Unambiguous prediction of human integrin transmembrane heterodimer interactions using only homologous sequences

Part of the interaction between the alpha- and beta-subunits of integrins is known to take place at the transmembrane (TM) domain, where both heteromeric and homomeric aggregates have been reported in vivo and in vitro. In a recent computational study, totally independent from biochemical or biophysical data, we explored the plausibility of various TM homo-oligomers using evolutionary conservation data as a filter for non-native interactions. We showed that several homodimeric and homotrimeric interactions for alpha- and beta-chains are evolutionarily conserved. We report herein the results of the application of the same exhaustive approach to the integrin heterodimer. We have studied all known human TM integrin alphabeta pairs, and we show unambiguously that two models of interaction are evolutionarily conserved. These two models are consistent with those proposed previously based on mutagenesis and crosslinking. Comparison with previous experimental data strongly supports that a glycophorin A-like model is an intermediate form of interaction between the resting state and the active form, where chain separation occurs. Surprisingly, these two models are also conserved when considering most of the possible alphabeta pair combinations, suggesting that specific pairing of integrins is not determined by the TM domain, which has remained unchanged in spite of the variety of known integrin functions. This fact highlights a common ancestral mechanism for signal transduction that has remained through evolution. In a broader context, our results show that it is possible to obtain correct and detailed interactions of alpha-helical heterodimers with total independence of experimental data.