Association of Genetic Markers with CSF Oligoclonal Bands in Multiple Sclerosis Patients

Objective to explore the association between genetic markers and Oligoclonal Bands (OCB) in the Cerebro Spinal Fluid (CSF) of Italian Multiple Sclerosis patients. Methods We genotyped 1115 Italian patients for HLA-DRB1*15 and HLA-A*02. In a subset of 925 patients we tested association with 52 non-HLA SNPs associated with MS susceptibility and we calculated a weighted Genetic Risk Score. Finally, we performed a Genome Wide Association Study (GWAS) with OCB status on a subset of 562 patients. The best associated SNPs of the Italian GWAS were replicated in silico in Scandinavian and Belgian populations, and meta-analyzed. Results HLA-DRB1*15 is associated with OCB+: p = 0.03, Odds Ratio (OR) = 1.6, 95% Confidence Limits (CL) = 1.1–2.4. None of the 52 non-HLA MS susceptibility loci was associated with OCB, except one SNP (rs2546890) near IL12B gene (OR: 1.45; 1.09–1.92). The weighted Genetic Risk Score mean was significantly (p = 0.0008) higher in OCB+ (7.668) than in OCB− (7.412) patients. After meta-analysis on the three datasets (Italian, Scandinavian and Belgian) for the best associated signals resulted from the Italian GWAS, the strongest signal was a SNP (rs9320598) on chromosome 6q (p = 9.4×10−7) outside the HLA region (65 Mb). Discussion genetic factors predispose to the development of OCB.


Introduction
The presence of oligoclonal bands (OCB) in the Cerebro Spinal Fluid (CSF) is a distinctive hallmark of Multiple Sclerosis (MS), found in 48 to 100% of patients in European populations [1]. The presence of a genetic influence on the OCB phenotype is suggested by its association, in several populations, with HLA-DRB1*15, which is also the strongest genetic susceptibility factor for the development of the disease. The Odds Ratio (OR) for OCB positivity ranges from 1.6 [2] to 3.4 [3], not dissimilarly from the OR observed for disease susceptibility. In spite of this indication and of all the efforts undertaken in the study of MS genetic susceptibility, the genetic association with the OCB phenotype was not extensively investigated so far. We aimed to explore the association between OCB and other genetic markers in Italian MS patients, both with a classical unbiased, genome-wide strategy and with a special attention for known MS susceptibility markers. This approach was worth attempting given the preliminary data on HLA-DRB1*15. We utilized an Italian population as a discovery sample for genome-wide study, and replicated the new findings in independent European cohorts from Scandinavia and Belgium.

Ethics Statement
The PROGEMUS study was approved by the Ethical Committee of the AOU Maggiore della Carità, Novara (Coordinating Centre) and those of the  [4], were ascertained by two Italian consortia, PROGEMUS (N = 1002), and PROGRESSO (N = 345). Patients born in Sardinia, of non-Italian ancestry, or refusing permission to DNA testing were not ascertained. A lumbar puncture was performed in 1115 patients (83% of the original cohort), 731 women and 384 men. Excluded patients (N = 232) were not statistically different from those included with regards to gender, clinical form (bout onset vs. primary progressive), age at onset, and MSSS [5] (data not shown). OCB positivity (OCB+) was defined as two or more bands present in CSF but absent in plasma at the same point in time. When two CSF analyses were performed, we considered the result of the last one. In Italy, OCB were assessed in laboratories pertaining to the network of Italian certified Neuroimmunology Laboratories and analyzed by isoelectric focusing and immunodetection of IgG. In Belgium, Norway and Sweden (apart from Stockholm hospitals)  CSF analysis was performed with agar electrophoresis before 2000-2 and with isoelectric focusing thereafter. In the Stockholm area, CSF analysis was done according to Olsson et al. [6]. All the samples and the clinical information were collected after written informed consent. The clinical characteristics of the Italian, Belgian, and Scandinavian MS patient samples are described in Table 1.

Genetic analysis a) overview
HLA-DRB1*15 and HLA-A*02 genotyping was performed in 1115 Italian patients, as previously described [7]. DRB1*03 genotypes were imputed by HLA*IMP software [8]. We tested 925 patients (814 OCB+, 111 OCB2) for association with the 52 non-HLA SNPs, shown to be associated with MS susceptibility in the recent large IMSGC-WTCCC2 genome wide association study (GWAS) [9]. In the same Italian group we calculated a weighted Genetic Risk Score (wGRS), as described by De Jager [10], using the ORs [9] for the 52 non-HLA MS susceptibility variants, and three HLA alleles (DRB1*15, DRB1*03 and A*02). Finally, we performed a GWAS with OCB status on the subset of Italian patients from the IMSGC-WTCCC GWAS study [9]. This sample included 562 (513 OCB+ and 49 OCB2) patients, genotyped with the Illumina 660-Quad platform. After quality control (QC) filtering [9], there were 504967 SNPs remaining. The genomic inflation factor was 1.009. The same (QC) filtering [9] has been applied to the Scandinavian and Belgium datasets. The association of each of the above markers with OCB status was evaluated comparing the allele frequency in OCB+ vs. OCB2 patients. All GWAS statistical analyses (QC, association and metaanalysis) were performed with PLINK software.    (Table S1) using data from Scandinavia and Belgium populations. (b) Meta-analysis for the rs9320598 SNP, the strongest signal after the meta-analysis of the combined results of all datasets (Italy, Scandinavia and Belgium). The forest plot summarizes the results obtained for the discovery Italian dataset, for the replication datasets from Scandinavia and Belgium, and the combined analysis calculated using the random and fixed-effect method. doi:10.1371/journal.pone.0064408.g002

d) GWAS analysis: Association and Meta-Analysis
Basic association testing was performed with the command line: plink -assoc -ci 0.95-reference-allele ,file.txt.. To select the covariates to be used in a logistic analysis, we checked the following potential confounders or interactors in the discovery dataset from the Italian population: gender, age of onset and MS clinical subtype (Primary Progressive vs. Bout Onset) for association with OCB status, using basic Chi-square test (for gender and clinical subtype) or Student's T test (for age of onset). None of these factors was differentially distributed in OCB+ vs. OCB2 patients, therefore none of these variables were selected as covariates.
Population stratification was evaluated both with the Genomic Control calculation (l), and by means of the quantile-quantile distribution. The lwas evaluated with PLINK, by adding -adjust to the basic association command line. The quantile-quantile distribution was visually analysed by Q-Q plot, generated by the R software ( Figure S1) The genomic inflation factor (l) was very low (1.009), indicative of a reduced population stratification. This is confirmed also by the quantile-quantile distribution ( Figure S1). On these bases, and on the bases of the previous analysis on covariates, we decided not to add Principal Components in our model, and therefore not to implement a logistic analysis. The Manhattan plot was created with the software Haploview, providing the.assoc file.
Genetic and clinical data for Scandinavian and Belgian patients were analyzed with a similar pipeline.
Meta-analysis was performed with PLINK software, basic option (string: plink -meta-analysis), that uses, for all datasets, the following information: SNP ID, chromosome, position, alleles, OR, standard error of OR and p-value. We calculated ORs and pvalues using both fixed and random effects models (reported in Table S1). Forest plot was instead created with R software, library ''meta''.

Results
Our results confirm that HLA DRB1*15 is positively associated with OCB+ (p = 0.03 OR = 1.6; 95%CL = 1.1-2.4); 313 patients were DRB1*15+/OCB+, 35 DRB1*15+/OCB2, 652 DRB1*152/OCB+, and 115 DRB1*152/OCB2. None of the 52 non-HLA MS susceptibility loci (Table S2) was associated with OCB status, except one SNP (rs2546890) near IL12B gene, with the MS risk allele conferring an OR of 1.45 (95% CL 1.09-1.92) for OCB positivity (p = 0.010). However, the combined effect of these markers plus three HLA MS risk alleles (DRB1*15, DRB1*03 and A*02) on OCB status (evaluated constructing a wGRS) was significant. In fact, the wGRS mean was significantly (p = 0.0008) higher in OCB+ (7.668) than in OCB2 (7.412) patients ( Figure 1) even after removing DRB1*15 (p = 0.0096). After removing the three HLA MS risk alleles, a similar trend, although non significant (p = 0.06), was still observed (wGRS mean: 6.784 in OCB+ vs. 6.686 in OCB2). Finally, to search for new associations, not yet correlated with disease susceptibility, we performed a GWAS with OCB status using the PROGEMUS/PROGRESSO subset of the IMSGC-WTCCC GWAS [9] as a discovery sample (Figure 2) to perform a GWAS with OCB status. Replication of the best signals in the Italian dataset was performed in silico using data from the previously published GWAS [9] from independent populations from Scandinavia (Norwegian-Danish-Swedish) (1367 OCB+ and 161 OCB2) and Belgium (317 OCB+ and 39 OCB2). To select SNPs for replication, an arbitrary significance threshold (p-value ,10 24 ) was chosen as a balance between sensitivity and specificity in order to ensure the detection of genuine associations that may not reach GWAS threshold in our sample set. A total of 89 SNPs were selected (Table S1). After the meta-analysis of the combined datasets (Italy, Scandinavia and Belgium), the strongest signal was a non-HLA SNP (rs9320598) on chromosome 6 (p = 9.4610 27 ) ( Figure 2, Table S1) outside the HLA region (65 Mb).

Discussion
The percentage of OCB2 patients (12%) in our Italian sample was in the range of that found in South Europe and OCB status was clearly associated with carriage of DRB1*15 as in other studies [1]. The intrathecal synthesis of immunoglobulins, as marked by OCB status, is observed in the preclinical phase of MS and the possibility of conversion from negative to positive is low [1], suggesting that there is a genetic trait predisposing to its development. Three findings from our data support this hypothesis: 1) DRB1*15 is associated with OCB+; 2) the combination of non-HLA markers associated with MS susceptibility also showed a trend for association with the presence of OCB as suggested by our wGRS, although most of its predictive power comes from DRB1*15; 3) after meta-analysis of the Italian best associated markers with two independent data-sets, we found a non-HLA SNP on chromosome 6 that almost reached GWAS significance. This SNP maps at 300 kb telomeric to C6orf167/MMS22L, a gene involved in the regulation of genome stability and DNA repair [11], and at 400 kb telomeric to KLHL32/KIAA1900, a gene with still uncharacterized function. Since this association did not reach genome-wide significance, further independent studies are needed to confirm this finding. Although it may be premature to assume that this locus is associated with OCB status, this and the other findings in our study suggest that genetic factors may be involved in the development of OCB. OCB status has been associated in turn with MS prognosis [12]; thus our findings could provide some clues to clarify the genetic predictors of disease prognosis.
One possible bias of our study is the selection of MS patients that underwent lumbar puncture; however their percentage was high and the clinical characteristics of those with CSF examination were similar to those without. We cannot exclude that the group of OCB negative patients represent a peculiar type of MS or even not MS cases. However, the clinical similarity of OCB+ and OCB2 patients argues against this [13].
In conclusion, our data support the hypothesis that HLA genes and possibly other non-HLA genes predispose to the development of OCB. Figure S1 Quartile-Quartile (QQ) plot for the GWAS analysis in the Italian population.