Cold Denaturation Unveiled: Molecular Mechanism of the Asymmetric Unfolding of Yeast Frataxin

What is the mechanism that determines the denaturation of proteins at low temperatures, which is, by now, recognized as a fundamental property of all proteins? We present experimental evidence that clarifies the role of specific interactions that favor the entrance of water into the hydrophobic core, a mechanism originally proposed by Privalov but never proved experimentally. By using a combination of molecular dynamics simulation, molecular biology, and biophysics, we identified a cluster of negatively charged residues that represents a preferential gate for the entrance of water molecules into the core. Even single-residue mutations in this cluster, from acidic to neutral residues, affect cold denaturation much more than heat denaturation, suppressing cold denaturation at temperatures above zero degrees. The molecular mechanism of the cold denaturation of yeast frataxin is intrinsically different from that of heat denaturation.

distances, where the columns are relative to the couples of atoms and the time dimension evolves with the rows. The legend contains the information about the couples of atoms.
The Euclidean distance evaluated for every combination of atoms and the results are stored in a standard XVG file. The list of atoms used as input to g_distmap is generated taking into account whether the amino acid residue is charged in at least one of the frataxin orthologs. Atoms selected to calculate relevant distances are the farthest side chain atoms least affected by internal rotation motions (Table S1). Table S1. Atoms used as labels for side chain position along MD trajectory Atoms are named according to PDB file notations; in the case of R, E and D atoms have been selected to remove ambiguities due to a) O1/O2 and b) NH1/NH2 exchanges.
For every couple of atoms, the Euclidean distance is analyzed along the time dimension and a plot is computed. An example of such data processing, leading to peaks of different shape for different neighbors of Ser126 in hFrata, is shown in Figure S1. The ratio of the area below the threshold distance and the total area of the peak is used to gauge the strength of the interaction in the network of charged residues. The network graph file was further generated by an ad hoc Python script and finally displayed using Gephi software. [3] Figure S1: Evolution of the distance between representative side chain atoms of three hFrata residues (pdb id 1ekg). Each histogram height is proportional to the number (#) of distances binned during the calculation of each frame of the 50 ns MD trajectory (sampling rate was of 1 ps). A) Evolution of the distance between Ser126 OG and Asp124 CG. B) Evolution of the distance between Ser126 OG and Lys135 NZ. Red arrows are positioned at 0.6 nm, the threshold distance used in graph analysis.
Parsing the distance map file. The distance map file is parsed using a custom Python 3 script. The algorithm evaluates the binned distances for every couple and uses a filter method to remove irrelevant interatomic distances from the analysis. In particular, only the couples of atoms which spend at least 1% of the time of the trajectory within 0.6 nm of distance are kept, the others are filtered out. A network graph is generated using this information and generating a weighted interaction map between residues. For every residue a node is generated inside the graph, and the edge weight between a couple of nodes is proportional to the time a couple of representative atoms stay below the threshold of 0.6 nm. In detail, given a plot where the x axis represents the binned distances and the y axis is the number of frames the binned interatomic distance has a certain value, the weight of the edge for a certain couple of atoms is proportional to the ratio between the area of the plot below 6Å and the total area.
The final GEXF output file for the Gephi software is generated using the NetworkX Python package. To obtain visually comparable images, a specific procedure was developed: one of the GEXF files was selected as the template in Gephi modifying the positions of the nodes according to the Force Atlas layout. This generated a new temporary GEXF file. The node position was then modified in the other GEXFs according to the frataxin alignment, obtained by Clustal Omega. [4] When a graph contained a node not available in the template, this was manually positioned and the process repeated using the new template.
Visual alignment of network graphs. To obtain visually comparable results, another Python 3 script has been developed. Indeed, using the software Gephi, it is possible to place a graph in a 2D plane in order to get clear results. The Python script allows to copy the coordinates of the nodes to other graphs, keeping sequence-related residues at the same position. A multiple alignment between the sequences of the Frataxin orthologs is initially performed using Clustal Omega. The result is analyzed by the script, which, using a sample PDB from the trajectory for every protein, performs a correlation between aligned atoms along different graph files. Using this information and the graph file generated by Gephi, the coordinates of the positioned nodes are copied to the nodes that still need to be correctly places. When a graph contains nodes that are not already placed because of an alignment gap between protein sequences, it is necessary to manually place these nodes and the script is run against the remaining graph files. Figure S2. Comparison of the graphs of the frataxins from yeast (Yfh1, panel A), bacteria (CyaY, panel B) and humans (hFrata, panelC). Residues are represented with their one letter symbol inside a circle and the numbering corresponding to the chosen pdb id, i.e 2fql, 1ew4 and 1ekg for Yfh1, CyaY and hFrata respectively. Only residues that are charged in at least one of the frataxin orthologs are shown in the maps. Circles corresponding to acidic residues (D, E) are colored in red, those of basic residues (K, R, H) are colored in blue. All remaining residues are inside blank circles. The color of edges reflects those of connected nodes. The yellow oval of Yfh1 comprises D101, E103 and E112, that of CyaY contains two negative residues (D31 and E33) and T42; which are flanked by a positive residue (K48). The residues of hFrata corresponding to the yellow oval of Yfh1 comprise one negative residue (D124), one neutral residue (S126) and one positive residue (K135).  To make the comparison easier, the spectrum of wild type Yfh1 shown in panel A was recorded in a solution containing 50 mM NaCl that shifts the equilibrium toward a completely folded protein. The spectra of mutants at low salt are very similar to that of the wt in medium ionic strength. The cross peaks of the wild-type protein [11c] are colored in red, those of the mutants are in black. All spectra were recorded at 25 °C.