Directed evolution to re-adapt a co-evolved network within an enzyme

Abstract

We have previously used targeted active-site saturation mutagenesis to identify a number of transketolase single mutants that improved activity towards either glycolaldehyde (GA), or the non-natural substrate propionaldehyde (PA). Here, all attempts to recombine the singles into double mutants led to unexpected losses of specific activity towards both substrates. A typical trade-off occurred between soluble expression levels and specific activity for all single mutants, but many double mutants decreased both properties more severely suggesting a critical loss of protein stability or native folding. Statistical coupling analysis (SCA) of a large multiple sequence alignment revealed a network of nine co-evolved residues that affected all but one double mutant. Such networks maintain important functional properties such as activity, specificity, folding, stability, and solubility and may be rapidly disrupted by introducing one or more non-naturally occurring mutations. To identify variants of this network that would accept and improve upon our best D469 mutants for activity towards PA, we created a library of random single, double and triple mutants across seven of the co-evolved residues, combining our D469 variants with only naturally occurring mutations at the remaining sites. A triple mutant cluster at D469, E498 and R520 was found to behave synergistically for the specific activity towards PA. Protein expression was severely reduced by E498D and improved by R520Q, yet variants containing both mutations led to improved specific activity and enzyme expression, but with loss of solubility and the formation of inclusion bodies. D469S and R520Q combined synergistically to improve k(cat) 20-fold for PA, more than for any previous transketolase mutant. R520Q also doubled the specific activity of the previously identified D469T to create our most active transketolase mutant to date. Our results show that recombining active-site mutants obtained by saturation mutagenesis can rapidly destabilise critical networks of co-evolved residues, whereas beneficial single mutants can be retained and improved upon by randomly recombining them with natural variants at other positions in the network.

Fig. 1

Trade-off between soluble TK expression and the specific activity towards propionaldehyde for a series of previously identified single mutants (), and the initially created double mutants obtained by recombining the singles (□). Double and triple mutants obtained from the new library guided by SCA, and also for D469T/R520Q, are shown as triangles (▵) and interpolated with a dashed line.

Fig. 2

Statistical coupling analysis (SCA) of the PP- and Pyr-domains of E. coli transketolase. (A) One monomer of the TK homodimer is shown as grey ribbons. Six clusters (each coloured differently), form a single large network at the dimer interface, as revealed by SCA of the combined PP–Pyr domains. (B) Nine-residues form a single co-evolved network (red surface) within the Pyr-domain. The top view shows the network within the whole Pyr domain (grey ribbons). The bottom view is rotated for a clearer view of only the networked residues (red surface) and interconnecting sequence (grey ribbons). The TPP cofactor is shown in sticks. Images were produced with PyMOL (http://www.pymol.org) and the E. coli TK structure 1QGD.pdb (Littlechild et al., 1995). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of the article.)

Fig. 3

Specific activity and soluble expression of single, double and triple mutants of D469S, E498D and R520Q. (A) Specific activities towards propionaldehyde, relative to wild type for soluble enzyme in sonicated clarified lysates. (■) Experimentally determined. (□) Expected from the additive accumulation of improvements for the single mutants. (B) Impact of mutations on (□) total protein expression, () soluble fraction and (■) final purified enzyme concentrations.

Fig. 4

Comparison of k_cat, propionaldehyde K_m and k_cat/K_m for the double mutant cycle of D469S and R520Q, at 50 mM HPA, 50 mM Tris–HCl pH 7.0.

Fig. 5

Effect of recombining R520Q or R520V with previously identified mutants D469T and D469Y upon their specific activities. (■) Experimentally determined with sonicated clarified lysates. (□) Expected from the additive accumulation of improvements for the single mutants.