This panel shows an array of bi-variate histograms for each possible pair of the top clique residues, even those that are not neighbors. Residue couplings are not explicit on this figure, which instead shows the initial data on joint Ginsenoside-Ro sequence identity used in the MI calculations. However, by selecting neighboring positions the bivariate histograms can be used to find which residue types are found at each position, and their frequencies. Also shown in Fig. 2B is a contact map for the top clique residues, which highlights the physical proximity of many of top clique residues. Some general observations can be made from Fig. 2B. For example, the top clique residues involve covariation between two or three different residue types. In some cases, the coevolving residue types that occur at a given Chloroquine Phosphate position are quite similar. Others coevolve as very different residue types, such as positions 374 and 430. One trend that might be expected from this data is that the physicochemical characteristics of the residue pair would be conserved at co-varying 
positions, despite changes in residue identity. However, such patterns are not readily apparent for these residues. This may be due to the nature of the structural interactions made by the top clique residues, which are predominantly between atoms in the side chain of one residue and the backbone of the other. The top clique residues fall on both sides of this interface, including residues from domains 3 and 4 of the protein. Indeed, these eight residues form an essentially contiguous residue patch, which spans the width of the domain interface. The remaining residues are found in domains 1, 2, and elsewhere in domain 3, generally near the center of the molecule, but not within the active site cleft. With the exception of K285, which makes ligand contacts in certain enzyme-substrate complexes, none of the top clique residues were of previously implicated functional significance in PMM/PGM. However, the domain 4 interface is a known site of conformational change of the protein, as first observed in crystallographic studies of enzymesubstrate complexes. For comparison with the results of the coevolutionary analysis, Fig. 3C shows the structure of P. aeruginosa PMM/PGM colored according to sequence conservation in the MSA. On this figure, it is clear that the most highly conserved residues cluster in the center of the molecule, in or near the active site. In particular, clusters of conserved residues are found in active site loops in domains 1 and 2 of the protein. These loops include residues involved in the phosphoryl transfer reaction and coordination of the Mg2+ required for enzyme activity. Only a few highly conserved residues are found near the domain 4 interface. Hence, the structural location of highly conserved residues and coevolving residues in PMM/PGM are distinct, although if considered as a group, they would tend to form a contiguous patch in/near the active site, and extending along the domain 4 interface. To investigate the functional role of the coevolving network, a number of the top clique residues were selected for site-directed mutagenesis to alanine. Proline and glycine residues were excluded from consideration, due to possible negative effects on protein folding. Of those remaining, residues were selected based on: i) their location at the interface between domain 4 and the rest of the protein; and, ii) a side chain involved in a hydrogen bond or ion pair interaction with another residue across the interface.