Cytochrome P450 enzymes are some of the most versatile enzymatic catalysts in nature. However, wild-type enzymes usually perform poorly with respect to activity, stability and selectivity toward non-natural substrates in industrial applications. These challenges can be solved by directed evolution. However, conventional combinatorial approaches such as DNA shuffling require high sequence homology in candidate genes, and lack a means to minimise disruption to protein structure.
The principal aim of the this study was to employ a protein structure-guided gene recombination (SGGR) to allow the combinatorial evolution of parental enzymes sharing low sequence identity to generate libraries with high protein integrity and diversity. Using available protein structures, the structural integrities of progeny proteins can be predicted by calculating essential interactions disrupted by the recombination events. Predicted optimal recombination sites with minimal disruption to the progeny structure will allow novel proteins to be generated by reusing pre-existing "building blocks" of parental structures.
Mosaic libraries were constructed by applying SGGR to CYP2D genes (CYP2D6, Cyp2d9 and Cyp2d22) and 1A genes (CYP1A1 and CYP1A2). Moreover, two CYP2D DNA shuffling libraries were created from these same three CYP2D parents and two expressible parents only (CYP2D6 and Cyp2d22) respectively, as controls for the CYP2D SGGR library. In the case of the CYP1A library, a CYP1A DNA shuffling library created from these same P450s previously in this laboratory was used as a control for the SGGR library.
Improvements were observed in both expression and activity preservation in the CYP2D SGGR library compared with the CYP2D DNA shuffling library; however the numbers of functional mutants are not sufficiently high to test whether there is a statistically significant improvement. According to data from the DNA shuffling library constructed with only CYP2D6 and Cyp2d22, it was found that Cyp2d9 showed a deleterious effect on hemoprotein expression and resulted in the low number of functional mutants.
The CYP1A SGGR library was created by an updated SGGR method. Fe(II).CO vs Fe(II)-difference spectroscopy revealed that 65% of the mutants from the SGGR library expressed significant hemoprotein with 46% showing higher expression than both parents, compared to 53% and 5% of mutants respectively in the CYP1A DNA shuffling library. Sequence analysis demonstrated superior genetic diversity and fidelity of the SGGR library: 8±2 crossovers and 1±1 spontaneous mutations were found per mutant (n=14) compared with 6Â±2 crossovers and 2Â±2 spontaneous mutations for the shuffled library. Catalytic diversity was assessed using the marker substrates: luciferins CEE, H, ME, and ME-EGE; 7-ethoxy- and 7-methoxyresorufin; indole and p-nitrophenol. Screening results demonstrate that SGGR can be used to artificially evolve P450s that combine a high level of structural integrity with an overall enhancement of catalytic activities. Moreover, structural modules have been identified to be associated with protein expression and particular protein properties.
The P450 redox partner, NADPH-dependent cytochrome P450 reductase (NPR), was also engineered in this study. P450 reductase has very high selectivity for NADPH with a strong bias against NADH. Due to the high cost of NADPH, a P450 reaction system that can use the cheap NADH as a cofactor is highly desirable.
To create a NADH-dependent P450 reductase, two engineering strategies were employed: rational domain recombination and semi-rational site-saturation mutagenesis. Structure analysis found that the FAD domain of NPR is structurally homologous to the FAD domain of cytochrome b5 reductase (b5R) containing a NADH binding domain instead. In the rational strategy, protein domains of NPR were exchanged with domains from b5R. Three chimeras were constructed: Chimera I was constructed by replacing the NPR FAD/NADPH binding domain with the b5R FAD/NADH binding domain; Chimera II was constructed by fusing the NPR FMN domain with the b5R FAD domain; Chimera III was constructed by replacing the NPR NADPH binding site with the b5R NADH binding site. Chimeras I and III showed up to a 300-fold increase in (kcat/KM)NADH/(kcat/KM)NADPH, while Chimera II revealed about a 14-fold decrease of KM towards NADH.
The second strategy was based on extensive site-saturation mutagenesis of essential residues in the nicotinamide-binding site, and essential residues in the 2'-phosphate binding site that stabilizes the NADPH bound complex. Two libraries were constructed by saturation mutagenesis of residues R597, K602 and W676 (conservative library) and S596-V603 and V675-S677 (extensive library) respectively. First tier screening of 24 randomly selected mutants from the extensive library using cytochrome c (cyt c) revealed that 20 mutants showed higher activity with NADH than NADPH at a single saturating coenzyme concentration. By contrast 3 of 24 randomly selected mutants from the conservative library showed higher turnover with NADH. The kinetic analysis of cyt c reduction revealed up to 1600-fold increases in (kcat/KM)NADH and 16000-fold increases in (kcat/KM)NADH/(kcat/KM)NADPH. However only two mutants showing enhanced NADH-dependent activity towards cyt c supported CYP2A6 activity towards the marker substrate coumarin. The best of these, mutant 1-003 (K602A and W676F), showed a 16-fold increase in (kcat/KM)NADH and a 24-fold increase in (kcat/KM)NADH/(kcat/KM)NADPH in cyt c reduction compared to wildtype. Further analysis of these mutants should reveal additional factors influencing coenzyme preference, and facilitate the further engineering of an efficient NADH-dependent NPR for biotechnological processes.