A predominant role for the HLA class II region in the association of the MHC region with multiple sclerosis Matthew R Lincoln 1,2,10 , Alexandre Montpetit 3,4,10 , M Zameel Cader 1,2 , Janna Saarela 5 , David A Dyment 1,2 , Milvi Tiislar 3 , Vincent Ferretti 3 , Pentti J Tienari 6 , A Dessa Sadovnick 7 , Leena Peltonen 5,8,9 , George C Ebers 1,2 & Thomas J Hudson 3,4 Genetic susceptibility to multiple sclerosis is associated with genes of the major histocompatibility complex (MHC), particularly HLA-DRB1 and HLA-DQB1 (ref. 1). Both locus and allelic heterogeneity have been reported in this genomic region 2,3 . To clarify whether HLA-DRB1 itself, nearby genes in the region encoding the MHC or combinations of these loci underlie susceptibility to multiple sclerosis, we genotyped 1,185 Canadian and Finnish families with multiple sclerosis (n = 4,203 individuals) with a high-density SNP panel spanning the genes encoding the MHC and flanking genomic regions. Strong associations in Canadian and Finnish samples were observed with blocks in the HLA class II genomic region (P < 4.9 x 10 13 and P < 2.0 x 10 16 , respectively), but the strongest association was with HLA-DRB1 (P < 4.4 x 10 17 ). Conditioning on either HLA-DRB1 or the most significant HLA class II haplotype block found no additional block or SNP association independent of the HLA class II genomic region. This study therefore indicates that MHC-associated susceptibility to multiple sclerosis is determined by HLA class II alleles, their interactions and closely neighboring variants. The MHC genomic region, on chromosome 6p21, has proved difficult to dissect because of its intense and variable patterns of linkage disequilibrium (LD), which may vary between populations 4 . In addition to the established association with HLA-DRB1, other MHC loci may have independent association, including TNF 5 and other loci in the HLA class III 6 and HLA class I 7,8 genomic regions and loci telomeric of genes encoding the classical MHC 2,3 . We genotyped two large independent cohorts from Canada and Finland with multiple sclerosis with a dense panel of SNPs spanning the MHC genomic region. The LD maps for these populations (and 30 trios from the Centre d’Etude du Polymorphisme Humain (CEPH) included in the HapMap project) are relevant to this and other studies of the MHC genomic region (Supplementary Fig. 1 online). Because SNP selection was done when the HapMap Project was beginning, in each population, B60% of the SNPs gave a successful assay with a minimum minor allele frequency of 2%. Although this was not unusual given the lack of prior validation for most SNPs, we observed a 12% conversion rate for 17 consecutive SNPs in a 150-kb window at the DR-DQ interval (with many carefully designed SNP assays show- ing patterns suggestive of null alleles caused either by unknown polymorphisms in the genotyping oligonucleotides or by copy- number polymorphism). Outside this window, the coverage obtained with the current set of markers is estimated to be high: 78% of 1,573 HapMap SNPs (release 16; minor allele frequency 4 10%) that are located within 2 kb of a gene in a region of 413 Mb encompassing the MHC genomic region are correlated to one of the working assays in this data set with r 2 4 0.5 (Supplementary Fig. 2 online). Finally, the LD maps for Canadian, Finnish and CEPH families are markedly similar. Using a previously published definition 9 , we identified 111 haplotype blocks spanning the MHC genomic region in both the Canadian cohort (52 MHC, 25 telomeric and 34 centromeric) and the Finnish cohort (50 MHC, 26 telomeric and 35 centromeric). We first tested haplotype blocks for association in order to reduce redundancy and the number of comparisons. We used a conditional logistic-regression framework 10,11 to test for disease association. To maximize the power of our test, we modeled only the main effects of each locus, ignoring interaction terms at this stage. We assessed 111 haplotype blocks in the Canadian cohort; 13 of these blocks showed evidence for association with multiple sclerosis after Bonferroni correction (Table 1 and Supplementary Table 1 online). Five of these blocks were clustered in the region encoding HLA class III, one was on the boundary between the regions encoding HLA class I Received 23 March; accepted 18 August; published online 25 September 2005; doi:10.1038/ng1647 1 Department of Clinical Neurology, Radcliffe Infirmary, University of Oxford, Oxford OX2 6HE, UK. 2 Wellcome Trust Centre for Human Genetics, University of Oxford, Oxford OX3 7BN, UK. 3 McGill University and Genome Quebec Innovation Centre, Montreal, Quebec H3A 1A4, Canada. 4 Departments of Medicine and Human Genetics, Research Institute of the McGill University Health Centre, McGill University, Montreal, Quebec H3G 1A4, Canada. 5 Department of Molecular Medicine, National Public Health Institute, Helsinki, Finland. 6 Department of Neurology, University of Helsinki, Neuroscience Programme, Biomedicum-Helsinki, Haartmaninkatu 8, PL700, Helsinki, Finland. 7 Department of Medical Genetics and Faculty of Medicine (Division of Neurology), University of British Columbia, Vancouver, British Columbia V6T 2B5, Canada. 8 Department of Human Genetics, University of California, Los Angeles, Los Angeles, California, USA. 9 Department of Medical Genetics, University of Helsinki, Finland. 10 These authors contributed equally to this work. Correspondence should be addressed to G.C.E. (george.ebers@clneuro.ox.ac.uk). 1108 VOLUME 37 [ NUMBER 10 [ OCTOBER 2005 NATURE GENETICS LETTERS © 2005 Nature Publishing Group http://www.nature.com/naturegenetics