Association results
The genome-wide association analysis was conducted in males and females, independently, in each cohort, i.e., Swedish (Supplementary Tables S1-S4), German (Supplementary Table S5-S8), and US-African American (Supplementary Tables S9-S10). The association results for these cohorts and supplementary Tables S1-S10 are reported in Additional File 1. The details of the discovery cohorts as well as the replication cohort in the UK Biobank, are shown in Table 1 A and B, respectively.
Meta-analysis of genome-wide association studies
Meta-GWAS on LS sex groups. The meta-analysis on LS sex groups performed by pooling together GWAS results from the Swedish and German cohorts revealed 766 SNPs in males and 683 SNPs in females with Pmeta<5×10-8. The lead SNPs were rs3130288 (OR=4.83, 95%CI=3.65-6.38, Pmeta=1.80×10-28) located in the 5’ of ATF6B in LS males, and rs9268219 (OR=3.69, 95%CI=2.87-4.74, Pmeta=2.04×10-24) located in an intron of C6orf10 in LS females. Several SNPs located outside the MHC and with Pmeta<5×10-5 were also identified, including rs116711574 (OR=0.24, 95%CI: 0.13-0.47, Pmeta=2.06×10-5) in an intron of HMP19 on chromosome 5 in LS males, and rs17713228 (OR=0.29, 95%CI: 0.17-0.48, Pmeta=2.21×10-6) located 68kb 3' of ZNF804A on chromosome 2 in LS females. Regional association plots for the lead meta-GWAS SNPs are illustrated in Figure 1 A-B. The forest plots highlighting the overall genetic effects in the LS sex groups are shown in Figure 2 A-B. The top 25 significant meta-GWAS SNPs in males and females are listed in Tables 2-3, respectively. The meta-GWAS results for LS males and females at Pmeta<5×10-8 are provided in Supplementary Tables S11-S12 in Additional File 2. Manhattan plots are shown in Supplementary Figure S9 A-B. Q-Q plots are shown in Supplementary Figure S9 C-D.
Results from the secondary meta-GWAS analysis including SNPs with PGC<0.05 in the Swedish cohort and SNPs with 0.05<PGC<0.1 in the German cohort identified 32 SNPs in LS males and 20 SNPs in LS females with Pmeta<5×10-8. The leading SNPs were rs3129763 (OR=2.05, 95%CI: 3.32-2.61, Pmeta=6.74x10-15) located 14kb 5' of HLA-DQA1 in LS males, and rs2517985 (OR=0.48, 95%CI:0.38-0.60, Pmeta=5.12×10-11) located in an intron of CCHCR1 in LS females. Results of secondary meta-GWAS for LS males and females are shown in Supplementary Tables S13-S14 in Additional File 2.
Meta-GWAS on non-LS sex groups. The meta-analysis on non-LS sex groups in the European cohorts (Sweden and Germany) identified 67 SNPs in males and 97 SNPs in females with Pmeta<5×10-8. The most significant SNP was rs1964995, located 36kb from the 3' of HLA-DRB5, and was the top signal for both sexes (OR=0.57, 95%CI:0.50-0.64, Pmeta=3.92×10-18 in males, and OR=0.58, 95%CI:0.51-0.65, Pmeta=5.00×10-20 in females). Complete results of the meta-GWAS analysis in the European cohorts for non-LS males and females are provided in Supplementary Tables S15-S16 in Additional File 3.
The meta-GWAS in the three cohorts comprising European and African ancestries identified 14 SNPs in males and 36 SNPs in females with Pmeta<5×10-8. The top signal was rs1964995 and was the same for both sexes (OR=0.56, 95%CI:0.50-0.63, Pmeta=1.46×10-22 in males, OR=0.66, 95%CI:0.60-0.72, Pmeta=2.86×10-19 in females). However, in non-LS females, rs1964995 had a high heterogeneity index with P<0.05, so the next top signal was rs7195 (OR=1.44, 95%CI:1.33-1.57, Pmeta=7.64×10-18) located in an intron of HLA-DRA. The top 25 significant meta-GWAS SNPs are listed in Tables 4-5, respectively. Forest plots highlighting the overall genetic effects in the non-LS sex groups are shown in Figure 3 A-D. Regional association plots for the lead SNPs are illustrated in Figure 4 A-B. The complete lists of meta-GWAS results in the non-LS sex groups are provided in Supplementary Tables S17-S18 in Additional File 3. Manhattan plots are shown in Supplementary Figure S10 A-B, respectively. Q-Q plots are shown in Supplementary Figure S10 C-D.
The secondary meta-analysis consisting of SNPs with PGC<0.05 in the Swedish and German cohorts and SNPs with 0.05<PGC<0.1 in the US-AA cohort revealed two genome-wide SNPs in non-LS males and females, respectively. Chiefly, in non-LS males rs2395153 (OR=0.69, 95%CI:0.61-0.77, Pmeta=4.77x10-10) located 5.9kb from the 5' of C6orf10 and rs1049550 (OR=0.67, 95%CI: 0.60-0.76, Pmeta=1.98x10-11) located in ANXA11 were found. In non-LS females, rs9275224 (OR=1.29, 95%CI:1.19-1.40, Pmeta=1.46x10-9) located 25kb from the 5' of HLA-DQB1 and rs4502931 (OR=0.79, 95% CI:0.73-0.86, Pmeta=3.29x10-8) located 5.9kb from the 5' of BTNL2 were observed. The complete lists of the secondary GWAS-meta results are shown in Supplementary Tables S19-S20 in Additional File 3.
Replication look-up of meta-GWAS SNPs in the UK Biobank
Given the lack of further characterization (i.e., LS and non-LS) on the sarcoidosis phenotype in the UK Biobank, we focus on the sex groups for the replication look-up. In the male group, 1,253 SNPs with Pmeta<5×10-5 in LS and non-LS were pooled together. Out of this set, 1,224 (97.6%) unique SNPs were identified and extracted from the UK Biobank using the GWAS summary statistics conducted in males. The lookup identified 723 SNPs with Bonferroni p-value of <4.0×10-5 (0.05/1,224). Moreover, after excluding SNPs with heterogeneity p-value<0.05, 715 SNPs were kept (Supplementary Tables S21 in Additional File 4).
In the female group, 1,356 SNPs with Pmeta<5×10-5 were pooled together from the GWAS meta-analyses of LS and non-LS. Out of this set, 1,316 (97%) unique SNPs were identified and extracted from the UK Biobank GWAS summary statistics conducted in females. The lookup identified 454 SNPs with Bonferroni p-value of <3.79×10-5 (0.05/1,316). Of note, since 9 SNPs had a heterogeneity p-value <0.05, these SNPs were excluded from the replication set, 445 SNPs were kept (Supplementary Table S22 in Additional File 4).
Gene-based analysis
Results from the gene-based analysis revealed different genomic loci in LS and non-LS sex groups. Mainly, 101 genes in LS males and 93 genes in LS females were identified at gene-based P-value<2×10-6. Genes shared (n=91) among LS males and females were located in the MHC region. These genes ranged from ZNF184 to PSMB9 (chr6: 27371788-32827628). The top 15 significant genes in the LS sex groups are shown in Tables 6-7. Complete results are shown in Supplementary Tables S23-S24 in Additional File 5.
In the non-LS sex groups, three genes (i.e., HLA-DRA and STK19 and ANXA11) in non-LS males and five genes (i.e., BTNL2, C6orf10, HLA-DRA, PPT2-EGFL8, and PPT2 in non-LS females were identified. Significant genes in the non-LS sex groups are shown in Tables 8–9. A Venn diagram illustrating significant loci between LS and non-LS sex groups is depicted in Supplementary Figure S11.
Expression quantitative trait loci (eQTLs) enrichment
eQTL enrichment using eQTL data of immune cells, whole blood, and lung on SNPs associated with LS and non-LS sex groups at Pmeta <5x10-8 identified several cis-eQTL SNPs (Supplementary Table S25 in Additional File 6). Specifically, the number of SNPs mapping immune cells, whole blood, and lung eQTL data were (40, 88, and 24) in LS males, (36, 79, and 19) in LS females, (4, 7, and 2) in non-LS males and (5, 1, and 3) in non-LS females. The lists of eQTL SNPs in LS and non-LS sex groups are shown in Supplementary Tables S26-S29 in Additional File 6.
A Venn diagram highlighting overlaps and differences in cis-eQTL SNPs in LS sex groups is shown in Figure 5 A-C. eQTL datasets are listed in Appendix 1 in the Supplementary Materials.
Gene enrichment and pathway analysis
In the LS, we identified significant pathway maps common to both sex groups and those specific to one group. Two pathways maps of interest were (1) immune response on the induction of the antigen presentation machinery by IFN-gamma (males: P = 2.39×10-12, 1.16×10-14, 2.04×10-14; females: P = 1.38×10-10, 1.80×10-13, 5.88×10-13) and (2) maturation and migration of dendritic cells in skin sensitization (males: P = 2.64×10-13, 1.31×10-11, 4.51×10-10; females: P = 2.06×10-13, 2.77×10-12, 2.54×10-10) (Supplementary Tables S30-S31 in Additional File 7).
In non-LS, pathways maps were only detected in whole blood eQTLs in males and in immune cells in females. Among these, pathway maps related to (1) immune escape mechanisms in prostate cancer (P=6.27×10-3) in males and (2) immune response induction of the antigen presentation machinery by IFN-gamma (P=2.39×10-8) in females were identified (Supplementary Tables S32-S33 in Additional File 7).
Moreover, a cross-comparison analysis showed various common and sex-specific pathway maps in LS. Results of the comparative analysis are shown in Supplementary Tables S34-S35 in Additional File 7.