The structures of partially-folded states appearing during the folding of a (βα) 8 TIM barrel protein, the indole-3-glycerol phosphate synthase from S. solfataricus (sIGPS), was assessed by hydrogen exchange mass spectrometry (HX-MS) and Gō-model simulations. HX-MS analysis of the peptic peptides derived from the pulse-labeled product of the sub-millisecond folding reaction from the urea-denatured state revealed strong protection in the (βα) 4 region, modest protection in the neighboring (βα) 1-3 and (βα) 5 β 6 segments and no significant protection in the remaining N-and Cterminal segments. These results demonstrate that this species is not a collapsed form of the unfolded state under native-favoring conditions nor is it the native state formed via fast-track folding. However, the striking contrast of these results with the strong protection observed in the (βα) 2-5 β 6 region after 5 s of folding demonstrates that these species represent kinetically-distinct folding intermediates that are not identical as previously thought. A re-examination of the kinetic folding mechanism by chevron analysis of fluorescence data confirmed distinct roles for these two species: the burst-phase intermediate is predicted to be a misfolded, off-pathway intermediate while the subsequent 5 s intermediate corresponds to an on-pathway equilibrium intermediate. Comparison with the predictions using a C α Gō-model simulation of the kinetic folding reaction for sIGPS shows good agreement with the core of structure offering protection against exchange in the on-pathway intermediate (s). Because the native-centric Gō-model simulations do not explicitly include sequencespecific information, the simulation results support the hypothesis that the topology of TIM barrel proteins is a primary determinant of the folding free energy surface for the productive folding reaction. The early misfolding reaction must involve aspects of non-native structure not detected by the Gō-model simulation.

Kinetic simulations of the folding and unfolding of the mammalian TIM barrel protein aldolase were conducted to determine if a minimalist monomeric Gō model, using the native structure to determine attractive energies in the protein chain, could capture the experimentally determined folding pathway. The folding order, that is, the order in which different secondary structures fold, between the Gō model simulations and that from hydrogen-deuterium exchange experiments, did not agree. To explain this discrepancy, two alternate variant of the basic Gō model were simulated: (1) a monomer Gō model with native contact energies weighted by a statistical potential (SP model) and (2) a monomer Gō model with native contact energies inversely weighted by crystallographic B factors (B model). The B model demonstrated the best agreement between simulation and experiments. The success of the B model is attributed to the ability of B factors to highlight local energetic frustration in the aldolase structure which results in weaker native contacts in these frustrated regions. The predictive success of the B model also reveals the potential use of B factor information for energetic weighting in general protein modeling.