Possible interactions between regions of allelic polymorphism in the alpha- and beta-chains of class II MHC molecules were examined by measuring the efficiency of surface expression and the reactivity with mAb of wild-type and recombinant A alpha A beta-chain pairs from the b, d, and k haplotypes. These studies revealed regions of polymorphism within the alpha- and beta-chains that interact with complementary regions in the other chain. Unexpectedly, almost all the variable segments of both the class II MHC alpha- and beta-chains either directly contributed to or were near sites of interchain interactions. The exception was the beta HV3 (hypervariable (HV] segment (residues 61-71), which appeared to neither participate in nor be affected by interchain interactions. This division of the MHC molecule into interacting vs independent regions of allelic structural variation suggests that mutagenesis experiments involving the beta HV3 segment can be analyzed in a straightforward manner, as such mutations appear unlikely to alter the conformation of other molecular segments. Furthermore, functions attributed to the beta HV3 segment either experimentally or by population analysis should have a high probability of transfer by beta HV3 exchange (either experimentally or evolutionarily), because epitopes assigned to this region of the molecule are not affected by sequences outside this segment. This is of special importance because of the apparent involvement of this region in defining a potential site of interaction with antigenic peptides and TCR. In contrast, almost all other variable segments of the MHC molecule appear to have the capacity to contribute to interactions involving at least one other variable segment. This suggests not only that the experimental analysis of the contributions of these regions to various functions requires a consideration of inter- and intrachain interaction, but also that the transfer of function by genetic exchange of these structurally dependent regions is unpredictable. Selection must therefore operate on these interacting HV segments in the context of the complete alpha beta heterodimer. These results support our earlier arguments for cis-co-evolution of alpha- and beta-chain polymorphism and the absence of selection for F1 (hybrid) class II molecules. Finally, asymmetries observed in the contributions of particular pairs of HV segments to the efficient expression of Ia alpha beta heterodimers provide a basis for understanding mechanistically how cis-co-evolution may have occurred.