Multiple sclerosis (MS) susceptibility demonstrates a complex pattern of inheritance. Haplotypes containing HLA-DRB1*1501 carry most of the genetic risk. Epidemiological evidence implicating epigenetic factors includes complex distortion of disease transmission seen in aunt/uncle-niece/nephew (AUNN) pairs. Unexpectedly, in AUNN families we found that allele frequencies for HLA-DRB1*1501 were different between the first and second generations affected. Affected aunts had significantly lower HLA-DRB1*15 frequency compared with their affected nieces (chi(2) = 9.90, P = 0.0016), whereas HLA-DRB1*15 frequency in affected males remains unaltered across the two generations (chi(2) = 0.23, P = 0.63). We compared transmissions for the HLA-DRB1*15 allele using a family-based transmission disequilibrium test approach in 1690 individuals from 350 affected sibling pair (ASP) families and 960 individuals from 187 AUNN families. Transmissions differed between the ASP and the AUNN families (chi(2) = 6.92; P = 0.0085). The risk carried by HLA-DRB1*15 was increased in families with affected second-degree relatives (AUNN: OR = 4.07) when compared with those consisting only first-degree relatives (ASP: OR = 2.17), establishing heterogeneity of risk among HLA-DRB1*15 haplotypes based on whether collateral parental relatives are affected. These observations strongly implicate gene-environment interactions in susceptibility and more specifically, that epigenetic modifications differentiate among human leukocyte antigen class II risk haplotypes and are involved in the determination of the gender bias in MS. These data strongly suggest that the female-specific increasing risk of MS is mediated through these alleles or adjacent variation. The comparison of transmission of the same allele in vertically affected pedigrees (AUNN) to collinear sibling pairs (ASP) may provide a useful screen for putative epigenetic marks.