bHLH-PAS transcription factors (TFs) orchestrate circadian timing, hypoxia responses, and neuronal development. Yet, how PAS domains encode partner specificity through their intrinsic dynamics remains unclear. Here, we present a comparative dynamics analysis across human bHLH-PAS heterodimers (CLOCK:BMAL1/NPAS2:BMAL1 and HIF/ARNT families, among others) using an all-atom elastic network model (ENM) normal-mode analysis with an additional projection-based method that decomposes self-coupled (domain internal) from directly coupled (partner influenced) motions. We validated ENM-derived self-coupled motions against those derived from explicit-solvent MD (1 μs CLOCK:BMAL1; 200 ns HIF-2α:ARNT), observing strong correspondence. Across complexes, self-coupled motions are more conserved than sequence or static structure and segregate PAS-A from PAS-B domains. Directly coupled motions cleanly divide complexes into CLOCK:BMAL1-type versus HIF:ARNT-type, which are not captured by sequence/structure alone. PAS-B domains are generally less flexible than PAS-A, yet show heightened, context-dependent flexibility at high-propensity interface regions, indicating asymmetric dynamic coupling across partners. A “crossover” HIF-2α:BMAL1 model further illustrates that component identity shapes domain-domain coupled dynamics, influencing the heterodimer-type. Finally, residue substitutions which perturb CLOCK:BMAL1 dimerization map to peaks in directly coupled RMSF, linking dynamic coupling to dimer stability. Together, these results suggest that PAS domains adapt to different binding partners through their conserved intrinsic dynamics, allowing complex-specific coupled motions. Our results pave the way for mapping residue-level, long-range couplings that govern PAS-domain heterodimerization, clarifying unresolved complexes, and aid predictions of promiscuous protein-protein interfaces.
Sudarsanam et al. (Sun,) studied this question.