An obvious way to unravel the apparently complex association between human leukocyte antigen (HLA) and rheumatoid arthritis (RA) is to reduce the heterogeneity of this multifactorial disease. Recently we have discovered that shared epitope (SE)-positive HLA-DRB1 alleles are exclusively associated with a subgroup of RA patients that test positive for auto-antibodies against cyclic citrullinated peptides. Further studies suggested that SE-positive alleles are classical immune response genes for the development of these antibodies. On the basis of these and other data we formulated a two-hit model for the pathogenesis of RA which incorporates a novel "citrullinated" SE hypothesis. About 5 years ago Zanelli et al. reported that HLA-DR6 (*1301 and *1302) and some other DR alleles that share the DERAA-sequence on amino acids 70-74 of their third hypervariable region are associated with protection from (severe) RA. Recently we corroborated these data in a large prospective cohort study and demonstrated that protection was observed both in the presence and in the absence of a SE susceptibility allele on the other haplotype. Finally we review the state of the art of the association of noninherited maternal HLA antigens with both susceptibility to and protection from RA.