Abstract
Objective More than 130 susceptibility loci for rheumatoid arthritis (RA) have been identified with genome-wide association studies. To investigate the genetic predisposition of Chinese patients to anticitrullinated protein antibody (ACPA)-positive RA, we carried out an exome sequencing study.
Methods Patients were recruited from 3 major public hospitals in Singapore: Tan Tock Seng Hospital (TTSH), Singapore General Hospital, and the National University Hospital. Controls came from an established exome collection and from the TTSH Health Control Biobank. All the participants were of Chinese descent. We performed whole-exome sequencing (WES) in 595 ACPA-positive patients with RA and 1281 controls and validated the candidate variants by genotyping 795 RA cases and 600 controls.
Results The discovery cohort yielded 73 susceptibility single-nucleotide variants (SNVs) that reached statistical significance. In the validation study with an independent cohort, 2 SNVs remained significant: PCNXL4 (P = 1.50 × 10−5) and DHRS7 (P = 6.02 × 10−5). The majority of known susceptibility foci were not captured by exome sequencing.
Conclusion In this WES study of ACPA-positive RA in Chinese patients, we discovered 2 new variants in PCNXL4 and DHRS7 associated with risk for RA.
Rheumatoid arthritis (RA) is a chronic systemic inflammatory disease that manifests as pain, swelling, and destruction in the joints. It affects 0.5% to 1% of the population and continues to cause morbidity and mortality in spite of improvements in disease management.1
A genetic cause of RA was first suggested based on its association with HLA-Dw4.2 The shared epitope hypothesis subsequently became the accepted explanation of this HLA association.3 This relationship was refined by the discovery that patients who have anticitrullinated protein antibodies (ACPA) are more likely to carry the shared epitope.4 Approximately 20 years ago, RA susceptibility variants outside HLA were increasingly reported.5,6 PTPN22, CTLA4 and PADI4 are the earliest non-HLA susceptibility genes discovered and remain the strongest association.7
After the Human Genome Project and HapMap were published in 2003, most of the disease-susceptibility genetic variants have been identified through array-based, genome-wide association studies (GWAS). In total, > 130 disease-susceptibility variants have been identified.8-10
The first GWAS of Han Chinese patients with RA, involving 952 cases and 943 controls, was published in 2014.11 The discovery set produced 32 single-nucleotide polymorphism (SNP) replications in a cohort of 2132 patients and 2553 controls. Three SNPs were identified: rs12617656 in an intron of DPP4 or CD26, rs12379034 in the coding region of cyclin-dependent kinase 5 regulatory subunit–associated protein 2 (CDK5RAP2), and rs1854853 in CCR6. Another Chinese GWAS with 1027 RA cases and 2879 controls in the discovery set and 709 cases and 1642 controls in the replication cohort was recently published.12 Five new susceptibility loci were found: IL12RB2, BOLL-PLCL1, CCR2, TCF7 and IQGAP1.
Since 2009, exome sequencing has become an important tool in research and clinical management of complex diseases.13 We hypothesize that the exome could harbor rare variants of large effect size, as these could directly affect protein function,14 although two-thirds of the known susceptibility variants are found in the intronic region.9 Here, we describe the results of a 2-stage genetic association study using exome sequencing in ACPA-positive Chinese patients with RA in Singapore. In the discovery stage, we analyzed the exomes in 595 cases and 1281 controls. After selecting the top susceptibility variants, we replicated the findings in an independent group of 795 RA cases and 600 controls.
METHODS
Study population. The patients were recruited from the 3 largest rheumatology centers in Singapore: Tan Tock Seng Hospital’s (TTSH) Department of Rheumatology, Allergy and Immunology15; the Singapore General Hospital’s Department of Rheumatology and Immunology16; and the National University Hospital’s Rheumatology Division.17 All participants were ≥ 21 years of age at study entry and fulfilled the 1987 American College of Rheumatology (ACR) revised criteria or the 2010 ACR/European Alliance of Associations for Rheumatology (EULAR) criteria for RA.18,19 We selected patients of Chinese descent to minimize the effect of ethnic heterogeneity.20 We also eliminated the stratification bias, in which patients of 1 ethnicity are overrepresented if there is a group predisposition to the disease, whereas the controls have a different ethnic admixture.21 Controls were Chinese volunteers without RA. Although susceptibility loci can be common to people of Asian and European ancestries, there are many that are unique to each group.10,22
Antibodies directed against citrulline-containing proteins were reported in the serum of patients with RA in 1964 but characterized only 30 years later.23,24 Subsequently, ACPA proved to be important not only for clinical utility but also for understanding the pathogenesis of the disease. ACPA-positive patients tend to develop more radiologic erosions and are more likely to experience work disruption.25,26 Susceptibility variants differ between ACPA-positive and ACPA-negative disease.27-30 As the proportion of ACPA-positive and ACPA-negative patients varies in different analyses, which could complicate the association analyses, we selected only ACPA-positive patients for this project.
The design of our study is consistent with recommendations in the literature.31 The 595 patients for exome sequencing were recruited from TTSH as part of the RA Disease Registry inaugurated in 2001. The 1281 genomes/exomes of healthy people came from a local database, the SingHealth Exome Consortium hosted by the SingHealth Duke–National University Singapore Institute of Precision Medicine. The 795 patients with RA in the confirmatory cohort came from the National University Hospital (300 patients), Singapore General Hospital (300 patients), and TTSH (195 patients). The 600 controls for the validation studies came from the TTSH Healthy Control Tissue Bank.
Sequence analysis. Library preparation of the genomic DNA was performed with the Nimblegen SeqCap EZ kit (Roche) and with Agilent SureSelect Human All Exon kit (Agilent Technologies). Products were purified using AMPure XP system (Beckman Coulter) and quantified using the Agilent high-sensitivity DNA assay on the Agilent Bioanalyzer 2100 system. The exome sequencing was performed by commercial providers using the IlluminaHiSeq2000 100PE platform.
Identification of candidate variants for validation. Sequence data were aligned to the hs37d5 human reference genome using Burrows-Wheeler Aligner (BWA)–maximal exact matches algorithm of BWA,32 followed by read duplication removal using Picard. Subsequently, using different modules (base recalibration and HaplotypeCaller) of Genome Analysis Toolkit (GATK) version 3, each sample was processed separately. Finally, all samples were processed together using GenotypeGVCF variants.33 This was followed by SNP filtering with “QualByDepth (QD) < 2.0 || FisherStrand (FS) > 60.0 || RMSMappingQuality (MQ) < 40.0 || MappingQualityRankSumTest (MQRankSum) < −12.5 || ReadPosRankSumTest (ReadPosRankSum) < −8.0” as the filtering criteria and indel using “QD < 2.0 || FS > 200.0 || ReadPosRankSum < −20.0,” as per Genome Analysis Toolkit (GATK; Broad Institute) best practice for hard filtering.34 These filtered variants were annotated by multiple passes through ANNOVAR.35
A merged dataset of 1876 samples (595 cases and 1281 controls), including SNPs with > 90% genotyping rate and samples having a call rate > 90%, was analyzed. After removing duplicates and first-degree relatives, 1809 samples remained. Candidate variants were prioritized through association analysis, selecting those with minor allele frequency (MAF) < 5%. Deviation from Hardy-Weinberg equilibrium in controls, Fisher exact test, and logistic regression were conducted with PLINK.36 We did not limit the variants to those predicted to change the protein sequence (nonsynonymous single-nucleotide variants [SNVs] and short insertions and deletions), as synonymous mutations can affect protein function.37
After standard quality-control measures, a logistic regression analysis (additive) was carried out to compare the exome sequencing reads of the patients and the healthy controls. Fisher exact test was performed in SNPs with allele counts of the minor allele < 10.
Validation analysis. Seventy-five variants were selected for testing in an independent sample of patients and controls. We used Sequenom MassARRAY QGE (John Hopkins Court) for the replication study. After determining the allele, we determined statistical significance using the chi-square test. We examined our results against previously reported SNPs.9,12 Sequenom assays each variant in separate experiments, so we should consider each variant as independently validated, unlikely sequencing in which a string of variants is determined in each run. Therefore, there is higher chance of variability because of multiple independent Sequenom runs.
RESULTS
Patient characteristics. A total of 546 (91.8%) of the 595 ACPA-positive patients in the discovery set and 648 (81.5%) of the 795 patients in the replication set were rheumatoid factor–positive. In the first group, 84.5% were female, the mean age was 56.9 (SD 11.8) years, and the mean age at diagnosis was 47.2 (SD 12.0) years. In the second group, 81.5% were female, the mean age was 58.8 (SD 13.0) years, and the mean age at diagnosis was 51.3 (SD 13.1) years.
There were 1300 controls in the discovery set and 600 controls in the validation set. In the first group, 81.8% were female and the mean age was 52.6 (SD 9.9) years. In the second group, 83.4% were female and the mean age was 32.7 (SD 9.9) years.
Exome sequencing of 595 patients with RA. The mean depth of coverage was 30×. We implemented the following variant quality-control filters before analysis: sample call rate > 90%, detection of related and duplicate samples, compliance with Hardy-Weinberg equilibrium, and exclusion of outliers. After principal component analysis, 47 outliers were identified and removed from the dataset. The genomic inflation factor λ was 1.02078. Principal component analysis is shown in Figure 1.
Principal component analysis showing substructure of the cohort. Outliers, shown encircled in red, were removed.
Identification of candidate variants for validation. We selected SNVs for validation based on the following criteria: the top 70 from logistic regression analysis, with the lowest P value on Fisher exact test. We excluded variants if the P value of Hardy-Weinberg equilibrium in controls exceeded 10−5. If the variants were in linkage disequilibrium, we selected only 1 for genotyping. The Manhattan plot is shown in Figure 2.
Manhattan plot showing the −log(P values) across the chromosomes.
Out of the 75 variants identified (Table 1), we could design primers for only 56 for the custom Sequenom Multiplex MassARRAY assay. A genotyping rate of 95% for samples was used to filter the data. Subsequently, variants were analyzed based on 2 inheritance models (de novo autosomal dominant and autosomal recessive).
Information on the 75 variants identified as possible susceptibility factors in the primary analysis.
Two SNVs remained statistically significant: 2 nonmajor histocompatibility complex (non-MHC) loci, PCNXL4 (P = 0.04) and DHRS7 (P = 0.03). The combined P values for these loci were 1.50 × 10−5 for PCNXL4 and 6.02 × 10−5 for DHRS7.
We examined a list of 106 known susceptibility SNPs (Supplementary Table, available from the authors upon request). Unfortunately, exome sequencing captured only 5 of the SNPs, and those that were captured did not show a statistically significant difference between patients and controls.
DISCUSSION
We sequenced the exomes of 595 Chinese ACPA-positive patients with RA from Singapore. We identified 75 SNVs, which we evaluated in an independent cohort of 795 patients with the same characteristics. We also verified 2 novel susceptibility variants, and we found that the genetic architecture of RA in patients of Chinese descent could be different from that in other populations. To our knowledge, this is the largest reported exome sequencing study for the pathogenesis of RA.
Limited exon sequencing studies in RA have produced mixed results. Sequencing the exons of 25 genes in 6 autoimmune diseases failed to reveal any disease-associated rare variants.38 Diogo et al sequenced the exons of 25 RA-associated genes in 500 RA cases and 650 controls of European ancestry and found 2 susceptibility variants, IL2RA and IL2RB.39 Bang et al failed to identify any rare variants after sequencing the exons of 398 genes, including 106 known RA loci, in 1217 Korean patients with RA and 717 controls.40 Li et al sequenced the exomes of 58 RA patients and 66 controls and identified 5 susceptibility genes: NCR3LG1, RAP1GAP, CHCHD5, HIPK2, and DIAPH2.41
In our study, we found 2 new susceptibility variants for RA. DHRS7 encodes a protein that belongs to the short-chain dehydrogenase/reductase superfamily. Its function is not well studied, though cortisone is shown to be one of its substrates.42 PCNXL4 has not been implicated in disease causation, and it encodes a protein whose function has not been well characterized. We think that the idea of Boyle et al applies here: the gene regulatory networks are sufficiently interconnected such that genes identified in association studies will ultimately be found to be related to the pathogenetic mechanisms.43
We were surprised by the paucity of coverage of known susceptibility loci with whole-exome sequencing (WES). Different populations are likely to have different susceptibility variants for RA, which was suspected from the study of a few genes and supported by further analysis. Polymorphisms in PADI4 and protein tyrosine phosphatase nonreceptor type 22 (PTPN22) genes are cited as examples of genetic differences between Asian and White patients with RA.22 Although coding variants within PTPN22 are associated with RA in patients of European descent,44 none of these were replicated in Chinese patients with RA.45,46 GWAS analyses of Chinese patients with RA do not report widespread replication of known loci.11,12 We confirm the lack of association of PTPN22 in our Chinese-descent cohort, as they were mostly monomorphic for this allele (rs2476601; MAF = 0.0007). In Koreans, 4q27, 6q23, CCL21, TRAF1/C5, CD40, and PTPN22 showed no correlation with RA susceptibility.47
This study highlights the limitations of exome sequencing. We examined 106 known susceptibility loci and found that our technique covered only 5 of them. Out of these, there was no statistical difference between patients and controls.
The strengths of this study are the application of large-scale exome sequencing to RA, the homogenous ethnicity and uniform positive ACPA status of the research participants, and the discovery and validation of novel susceptibility loci.20 Although almost 1400 patients were included in this work, larger datasets should be assembled to explore rare variants with small effects.
To conclude, in this WES study of ACPA-positive RA in Chinese patients, we discovered 2 new variants in PCNXL4 and DHRS7 associated with RA risk. We did not find rare coding variants to account for the missing heritability of RA in the exome.
ACKNOWLEDGMENT
We thank Ms. Ying Qi Choong, Ms. Jocelyn Gay, and Ms. Joo Yong Ong for data collection and management. We also acknowledge the following members of the Singapore Rheumatoid Arthritis Study Group who contributed patient data and samples to the project: Grace Yin Lai Chan, Madelynn Tsu-Li Chan, Faith Li-Ann Chia, Hiok Hee Chng, Hwee Siew Howe, Li Wearn Koh, Kok Ooi Kong, Weng Giap Law, Tsui Yee Lian, Xin Rong Lim, Mung Ee Loh, Mona Manghani, Yu Hor Bernard Thong, Department of Rheumatology, Allergy and Immunology, Tan Tock Seng Hospital, Singapore; Jiacai Cho, Dow Rhoon Koh, Tang Ching Lau, Anita Yee Nah Lim, Anselm Mak, Amelia Santosa, Frank Sen Hee Tay, Gim Gee Teng, Division of Rheumatology, National University Hospital, and Department of Medicine, Yong Loo Lin School of Medicine, Singapore; Aisha Lateef, Division of Rheumatology, National University Hospital, and Woodlands Health, Singapore; Aziman bin Ya’akub, Li-Ching Chew, Tyng Yu Chuah, Kok Yong Fong, Cassandra Hong, Annie Nee Hui Law, Warren Fong, Ying Ying Leung, Andrea Hsiu Ling Low, Sue-Ann Pei Lun Ng, Swee Cheng Ng, Nur Emillia binte Roslan, Yih Jia Poh, York Kiat Tan, Tze Chin Tan, Julian Thumboo, Siaw Ing Yeo, Jon Kah Choun Yoong, Department of Rheumatology and Immunology, Singapore General Hospital, Singapore.
Footnotes
CONTRIBUTIONS
KPL: conceptualization, methodology, project administration, software, formal analysis, funding acquisition, writing - original draft, writing - review & editing; MYY: data curation, project administration, investigation, formal analysis, writing - review & editing; ETK: conceptualization, funding acquisition, writing - review & editing; PPMC: conceptualization, funding acquisition, writing - review & editing; ML: conceptualization, funding acquisition, writing - review & editing; CTN: conceptualization, writing - review & editing; CMW: data curation, writing - review & editing; LLG: conceptualization, writing - review & editing; SHHL: data curation, writing - review & editing; PD: data curation, writing - review & editing; GYMC: data curation, writing - review & editing; JWLT: conceptualization, funding acquisition, writing - review & editing; WH: formal analysis, writing - review & editing; MLC: formal analysis, writing - review & editing; VK: methodology, software, formal analysis, resources, writing - review & editing; SD: conceptualization, methodology, investigation, software, formal analysis, resources, funding acquisition, writing - review & editing.
FUNDING
This study is supported by the National Healthcare Group Small Innovative Grant (SIG/15036), Centre Grants (CG12Aug17 and CGAug16M012), and an individual research grant (NMRC/CG/017/2013) awarded by the National Medical Research Council, Ministry of Health, Singapore.
COMPETING INTERESTS
The authors declare no conflicts of interest relevant to this article.
ETHICS AND PATIENT CONSENT
The study was approved by the institutional review board (DSRB 2015/00582). Informed consent was obtained from each volunteer.
- Accepted for publication November 8, 2024.
- Copyright © 2025 by the Journal of Rheumatology
