Known SLE associations found in the Thai population
In the discovery dataset, the association studies were initially performed using healthy controls (n = 1,606) and SLE patients (n = 487) collected from King Chulalongkorn Memorial Hospital. Regarding the result, we found that variants at the HLA class II regions were strongly associated with SLE (p-value < 5E-08). Similarly, GWAS from 405 SLE cases and 1,590 non-immune mediated disease controls found variants at the HLA class II regions reached the genome-wide significant threshold (p-value < 5E-08). Our findings were consistent with previous reports in other ethnic groups . Inflation factors from both datasets were calculated as reported in Supplementary figure 1.
Subsequently, a meta-analysis of the two Thai GWAS was carried out, and we systematically examined associations across the 90 known SLE-associated loci, which were collected from the GWAS Catalogue (https://www.ebi.ac.uk/gwas/) and previous review articles . Of these loci, the HLA-DQA1, HLA-DRB1, STAT4, FAM167A-BLK, and GTF2I loci have reached the genome-wide significant threshold (p-value < 5E-08; Figure 1B, Table 2) in Thai population, and the variants at the PROS1C1, NOTCH4, HCP5, C6orf10, TAP2, TNFSF4, RasGRP3, TERT, TNPO3-IRF5, CXCR5, GPR19, SLC15A4, and ITGAM loci showed suggestive evidence of associations with SLE (p-value < 5E-05, Supplementary Table 1). These loci have been found in several ancestries, including Han Chinese, Korean, North American, European, African, and Hispanic populations [30, 31].
We noticed that some of the previously characterized nonsynonymous polymorphisms also showed certain evidence of association (p-value < 0.05) in Thai population, such as rs11235604 (ATG16L2, R58W), rs13306575 (NCF2, R395W), rs1990760 (IFIH1, A946T), rs3734266 (UHRF1BP1, Q454L), rs2841280 (PLD4, E27Q) and rs2230926 (TNFAIP3, F127S). Details of these associations were summarized in Table 3. All significant variants were calculated for Hardy-Weinberg equilibrium, as reported in Supplementary Table 2.
Identification of novel loci associated with SLE
Excluding the variants at the known SLE-associated loci, we discovered a novel variant on FNB2 (rs74989671, OR=1.54, p-value=1.61E-08) specifically associated with SLE in Thai population (Figure 1B and Figure 2A, Table 2) when comparing the association in Europeans (OR=0.998, p-value=0.979) and in Chinese populations (OR=0.982, p-value=0.692) . Further analyses based on different genetic inheritance models suggested that the disease risk was associated with the copy number of risk alleles that the individuals carried (additive model) (Table 4). Three SNPs on FBN2 loci (rs74989671, rs35983844, rs6595836) showed linkage disequilibrium (LD r2 = 0.82) (Figure 2B, Supplementary Table 1). Of these variants, rs74989671 was found to locate within the peak of H3K36me3 derived from CD14 positive monocytes and H3K4me1 (associated with active enhancers) derived from the primary T cells (Figure 2C).
In addition, we found variants at the ATP6V1B1, MIR4472-2, MYO5C, ADCY5, and DGKG, showing suggestive evidence of associations with SLE in Thai population (p-value < 5E-05) (Supplementary Figure 2, Supplementary Table 1). Though these polymorphisms are likely to associate with Thai SLE patients, an independent GWAS dataset of SLE patients with Thai background is needed for further validation.
In silico functional annotation of SLE-associated variants in Thai population
To understand the biological meaning underlying the SLE-associated loci in the Thai population, we performed the pathway analysis using the SNP-nexus program . Variants with p-value < 5E-05 were involved in this study. Notably, we found that 50% of all-variants were located within the coding region, by which 10% is non-synonymous polymorphisms. Pathway analysis results revealed that SLE-associated variants were highly enriched in the regulation of interferon signalling, PD-1 signalling, MHC-class II antigen presentation, TCR/BCR signalling, cytokine signalling, TNF-signalling, NOTCH4 signalling, calcium-activated potassium channels, and cell-cell junction organization pathways. Furthermore, we found that extracellular matrix organization was significant in our results (Figure 3). It indicated that Thai SLE patients might have a higher risk of fibrosis-associated inflammation.
Polygenic risk score prediction for the individuals
To apply the GWAS result to predict the Thai SLE outcome, we also tested the hypothesis of whether the PRS models trained by individuals with Chinese ancestry could be applied for Thai SLE patients. We calculated PRS for individuals in the Thai GWAS, based on the training data from the Chinese population (2,618 cases and 7,446 controls) . Significantly, the PRS for SLE cases were higher than controls (mean difference = 0.89; p-value = 2.2E-16; Figure 4A), and the area under the receiver-operator curve (AUC) achieved 0.76 for this predictor. This analysis indicated the potential application for the PRS in the Thai population, based on the results from other Asian populations. Regarding the analysis, this might be a clue for predicting an outcome of SLE clinical characteristics in Thai SLE patients, and it is a good source for further genetic analysis to identify actual SLE pathogenesis in the different ancestry.