Unfolding and Refolding the Engrailed Homeodomain

Structures of EnHD showing the Native, Native′, Transition, and Denatured States from the T_m simulations

The engrailed homeodomain (EnHD) is of interest to our lab because it is an ultra-fast folding and unfolding protein and its folding/unfolding dynamics have been extensively characterized both experimentally and computationally. That it is ultra-fast folding is important because molecular dynamics (MD) can currently only probe protein motion on scales up to microseconds. The experimental folding data provide properties we can compare to in order to validate our simulations. Using the lab's extensive set of MD analysis tools, I am analyzing the folding and unfolding pathway of the protein under a variety of different conditions.

My first project in the lab was directly comparing the unfolding and refolding pathways of EnHD in simulations around the protein's melting temperature (T_m = 52 °C). I identified a Native′ state characterized by a small shift in helx III, which was populated between the Native and Transition States. In five of the seven transient unfolding events, the unfolding pathway was identical to the folding pathway based on RMSD and order of contacts gained/lost. In the other two, the pathways deviated, but the transition states were were still the same. The 13 transition state ensembles identified agreed well with previously identified transition state structures and experimental Φ-values. This study provided evidence for the hypothesis that protein folding obeys the principle of microscopic reversibility and that the ensemble of unfolding and refolding pathways is one and the same.

The full folding pathway of EnHD was observed in one of 46 simulations, totalling 14.9 μs, providing an atomic-level view of refolding. The native state was identified using a 35-property multidimensional property space reaction coordinate to compare to 1.2 μs of native state simulation data. The folding pathway mirrored the unfolding pathway from a high temperature unfolding simulation. In the 45 simuations where EnHD did not reach the native state, the protein was trapped in conformations stabilized by nonnative salt bridges. These traps provide a structural explaination for the accumulation of the previously identified folding intermediate.

I am continuing to compare the folding pathway of EnHD in these single-molecule simualations with "test tube" multi-molecule simulations. I plan to identify whether intermolecular interactions affect the un/refolding pathway.

Dynamics of Natural vs. Engineered Proteins

Structures of EnHD (left) and UVF (right) showing buried residues that were mutated from polar (cyan) to hydrophobic (green) to thermostabilize UVF

In 2007, Shah et al. [ref] designed two thermostabilized variants of EnHD. The design strategy employed their in-house algorithm to place only hydrophobic residues at buried positions in the protein and polar/charged residues on the surface. The resulting proteins, UVF and UMC had 22 and 24% identity to EnHD, respectively, and were stabilized up to T_m > 99 °C and ΔG_U = 2.3 kcal/mol. EnHD, on the other hand has T_m = 52 °C and ΔG_U = 1.8 kcal/mol. Using MD simulations, I discovered that the backbone of EnHD was more rigid than the thermostabilized proteins. In addition, UVF and UMC made many more nonnative contacts than EnHD while still maintaining a structure that satisfied the NOEs previously measured by NMR. While UVF and UMC were designed to maximized enthalpy upon folding, it seems that they also benefit from increased entropy.

Folding Pathways of a Pair of Proteins with 88% Identity but Different Topologies

G_A88 (left) and G_B88 (right) have 88% sequence identity but all-α and α/β folds, respectively. Their 7 differing residues are shown in sticks.

The Paracelsus Challenge [ref] was issued to encourage scientists to convert one protein fold to another by changing <50% of the original sequence. Alexander et al. [ref] designed such a pair of proteins based on the A and B domains of Protein G with only 7 of 56 residues differing, or 88% sequence identity. In collaboration with the Gianni Group, I performed unfolding MD simulations of the two proteins to investigate when in the folding pathway the proteins committed to their all-α or α/β native topology. The simulations indicated, in agreement with experimental results from the Gianni Group, that G_B88 had some loose native structure present in the denatured state stabilized by salt bridges in the β-hairpins that was lost at low pH. The denatured state of G_A88, on the other hand, did not contain these salt bridges, was uneffected by low pH, and instead had some fluctuating α-helix. Nucleation of native structure occured early in the denatured state to determine the final folded structures.

I am currently identifying mutations, with the aid of MD simualtions, to be tested experimentally in order to see whether the specific interactions in the denatured state that were predicted by MD in the previous study were correct.