Variation of the Myelin Oligodendrocyte Glycoprotein gene is not primarily associated with multiple sclerosis in the Sardinian population

Background Multiple sclerosis (MS) is consistently associated with particular HLA-DRB1-DQB1 haplotypes. However, existing evidence suggests that variation at these loci does not entirely explain association of the HLA region with the disease. The MOG locus is a prime positional and functional candidate for such additional predisposing effects but the analysis is complicated by the strong, albeit labyrinthine pattern of linkage disequilibrium in the region. Here we have assessed the association of MOG variation with MS in the Sardinian population to see if it represents an independent contributor to MS predisposition. Results After re-sequencing the MOG gene in 21 healthy parents of MS patients we detected 134 variants, 33 of which were novel. A set of 40 informative SNPs was then selected and assessed for disease association together with 1 intragenic microsatellite in an initial data set of 239 MS families. This microsatellite and 11 SNPs were found to be positively associated with MS, using the transmission disequilibrium test, and were followed up in an additional 158 families (total families analysed = 397). While in these 397 families, 8 markers showed significant association with MS, through conditional tests we determined that these MOG variants were not associated with MS independently of the main DRB1-DQB1 disease associations. Conclusion These results indicate that variation within the MOG gene is not an important independent determinant of MS-inherited risk in the Sardinian population.


Background
Multiple sclerosis (MS) is a serious chronic inflammatory and demyelinizating disorder of the central nervous system which results from an autoimmune attack on components of the oligodendrocyte cell [1]. The disease is more common in European and European-derived populations. Within Europe it shows a north-south gradient, with the notable exception of the Mediterranean island of Sardinia, which has one of the highest prevalences worldwide [2]. The disease risk for a monozygotic twin of an affected patient is about 30%, showing a quick fall-off rate with decreased genetic relatedness to affected individuals. These data, and the increasing incidence of disease reported in some populations over the last few decades [3,4], suggest that the chance of this inflammatory process occurring depends on the complex interplay between a polygenic trait and unknown environmental factors influencing the penetrance of susceptibility genes [5,6]. MS has been found to be consistently associated with specific HLA class II variants and notably with the HLA-DRB1*1501-DQB1*0602 haplotype, which represents the main risk factor for disease occurrence in different ethnic backgrounds [7][8][9][10][11]. The relative contribution of variation at the DRB1 and DQB1 loci to disease predisposition is still not completely clear although some studies based on cross-comparing rare HLA DRB1-DQB1 haplotype splits have indicated that the main contribution comes from variation at the DRB1 locus [11]. Analysis of large data sets from Finland and Canada have also suggested that in those populations, any important additional modifiers of MS susceptibility were likely to be contained in the regions close to DRB1 [12]. In Sardinia, the HLA-DRB1*1501-DQB1*0602 haplotype is rare but is still significantly positively associated with MS together with an additional 4 haplotypes including DRB1*1303-DQB1*0301, DRB1*0405-DQB1*0301, DRB1*0301-DQB1*0201 and DRB1*0405-DQB1*0302 [13]. Some of these haplotypes, such as DRB1*0301-DQB1*0201 are also associated with MS in some non-Sardinian populations [11,14,15] but not in others [8,9,[16][17][18][19][20]. The presence of HLA non-DRB1-DQB1 predisposing effects related to a different distribution of extended HLA-DRB1*0301-DQB1*0201 haplotypes in different populations can explain these findings. Indeed, some studies have suggested that within the HLA region there are further independent predisposing effects determined by as yet unidentified non-DRB1-DQB1 variants [13,[21][22][23][24]. However, strong linkage disequilibrium (LD) between the variants contained in the HLA region makes it difficult to detect which polymorphisms, outside the exon 2 sequences of the DR/DQ loci but within the HLA region, further influence disease risk. Within the HLA region, the MOG gene is a prime candidate for additional MS associations. This gene is located, 2.9 Mb telomeric of the DRB1 locus, in a chromosome interval which has shown some evidence of association with MS independent of DRB1-DQB1 [13]. Furthermore, the rodent ortholog of this gene encodes for an autoantigen which triggers autoimmune responses in experimental models of disease [25,26]. Moreover, in MS patients both T-cell and antibody responses against this protein have been detected [27,28] and an aetiologic role of anti-MOG antibodies has been suggested in acute lesions of MS patients [29], although this role is controversial [30,31]. These findings could be consistent with a model in which specific polymorphisms in the MOG gene could determine amino acid variation or differences in the level of expression of this protein in the central nervous system and affect immune responses against it, thus acting as primary aetiologic determinants of disease pathogenesis.
A few studies have tested the association between MOG polymorphisms and disease, with conflicting results. Some of these studies indicated a potential role of MOG [32,33] while others failed to find any evidence of primary association with MS [34,35,12]. However, none of these studies analysed the MOG gene comprehensively, nor assessed the association of genetic variation of this gene in the context of extended DRB1-DQB1-MOG haplotypes.
The aim of this study was to establish the SNP content of the MOG gene in the Sardinian population and use this information together with appropriate conditional association tests to assess the individual contribution of the MOG gene to MS inherited risk to determine if it is an important independent risk locus.

Results and discussion
In order to assess the association of MOG variation with MS we first established polymorphism content of the MOG gene in the Sardinian population by resequencing the gene in 21 healthy parents of MS patients. The primers used are listed in Table 1, and technical details are reported in the Methods section. We detected 134 SNPs, 101 of which had already been reported [36] and 33 were novel. A set of 40 informative SNPs was then selected to be tested for disease association (see Methods). These were initially genotyped in a data set of 239 MS families. To identify the haplotype-tagging patterns, we computed the intermarker D' and r2 values of these 40 informative SNPs, computed for 478 parents of the MS patients, shown in Figure 1. The physical positions, with base 1165 corresponding to the translation start site of MOG, and other relevant information are indicated in Table 2. D' values indicate that these SNPs are part of the same linkage disequilibrium block (average 0.97 between contiguous marker pairs) that, as we have previously shown, encompasses HLA class I and III sub-regions and also includes the HLA class II DR and DQ loci [37,13]. As expected, considering selected SNPs were chosen in order to avoid genetically redundant markers, the r2 values showed relatively low values with an average of 0.13 between contiguous marker pairs ( Figure 1).
We then assessed disease association of the 41 variants of interest, including these 40 SNPs of interest and the MOG51 microsatellite, in a sample set of 239 families using the extended transmission disequilibrium test (ETDT). Without taking linkage disequilibrium with variation at the HLA-DRB1-DQB1 loci into account, 11 SNPs -SNP07 (novel), SNP06 (novel), SNP05 (novel), rs2071652, rs2857767, SNP04 (novel), rs2535260, rs2535261, SNP03 (novel), rs2252734, rs29228 and the MOG51 microsatellite showed evidence of association at a significance level of 5% with MS (data not shown). We then evaluated MS association of these markers in an additional 158 MS families (total n = 397 families). In this larger sample set, 8 markers showed some evidence of association at 5% significance level ( Figure 2) with one haplotype defined by variants at rs2071652 (allele C), rs2857767 (allele C) and rs2535260 (allele C) showing the strongest disease association (P = 3.29 × 10 -6 ). This level of significance was slightly lower than that shown by association of HLA-DRB1 (P = 7.9 × 10 -7 ) and -DQB1 in the same sample set (P = 1.3 × 10 -7 ). The transmission data of these variants individually and in the context of the three locus DRB1-DQB1-MOG haplotypes are shown in the online appendix along with the D' values among MOG variants and the various DRB1-DQB1 haplotypes. The eight disease associated variants are essentially part of two main LD blocks formed, on the one hand by rs2071652, rs2857767 and rs2535260 and on the other hand by the other five variants. From these data it is evident that while SNP07, SNP06; SNP04, SNP03 are in positive LD only with the DRB1*0301-DQB1*0201 haplotype, SNPs rs2071652, rs2857767 and rs2535260 occur frequently (albeit at different degrees of positive LD from strong to modest) and show a significant association with more than one HLA-DRB1-DQB1 haplotype. The cooccurrence of these MOG variants with different DRB1-DQB1 haplotypes could just be a chance event or could be due to any of the various forces which impact LD [37]. Importantly, we also assessed the disease association of MOG alleles which individually showed significant evidence of association with MS together with the other DRB1-DQB1 haplotypes grouped (that is, all DRB1-DQB1 haplotypes which did not show any evidence of significant association with MS). These extended DRB1-DQB1-MOG haplotypes showed a consistent negative transmission from parents to affected patients (see Additional file 1). This suggested that MOG disease associations were most likely driven by DRB1-DQB1 variation.
Next, we wanted to directly assess whether these MOG associations were independent from DRB1-DQB1 associations. To this end, we analysed the overall effect of allelic and haplotype variation at the MOG gene using the conditional ETDT (CETDT) [38]. Using this conditional analysis, which takes into account linkage disequilibrium between overall MOG variation and overall variation at the HLA-DQB1 and -DRB1 loci, MOG associations were no longer significant. Conversely, the overall associations of DRB1-DQB1 haplotypes, taking into account the most associated MOG haplotype, gave a P value of 9.7 × 10 -3 .
We also evaluated the relative transmission of the MOG alleles and haplotypes according to which specific DRB1-DQA1-DQB1 haplotypes they were on ( Figure 3). In order to carry out these analyses we used the haplotype method (HM)-TDT (Methods) [39]. Using this test we initially evaluated association of specific variants of MOG, taking into account linkage disequilibrium with the HLA-DRB1-DQB1 haplotypes for which there is prior evidence of association with MS in the Sardinian population (i.e. DRB1*1303-DQB1*0301, DRB1*0405-DQB1*0301, DRB1*0301-DQB1*0201, DRB1*1501-DQB1*0602 and DRB1*0405-DQB1*0302 haplotypes) [13]. Again, allelic variation at the MOG locus did not significantly affect the transmission of these predisposing DRB1-DQB1 haplotypes ( Figure 3). Also the MOG haplotype defined by rs2071652, rs2857767 and rs2535260 did not produce significant evidence of primary association when analysed with the HM-TDT conditional on DRB1-DQB1 positively associated haplotypes (data not shown). Moreover, we also conducted the HM-TDT to assess whether these MOG variants were influencing association of the non-predisposing DRB1-DQB1 haplotypes grouped. No significant heterogeneity in the association of DRB1-DQB1 haplotype conditional on MOG was seen, again suggesting that the initial observations of association with MS risk were likely to be only a consequence of LD with the DR-DQ mediated MS risk. This is shown in Figure 3, Graph F.

Conclusion
From our results we can conclude that variation at the MOG gene does not provide a significant independent contribution to the inherited risk for MS in the Sardinian population. The data suggests that the protein product of MOG is unlikely to represent the primary autoantigen in human MS.

Study participants
In the study we considered a sample set of 397 independent MS families, which included 391 trios families (one affected child and both healthy parents), 4 multiplex families (more than one affected sibling and their healthy parents) and 2 vertical families (one parent and one affected offspring), all being at least third generation Sardinian. This data set of 397 MS families is a proper subset of the 490 Sardinian MS families described previously [13]. All patients met the criteria of definite MS [33]. The average age of onset for MS was 28.9 years (range 6 -51 years), and the female/male ratio was 2.3. The disease course was relapsing-remitting in 65%, secondary progressive in 27% and primary progressive in 8% of patients. All families had previously been analysed at the DRB1-DQA1-DQB1 loci [13].

Power calculations
To examine the feasibility of verifying whether variants in the MOG gene, or in strong LD with it could explain the association peak that was previously observed in Sardinia [13], we focused on the parent-affected child trios that we had already studied during the initial examination of the role of the HLA in MS susceptibility. We postulated that our search would be for a fairly common variant (in line with the frequency of the predisposing DRB1-DQB1 hap-lotypes) with a maximal relative risk of 1.9 and using a disease prevalence of 0.0014 (corresponding to the prevalence in Sardinia). The strength of LD between the DRB1-DQB1 locus and MOG was estimated at a mean value of r 2 = 0.3 as previously shown [13]. Purcell and Sham power calculations for our 391 trios gave for an alpha value of 0.10 a power of 0.83 [44]. It should be noted that for the conditional analyses, in the case of conditioning on the absence of high-risk HLA haplotypes, the trio size is 121 and the corresponding power is reduced to 0.42 for the same alpha criteria.

MOG polymorphism analysis
We resequenced the MOG gene in 21 parents of MS patients heterozygous for the 226 bp allele at the MOG51 microsatellite locus, which was found to be significantly associated with MS [13]. A set of 40 informative SNPs was selected and dot-blot genotyped in the family sample set, using the set of primers listed in Table 1 and the SSO probes reported in Table  2. Selection of these SNPs was based on the patterns of LD (in order to provide good map coverage and avoid genotyping and testing genetically redundant variants), on the position in the gene (in order to maximise the chance of testing functionally relevant variants) and on the minor allele frequency (MAF) (in order to exclude particularly rare variants for which we did not have sufficient statistical power in the subsequent association stage).
In addition to these SNPs, in this manuscript we also assessed the association of MS with 1 intragenic microsatellite, MOG51 [34], whose primer sequences have been previously reported [13].

Statistical analysis
Single-point disease association of individual MOG variants as well as HLA-DRB1, DQB1 alleles was evaluated using ETDT [45]. This test takes into account the transmission or non-transmission of alleles of a locus relative to those present on the other parental chromosome. In the case of multiallelic markers, the ETDT takes multiple alle-  [13]. Association is presented as the negative log of the P-value for the extended transmission disequilibrium test (ETDT). The bold horizontal line corresponds to the threshold significance of 0.05 For all analyses performed in this manuscript, haplotypes were established following the co-segregation of alleles within families and using computer programmes written by F. Dudbridge and B. Koeleman. Only haplotypes showing definitive parental genotype data are considered in the analyses shown in this paper. In those families with more than one affected patient, only the probands were evaluated.

Competing interests
The author(s) declare that they have no competing interests.