Abstract
Voltage-gated cation channels are crucial membrane proteins responsible for the electrical activity in excitable nerve, muscle, and cardiac tissue. These channels respond to changes in the membrane potential via conformational changes in their voltage-sensing domains (VSDs) that lead to the opening and closing of the ion conduction pore. Since alternative states of the VSDs are difficult to capture via experimental methods, we investigated the application of AlphaFold2 and subsampling of its multiple sequence alignment input to computationally predict structures across a range of intermediate and endpoint states. By generating 600 models for 32 members of the voltage-gated cation channel superfamily, we show that AlphaFold2 is capable of predicting diverse structures of the VSDs that could represent activated, deactivated, and intermediate conformations with more diversity seen for some VSD families compared with others. Modeling the full sequence of pseudo-tetrameric channels also produced a range of heterogeneous states in the pore and intracellular regions representative of local conformational changes and key secondary structural transitions. However, we observe that the global conformational coupling is limited across models, as different functional domains adopt physiologically incompatible states. Although short molecular dynamics simulations of a subset of the models suggest they are structurally plausible conformations, there are some incongruities between certain generated models and resolved cryo-EM structures. Further validation is required to confirm their structural and functional relevance.