Introduction

Rheumatoid arthritis (RA) is a complex autoimmune disease characterized by chronic inflammation of multiple joints, leading to progressive destruction to articular cartilage and bone. RA is strongly tied to the patients’ genetic makeup. The heritability of RA approaches 65% [1]. Extensive efforts including numerous genome-wide association studies (GWASs) so far have dramatically escalated the rate of discovery of RA-associated variants [2–4]. Recently, a genome-wide association study meta-analysis in a total of >100,000 subjects of European and Asian discovered 101 RA risk loci [5]. The SNPs identified to date, however, collectively only explain a modest proportion of the total heritability. One of possible reasons is that the traditional SNP-based GWAS used stringent thresholds of significance to control errors for the multiple testing, which resulted in a large number of SNPs with potential effects being filtered out and ignored. To help address this issue, several methods of combining P values to guide gene-level association studies were established [6–8]. Among these methods, GATES, a Simes test extension, is considerably efficient but faster and more convenient [9]. Indeed, recent studies have supported the high efficiency of gene-based association analysis in detecting disease-susceptibility genes [10–14], but currently no gene-based association study was performed to detect more novel genes for RA.

Obvious evidence has supported that substantial genetic heterogeneity exists in underlying autoimmunity among different ethnic populations. For example, the prevalence of RA is estimated to be 0.5–1.0% worldwide. However, a higher prevalence exists in populations of European ancestry than those of Asian ancestry. Among the genetic predisposition factors identified to date, HLA-DRB1 gene is the most major determinant of RA genetic predisposition among multiple ethnic studies. But in more often situations the genes identified contributed to RA with an ethnic-specific pattern, especially for the non-HLA susceptibility genes, for example, PTPN22 gene in European populations [15,16] and PADI4 gene in Asian populations [17,18]. The detected ethnic-specific pattern may come from the inherent genetic specific differences across different ethnic populations [19,20] and also probably come from sampling biases or a lack of statistical power in the association analyses. In the era of GWASs, integrating original research results from multiethnic studies greatly improve the statistical power to uncover unknown genetic predispositions and clarify their differences in genetic background among ethnicities [21].

Therefore, based on the publicly available large RA datasets [5], this study performed high powerful gene-based association analysis to detect unknown susceptibility to RA and addressed the ethnic differences in genetic susceptibility to RA between European and Asian populations.

Materials and Methods

Differential Expression Verification of RA Associated Genes

We performed differential expression analyses for the ‘novel’ RA-associated genes identified by gene-based association study. First, we downloaded four publicly available expression datasets from GEO Datasets (www.ncbi.nlm.nih.gov/geo). These data were released in RA-related studies conducted in Caucasian subjects (GSE55235, GSE55457 and GSE15573) [24,25] and in Asian subjects (GSE17755) [26], respectively. Details on sample quality control, experiment procedures and data analyses including normalization of raw data were described in the original publications. Second, the means of the interested gene expression signals were singled out from the four datasets. Third, comparisons of mean gene expression signals between RA cases and controls were conducted separately in the four datasets through Independent-Samples T Test. P value < 0.05 was considered as significant. If the significant differential expression of one gene was verified in at least three GEO datasets, it would be determined as a ‘highly verified’ gene.

Next, the secretory genes were selected from the ‘highly verified’ ‘overlapped’ RA-associated genes for ELISA testing in our in-house sample (plasma of 25 RA patients and 13 age- and sex- matched health controls) using commercially available ELISA kits (Enzyme-linked Biotechnology Co., Ltd., Shanghai, China) according to the manufacturers’ protocols. Comparison of plasma concentrations between RA patients and controls was performed using a Mann-Whitney test. P value < 0.05 was considered significant. All patients fulfilled the American College of Rheumatology 1987/2010 revised criteria for diagnosis of RA, the average disease activity score (DAS28) of whom was 5.71. The study was approved by the Scientific Ethical Committee of the First Affiliated Hospital, Soochow University and followed the tenets of the Declaration of Helsinki. Participants in this study all provided their written informed consent.

The flow chart of data analysis is shown below in Fig 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. The flow chart of data analysis.

European-specific and Asian-specific multivariate gene-based association tests were conducted separately by using extended Simes procedure (GATES) [9], KGG 2.5, using raw data of 18 European GWASs and 4 Asian GWASs. The 221novel genes were screened from 402gene-based detected genes. Among the 221genes, the differential expression of 105 genes was verified at least in one of four GEO datasets. The differential expressions of 20 genes were verified at least in three of four GEO datasets. Three genes encoding secretory proteins were selected from the 11 ‘highly’ verified ‘overlapped’ genes. GWASs: genome-wide association studies; PheGenI: Phenotype-Genotype Integratorhttp://www.ncbi.nlm.nih.gov/gap/phegeni/;GEO: gene expression omnibus.

https://doi.org/10.1371/journal.pone.0167212.g001

Results

Differential Expression Analyses of ‘Novel’ Detected RA Associated Genes

In the peripheral blood mononuclear cells (PBMCs) and synovial tissue of European or Asian RA patients, t-test showed that a total of 105 genes including the 37 ‘overlapped’ genes, 41 ‘European-specific’ genes and 27 ‘Asian-specific’ genes have differential expression signals (P value < 0.05) in at least one of the four functional studies (S4 Table). Especially, 20 genes including 11 ‘overlapped’ (ABCF1, FLOT1, HLA-F, IER3, TUBB, ZKSCAN4, BTN3A3, HSP90AB1, CUTA, BRD2, HLA-DMA), 5 ‘European-specific’ (PHTF1, RPS18, BAK1, TNFRSF14, SUOX) and 4 ‘Asian-specific’ (RNASET2, HFE, BTN2A2, MAPK13) genes were differentially expressed between RA patients and health controls in three studies or four studies (Table 1 and S4 Table), and these genes were regarded as ‘highly verified’ RA-associated genes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. The 20 ‘highly verified’ RA-associated genes newly identified by gene-based association study.

https://doi.org/10.1371/journal.pone.0167212.t001

Further, we selected three genes (FLOT1, HLA-DMA and TUBB) that encode secretory proteins from the above 11 ‘highly verified’ ‘overlapped’ genes to test if there are differential expressions in protein level by using ELISA testing in plasma. As we expected, protein levels of FLOT1 and HLA-DMA were significantly lower in RA patients compared with health controls, but not significant for TUBB (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Clinical characteristics and ELISA test results of all patients and controls.

https://doi.org/10.1371/journal.pone.0167212.t002

Gene Set Enrichment Analysis and Network Pathway Analysis

For the 221 newly identified RA-associated genes, 23 GO terms and three KEGG pathways (hsa05322: Systemic lupus erythematosus, hsa05034: Alcoholism and hsa05203: Viral carcinogenesis) were significantly enriched after Bonferroni correction (S5 and S6 Tables). Most of the significant GO terms and pathways were related to the histone gene cluster on chromosome 6 which were enriched in ‘Asian-specific’ genes. The PPI among the newly identified RA-associated genes were showed in Fig 2. The most visible gene set is mainly composed by histone 1H family both in 221 total novel genes and 74 ‘Asian-specific’ genes. Most of the ‘highly verified’ RA-associated genes such as TUBB, HSP90AB1, RPS18, BRD2, PHTF1, MAPK13, BAK1, HLA-F, IER3, RNASET2, HLA-G, ZKSCAN4 and HFE were showed in the STRING Network Visualization.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. PPI network analysis for the 221‘novel’RA-associated genes.

This is the confidence view of protein-protein interactions produced by STRING for(A) 221 total, (B) 71 overlapped, (C) 76 European-specific and (D)74 Asian-specific gene-based RA-associated genes whose integrated scores are bigger than 0.4. The disconnected nodes are not shown in the figure. Stronger associations are represented by thicker lines. The most visible gene set is mainly composed by histone 1H family both in (A) and (D).

https://doi.org/10.1371/journal.pone.0167212.g002

Discussion

In this study we performed the gene-based GWASs association tests using the publicly available datasets of the largest combining GWASs. The gene-based analysis has the following advantages: 1) genes, not SNPs, are thought to be the functional units in the genome; 2) genes rather than SNPs are highly consistent across diverse populations; 3) gene-based analyses rather than SNP-based analyses in GWASs can alleviate the multiple testing burden and thus improves the statistical power to detect significant genes; 4) candidate genes identified by gene-based association study are directly suitable for further pathway and network-based analysis. When doing SNP-based study, KGG prioritizes SNPs through a knowledge-based weighting method which can maximize the potential power of association tests while controlling false positive discoveries rate and thus could detect more candidate genes. The gene-based association study identified 402 RA-susceptibility genes even after very strict Bonferroni corrections. More importantly, after excluding the known RA-associated genes, the present study discovered 221 ‘novel’ RA genes. Near half of the 221 novel genes (105 genes) had significant differential expression signals between RA patients and health controls in the next functional validation tests, among which twenty genes were highly verified. All these evidences highlighted the relatively higher power for the gene-based association analysis.

An important topic of this study is the ethnogenetic homogeneity and heterogeneity in RA etiology. We provide evidence of 71 ‘overlapped’ RA risk genes in Asian and European individuals. Among them, 37 genes have differential expression signals (P value < 0.05) in synovial tissues or PBMCs of RA patients of Asian and European, and,11 genes (ABCF1, FLOT1, HLA-F, IER3, TUBB, ZKSCAN4,BTN3A3, HSP90AB1, BRD2, HLA-G and HLA-DMA) are highly verified in three or four functional studies. These observations support the view that the genetic risk of RA is shared, in general, among Asians and Europeans [5,27]. We also highlight apparent differences across ethnic groups. First, there are 74 ‘Asian-specific’ and 76 ‘European-specific’ RA risk genes detected by our gene-based association study, suggesting that ethnic variation should be considered in RA etiology. Second, the ‘Asian-specific’ RA risk genes are clustered together within chromosome 6 while the ‘European-specific’ RA risk genes are scattered across multiple chromosomes, which means that multiple risk genes scatted in the genome may contribute RA pathogenesis even if they are not the primary causes, and, Europeans may have more diverse genetic heterogeneity in RA etiology. Third, more than half of the newly identified ‘Asian-specific’ genes are histone 1H family genes, which accounts for less than one-tenth in ‘European-specific’ genes. It is commonly known that histones play a central role in transcription regulation, DNA repair, DNA replication and chromosomal stability. However, there are few reports about the relationship of histone family and RA. It is a novel finding that the histone 1H family is associated with RA in Asian population. Fourth, although a total of 27 ‘Asian-specific’ and 41 ‘European-specific’ newly identified genes are differentially expressed between RA patients and controls, only two ‘European-specific’ genes, PHTF1 and TNFRSF14, are validated by all the four functional studies, and, only PHTF1 shows an opposite RA/control ratios of mean expression value between Europeans and Asians. This hints that we might need to consider both the tissue specificity and race specificity when making functional verification tests.

Another interesting finding is that most of the ‘highly verified’ RA-associated genes might have potential connections with RA pathogenesis. For instance, BRD2, the ‘European-specific’ RA-associated gene, is directly connected with the histone 1H cluster in the confidence view of STRING. Although the functional relationship of BRD2 and RA is unclear till now, it is reported that Bromodomains (BRDs) are protein interaction modules that exclusively recognize acetylation motifs [28] and there is a structural basis for deciphering the histone code by BRD2 through the binding with a long segment of the histone H4 tail and then presumably prevent erasure of the histone code during the cell cycle [29]. As for HLA-DMA, it is another highly verified RA risk gene both in European and Asian populations. It plays a critical role in catalyzing the release of class II HLA-associated invariant chain-derived peptides from newly synthesized class II HLA molecules and freeing the peptide binding site for acquisition of antigenic peptides [30]. Given that a striking association is found between RA and particular HLA-DRB1, it seems to be a good candidate allele involved in RA pathogenesis [31]. However, it is previously reported that the HLA-DM (DMA and DMB) genes do not have any influences on their own to genetic susceptibility to RA [32,33]. More in-depth work is necessary to determine whether HLA-DMA is indeed associated with RA. With regard to PHTF1and BTN3A3, the highly verified ‘European-specific’ and ‘overlapped’ RA risk gene, no direct evidence has been reported till now that they are involved in RA etiology. PHTF1 (putative homeodomain transcriptional factor), a putative homeobox gene located at band 1p11-p13 of the human genome, may play a role in transcription regulation. It encodes a membrane protein abundantly expressed in male germinal cells [34]. The rs6679677 (PHTF1-PTPN22) is reported as a susceptibility factor for autoimmunity in diabetes type 1 [35,36] while PTPN22 is a well-known RA risk gene. BTN3A3 (Butyrophilin, Subfamily 3, Member A3), also called CD277, belongs to the B7 family members and is expressed in various immune cells such as T and NK cells [37]. BTN3A3may act as one of the inhibitors of co-stimulation for T lymphocyte priming, similar to CTLA-4 [38]. It also found that SNPs near the butyrophilin genes (BTN3A3/BTN2A1) are associated with variations in IFN-γ secretion [39]. As for FLOT1 (flotillin-1), another gene verified by our ELISA test, its important roles in promoting tumorigenesis and progression of several cancers like non-small cell lung cancer, breast cancer and hepatocellular carcinoma have been recently reported [40,41]. The function of flotillin 1 in RA development has not been determined. However it is found that FLOT1 can activate tumor necrosis factor-alpha (TNF-α) receptor signaling and sustain activation of NF-kappa B in esophageal squamous cell carcinoma cells [42]. Taken together, the above evidence mentioned supports that the ‘highly verified’ RA-associated genes are worth in-depth study. Further studies are needed on a number of issues including how histone 1H genes relate to RA, whether the newly identified candidate genes especially those highly verified genes truly relate to RA etiology, and, if any, what functional relationships are between these genes and RA.

Our study has several limitations. The gene-based association analyses of combining GWASs did not include the SNPs in X/Y chromosomes or that could not be recognized by KGG, thus the significant genes might not be fully detected. The sample size of our functional differential expression analyses was relatively small. Since only plasma for the subjects is available for us and our budget is limited, we could only select three secretory genes from the eleven ‘highly verified ‘‘overlapped’ RA-associated genes for ELISA test and left a long list of candidate genes to be tested in protein level.

In conclusion, using the gene-based association research strategy, our study identified a long list of novel RA associated genes and also addressed their ethno-genetic homogeneity and heterogeneity in European and Asian populations. Our findings point to the involvement of novel genes and pathways in the pathogenesis of RA, and provide more insights into ethnic differences in genetic susceptibility to RA between European and Asian populations.

Supporting Information

Acknowledgments

We are grateful to Dr Xing, Cheng for polishing the article.