Hinge-shift mechanism as a protein design principle for the evolution of β-lactamases from substrate promiscuity to specificity

TEM-1 β-lactamase degrades β-lactam antibiotics with a strong preference for penicillins. Sequence reconstruction studies indicate that it evolved from ancestral enzymes that degraded a variety of β-lactam antibiotics with moderate efficiency. This generalist to specialist conversion involved more than 100 mutational changes, but conserved fold and catalytic residues, suggesting a role for dynamics in enzyme evolution. Here, we develop a conformational dynamics computational approach to rationally mold a protein flexibility profile on the basis of a hinge-shift mechanism. By deliberately weighting and altering the conformational dynamics of a putative Precambrian β-lactamase, we engineer enzyme specificity that mimics the modern TEM-1 β-lactamase with only 21 amino acid replacements. Our conformational dynamics design thus re-enacts the evolutionary process and provides a rational allosteric approach for manipulating function while conserving the enzyme active site.


Fig. S1
Hinge location in GNCA and TEM-1 β-lactamases are shown using %DFI scores in a their color-coded cartoon representations. The residue positions colored red are highly flexible whereas those colored blue are least flexible, i.e., rigid. The top and bottom panel shows the same protein with different perspective. The residue positions shown as spheres are some of the residues who have retained their rigidity through evolution. On the other hand, we have highlighted residue positions which have undergone hinge-shifts through evolution. We observe a number of residues around the regions with retained rigidity where a shift in their rigidity is observed. This implies a rewiring of the dynamical network of interactions in the protein. The pairwise dynamic coupling of the selected common and sequentially non-conserved hinges comprising (A) X mutation set in GNCA and TEM-1 with other non-common and sequentially non-conserved hinges in GNCA and TEM-1, and (B) Y mutation set in GNCA and TEM-1 with other non-common and sequentially non-conserved hinges in GNCA and TEM-1. The hinge residues selected for the two sets are strongly coupled (%DCI > 0.8) to other such non-common and non-conserved hinge residues in both GNCA and TEM-1. Calculated data are provided as a Source Data dci_profiles file.

Fig. S3
Selection criterion for the substitutions in set Y. The pairwise dynamic coupling of all the common and sequentially non-conserved (NC) hinges comprising with other non-common and sequentially non-conserved (NN) hinges in GNCA β-lactamase is shown. Common and sequentially nonconserved hinges which exhibit a higher coupling (%DCI > 0.8) with other non-common and sequentially non-conserved hinges are selected for substitutions in set Y. For this analysis, we selected such residues from each region in the protein. Therefore, for example, out of residues 262, 263 and 265, only 262 and 263 are selected due to their high coupling with the maximum number of NN hinges; out of 244, 245 and 246, only 244 and 246 are selected, and similarly for others. However, 182 is the only NC hinge in its vicinity which showed higher coupling with few of NN hinges (44 and 260), as a result we included it in the set Y. Calculated data are provided as a Source Data dci_profiles file.  modelling all mutations at all the NN hinges and then obtained its DFI profile. Comparison of the DFI profile of GNCA-AllNN with the that of wildtypes GNCA and TEM-1 β-lactamases is shown through the PCA analysis based on their lowest two principle components (a) and through the differences in in their flexibility profiles per residue positions, shown as a plot (b). We observe that mutating all the NN hinges significantly alter the dynamics of the GNCA mutant. Particularly comparing the DFI profiles of each position reveals that mutating all of the NN hinges (broken black line) impacts dynamics of the regions around the catalytic site 70, 166 and 234, suggesting that these substitutions could be damaging to the function. On the other hand, the impact of the mutation from set X, which contains only a select number of non-common and non-common hinges, on dynamics in these regions also impact the dynamics but not so severe (black solid line). This analysis also suggests that substitution of NN sites alone is not enough to modulate function. Calculated data are provided as a Source Data dfi_profiles file.

Fig. S6
The effect of mutations from set Y with (GNCA-Y) and without the mutation T182M (GNCA-Y_wo_182). It can be observed by clustering their DFI profiles with the wildtypes GNCA and TEM-1 β-lactamases (a) and also by comparing the differences in their DFI profiles per position (B). From (A), we observe that the mutations from set Y without the T182M mutation shifts the dynamics of the GNCA mutant farther away from those of wildtypes. This also provides a strong evidence about the importance of T182M mutation in its dynamics. This can also be observed from (b) where we observe that the variant (broken black line) exhibits a very different dynamics particularly in several regions 78-110, around residue 200 and 240-260 when T182M mutations is not present. On the other hand, the impact of the mutation set Y on dynamics in these regions is not so severe when we incorporate the T182M mutation (black solid line). Calculated data are provided as a Source Data dfi_profiles file.

Fig. S7 Selection criterion for the substitutions in set Z. (a)
The coupling of residues selected for substitutions in set Z with the non-common and sequentially non-conserved hinges. The positions that exhibit higher coupling (%DCI > 0.8) with the non-common and sequentially non-conserved hinges are selected for mutations. Apart from this, these residues also have a medium flexibility (0.3 < %DFI < 0.5) and are distally located from the active site (> 8 Å). (b) In addition, the residues in set Z are also coupled to the active site as shown by the cartoon representation of GNCA β-lactamase where each residue is color coded based on corresponding dynamic coupling with the active site ( %DCI score). Red colored residues are the highest coupled (%DCI=1) and white are the least coupled (%DCI = 0). The residues in set Z are shown as sticks. We can observe that these also exhibit high dynamic coupling with the active site (%DCI > 0.8). Calculated data are provided as a Source Data dci_profiles and dfi_profiles file. performing mutations only closer to the catalytic site were derogatory for the function as both of them have made mutants ineffective for degradation of both the antibiotic benzyl-penicillin and cefotaxime (see Table 1).

Fig. S9
Comparison of the percentile ranking of the average DFI profile obtained using covariance matrices calculated from time windows of different sizes (25ns, 50ns, 75ns and 100ns). These time windows are obtained by sampling the trajectory through a moving window which is shifted by 25ns. If the %DFI profile obtained from different window sizes give a consensus result for the low flexibility region as well as a very low discrepancy of high flexibility region, then the covariance matrices are considered to be sampling from the converged dynamics of a single well. Here we have shown an example of %DFI profiles sampled from different windows from a 1000ns simulation of GNCA-XYZ. The observed high consensus between their DFI profiles give us the confidence that simulation is converged, and the obtained DFI profile reflects the flexibility profile from a well equilibrated single energy well.  Table 1. Calculated data are provided as a Source Data raw_rates file.