今天推送的文章发表在ACS Catalysis的“Enhancing the Imine Reductase Activity of a Promiscuous Glucose Dehydrogenase for Scalable Manufacturing of a Chiral Neprilysin Inhibitor Precursor”，通讯作者为美国Codexis公司的Xiang Yi。
Glucose dehydrogenase (GDH) has been widely used to recover NAD(P)H from NAD(P)+ while oxidizing d-glucose to gluconolactone. As a prototype member of the short-chain dehydrogenase/reductase (SDR) family, GDH is considered a highly substrate-specific and monofunctional enzyme. Recently, GDH (Rd1bb enzyme) has been found to be a catalyst for the asymmetric reduction of cyclic imide salts. However, sequence alignment and structural comparisons of the Rd1bb enzyme and known IRED showed little similarity. The understanding of the promiscuous imide reductase activity of GDH is apparently limited, and the reactivity to substrates other than cyclic imine salts has not been studied so far.
In the present study, the authors report the imide reductase activity of GDH that was discovered incidentally. In a recent enzymatic evolution program in which amination of α-ketoacid2a was reduced with (S)-alanine ethyl ester (1a) to obtain intermediate 3a in the synthesis of neperisine inhibitors, a drug candidate for the treatment of refractory hypertension (Scheme 1), it was observed that the GDH enzyme could not only catalyze the recycling of NAD(P)H, but could also exhibit IRED-like activity in its own right, acting as a catalyst in reductive amination reactions. Conversion of GDH enzymes to IRED using state-of-the-art CodeEvolver directed evolution technology. The final enzyme catalyzes the desired reductive amination reaction and achieves the predetermined efficiency and selectivity goals.
1
The iminoreductase activity of GDH Rd1bb was found
The authors performed an initial enzymatic evolution procedure for the synthesis of the desired product 3a from the unnative substrates α-ketoacid 2a and (S)-alanine ethyl ester (1a) on Codexis IRED mutants derived from sterol dehydrogenases (ODHs). The enzymatic reaction is coupled to a GDH cofactor recovery system using glucose as the terminal reducing agent to achieve a high conversion rate of the IRED reaction (Scheme 1). GDH Rd1bb, an engineered GDH from Bacillus subtilis with significantly improved thermal stability and glucose dehydrogenase activity, is used as a cofactor recovery system. In the absence of IRED enzymes, trace amounts of product signal were detected by LC-MS analysis. The Rd1bb-dependent reaction was further tested by testing two additional cofactor recovery systems, phosphite dehydrogenase (PDH) and sodium phosphite or NADH alone. No product signal was detected in the reaction without Rd1bb, confirming that Rd1bb was indeed catalyzing the reductive amination reaction. Finally, the dose-response of Rd1bb is measured in a reaction that does not contain the initial IRED enzyme, confirming that Rd1bb shows IRED-like reactivity.
2
GDH Evolution Round 1: Evolution of GDH responsiveness to IRED
The promiscuous imide reductase activity of Rd1bb exhibits a low but exclusive preference for the desired (S,S)-diastereomer, compared to IRED-catalyzed reactions, which exhibit a strong preference for undesirable (S,R)-products. Therefore, increasing activity while maintaining the high diastereoselectivity of Rd1bb is the goal of the evolutionary program. To maximize screening speed, all libraries were screened using a highly efficient RapidFire MS method that was selective for the desired product 3a in each round of evolution. In order to ensure the desired diastereoselectivity of product 3a, the resulting mutant activity was further verified by LC-MS.
The authors designed and constructed a saturation mutagenesis (SM) library that targeted 33 locations around the glucose-binding pocket (Figure 1), yielding approximately 400 mutants with 37% coverage. Screening was performed in the presence of 5 equivalent amine 1a using PDH and sodium phosphite as a stand-alone NADH recirculation system, or at concentrations of 6 and 20 g/L of NAD+ and glucose in the absence of an additional NADH circulation system at a substrate 2a concentration of 1.25 g/L. The latter screening setup assumes that some of the newly created GDH mutants retain their ability to oxidize glucose while acting simultaneously with IRED. Both screening strategies have been successful. During the high-throughput screening (HTP) process, it was found that many hits exceeded the initial Rd1bb enzyme.
Most of the mutant residues of beneficial mutations are located within glucose-binding pockets, such as L95A or V, E96A or C, H147I, S, a or Q, F155N or a and F256A, V, T, S, Q or L, or in close contact with NAD+, such as A159C or T (Figure 1). Notably, F256 from the C-terminal ring of adjacent monomers contributes to the increase in activity. Multiple beneficial mutations at this location highlight its critical impact on IRED activity. Typically, mutations to small and/or flexible residues, which may increase the size of the binding pocket, accommodating larger ketone and amine substrates. The results of the first round of evolution proved that the imide reductase activity of GDH can be increased. Although the total activity is still low, multiple mutants show significant improvement in activity, maintaining >99% diastereoselectivity for the desired (S,S)-diastereomer. The H147I mutant with excellent performance was selected as the template for the second round of mutations.
3
GDH Evolution Rounds 2-4: Inhibition of Ketone Reduction
During the initial screening process, the reduction of the α-ketoacid 2a substrate exceeded the required imide reductase activity, resulting in the formation of a large number of alcohol byproducts4 (Figure 2A). There are three sources of alcohol formation: GDH enzymes, unknown background reactivity of lysates, and direct chemical reduction of NADH. In the validation phase of the second round of trials, the risk of NADH non-catalytic reduction was reduced by reducing the concentration of NAD+ cofactor from 1.5 g/L to 0.1 g/L, which is suitable for large-scale production. The amount of alcohol by-products of the identified hotspot mutant Rd3bb was reduced from 50% to approximately 20%, and this mutant was selected as the Round 3 mutation template (Figure 2A). To compensate for the reduced cofactor concentration and maintain sufficiently high cofactor recovery, Rd1bb and glucose replaced PDH and sodium phosphite as recovery systems from round 3 onwards due to the higher turnover.
The goal of screening in rounds 3 and 4 is to simultaneously increase activity and decrease alcohol formation to guide library design and hit selection. By analyzing the HTP screening data by plotting a fold increase in product activity versus alcohol by-products, hits with excellent chemoselectivity can be easily identified, which is manifested by an increase in the product-to-alcohol ratio accompanied by an increase in activity (Figure 2B,C). Sequence activity and sequence chemoselectivity analysis was further performed by selecting the top hit (Figure 2B, above the dashed line) by direct comparison with the template sequence (hits for mutant libraries) or MOSAIC (hits for combination libraries), a software used to deconvolute and predict the effects of mutations on desired IRED activity and chemoselectivity, respectively. Through this work, mutations that favor increased activity and improved chemoselectivity were identified (Figure 2D, red) and further used to construct the next round of combinatorial libraries. Rounds 3 and 4 had the highest hits of Rd4bb and Rd5bb, respectively, with further inhibition of alcohol by Rd4bb as low as 10%, and complete elimination of alcohol by Rd5bb, while product activity continued to rise and dominate (Figure 2A). From round 4 onwards, the formation of alcohol 4 as a by-product was no longer detected in the screening and validation reactions.
Notably, the mutation identified in Rd5bb (I147R) is a reversal of the key mutations (Rd1bb and H147I) in the first round of evolution that abolishes the glucose dehydrogenase activity of GDH. His147, together with Tyr158 and Ser145, is thought to constitute a catalytic ternary structure responsible for facilitating hydride transfer from the C1 atom of glucose to the nicotinamide moiety of NAD+ through a relay mechanism. It is speculated that the residue at position 147 also plays a key role in the catalytic ketoreductase function of GDH. The I147R mutation results in no ketone reductase activity of GDH against substrate 2a and simultaneously increases its imide reductase activity against product 3a. Therefore, any library variant carrying this mutation is unlikely to produce alcohol by-products in rounds 5-8 of screening4.
4
GDH Evolution Rounds 5-8: Further improvements for large-scale manufacturing
While significant progress has been made in the required activity and chemoselectivity in the five rounds of evolution, the imide reductase activity of the evolved GDH enzyme is not sufficient for large-scale production of product 3a. At 30°C, 50 g/L α-ketoate 2a, 5-fold molar excess of (S)-alanine ethyl ester (1a), 0.1 g/L NAD+, 1 g/L Rd1bb, and 35 g/L d-glucose, only about 40% conversion of Rd5bb under 5 g/L enzyme load was achieved within 15 hours. Therefore, under typical manufacturing conditions, especially at higher concentrations of substrate 2a and at elevated temperatures to reduce excess co-substrate amine 1a, the authors underwent further evolution to optimize performance, resulting in higher activity and reaction rates (Table 1). The IRED activity of the evolved GDH enzyme continued to increase until the last round. Over the course of eight rounds of evolution, 20 libraries and more than 16,000 mutants were screened, targeting 99% of all positions and 60% of all mutations. In each round of evolution, new diversity is introduced by site-saturation mutagenesis libraries by targeting specific regions of the enzyme, and beneficial diversity identified by HTP screening is recombined into the combinatorial libraries of the next round. The hit rates of combinatorial libraries are often shown to outperform site-mutagenic libraries, demonstrating that the accumulation of proven beneficial diversity provides incremental improvements in a combinatorial manner.
During the evolutionary process, GDH became more hydrophilic, and the expression and solubility of GDH also increased, especially during the 4th and 5th rounds of evolution, when the library began to target the surface position. The final mutant, Rd9bb, carries 29 mutations (Figure 3A) and increases cumulative activity by approximately 6.3×106 compared to the initial Rd1bb enzyme (Figure 3B).
The authors further investigated the effect of key process parameters on the imide reductase activity of GDH mutants. These include pH, cofactor preference of NAD+ over NADP+, stoichiometry between α-ketoacid 2a and (S)-alanine ethyl ester (1a), thermal stability of the enzyme, and reaction temperature. Based on these results, early process development with Rd6bb was carried out using (S)-alanine ethyl ester (1a) as the corresponding hydrochloride. Under the optimal conditions, the complete transformation to the target product 3A was achieved within 3 hours. Product 3a, which is a single stereoisomer, is provided by HPLC with a separation yield of 94% and a purity of >99.9, with an ER of >99.9:0.1 and a DR of >99.9-0.1.
5
Evolution of glucose dehydrogenase and regulation of mixed IRED activity
For any confounding enzyme, the side reaction, i.e., imide reductase activity in the case of GDH, is slow relative to the natural catalytic activity of the main reaction pathway. This study shows that applying appropriate mutagenesis and screening strategies, it is possible to transform a previously confounding activity into a future primary activity. With the exception of the initial Rd1bb mutant, the evolutionary template did not show any measurable dehydrogenase activity when incubated with glucose as a substrate and NAD+ as a cofactor (Figure 4A). This observation inferred that GDH lost its dehydrogenase activity during the first round of evolution in which the glucose-binding site became the target. Although most mutations at the glucose binding site did not have a beneficial effect on the enzyme's IRED capacity, changes in the three residues E96, H147, and F155 in the glucose binding pocket appeared to have an effect, with a 2-4-fold increase in IRED activity (Figure 4B).
In addition, another method to identify key residues for transition from GDH to IRED is based on the function of the GDH recovery cofactor NAD+. IRED mutants E96, H147, and F155, which exhibit low IRED activity in the absence of a cofactor recovery system but good activity in the presence of a cofactor circulation system, lose their native GDH activity, and H147A or H147S mutations play a critical role in eliminating dehydrogenase activity and promoting IRED activity. F155N does not affect the oxidation of glucose by Rd1bb, probably because the location is far from the catalytic site of glucose oxidation (Figure 1). In addition, mutant F155N may reduce steric hindrance and/or provide more charge interactions, leading to increased regulation of imine intermediates by enzymes and concomitant increased IRED activity.
The authors resolved the crystal structure of the NAD+-bound Rd1bb enzyme and used it as a template to model the H147S and H147I mutants. Docking simulations of glucose were performed using SMINA to further investigate the structural and functional roles of these key residues. In the Rd1bb model, glucose adopts the conformation of hydrogen bonding of each hydroxyl group with adjacent amino acids including Tyr158, Glu96, Lys199, Asn196, His147, and Ser145. Hydrogen bonds form between H147 and the C6 hydroxyl group of glucose, and His147 itself also forms π–π interactions with Trp152, potentially further stabilizing the orientation of His147 (Figure 5A). As a result, glucose is directed to the position of reaction with NAD+. His147, along with Tyr158 and Ser145, has been reported to act as a catalytic triad, initiating a catalytic relay mechanism and ultimately leading to the transfer of hydride from the C1 atom of glucose to the nicotinamide moiety of NAD+, which is located below glucose in the active site. When H147 mutates to serine (Figure 5B) or isoleucine (Figure 5C), hydrogen bonds are lost and the π–π interaction with Trp152 disappears, which both weakens the interaction with glucose and may lead to the loss of catalytic function. H147S shows complete elimination of dehydrogenase activity, while H147I still acts as a dehydrogenase, albeit to a much lower extent than Rd1bb. The interfacial scores of glucose and protein were calculated to quantify hydrophobic interactions. The interface score of the H147S mutant increased by 1.43 Rosetta energy units (REU) relative to Rd1bb, suggesting a decrease in hydrophobic interactions, while the interface score of glucose with the H147I mutant was almost the same as that of the wild type (−0.15 REU), suggesting that the hydrophobic contact of isoleucine may have contributed to the placement of glucose in the production location and thus partially maintained its dehydrogenase activity.
To further understand the changes at the structural level due to evolution, the X-ray crystal structure of the Rd9bb enzyme was also determined with NAD+ binding. The superposition of Rd1bb with Rd9bb shows only a small difference in the coordinates of the overall structure (rmsd of 0.43 Å). Except for mutations between the two sequences, where there is a clear difference in electron density, there is little change in the active sites of the two proteins. A slight shift of the helix consisting of residues 194 to 201 in the mutants (I195G, K199R, F200W) may contribute to the larger substrate that accommodates the Rd9bb enzyme. Proteins have a weak C-terminal density, so it is not clear how these proteins interact with substrates. Since the structure of Rd9bb bound to product 3a was not obtained, the authors performed a docking simulation of glucose with Rd1bb and gave the predicted hottest poses using the SMINA program (Figures 1 and 5). The docked glucose molecule may form multiple hydrogen bonds with various residues in the active site of Rd1bb. In this pose, the C1 atom of glucose is located at a distance of 3.5 Å from the nicotinamide ring and has the potential to donate hydride to NAD+.