Association of SNPs of CD40 Gene with Multiple Sclerosis in Russians

Multiple sclerosis (MS) is a serious, incurable neurological disease. In 2009, the ANZgene studies detected the suggestive association of located upstream of CD40 gene in chromosome 20q13 (p = 1.3×10−7). Identification of the causal variant(s) in the CD40 locus leads to a better understanding of the mechanism underlying the development of autoimmune pathologies. We determined the genotypes of rs6074022, rs1883832, rs1535045, and rs11086996 in patients with MS (n = 1684) and in the control group (n = 879). Two SNPs were significantly associated with MS: rs6074022 (additive model C allele OR = 1.27, 95% CI = [1.12–1.45], p = 3×10−4) and rs1883832 (additive model T allele OR = 1.20, 95% CI = [1.05–1.38], p = 7×10−3). In the meta-analysis of our results and the results of four previous studies, we obtain the association p-value of 2.34×10−12, which confirmed the association between MS and rs6074022 at a genome-wide significant level. Next, we demonstrated that the model including rs6074022 only sufficiently described the association. From our analysis, we can speculate that the association between rs1883832 and MS was induced by LD, whereas rs6074022 was a marker in stronger LD with the functional variant or was the functional variant itself. Our results indicated that the functional variants were located in the upstream region of the gene CD40 and were in higher LD with rs6074022 than LD with rs1883832.

Association with the CD40 gene region has been implicated in a number of human autoimmune diseases such as systemic lupus erythematosus (SLE) [16], rheumatoid arthritis (RA) [17], Crohn's disease [18], and Grave's disease [19]. The participation of CD40 in autoimmune processes is clearly demonstrated in experimental animal models. A non-obese diabetic (NOD) mouse line is one of the most studied models of autoimmune diabetes. The disruption of CD40-CD40L interactions by antagonistic antibodies to CD40L prevents the development of diabetes in NOD mice, which confirms the critical function of the CD40-CD40L complex in the development of this disease [20].
In the mouse line K/BxN, RA spontaneously develops with many characteristics similar to the clinical course of RA in humans. CD40 knockout (K/BxN-CD40-/-) mice do not develop RA [21]. If a CIA mouse model of RA is treated with antagonistic antibodies such as CD40L mAb before the onset of collageninduced arthritis, the mouse do not develop RA [22] or the disease decreases in severity [23]. However, if the therapy is started after the development of arthritis, no improvement occurs [25,26].
CD40 contributes to the susceptibility to human autoimmune diseases, in which the B and T cell pathways play key roles. The role of CD40-CD40L interactions is identified in the development of type-1 diabetes. The CD40 signaling pathway induces the production of proinflammatory cytokines in the islet cells of primates and humans [24]. Increased expression of CD40L + cells has been observed in the brains of patients with MS [25].
Also, the expression of CD40 in keratinocytes and endothelial cells in psoriatic plaques, as well as the increased expression of CD40L in the peripheral blood T cells [26], have been shown in patients with psoriatic arthritis. An elevated level of circulating CD40L has been found in patients with RA, SLE, and Sorgena syndrome during exacerbation [27]. In fact, the interaction between CD40 and CD40L triggers the immune response.
The potential association of the CD40 region with MS is in accordance with the theory that MS has an autoimmune origin. Data from animal models also suggest similarities among the molecular mechanisms underlying the development of various immune disorders. Identification of the causal variant(s) in the CD40 locus will lead to a better understanding of the mechanism underlying the development of autoimmune pathologies. This study aimed to replicate a previously reported association of rs6074022 with MS in a Russian population and to study the association of MS with some other SNPs of the CD40 region.

Patients
Among a group of Russian-ethnicity patients with MS, 1684 people (1124 women and 560 men; mean age6SD = 36.7611.2 years) were included based on the McDonald criteria for MS [28]. This study was approved by the ethics committees of all participating centers. All participants signed a written informed consent.

Selection of SNPs
The CD40 region has two blocks of linkage disequilibrium (LD) according to the result of the HapMap 3 panel for Utah residents of Northern and Western European descent and for Toscans in Italy ( Fig. 1). In our study, we selected the following four SNPs: rs6074022 and rs1883832 from the first block of LD, rs11086998 from the second block, as well as rs1535045 from the space between the two blocks.
The two SNPs previously associated with MS lie in the first block of LD. SNP rs6074022 T-.C was associated with MS at a genome-wide suggestive level (p = 1.3610 27 ) [7]. SNP rs1883832 C-.T is located in the 21 position relative to the start of the transcription of CD40 gene. In previous candidate gene studies, the association of the minor allele T of SNP rs1883832 with a number of diseases has been examined. Such diseases include lymphoma [29], non-Hodgkins lymphoma [30], osteoporosis [31], multiple sclerosis [18], Crohn's disease [18], sporadic breast cancer [32], Behcet's syndrome [33], and rheumatoid arthritis [34].
In the second block, we selected SNP rs11086998 based on its potential function. The presence of minor G allele in rs11086998 leads to a proline-to-alanine substitution (P227A) in the intracellular region of CD40. The P227A site is 3 amino acid residues proximal to the TRAF6 binding site, and leds to increased TNF-a and IL-6 production in murine cell lines [35]. Notably, the rs11086998[G] allele has rather different frequencies in different ethnic groups, from 29% in Mexicans and South Americans to ,2% in others. SNP rs1535045 was chosen to tag the gap between the two blocks of LD.
Genotyping DNA was extracted from venous blood using standard procedures, including selection and lysis of blood cells, the hydrolysis of proteins with proteinase K, DNA purification by extraction with phenol-chloroform, and the DNA precipitated with ethanol. Genotyping of SNPs rs6074022, rs1883832, rs1535045 and rs11086996 was performed by TaqMan real-time PCR (ICBFM SB RAS, Novosibirsk, Russia).

Statistical data analysis
Tests for the Hardy-Weinberg equilibrium were performed using the DeFinetti program available from the website of the Institute of Human Genetics (Munich, Germany; http://ihg2. helmholtz-muenchen.de/cgi-bin/hw/hwa1.pl). Associations of genotype with the disease was studied using logistic regression analysis, as implemented in ''glm'' function of the R package for statistical analysis (www.r-project.org). The likelihood ratio test (LRT) was used to test the statistical hypotheses and the Akaike Information Criterion (AIC) was used to decide on ''best model'' describing the association of MS with SNPs. Meta-analysis and Qtest were carried out using the 'rmeta' package for R (http://cran. r-project.org/web/packages/rmeta/rmeta.pdf). Haplotype analysis was carried out using the 'haplo.stats' package for R (http:// cran.r-project.org/web/packages/haplo.stats/haplo.stats.pdf). Results were considered statistically significant for all statistical calculations if P,0.05.

Results
We determined the genotypes of rs6074022, rs1883832, rs1535045, and rs11086996 in patients with MS and in the control group (Table 1). The call rate was$99.9% for all SNPs. Five patients were excluded from further analysis because of missing data in at least one locus. Genotypic distribution did not significantly deviate from the Hardy-Weinberg equilibrium expectations for all four studied SNPs in the MS and control groups. The Q-test for heterogeneity of the minor allele frequency between sub-samples from different cities did not show significant differences (Table S1).
Additionally, we performed stratified analysis to exclude a potential confounding by genetic substructure in our study. Subsamples from each city were assigned a stratum in the stratified analysis ( Figure S1). The summary OR, its 95% confidence interval, significant level, and p-level of heterogeneity are shown in Table S2 for each SNP. The results of stratified analysis were in accordance with the association analysis in the entire group. Despite the smaller sample size (the Kemerovo and Tomsk groups were excluded in the meta-analysis because they contained only cases), we observed in the joint analysis that rs6074022 (p = 0.02) and rs1883832 (p = 4610 23 ) were associated with MS. The risk alleles were the same as those found in the association analysis in the entire group.
We also analyzed the association of clinical sub-phenotypes of MS (RRMS, PPMS, SPMS, and CIS) with all SNPs (Table S3). After correction for multiple testing RRMS-the most common  sub-type-was associated with rs6074022 (OR = 1.26, p = 0.0008) and with rs1883832 (OR = 1.24, p = 0.003). We performed a haplotype analysis for rs6074022 and rs1883832. The haplotype frequency, OR, its 95% confidence interval, and significant level are shown in Table 2 We estimated the LD between the studied SNPs of CD40 gene ( Figure S2). Relatively high values of D9 were detected in all SNP pairs. However, the r 2 between SNPs was weak at ,0.1, except for rs6074022-rs1883832 (r 2 = 0.59).
We used the LRT to compare three models of association of rs6074022 and rs1883832 with MS: the general model (where we estimated the effects of genotypes of both SNPs rs1883832 and rs6074022) and two nested (effects estimated for only one SNP). According to the LRT, the nested model including rs6074022 only did not significantly differ from the more general model, including both SNPs rs6074022 and rs1883832 (p = 0.99), whereas the nested model including rs1883832 only was significantly worse than the general model (p = 0.01). The model including rs6074022 only was also the best according to the AIC (AIC = 3284.4, 3282.4, and 3288.4 for the general, rs6074022, and rs1883832 models, respectively). We conclude that the association of MS with the CD40 locus can be described in terms of the involvement of rs6074022 only.
We also performed a meta-analysis of our results with previously published data on the association between rs6074022 and MS: GenMSA (NL), GenMSA (US), GenMSA (CH) [1], IMSGC (UK), IMSGC (US) [3], BWH/TT [2], and ANZgene [7]. Table 3 summarizes the results of previous studies used for this metaanalysis. In the meta-analysis (Figure 2), the total OR for all studies was 1.17 (95% CI = 1.10-1.23) with a statistical significance of p = 2.24610 212 . The heterogeneity test (Q-test) did not find significant differences between the studies (x 2 (7) = 12.16, p = 0.10). These data confirmed the association of marker locus rs6074022 and MS at a level of significance accepted for GWASs.

Discussion
CD40-CD40L is reportedly a common link in the pathogenesis of autoimmune diseases. This hypothesis is supported by the established role of the CD40-CD40L interaction in the development of several autoimmune conditions in animal models [20][21][22][23] and by the association of CD40 SNPs with the risk for a number of autoimmune diseases [16][17][18][19]. Moreover, the successfully completed Phase 1 of the clinical trial of SLE treatment by CD40L (http://www.biogenidec.com/research_product_pipeline. aspx?ID = 5778) provides evidence of the important role of the CD40-CD40L interaction in the pathogenesis of SLE. The CD40-CD40L interaction is known to result in the switch to antigen-specific Th2 type response [36]. Therefore, the CD40-CD40L complex is an extremely attractive and promising target for the development of drugs for suppressing autoimmune attack. Abbreviations: 95% CI, 95% confidence interval; OR, odds ratio; Sample frequency-haplotype frequency in MS and control groups together; empirical p-value-p-value of association haplotype with MS; *-marked the risk allele from association analysis for each SNPs alone. Analysis was performed using logistic regression. doi:10.1371/journal.pone.0061032.t002 Table 3. Meta-analysis of association between rs6074022 C allele and MS.  However, for the successful creation of drugs, the molecular mechanism underlying the initiation of autoimmune inflammation through CD40-CD40L in humans must be studied.
In this work, we aimed to replicate in the Russian population the previously reported association of rs6074022[C] with MS [7]. We observed a statistically significant association of the allele rs6074022[C] with the development of MS (per C allele OR = 1.27, CI = 1.12-1.45, p = 3610 24 ). Our results were in accordance with those of ANZgene (Table S4). The minor allele C of rs6074022 had a similar frequency both in the case and control groups. Also, the minor allele was the risk-associated allele in both studies. SNP rs1883832 was less significantly associated with MS than rs6074022 in both studies.
In a previous meta-analysis of GWASs, rs6074022 has been implicated at a suggestive level at most [37]. In the meta-analysis of our results and the results of four previous studies, we obtain the association p-value of 2.24610 212 , which confirmed the association between MS and rs6074022 at a genome-wide significant level.
Our findings of the association between MS and CD40 gene are in accordance with the hypothesis of the autoimmune nature of MS. However, the molecular mechanism underlying the involvement of SNPs of CD40 gene in the autoimmune processes was unclear for MS and other autoimmune diseases. The first step in solving this problem is to determine functional SNPs in the gene.
Haplotype  26 ). Each haplotype contains one of the ''risk'' alleles as identified in single-SNP analyses. Interestingly, the haplotype containing both risk alleles was not associated with MS in our analysis. We can speculate that this finding is consistent with the hypothesis of two functional polymorphisms located close to the marker SNPs, one increasing the risk of developing MS whereas the second being protective.
In our study, both SNPs rs6074022 and rs1883832 were significantly associated with MS (per C allele OR = 1.27, CI = 1.12-1.45, p = 3610 24 ; per T allele OR = 1.20, CI = 1.05-1.38, p = 7610 23 ). We also demonstrated that the model including rs6074022 only sufficiently described the association. A large proportion of polymorphisms located in the same block of LD with a causal variant are likely to show an association with a disease. However, the polymorphisms that are in greater LD with the functional variant should on average have a stronger association. From our analysis, we can speculate that the association between rs1883832 and MS is induced by LD, whereas rs6074022 is a marker in stronger LD with the functional variant or is the functional variant itself. We may speculate that the functional variant(s) is likely to be located in the upstream region of the gene CD40 and is in higher LD with rs6074022 than with rs1883832.

Summary
Our results confirmed the association of SNP rs6074022 with the risk of MS development. Our estimates suggeste that the functional variant(s) is located in the upstream region of the gene. Further empirical studies are required to find the functional variant.