GARFIELD (GWAS analysis of regulatory or functional information enrichment with linkage disequilibrium (LD) correction)²² to explore whether the set of genetic variants shared across vasculitides colocalised with tissue-specific regulatory features extracted from the Roadmap Epigenomics project. Specifically, we selected nine histone marks related to active promoters, enhancers and active (or at least accessible) genes (H2BK20ac, H3K27ac, H3K4me1, H3K4me2, H3K4me3, H3K9ac, H3K4ac, H3K79me1, H2BK15ac). Briefly, GARFIELD quantifies functional annotation enrichment for associated SNPs at various p value thresholds (p≤1E-08, p≤1E-07, p≤1E-06, p≤1E-05) while taking into account LD, minor allele frequency and distance to the nearest transcription start site. In addition, from the wide catalogue of tissues and cell types included in the Roadmap Epigenomics project, we focused on immune cells. To adjust for multiple testing correction, we used the R code Garfield-Meff-Padj.R provided by GARFIELD to calculate an enrichment p value threshold (p≤6.12E-04) adjusted on the effective number of annotations (Meff=81.66).

Drug repurposing analysis

Finally, we used the database DrugBank (V.5.0)²³ to explore whether proteins encoded by the prioritised genes are targets for approved, clinical trial or experimental drugs. In order to select the most potentially promising drugs for vasculitis, we focused on the mechanism of action of the different drugs as reported in DrugBank and we also performed a manual literature search using Pubmed and ClinicalTrials.gov.

Next, to calculate if the prioritised genes obtained through functional annotation were significantly enriched in targets of drugs indicated to autoimmune diseases, we applied Fisher’s exact test. For this analysis, we used gene-products of the prioritised genes in the previous analyses and unrelated to any of the analysed diseases, drug targets indicated for autoimmune diseases and coding genes of the genome that are potentially druggable (online supplemental table 3).

Results

Cross-phenotype meta-analysis

Summary statistics data of a total of 8467 patients diagnosed with any of the 8 types of vasculitis comprised in the current study and 29 795 healthy individuals were included in the analysis.

After cross-disease meta-analysis and excluding the HLA region, 85 genetic variants at 12 genomic regions were associated with two or more vasculitides at the genome-wide level of significance (figure 1). Based on LD, 14 genetic variants within these loci were independently associated (online supplemental table 4). Interestingly, 12 of these SNPs represented new associations for some of the phenotypes included in the best subset (table 1 and online supplemental table 4), 9 of them showing similar effects in all the diseases contributing to the association signal and three showing opposite effects across diseases (figure 2).

Figure 1.

Manhattan plot showing the results of the cross-phenotype meta-analysis. The −log10 of the p values are plotted against their physical chromosomal position. The red line represents the genome-wide level of significance (p<5×10⁻⁸). Loci reaching the significant threshold are annotated in the plot. Loci representing new shared risk loci in vasculitis are highlighted in bold.

Table 1.

Independent genetic variants reaching genome-wide significant level in the subset-based meta-analysis and representing novel risk loci shared across vasculitides

Region	Base pair	SNP	Gene	A1	P value	Contributing disease	OR (95% CI)
1p31.3	67 751 193	rs11209039	IL23R/IL12RB2	G	6.45E-10	BD EGPA+KD	0.81 (0.76 to 0.87)
2q33.2	204 689 376	rs62184865	CTLA4	T	2.00E-08	AAV EGPA+	1.72 (1.43 to 2.08)
3p21.31	46 208 310	rs2087726	CCR3	G	1.04E-08	BD IgAV	0.81 (0.75 to 0.87)
4p16.2	824 988	rs4690319	CPLX1	A	4.72E-09	BD TAK	0.82 (0.77 to 0.88)
5q31.1	131 540 875	rs128738	P4HA2	T	2.78E-09	EGPA- GCA	1.32 (1.20 to 1.44)
5q31.1	131 797 547	rs6894249	IRF1	G	7.81E-09	EGPA+ EGPA- | BD KD TAK	1.40 (1.21 to 1.61) | 0.85 (0.83 to 0.94)
5q33.3	158 777 001	rs7725339	IL12B	T	1.84E-10	BD EGPA+ TAK	1.24 (1.15 to 1.32)
5q33.3	158 834 367	rs60689680	IL12B*	T	4.62E-08	IgAV TAK | EGPA-	1.32 (1.19 to 1.45) | 0.80 (0.69 to 0.95)
6q26	161 143 608	rs4252120	PLG	C	2.65E-08	EGPA- GCA	1.28 (1.17 to 1.39)
8p21.2	27 219 987	rs73223431	PTK2B	T	1.66E-08	EGPA+TAK	0.73 (0.66 to 0.82)
10q21.2	64 396 042	rs10995248	ZNF365/ADO	T	2.10E-08	AAV BD | EGPA+GCA IgAV	1.16 (1.08 to 1.24) | 0.84 (0.77 to 0.91)
21q22.2	40 465 178	rs2242944	chr21q22	A	1.30E-10	EGPA- TAK	0.75 (0.68 to 0.82)

*Novel signals within previously known loci. Diseases included in the best subset and for which identified associations have not been previously reported are shown in bold.

A1, alternative allele used in the logistic regression; AAV, ANCA-associated vasculitis; BD, Behçet’s disease; EGPA+, ANCA-positive eosinophilic granulomatosis with polyangiitis; EGPA-, ANCA-negative eosinophilic granulomatosis with polyangiitis; GCA, giant cell arteritis; IgAV, IgA vasculitis; KD, Kawasaki’s disease; TAK, Takayasu arteritis.

Figure 2.

Novel risk loci shared across vasculitides. Effect of the independent genetic variants reaching genome-wide significant level in the subset-based meta-analysis is shown. Circles represent the analysed phenotypes of vasculitis. AAV, ANCA-associated vasculitis; BD, Behçet’s disease; EGPA, eosinophilic granulomatosis with polyangiitis; GCA, giant cell arteritis; KD, Kawasaki disease; IgAV, IgA vasculitis; TAK, Takayasu’s arteritis.

Notably, two of these SNPs, rs4690319 and rs62184865, annotated by proximity to CPLX1 and CTLA4, respectively, have not been previously associated at the genome-wide level of significance with any vasculitis, thus representing novel susceptibility loci. In addition, the remaining 10 signals have been previously associated with only 1 of the diseases included in the best subset and, therefore, they can be considered as new shared risk loci.

Moreover, our analysis also identified two independent signals within the IL12B region, a known susceptibility locus for TAK. Three diseases, BD, ANCA-positive EGPA and TAK, contributed to the association observed for the rs7725339 genetic variant, which is in complete LD with the SNP previously associated with TAK (r²=1).¹² In addition, rs60689680, which is not linked to the TAK-associated variant (r²=0.007), also emerged as a common susceptibility locus. Notably, this last SNP showed opposite effects across the diseases contributing to its association, with a risk effect in TAK and IgAV and a protective effect in ANCA-negative EGPA. In addition, two independent signals (r²=0.05) were identified within the 5q31.1 region. The first one (rs128738), previously associated with GCA and annotated to the P4HA2 gene, was also associated with ANCA-negative EGPA. The second one (rs6894249), previously associated with ANCA-negative EGPA and annotated to IRF1, appeared to be a common risk locus for a high number of vasculitides but with opposite effects across them (risk to ANCA-positive and ANCA-negative EGPA and protection to BD, KD and TAK). Finally, the signal located at the 10q21.2 region also revealed opposite effects. This locus, previously associated with BD, conferred risk to BD and AAV and protection to ANCA-positive EGPA, GCA and IgAV.

In addition, while AAV (considering MPA and GPA patients as a whole) was found to contribute to only two of the newly identified shared associations (table 1), meta-analysis considering MPO-positive and PR3-positive AAV as independent phenotypes allowed identifying a specific contribution of these subgroups to three of the identified common signals (table 2 and online supplemental table 5). Specifically, MPO-positive AAV was involved in the associations observed within the 1p31.1 and 8p21.2 loci whereas PR3-positive AAV contributed to the shared association identified at 3p21.31. Moreover, this analysis also yielded three new pleiotropic loci in vasculitis, 1q32.1 and 6q23.3, previously associated with BD and annotated to IL10 and IFNGR1, for which PR3-positive AAV and MPO-positive AAV, respectively, were contributing to the association, and 2q13, previously associated with EGPA and annotated to BCL2L11, for which PR3-positive AAV was also included in the best subset (table 2, online supplemental table 5 and figure 2).

Table 2.

Shared associations across vasculitides considering MPO and PR3-autoantibody specificity in patients with ANCA-associated vasculitis

Region	Base pair	SNP	Gene	A1	P value	Contributing disease	OR (95% CI)
1p31.3	67 744 601	rs6660226	IL23R/IL12RB2	A	6.78E-11	MPO+_AAV | BD EGPA+KD	1.39 (1.10 to 1.75) | 0.79 (0.74 to 0.85)
1q32.1	206 945 311	rs3024490	IL10*	A	1.49E-08	PR3+_AAV BD	1.24 (1.15 to 1.34)
2q13	111 905 867	rs144569746	BCL2L11*	T	1.11E-08	PR3+_AAV EGPA+EGPA-	1.68 (1.40 to 2.00)
3p21.31	46 208 310	rs2087726	CCR3	G	4.94E-09	PR3+_AAV BD IgAV	0.82 (0.77 to 0.88)
6q23.3	137 514 790	rs4896243	IFNGR1*	C	1.59E-08	MPO+_AAV BD	1.23 (1.15 to 1.32)
8p21.2	27 219 987	rs73223431	PTK2B	T	3.23E-08	MPO+_AAV EGPA+TAK	0.76 (0.69 to 0.84)

Diseases included in the best subset and for which identified associations have not been previously reported are shown in bold.

*New shared associations identified in the meta-analysis considering MPO-positive and PR3-positive AAV as independent phenotypes.

A1, alternative allele used in the logistic regression; AAV, ANCA-associated vasculitis; BD, Behçet’s disease; EGPA+, ANCA-positive eosinophilic granulomatosis with polyangiitis; EGPA-, ANCA-negative eosinophilic granulomatosis with polyangiitis; IgAV, IgA vasculitis; KD, Kawasaki’s disease; TAK, Takayasu arteritis.

Functional annotation of pleiotropic variants

FUMA annotation of the shared genetic variants showed that none of the lead SNPs were coding, whereas five of their proxies (r² ≥0.6) were exonic, including three non-synonymous variants, mapping at BCL2L11, IL3 and PLG, and two synonymous variants, mapping at PLG. Of these missense variants, only the SNP at BCL2L11 was annotated as possibly damaging according to SIFT and Polyphen scores (online supplemental table 6).

Since all the lead SNPs, and most of their proxies, were located in non-coding regions, we annotated them using FUMA to determine their overlap with functional regulatory elements as well as to prioritise their most probable target genes. Notably, all but one of the shared genetic variants overlapped with predicted regulatory regions in immune cells, whole blood and/or arterial tissue, whereas 69% overlap with at least three functional annotations (figure 3). Remarkably, 14 out of the 16 shared SNPs, or their proxies, appeared to act as eQTLs affecting gene expression levels (figure 3 and online supplemental table 7). Based on this annotation, FUMA prioritised 182 genes, of which 159 were protein-coding, as potential causal genes in vasculitis (table 3).

Figure 3.

Functional annotation of the identified pleiotropic variants. Colours indicate both lead and proxy polymorphisms overlapping with the different regulatory elements analysed. CI (yellow), chromatin interactions; E (orange), enhancers; P (blue), promoters; eQTL (green), expression quantitative trait loci.

Table 3.

Prioritised causal genes for the pleiotropic loci based on functional annotation by FUMA

Region	Gene	Lead SNP	Nominated genes by proximity	Nominated genes by eQTL	Nominated genes by sQTL	Nominated genes by CI	Nominated genes by eQTL+CI	Nominated genes by eQTL+promoter
1p31.3	IL23R/IL12RB2	rs11209039	IL12RB2	IL23R, MIER1, SLC35D1		C1orf141
1q32.1	IL10	rs3024490	IL10	IL10, IL19, IL24, FAIM3		IL19, FCMR, PIGR, FCAMR, C1orf116, YOD1, PFKB2, CR1, CR2, MAPKAPK2	IL19	YOD1, PFKB2, CR1, CR2, MAPKAPK2
2q13	BCL2L11	rs72836352	ACOXL, BCL2L11	BCL11A, BCL2L11, C3AR1, CPVL, CSFR1, FAM46A, LARGE, MS4A14, MS4A7, MTSS1, PEA15, TNFRSF8		ACOXL		BCL2L11
2q33.2	CTLA4	rs62184865	CTLA4	CTLA4
3p21.31	CCR3	rs2087726	CCR1, CCR3	ABCB1, AGAP1, AUTS2, CAMTA1, CCR1, CCR2, CCR3, CCR5, CCR9, CCRL2, CFH, COLQ, CXCR6, CYP2E1, CYTH3, DPP4, ELOVL4, FYCO1, GZMK, IL7R, KDSR, KLRB1, LTK, LZTFL1, NCALD, RORA, RORC, SACM1L, SLAMF1, SMAD3, SYTL2, TDGF1, XCR1	LZTFL1, FYCO1, CCR3	CCR1, CCR2, CCR5, CCR9, CCRL2, CXCR6, FYCO1, LZTFL1, SACM1L, XCR1	CCR1, CCR2, CCR5, CCR9, CCRL2, CXCR6, FYCO1, LZTFL1, SACM1L, XCR1
4p16.2	CPLX1	rs4690319	CPLX1	CPLX1, CRIPAK, DGKQ, GAK, IDUA, MFSD7, MYL5, PCGF3, TMEM175, UVSSA
5q22.1	TSLP	rs1837253	TSLP
5q31.1	P4HA2	rs128738	C5orf56, CSF2, IL3, P4HA2	ACSL6, C5orf56, CDC42SE2, FNIP1, HINT1, LYRM7, P4HA2, RAD50, RAPGEF6, SLC224A, SLC22A5	P4HA2, CDC42SE2, HINT	C5orf56, PDLIM4, RAD50, RAPGEF6, SLC22A5	C5orf56, RAD50, RAPGEF6, SLC22A5
5q31.1	IRF1	rs6894249	C5orf56, IRF1, P4HA2	ACSL6, AFF4, ALDH1A1, APBA2, C1QA, C1QB, C5orf56, CACNA2D3, CTSL, CYP2S1, CYP4F22, DDX58, EPB41L3, FAM26E, IL15, IL5, IRF1, KCNMA1, LAP3, LGALS3BP, MYOF, P4HA2, PDE7B, RAD50, SLC22A4, SLC22A5, STAT1, TCN2, TRANK1, TSPAN2, VAMP5, WARS	P4HA2, RAD50, IRF1, SLC22A4	AFF4, C5orf56, CCNI2, GDF9, HSPA4, IL13, IL4, IL5, IRF1, KIF3A, LEAP2, PDL1M4, RAD50, SEPT8, SHROOM1, SLC22A5, SOWAHA, UQCRO, ZCCHC10	AFF4, C5orf56, IL5, IRF1, RAD50, SLC22A5	IL5, IRF1
5q33.3	IL12B	rs7725339	IL12B	NSMCE1, RNF145		RNF145, UBLCP1	RNF145
5q33.3	IL12B	rs60689680		RNF145		ADRA1B, C5orf54, RNF145, TTC1, UBLCP1	RNF145
6q26	PLG	rs4252120	PLG			AGPAT4, MAP3K4, PLG
6q23.3	IFNGR1	rs4896243	IFNGR1, IL22RA2	IFNGR1		IFNGR1, TNFAIP3, IL20RA	IFNGR1	IFNGR1
8p21.2	PTK2B	rs73223431	PTK2B	CHRNA2, CTNNA1, EPHX2, PTK2B, TRIM35	PTK2B	CLU, PTK2B, TRIM35	PTK2B, TRIM35	PTK2B, TRIM35
10q21.2	ZNF365/ADO	rs10995248	ZNF365	ADO, EGR2, RTKN2		ADO, ARID5B, EGR2, JMJD1C, NRBF2	ADO, EGR2
21q22.2	chr21q22	rs2242944		ETS2, HMGN1, LCA5L, WRB		ETS2	ETS2

Genes nominated by eQTL + promoter and also with CI are highlighted in bold

CI, chromatin interaction; eQTL, expression quantitative trait loci; sQTL, splicing quantitative trait loci.

Regulatory element enrichment

Considering that most of the shared genetic variants overlapped with regulatory regions, we applied GARFIELD to determine whether the set of pleiotropic polymorphisms was enriched in tissue-specific histone modifications. We detected a total of 60 significant enrichments (p≤6.12E-04) in all the analysed immune cell types (figure 4). The most significant enrichments were found for H2BK15ac (mapping accessible genes), in NK cells (p=1.07E-07) and monocytes (p=9.37E-07); H2BK20ac (mapping enhancer regions), in monocytes (p=1.46E-07), B cells (p=7.98E-07) and CD4+T cells (p=2.72E-06); and H3K4ac (mapping promoter regions), in B cells (p=1.55E-06) and CD4+T cells (p=2.11E-06). Specific enrichment in these cell types was found for at least five of the nine analysed histone marks.

Figure 4.

Results of the histone mark enrichment analysis of the set of pleiotropic variants. First column shows the analysed cell-types. The remaining columns denote the analysed histone modifications, including histone marks associated with enhancers (red colour), promoters (orange colour) and accessible genes (green). Results of the enrichment analysis are represented in a scale-based colour gradient depending on the p value. Blue indicates enrichment and white indicates no statistical significance.

Similar results were found when enrichment analysis was performed based on the results of the meta-analysis considering MPO-positive and PR3-positive AAV as independent phenotypes (online supplemental table 8).

Drug-repurposing candidates in vasculitis

Finally, we performed a drug repurposing analysis with the aim of identifying new potential therapies for the analysed vasculitides. Proteins encoded by 33 of the 159 prioritised coding-genes were targets of both pharmacologically active drugs indicated for any disease and drugs with unknown mechanisms of action but currently indicated for immune-mediated diseases (online supplemental table 9). In total, we identified 103 drugs that could be potentially repurposed in vasculitis (online supplemental table 9), 13 of which are currently indicated for immune-mediated disorders (table 4).

Table 4.

Drugs indicated for immune-mediated disorders that could be potentially repurposed in systemic vasculitis

Associated genomic region	Target gene	Drug	Type	Indication	Potential clinical application
2q33.2	CTLA4	Abatacept	Fusion proteins	RA, PsA, JIA	AAV EGPA+
3p21.31	LTK	Fostamatinib	Small molecule	Chronic immune thrombocytopenia	BD IgAV
4p16.2	GAK	Fostamatinib	Small molecule	Chronic immune thrombocytopenia	BD TAK
	IDUA	Chondroitin sulfate	Small molecule	Primary osteoarthritis
5q31.1	C1QA/C1QB	Daclizumab	mAb	MS	EGPA+EGPA BD KD TAK
	CTSL	Fostamatinib	Small molecule	Chronic immune thrombocytopenia
	IL13	Tralokinumab	mAb	Atopic dermatitis
	IL5	Mepolizumab	mAb	EGPA, Asthma, Hypereosinophilic syndrome
		Reslizumab	mAb	Asthma
5q33.3	IL12B	Ustekinumab	mAb	Ps, PsA, CD, UC	BD EGPA+TAK
		Tildrakizumab	mAb	Ps
		Risankizumab	mAb	Ps, PsA, CD
6q26	MAP3K4	Fostamatinib	Small molecule	Chronic immune thrombocytopenia	EGPA- GCA
6q23.3	TNFAIP3	Mesalazine	Small molecule	UC	MPO-AAV BD
		Sulfasalazine	Small molecule	UC, RA
8p21.2	PTK2B	Leflunomide	Small molecule	RA	EGPA+TAK
		Fostamatinib	Small molecule	Chronic immune thrombocytopenia

AAV, ANCA-associated vasculitis; BD, Behçet’s Disease; CD, Crohn’s disease; EGPA+, ANCA-positive EGPA; EGPA-, ANCA-negative EGPA; EGPA, eosinophilic granulomatosis with polyangiitis; GCA, giant cell arteritis; IgAV, IgA vasculitis; JIA, juvenile idiopathic arthritis; KD, Kawasaki’s disease; MS, multiple sclerosis; Ps, psoriasis; PsA, psoriatic arthritis; RA, rheumatoid arthritis; TAK, Takayasu’s arteritis; UC, ulcerative Colitis.

Interestingly, drug enrichment analysis revealed that the set of prioritised coding-genes was significantly enriched in pharmacological active drug targets currently used in the treatment of autoimmune diseases (OR=2.19 (95% CI 1.10 to 3.97); Fisher’s exact test p=1.57E-02) (online supplemental table 3).

Discussion

In this study, we performed a comprehensive analysis of the genetic overlap across the major forms of systemic vasculitis. The results greatly improve our knowledge regarding the genetic architecture of these conditions. We identified 15 new shared risk loci, most of which appear to affect disease by regulating gene expression levels, and prioritised potential causal genes based on functional annotation of pleiotropic polymorphisms.

Although all the lead pleiotropic SNPs, and most of their proxies, were located in non-coding regions, functional annotation also showed some non-synonymous variants, located at BCL2L11 (2q13) and PLG (6q26), in almost complete LD with the most associated polymorphisms. Considering their potential direct effect on protein function, they could be considered as the most probable causal variants within these loci, thus supporting the role of PLG, involved in vascular remodelling and angiogenesis, and BCL2L11, encoding a Bcl2 family member involved in apoptosis and immune homeostasis, as causal genes in vasculitis.

In addition, this strategy allowed us to identify two shared loci that have not been previously associated with any vasculitis at the genome-wide significance level. For one of these loci (lead SNP rs62184865), functional annotation prioritised CTLA4, a common genetic risk locus in autoimmunity involved in the negative regulation of T cell activation, as the most probable causal gene. Specifically, AAV (including MPA and GPA) and ANCA-positive EGPA contributed to this association signal in our cross-disease meta-analysis. This is in line with the results of previous GWAS suggesting a role of CTLA4 in the susceptibility to MPA and GPA^{11 24}; however, no genomic association was found in these studies. Therefore, our findings provide for the first time genomic evidence of association between this locus and AAV, adding ANCA-positive EGPA to the list of diseases in which this locus plays a pleiotropic role. Regarding the other new vasculitis-associated locus (lead SNP rs4690319), 10 genes were prioritised (CPLX1, CRIPAK, DGKQ, GAK, IDUA, MFSD7, MYL5, PCGF3, TMEM175, UVSSA) based on eQTL overlap. However, no obvious role in the immune response or any process directly related with the pathogenesis of vasculitis has been described for these genes, to date. Further studies are warranted to fully understand how this locus might be affecting the pathophysiology of the associated vasculitides. It should be noted that this locus has not been associated with any immune-mediated disease so far, which suggests a specific role in vasculitis susceptibility. This is also the case of the associations at the P4HA2, PLG, IL10 and IFNGR1 loci, which have been previously reported in a single disease, the first two in GCA⁹ and the last two in BD.^{13 14} Our results now indicate that these risk factors are common to GCA and ANCA-negative EGPA, in the case of P4HA2 and PLG, and shared between BD and PR3-positive AAV, in the case of IL10, and between BD and MPO-positive AAV, in the case of IFNGR1. Interestingly, the P4HA2 variant has also been associated with eosinophil count,²⁵ a crucial cell type in the pathophysiology of EGPA, which would explain its role in the susceptibility to this vasculitis.

Moreover, two independent signals at the 5q33.3 locus were also identified, one of them (lead SNP rs7725339) previously associated with TAK and annotated to IL12B,¹² and now also associated with BD and ANCA-positive EGPA, and another one (lead SNP rs60689680) that represents a novel signal in vasculitis and appeared to confer risk to TAK and IgAV and protection to ANCA-negative EGPA. Surprisingly, functional annotation of both signals pointed to RNF145, encoding an E3 ubiquitin ligase, as the most likely causal gene. This gene was recently shown to activate the NF-κB signalling pathway and to promote the transcription of IL-8, a chemotactic factor that attracts neutrophils to the site of inflammation and that is involved in angiogenesis.²⁶ Interestingly, increased levels of this chemokine have been reported in all the vasculitides contributing to the association signals, EGPA,²⁷ IgAV,²⁸ TAK²⁹ and BD.³⁰ Nevertheless, prioritisation based on positional mapping also nominated IL12B as the target gene of the signal originally associated with TAK. This gene encodes the p40 subunit common to IL-12 and IL-23, two cytokines that are crucial in Th1 and Th17 responses, respectively. Therefore, given the relevant role of this gene in inflammation and the proven usefulness of prioritisation based on physical proximity,³¹ it could be possible that this variant influences vasculitis by affecting both RNF145 and IL12B.

Genetic variants at nine additional loci, previously associated with a single vasculitis, emerged as new shared associations. Although several of these pleiotropic loci included multiple prioritised genes, two or less genes showed the strongest evidence of causality in five of these loci, 5q31.1, 6q23.3, 8p21.2, 10q21.2 and 21q22.2. These highly probable causal genes included IL5, IRF1, IFNGR1, PTK2B, TRIM35, ADO, EGR2 and ETS2. Both IL5 and IRF1 were linked to the association signal at 5q31.1 (lead SNP rs6894249), which conferred risk to ANCA-positive and ANCA-negative EGPA and protection to BD, KD and TAK. Notably, IL5 encodes a Th2 cytokine involved in the regulation of growth, activation, recruitment and survival of eosinophils,³² whereas IRF1 encodes a transcription factor that promotes transcription of genes involved in both innate and acquired immune responses.³³ Considering this, the opposite effect observed across vasculitides could be due to the different biological consequences of the shared genetic variant in different cell types. Indeed, the lead SNP (rs6894249) affects IL5 expression levels in Th1/17 cells and IRF1 levels in monocytes (online supplemental table 7). Therefore, it could be hypothesised that this signal influences EGPA pathogenesis by regulating IL5 in one cell type, but affects BD, KD and TAK by regulating IRF1 in another.

Along with the 5q31.1 locus, 10q21.2 was also found associated with multiple vasculitides, including the already known association with BD,¹³ but also with AAV, to which it confers risk, and GCA, IgAV and ANCA-positive EGPA, in which has a protective effect. This locus was also mapped to two different genes, ADO, encoding a dioxygenase involved in amino acid metabolism and EGR2, which plays a crucial role in regulating inflammation in different cell types and has been involved in autoimmunity.^34–36 Indeed, a recent study has described a role of EGR2 in the impaired frequency and function of type 1 regulatory T (Tr1) cells detected in IgAV, one of the vasculitides contributing to this association.³⁷ Therefore, EGR2 seems to be a more plausible candidate to be involved in vasculitis pathogenesis. Regarding the 8p21.2 locus, two genes with a role in the immune response, PTK2B and TRIM35, were prioritised. The protein encoded by PTK2B is involved in promoting T and B cell adhesion and migration,^{38 39} whereas TRIM35 plays a role in the innate immune response.⁴⁰ The association at the 6q23.3 region was detected when analysing AAV patients according to autoantibody specificity. Two phenotypes contributed to this signal, BD, previously associated with this locus¹³ and MPO-positive AAV. The IFNGR1 gene, which plays a crucial role in inflammation, emerged as a highly probable causal gene within this region. Finally, and supporting previous results, a single gene, ETS2, was linked to the 21q22.2 locus,¹² which was associated with TAK and ANCA-negative EGPA according to our results. It has been described that this transcription factor has anti-inflammatory functions, however, ETS2 is also involved in the differentiation, activation and survival of macrophages,⁴¹ and in promoting angiogenesis.⁴²

It should be noted that our study did not replicate the association between the KDM4C polymorphism rs16925200 and several vasculitides (GCA, TAK, IgAV and AAV) previously reported in a meta-Immunochip study.⁸ Although several of these diseases contributed to this signal in our study, namely IgAV and AAV, this association did not reached genome-wide significance (p=0.036). Taking into account, these contradictory results, further studies are needed to clarify the role of KDM4C as a common risk locus for vasculitis.

Enrichment analysis of shared genetic risk variants in histone marks related with active gene transcription evidenced a specific regulatory role of these polymorphisms in monocytes and NK, B and CD4+T cells, thus suggesting that these cell types are especially relevant for the common pathogenic mechanisms influencing vasculitis. It is worth highlighting that NK cells showed the strongest enrichment signal. These results support previous knowledge about the pathogenesis of vasculitis and suggest that special attention should be paid to NK cells, whose role in vasculitis is less well understood.

It has become increasingly evident that GWAS findings are useful in identifying opportunities for repurposing existing drugs.⁴³ In this regard, based on the results of our cross-phenotype analysis, we identified several drugs that could be potentially repositioned for vasculitis, some of which are especially promising, since they are already indicated for the treatment of other immune-mediated disorders. Indeed, two of these drugs, abatacept, a fusion protein consisting of the Fc region of IgG1 and the extracellular domain of CTLA4 that inhibits T cell costimulation, and ustekinumab, a monoclonal antibody against the p40 subunit encoded by IL12B, are currently in clinical trials in some of the vasculitides contributing to these associations, GPA (NCT02108860), TAK (NCT04882072) and BD (NCT02648581). It is also interesting that several of the prioritised genes were target for the same drug, Fostamatinib, a tyrosine kinase inhibitor, which would support the potential clinical utility of this drug in the treatment of the different vasculitides contributing to these associations.

Despite the large number of individuals analysed in this study, the fact of meta-analysing data from different studies and that were, therefore, genotyped with different platforms, imputed using different reference panels and filtered based on slightly different quality criteria, may affect the overlap of variants, thus decreasing the statistical power of the analysis. Therefore, the existence of additional shared associations in vasculitis that were not detected here cannot be ruled out. In addition, through in silico functional annotation, we have been able to prioritise several genes as potentially causal; however, experimental validation of these results is essential to confirm the role of these loci in vasculitis susceptibility.

In conclusion, through the largest and most comprehensive cross-phenotype study performed to date in vasculitis, we provided new insights into the shared genetic component across vasculitides, revealing common biological mechanisms and new therapeutic options that could be explored for the treatment of these diseases.

Data availability statement

Data are available on reasonable request.

Ethics statements

Patient consent for publication

Not applicable.

Ethics approval

This study was approved by the CSIC Ethic Committee and the Ethic Committee of Research of the Granada Province (CEIM/CEI). These organisations do not provide a reference number. Participants gave informed consent to participate in the study before taking part.