1. Introduction
Amino sugars are constituents of various biomacromolecules in primary metabolism such as glucosamine for the biosynthesis of structural polysaccharides. Additionally, they serve as precursors to microbial secondary metabolism, contributing to the synthesis of antibiotics including macrolides, anthracyclines, glycopeptides, and polyenes [
1], [
2]. Extensive research has been performed toward their synthesis owing to the utility of amino sugars and their wide-ranging biological activities; however, challenges abound owing to the inherent low stereoselectivity of chemical synthesis [
3], [
4]. Sugar aminotransferases (SATs) are enzymes that are involved in the synthesis of amino sugars and belong to the pyridoxal 5′-phosphate (PLP)-dependent aspartate aminotransferase type I (AAT-I) superfamily [
5]. These enzymes catalyze the installation of amino groups on a keto-sugar substrate with high stereoselectivity [
6]. SATs are becoming invaluable tools for amino sugar synthesis in pharmaceutical and chemical synthesis [
7], [
8], [
9].
Research on various SATs has deepened our understanding of their biological functions and how amino sugars are used in natural product biosynthesis [
2]. Additionally, structural characterizations of numerous SATs from different bacteria (related to different natural product biosynthetic pathways) have been reported [
10], [
11]. These include WecE from
Escherichia coli (
E. coli) [
12], DesV from
Streptomyces venezuelae [
13], and RbmB from
Streptomyces ribosidificus [
14], which are representative of the enzyme subgroups acting on nucleoside diphosphates (NDP)-4-keto sugars, NDP-3-keto sugars, and
scyllo-inosose, respectively.
Valienamine is a representative pseudo-sugar and is a functional component in various bioactive pseudo-oligosaccharides such as the antidiabetic drug, acarbose, and crop protectant, validamycin A [
15], [
16]. Valienamine has a glucose-analogous structure and is acknowledged for its excellent glycosidase inhibitory properties and as a functional precursor for new drug discovery [
17]. The unique multi-chiral structure implies that chemical methods are not suitable for valienamine synthesis owing to low stereoselectivity [
18]. Currently, valienamine is produced via a multi-step semi-biosynthetic process, in which validamycin A is initially generated by
Streptomyces hygroscopicus 5008 and subsequently degraded by
Flavobacterium saccharophilum or other bacteria [
19], [
20]. This complex process involves the intricate pathways of two different bacteria and is inefficient, forming many byproducts, owing to the poor specificity of the C−N−C cleavage site [
19], [
20]. However, a suitable chiral precursor for valienamine is valienone, an intermediate of validamycin A [
21].
Previously, we explored the transamination activity of WecE, which catalyzes amino group installation on TDP-4-keto-6-deoxy-
D-glucose (TDP-Glc4O) in
E. coli [
12], from its natural substrate TDP-Glc4O to the non-natural substrate valienone to produce valienamine with excellent stereospecificity (enantiomeric excess value > 99%); this reaction was accomplished via a single simple transamination reaction (
Fig. 1) [
7]. Similar to most potential candidate enzymes [
22], WecE displayed inherent drawbacks, including low thermostability (a poor half-life (
t1/2) of 2.07 min at 40 °C) and insufficient activity toward the non-natural substrate valienone, which limited the catalytic efficiency of this promising approach.
The catalytic activity of enzymes may be improved by the combinatorial active-site saturation test (CAST) and iterative saturation mutagenesis (ISM) strategy, which creates mutagenesis libraries for directed evolution at active site residues [
23]. CAST-ISM has been applied to a few ω-aminotransferase subfamily enzymes, for example, Novick et al. [
24] performed evolution at residues around the substrate in the active site of an amine transaminase, ATA-217 from
Vibrio fluvialis JS17, substantially improving its activity toward a chiral sacubitril precursor. Furthermore, Jia et al. [
25] subjected the enzymatic residues surrounding a large catalysis pocket of an (
R)-selective transaminase,
Capronia epimyces TA, to saturation and combinatorial mutagenesis, and obtained a positive mutant with notable improvements in both activity and thermostability. To the best of our knowledge, no reports exist on stability-activity engineering of SATs using CAST or other methods.
In this study, we investigated a customized CAST-ISM approach aiming to identify SAT WecE mutants that exhibit both enhanced stability and activity toward the non-natural substrate valienone. Initially, potential stability-activity hotspots in the WecE active site were targeted to improve the efficiency of direct evolution. Thereafter, the CAST library was constructed, based on multiple assessments including the evolutionary conservation and relative folding energy of mutations. Briefly, 14 hotspots of 57 residues in the active site of WecE, related to improving the stability-activity, were identified and targeted, generating mutants with improved stability and activity that outperformed the wild-type (WT) WecE. Conceptually, our study illustrates a rational and generalizable scheme to explore the hotspots to improve the stability and activity toward SATs. This approach demonstrates promise for the future development of similar engineering strategies aimed at improving enzyme performance in biocatalysis and synthetic biology applications.
2. Materials and methods
2.1. General materials
The pET28a expression vector and competent cells of E. coli BL21 (DE3) were acquired from Novagen (Germany) and Vazyme Biotech Co., Ltd. (China), respectively. Gene amplification and mutant library construction were performed using pfu DNA polymerase supplied by Takara Bio Inc. (China). Valienone was synthesized by WuXi AppTec Company (China). Valienamine standard was purchased from Shanghai Haixiang Pharmaceutical Development Ltd. (China). All other chemicals were of reagent grade or higher and were procured from Sigma-Aldrich Corporation (USA).
2.2. Hotspot identification
2.2.1. Evolutionary conservation analysis
The HHblits algorithm [
26] was used to search for homologs within the UniRef30 database [
27], resulting in a total of 94 680 entries. These entries were subsequently aligned using the Clustal Omega online tool, which is supported by EMBL-EBI [
28]. The phylogenetic tree was constructed using MEGA X software (v10.0.2) [
29], following which the evolutionary conservation analysis was conducted using the ConSurf Server [
30] with the best fit amino acid substitution model. Conservation scores were calculated using the Bayesian method, generating a multiple-sequence alignment comprising 2105 homologous sequences. Each residue in the alignment was assigned a conservation score ranging from 1 to 9 with higher scores indicating greater conservation. The interaction analysis for the LLP internal aldimine intermediate in WecE from Protein Data Bank (PDB ID: 4PIW) was conducted using the ligand-protein interaction analysis tool LigPlot
+ [
31].
2.2.2. Stability effect assessment
The saturation mutations at the 39 variable residues in the active site of WecE were assessed for stability effect. The relative free energy of folding energies of each substitution (ΔΔ
G = Δ
Gsubstitution−Δ
GWT) was calculated by the FoldX algorithm that uses an empirical effective force field (
$\text{ }\!\!\Delta\!\!\text{ }\!\!\Delta\!\!\text{ }{{G}_{\text{FoldX}}}$) and a backbone-based folding program ABACUS2 using statistical energy functions (
$\text{ }\!\!\Delta\!\!\text{ }\!\!\Delta\!\!\text{ }{{G}_{\text{ABACUS}2}}$) via the integrated parallel DDGScan workflow [
32]. The
$\text{ }\!\!\Delta\!\!\text{ }\!\!\Delta\!\!\text{ }{{G}_{\text{FoldX}}}$ and
$\text{ }\!\!\Delta\!\!\text{ }\!\!\Delta\!\!\text{ }{{G}_{\text{ABACUS}2}}$ values of the saturation substitutions were min−max normalized [
33] and averaged using SPSS Statistics software v26.0 (IBM, USA). The
Z-score, a statistical measure that quantifies the number of standard deviations a data point is from the mean of a dataset [
33], for the average mutation folding energies of the saturation substitutions at each site (
$\overline{\text{ }\!\!\Delta\!\!\text{ }\!\!\Delta\!\!\text{ }{{G}_{\text{FoldX}}}}$ and
$\overline{\text{ }\!\!\Delta\!\!\text{ }\!\!\Delta\!\!\text{ }{{G}_{\text{ABACUS}2}}}$) was calculated to compare with the average of the dataset using SPSS® Statistics software v26.0.
2.3. Enzyme evolution and mutant characterization
2.3.1. Site-directed mutagenesis and screening
The gene encoding WecE was amplified from the genome of
E. coli MG1655 using polymerase chain reaction (PCR) and ligated into the pET28a expression vector using
NheI/
XhoI restriction enzymes to serve as the originating evolution template. The site-saturation mutagenesis (SSM) libraries were constructed using NNK degenerate primers (Table S1 in Appendix A) through whole-plasmid PCR [
34]. The transformants were cultured and screened in 96-well microplates using a high-throughput method based on a sugar aminotransferase (SAT) and
L-glutamate dehydrogenase (
L-GDH)-coupled system (Scheme S1 in Appendix A) [
35], [
36]. Following the incubation of cell-free extracts at 37 °C for 30 min, the target transamination reaction containing 1 mmol∙L
−1 valienone, 5 mmol∙L
−1 glutamate (Glu), 50 μmol∙L
−1 PLP in 20 mmol∙L
−1 potassium phosphate buffer (pH 7.5) was coupled with an indicator reaction (0.5 unit (U)
L-GDH, 0.5 mmol∙L
−1 nicotinamide adenine dinucleotide, reduced form (NADH), and 5 mmol∙L
−1 NH
4Cl) within the 96-well plates. The indicator reaction was initiated after the transamination reaction was performed for 2 h. The consumption of NADH was monitored by the decrease in absorbance at 340 nm as the reporter signal, using a multiscan spectrum microplate spectrophotometer (SpectraMax®M5 Multimode Plate Reader; Molecular Devices, USA).
2.3.2. Protein expression and purification
The recombinant strains of WT WecE and the mutants were cultured in LB medium (1% tryptone, 0.5% yeast extract, 1% NaCl, and 50 μg∙mL−1 kanamycin) at 37 °C with shaking (220 r∙min−1) until the optical density at 600 nm (OD600nm) reached 0.5. Protein expression was induced by adding 0.5 mmol∙L−1 isopropyl-β-D-thiogalactopyranoside (IPTG) and incubating for 12 h at 37 °C. The cells were collected through centrifugation (4 °C, 15 min, 12 000 r∙min−1) and lysed by sonication in 20 mmol∙L−1 potassium phosphate buffer (pH 7.5). The soluble proteins were isolated using Ni2+-chelating affinity chromatography and eluted with 200 mmol∙L−1 imidazole in 20 mmol∙L−1 potassium phosphate buffer (pH 7.5). The expression and purification of proteins were confirmed by sodium dodecyl-sulfate polyacrylamide gel electrophoresis (SDS-PAGE), and the protein concentration was quantified using a NanoDrop Lite Spectrophotometer (Thermo Fisher Scientific Inc., USA) based on ultraviolet (UV) absorption.
2.3.3. Activity and kinetic assay
The transamination activity of WecE and its mutants toward the non-natural substrate valienone was determined in a reaction mixture containing 10 mmol∙L
−1 valienone, 40 mmol∙L
−1 L-glutamine (Gln), 50 μmol∙L
−1 PLP, and 1 mg∙mL
−1 of the purified enzyme in a 20 mmol∙L
−1 potassium phosphate buffer (pH 7.5). A control mixture that excluded the enzyme was also prepared. The reactions were conducted at 37 °C for 2 h and were terminated using an equal volume of methanol, followed by centrifugation of the mixture at 12 000 r∙min
−1 for 10 min to detect the product. The transamination product of the reaction, that is, valienamine, was derivatized with
o-phthalaldehyde (OPA) in borate buffer (pH 9.0) at 25 °C for 30 s. The derivatized product was then separated and quantified on an Eclipse XDB-C18 column (5 μm, 4.6 mm × 150.0 mm) using 22% acetonitrile, which was monitored at an emission wavelength of 445 nm and an excitation wavelength of 340 nm using a fluorescence detector on an Agilent 1200 Infinity LC System [
37].
Kinetic constants of both WT WecE and the mutants toward valienone were determined at a fixed concentration of the amino donor in a 100-μL reaction mixture (0.85 mg∙mL
−1 purified enzyme, 5 mmol∙L
−1 L-Glu, and 5 μmol∙L
−1 PLP) using NADH-dependent
L-GDH as a coupling enzyme (0.5 U
L-GDH, 0.5 mmol∙L
−1 NADH, and 5 mmol∙L
−1 NH
4Cl) at pH 7.5 and 25 °C in 96-well plates [
35]. Initial velocities were determined with the valienone concentration gradient from 0.01 to 1.50 mmol∙L
−1. The oxidation of NADH was continuously monitored at 340 nm. One unit was defined as the amount of enzyme capable of reducing 1 μmol of NADH per minute [
12], and the kinetic parameters
KM and
kcat, which represent the affinity of the enzyme for its substrate and the catalytic efficiency of the enzyme, respectively, were determined by performing nonlinear regression analysis of the Michaelis-Menten equation using Origin 2021 software (OriginLab, USA).
2.3.4. Stability characterization
The melting temperatures (
Tm) of WecE and the mutants M3 and M4 were determined using a differential scanning fluorimetry (DSF) method [
38]. Aliquots (20 μL) of each mixture, containing 0.2 mg∙mL
−1 enzymes and 1 μL 1× SYPRO Ruby orange dye in a final concentration, were pre-incubated at 4 °C for 20 min. The initial temperature was set at 25 °C and increased to 85 °C at a rate of 2 °C∙min
−1. Fluorescence was detected using real-time PCR with an excitation wavelength of 490 nm and emission wavelength of 580 nm. The
Tm values were obtained by determining the inflection point of a fit of the fluorescence signal to a Boltzmann sigmoidal fit using Origin 2021 software (OriginLab).
The
T5015 value represents the temperature at which enzyme activity is reduced to 50% after a 15-min heat treatment [
39]. To determine this value, the purified protein was heated by incubating the enzyme in 0.2-mL PCR tubes using a programmable thermal cycler for precise temperature control in the range of 35 to 50 °C. Enzymes (1 mg∙mL
−1 in 20 mm potassium phosphate buffer, pH 7.5) were heated at different temperatures for 15 min and cooled at 4 °C for 10 min. The residual activities were assayed after 2-h incubation at 37 °C in the reaction mixture. The
T5015 value was obtained by determining the inflection point of the residual activities at specific temperatures using a Boltzmann sigmoidal fit with Origin 2021 software (OriginLab).
The
t1/2 value refers to the half-life of enzyme inactivation [
40]. To determine the
t1/2 value, both WT WecE and the detected mutants M3 and M4 (2 mg∙mL
−1) were incubated at 40 °C for various time intervals ranging from 0 to 20 min and then cooled on ice for 10 min. The residual catalytic activities of the enzymes were assayed at 37 °C, as described previously in Section 2.3.3. The first-order rate constants (
kd) were determined by linear regression of ln (residual activity) versus the incubation time (
t) using Origin 2021 software (OriginLab).
t1/2 at 40 °C was obtained using the following equation:
t1/2 = ln2/
kd [
40].
2.3.5. Preparative-scale assay
A 50 mg preparative-scale enzymatic synthesis of valienamine was conducted using purified mutant M3, M4, or WT WecE at 10 mg∙mL
−1, 20 mmol∙L
−1 valienone (52.6 mg), 80 mmol∙L
−1 L-Gln, and 100 μmol∙L
−1 PLP in a 15-mL reaction system. The reaction was performed at 37 °C and 50 r∙min
−1 in 20 mmol∙L
−1 potassium phosphate buffer with a pH of 7.5. The progress of the reaction and formation of the valienamine product were monitored every 2 h for a period of 12 using OPA pre-column derivative high performance liquid chromatography (HPLC) analysis [
37].
2.4. Molecular dynamics simulation and analysis
To analyze the structure of WT WecE and its mutants, the homo-dimeric structure of WT WecE was obtained from the Protein Data Bank (PDB ID: 4ZAH). The homo-dimeric models of the mutant M3 and M4 were constructed on the SWISS-MODEL online server using the 4ZAH dimer as a template. The structural validation of the model was conducted using the protein structure validation tool MolProbity [
41], GMQE, and the QMEANDisCo Global on SWISS-MODEL. The external aldimine intermediate with pyridoxamine 5′-phosphate (PMP), valienone-PMP, was constructed using Discovery Studio 3.5 (Accelrys, USA). The 100 ns molecular dynamics (MD) simulations were performed using Amber 2022 under the ff14SB and GAFF2 force fields of protein (WT WecE and the mutants M3 and M4) and ligand (valienone-PMP), respectively.
The enzyme-intermediate complex structures were visualized and the electrostatic potential was calculated using PyMol (v2.6.0). Ligand contact surface area (LCSA) calculations were performed using the Analyze Protein Interface module of Discovery Studio 3.5 (Accelrys). The root mean square deviation (RMSD), root mean square fluctuation (RMSF), average configuration, catalytic distance measurement, hydrogen bond (HB) interactions at the dimer interface and enzyme-intermediate, solvent-accessible surface area (SASA), and the radius of gyration (RoG) during the simulation, were calculated using the CPPTRAJ suite of Amber 2022. The molecular mechanics Poisson-Boltzmann surface area (MM/PBSA) receptor energy and binding energy of the valienone-PMP transition state were calculated using MMPBSA.py of Amber 2022. Salt bridge (SB) interactions were analyzed using VMD 1.9.4, with an oxygen-nitrogen distance cutoff at 3.2 Å. The structural characteristics of the average configuration during the simulation, including the interface area, volumes of substrate binding sites, and the interactions of dimer interface, were analyzed using PDBsum [
42].
3. Results
3.1. Variable residues identification in the active site
We initially analyzed the residues in the active site of WecE by conducting an evolutionary conservation assessment using the ConSurf server, which helped determine the importance of residue positions based on the phylogenetic relations between homologous sequences [
30], and a conserved cofactor interaction analysis using LigPlot
+ [
31] on the crystal structure of WecE (PDB ID: 4ZAH and 4PIW) [
43]. The binding region of WecE for the cofactor PLP had a higher number of conserved residues compared with that of its binding region for TDP-Glc4O (
Fig. 2(a)). Moreover, all the residues that interacted with PLP via covalent and HBs in the bound crystal structure were predicted to have conservation scores of 8 or higher (
Fig. 2(b); Figs. S1 and S2 in Appendix A). Therefore, we defined a conservation score of 8 as the threshold for distinguishing variable from invariable residues in the active site of WecE.
According to the well-established transamination mechanism [
44], the transamination reaction consists of two parts, each of which involves an external aldimine comprising an amino donor and acceptor, respectively (Scheme S2 in Appendix A). Based on the crystal structure of the external aldimine transition state TDP-Glc4O-PMP in 4ZAH [
43], we constructed and docked the external aldimine valienone-PMP transition state into the active site of WecE (
Fig. 2(c)). Based on the contact distance of residues in the protein [
45], there were 57 residues within 8 Å of the bound valienone-PMP (Fig. S3 in Appendix A). The distribution of conservation scores indicated that 18 out of the 57 residues in contact with valienone-PMP had scores of 8 or higher (
Fig. 2(c)). These residues were excluded from the candidate evolution residues in the active site to minimize the generation of deactivating mutations in the CAST library.
3.2. Stabilizing effect assessment for variable residues
We next identified candidate residues that could improve stability among the aforementioned 39 variable residues. According to the principle that proteins with lower folding free energy typically have higher stability [
46], we established the relative free folding energy
$\left( \text{ }\!\!\Delta\!\!\text{ }\!\!\Delta\!\!\text{ }G \right)$ based on the protein structure before conducting CAST engineering. The stability effects of
$\text{ }\!\!\Delta\!\!\text{ }\!\!\Delta\!\!\text{ }G$ landscapes for all possible saturation mutagenesis of the 39 residues were calculated using two prediction tools that utilize different force fields, FoldX [
47], [
48] and ABACUS2 (a backbone-based amino acid-usage-survey) [
49]. A total of 1560 mutation folding energies (
$\text{ }\!\!\Delta\!\!\text{ }\!\!\Delta\!\!\text{ }{{G}_{\text{FoldX}}}$ and
$\text{ }\!\!\Delta\!\!\text{ }\!\!\Delta\!\!\text{ }{{G}_{\text{ABACUS}2}}$) of saturation substitutions were calculated and normalized by min−max normalization [
33] for the 39 residues (39 × 20 × 2), with the predicted
$\text{ }\!\!\Delta\!\!\text{ }\!\!\Delta\!\!\text{ }G$ distribution ranging from 0 to 1 (
Fig. 3(a)).
The average saturation mutagenesis free energy
$\overline{\text{ }\!\!\Delta\!\!\text{ }\!\!\Delta\!\!\text{ }G}$ of each site was calculated to compare the overall predicted stability effect of the individual residues. The
Z-scores [
33] for the average mutation folding energies of the two statistical groups
$(\overline{\text{ }\!\!\Delta\!\!\text{ }\!\!\Delta\!\!\text{ }{{G}_{\text{FoldX}}}}$ and
$\overline{\text{ }\!\!\Delta\!\!\text{ }\!\!\Delta\!\!\text{ }{{G}_{\text{ABACUS}2}}}$) were also calculated and compared. Sites with a
Z-score of
$\overline{\text{ }\!\!\Delta\!\!\text{ }\!\!\Delta\!\!\text{ }{{G}_{\text{FoldX}}}}$ or
$\overline{\text{ }\!\!\Delta\!\!\text{ }\!\!\Delta\!\!\text{ }{{G}_{\text{ABACUS}2}}}$< 0, indicating that the
$\overline{\text{ }\!\!\Delta\!\!\text{ }\!\!\Delta\!\!\text{ }{{G}_{\text{FoldX}}}}$ or
$\overline{\text{ }\!\!\Delta\!\!\text{ }\!\!\Delta\!\!\text{ }{{G}_{\text{ABACUS}2}}}$ value was lower than the average of the test dataset, were selected as potential stabilizing residues because a lower average mutation folding energy is expected to correlate with higher stability [
46]. These analyses identified 20 FoldX-stabilizing residues based on empirical effective force field (EEEF) [
48] and 19 ABACUS2-stabilizing residues using unique statistical energy functions (SEF) [
49] (
Fig. 3(b)). Considering the two folding energy evaluations simultaneously, 14 residues were categorized as potential variable stabilizing “hotspots” with
$\overline{\text{ }\!\!\Delta\!\!\text{ }\!\!\Delta\!\!\text{ }{{G}_{\text{FoldX}}}}$ and
$\overline{\text{ }\!\!\Delta\!\!\text{ }\!\!\Delta\!\!\text{ }{{G}_{\text{ABACUS}2}}}$ values lower than the test dataset average (
Z-score < 0).
3.3. Site-directed saturation mutagenesis (SSM) on pre-targeted hotspots
Upon identifying 14 potential key residues for engineering in the WecE active site, we used a degenerate NNK codon approach (N = A/C/G/T and K = G/T; Table S1) to construct SSM [
50] libraries for these 14 residues using WT WecE as the original template (
Fig. 4(a); Table S2 in Appendix A). Following the incubation of cell-free extracts at 37 °C for 30 min, the mutant libraries were screened for improved stability and transamination activity toward valienone using an NADH-dependent
L-GDH-coupled high-throughput method in a 96-well microplate format (Scheme S1) [
35], [
36]. The stability of the mutants was tested by measuring the thermal denaturation
Tm [
38], the
t1/2 of enzyme inactivation at 40 °C [
40], and temperature at which enzyme activity is reduced to 50% after a 15-min heat treatment (
T5015) [
39]. Additionally, the transamination activity of the purified mutants toward valienone was assayed
in vitro with the product of valienamine detected using a pre-column OPA derivatized HPLC [
7], [
37].
The data assessment identified residues Y321, K209, V318, and F319 as hotspots that apparently synergistically enhanced stability and activity (
Fig. 4(a)). The most effective activity improved mutants at each site were Y321F, K209F, V318R, and F319V, which displayed activity increases of 4.82-, 3.52-, 2.37-, and 1.87-fold, respectively, compared with that of WT WecE (
Fig. 4(b); Fig. S4 in Appendix A). Correspondingly, the
Tm values increased by 1.28 to 5.03 °C,
T5015 values increased by 1.52 to 4.93 °C, and
t1/2 at 40 °C increased by 4.29- to 39.83-fold compared with those of WT WecE (Table S3 in Appendix A). Among the four mutants, Y321F and V318R were preferred for contributing to the activity, while F319V and K209 were preferred for contributing to the stability (
Fig. 4(b)).
3.4. Iterative cycles to enhance stability and activity effects
Three additional rounds of evolution were performed according to the ISM strategy [
51] to maximize the potential synergy of each of the four stability-activity sites. The Y321F (M1) mutant, which exhibited the highest increase in activity, was chosen as the starting template for iterations in the improved activity sequential order of K209, V318, and F319. The parent template of each iterative round and WT were used as controls during the iterative library screening to determine if the effects of the stability-activity trade-off resulted in a reduction in activity. The mutants that displayed the greatest improvement during each round of iteration were named M2-M4. Each of these mutants was purified, and their stability parameters and kinetic constants were determined using valienone as the substrate (
Table 1; Figs. S5-S8 in Appendix A).
Each of the four top-performing mutants from each iterative cycle significantly improved stability and activity toward valienone compared with WT WecE. Furthermore, the iterated multiple-site mutations exhibited a cooperative non-additive effect. The two mutants, M3 and M4, demonstrated 83.04- and 641.49-fold enhancements in
t1/2 at 40 °C, respectively (
Table 1; Figs. S5-S7 in Appendix A). Additionally, in the
in vitro assay for valienamine synthesis, M3 exhibited a 37.77-fold increase in activity, while M4 displayed a 31.37-fold increase compared with that of WT WecE (Fig. S4). A trade-off between stability and activity was observed during the iterative rounds of the four beneficial sites (
Fig. 5(a)). The stability parameters and kinetic constants indicated that the best stability mutant, M4, exhibited a 7.73-fold improvement in
t1/2 at 40 °C along with a 11.3% decrease in
kcat/
KM toward valienone compared with those of M3 when iterated with the mutation of F319V. Conversely, the best activity mutant M3 displayed a 2.93-fold increase in
kcat/
KM, accompanied by a 16.7% decrease in
t1/2 at 40 °C, compared with those of M2, when iterated with V318R (
Fig. 5(a) and
Table 1).
We subsequently performed a 50 mg preparative-scale reaction, monitoring the valienamine production every 2 h over a 12-h window using OPA pre-column derivative HPLC [
37]. The progress curves of both M3 and M4 displayed significant increases in valienamine production over 1-8 h and remained stable for an additional 4 h. At the end of the 12-h reaction, a 20.03- and 17.67-fold increase in valienamine production compared with that of WT was observed for M3 and M4, respectively (
Fig. 5(b)).
3.5. Elucidating the mechanism underpinning enhanced stability
To investigate the mechanism underlying the enhanced stability and activity, homo-dimeric models of the most significantly activity-improved mutant, M3, and the most substantially stability-enhanced mutant, M4, were constructed using SWISS-MODEL. WT WecE (PDB ID: 4ZAH) was used as the template. The models were assessed using the protein structure validation tool MolProbity [
41], which indicated that 96.79% and 96.52% of the total residues in the models M3 and M4, respectively, were regarded as Ramachandran favored implying that the models were of sufficient quality (Figs. S9 and S10 in Appendix A). MD simulations were performed, based on these models (Fig. S11 in Appendix A) to examine the structural characteristics and interactions involved in the dimer interface and with the external aldimine transition state of valienone-PMP.
The average M3, M4, and WT simulated structures with valienone-PMP were calculated after each 100-ns MD simulation, and the structural characteristics were analyzed using PDBsum [
42]. Compared with those of WT, the mutants M3 and M4 exhibited an increase in the interface areas and a decrease in the volume of the corresponding substrate binding sites owing to changes in shape and electrostatic potential (
Fig. 6(a); Figs. S12-S14 in Appendix A). The interface interactions of the average configuration of the mutants M3 and M4 were significantly strengthened with the increase in the SB and HB interactions (Figs. S12-S14). These findings support the idea that strengthening the interface interactions, which contribute to inter-subunit assembly, often enhances the stability of multimeric enzymes and protects proteins from disintegration [
52], [
53].
During the MD simulations, the SB interactions were analyzed and compared among WT WecE, and the mutants M3 and M4. The number of SBs was higher in the M3 and M4 mutants compared with that in WT WecE (
Fig. 6(b)). The SB interactions in both WT and the mutants involved the same number of residues. However, the only residue that differed between the two was K209, which formed the SB in WT but did not participate in the SB interaction when it mutated to F209 in mutants M3 and M4. Conversely, V318 did not form the SB in the WT and exhibited new SB interactions upon mutation to R318 in M3 and M4 (Fig. S15).
The analysis of interface HB interactions indicated that both the total number of interface HB interactions and the number of residues involved in HB interactions increased in M3 and M4 compared with those in WT (
Fig. 6(b)), particularly in the 307 to 321 regions (Fig. S16 in Appendix A). Residues 209 and 319 were engaged in HB interactions in both WT and the mutants M3 and M4, whereas Y321F and V318R instigated novel HB interactions at the interface in the mutants M3 and M4 (
Fig. 6(c) and Fig. S16). Finally, we discovered that the RoG, SASA, and MM/PBSA receptor energy values, and the free energy of the protein which represents the total internal interactions of the protein, of M3 and M4 decreased compared with those of WT, which is consistent with the observed enhancement in stability (
Fig. 6(d)).
3.6. Unraveling the mechanism underpinning enhanced activity
Previous studies of transamination reactions have emphasized that nucleophilic attack on the external aldimine by a catalytic lysine residue requires the reacting atoms to be positioned in close proximity [
44], [
54]. The conserved catalytic residue K181 (Fig. S17 in Appendix A), which forms an internal Schiff base with PLP in the crystal structure, serves as the nucleophilic attack residue within the active site of WecE [
43]. Our MD simulations facilitated the calculation of the catalytic distances from the ε-amino group of K181 to the N-atom of the C=N bond of the valienone-PMP intermediate for WT, M3, and M4. In essence, M3 and M4 exhibited a reduced catalytic distance in comparison to that of WT WecE (
Fig. 7(a); Fig. S18 in Appendix A), a finding that aligns with that which indicates that a diminished catalytic distance is conducive to nucleophilic attack [
44], [
54].
In terms of the HB interactions throughout the MD simulation (
Fig. 7(b)), the activity of M3 and M4 demonstrated a positive correlation with the aggregate number of HB interactions between K181 and valienone-PMP (
Fig. 7(c)), which agrees with the catalytic distance analysis. The number of HB interactions between valienone-PMP and the K209F and Y321F residues exhibited a decrease in the mutants relative to those of WT (
Fig. 7(c); Fig. S19 in Appendix A). This observation aligns with the residue characteristics, indicating that the mutation of the polar tyrosine and positively charged lysine at positions Y321 and K209 to the hydrophobic phenylalanine would instigate a shift from hydrophilic to hydrophobic interactions at these sites [
55]. This shift increases the hydrophobicity of these sites and consequently attenuates the interaction with the high polar polyhydroxylated valienone-PMP.
We also assessed the MM/PBSA binding energy, which serves as an indicator of the interaction strength and binding stability between the protein and ligand [
56]. Additionally, we examined the LCSA. The calculated MM/PBSA value for the valienone-PMP binding energy decreased for M3 and M4 in comparison to that of WT. Further, M3 exhibited a lower predicted binding energy than that of M4, which is consistent with the larger increase in catalytic activity toward valienone compared with that of M4. Furthermore, the increased LCSA (Table S4 in Appendix A), which indicates the contact area between the dimer interface and ligand, also revealed that the interactions between the mutants and valienone-PMP strengthened (
Fig. 7(d)).
4. Discussion
Facile and economical access to bioactive chiral amines is desired in chemistry, biology, and medicine [
57]; however, such access is hindered by inherent drawbacks such as low thermostability and catalytic activity that can lead to inactivation [
22]. Protein engineering is a powerful method for modifying the properties of enzymes including activity and stability [
58]. However, a widespread negative correlation exists between stability and catalytic activity in enzyme evolution, a phenomenon often referred to as the “stability-activity trade-off” [
59], [
60], thus, fewer examples of successful simultaneous engineering of stability and activity abound compared with the engineering of single properties [
61].
On the lookout for enzymes with both high stability and activity, several researchers have focused on engineering interface residues positioned at a considerable distance from the active site to protect activity. For example, the evolution of multimeric lysine decarboxylase, CadA [
62] and the thermostability and activity evolution of a transaminase, At-ATA, from
Aspergillus terreus [
63]. Alternatively, other researchers have begun by using highly stable enzymes from thermophilic organisms as the initial template for efforts to enhance catalytic activity [
64]. The active site of homo-dimeric WecE assemblies with both monomers contributing functional residues, and the interconnected active site regions of WecE also form part of the dimer interface [
43], [
65]. As the WecE active site contains both catalytic and interface-mediating residues, we postulated that it should be possible to enhance activity and stability simultaneously by introducing mutations in this region.
Accurate pre-targeting of hotspots is acknowledged as being a prerequisite step for the CAST-ISM process [
58]. Numerous reports support the notion that mutagenesis targeting non-conserved residues is more permissive of substitution, which minimizes deleterious mutations and improves the quality of CAST libraries [
66]. When using evolutionary conservation as an indicator to protect the essential catalytic residues in the active site, there is a lack of a standard to define the variability simply based on the sequence-based conservation scores from low to high (for example when using the 1 to 9 categories of the ConSurf server, with higher scores indicating greater conservation,
Fig. 2(a)). HB interaction analysis of WecE with the cofactor PLP in the crystallized structure (PDB ID: 4PIW) indicated that the residues involved in PLP binding were highly conserved (conservation score ≥ 8) (
Fig. 2(b)). Thus, we defined the residues with scores lower than 8 as the variable residues. This scheme provided sequence-, structure-, and function-based multidimensional criteria for evaluating the variability of the candidate residues.
In addition, mutations resulting in lower relative free folding energy (ΔΔ
G) may contribute to enhanced protein stability [
46]. For WecE, we used virtual saturation mutagenesis with two algorithms (aiming to increase accuracy) for 39 variable residues out of 57 residues within 8 Å of the bound valienone-PMP in the active site, resulting in 1560 (39 × 20 × 2) ΔΔ
G values. Notably, to our knowledge, no definitive threshold exists for assessing the stabilizing effect of mutations based on the ΔΔ
G value alone. The current method involves selecting sites with ΔΔ
G values < 0 for low energy, which benefits stability [
67], [
68]. For saturation mutagenesis at a particular site, the ΔΔ
G values generated by 20 mutagenesis will diverge over a wide range and cannot be measured by a single positive or negative value, which would be necessary to make rational decisions regarding any stabilizing tendency for that residue.
We have explored a possible solution for this in the present study: Specifically, we normalized the ΔΔ
G values first, to give the data a uniform dimension and preserve the relative relationships between the original data [
33] (Figs. S20(a) and (b) in Appendix A). The mean mutation folding free energy (ΔΔG¯) for the substitutions at each site was calculated and compared with the value for other sites to identify candidate stabilizing residues. Finally, the
Z-score values of the 39 residues were calculated to describe the relationship of each ΔΔG¯value to the mean of the dataset of value (ΔΔG¯¯). The residues with negative
Z-scores have a lower mutation energy than average; these were defined as candidate stabilizing residues as they were predicted to be relatively more conducive to improving stability (Fig. S20(c) in Appendix A).
We determined the threshold for variable residues based on the conserved cofactor-interacting residues in the crystal, and evaluated the stabilizing effect by comparing the average mutation energy of the saturated mutagenesis for each site. The resulting stabilizing landscape was assessed based on residues having values lower than the dataset average as the cutoff. The assessments of variability and stabilizing effects allowed us to informatively pre-select 14 potential hotspots for stability and activity among the 57 active site residues of WecE. This enabled us to significantly reduce the screening scale to less than 2000 clones [
69] for our CAST and ISM tasks, compared to the random mutagenesis library of the traditional directed evolution [
70].
5. Conclusions
In this study, we investigated a customized CAST-ISM approach and successfully developed SAT WecE mutants that exhibit both enhanced stability and activity for the non-natural substrate valienone, in comparison to those of the WT enzyme. This simultaneous improvement is noteworthy owing to the trade-off between stability and activity typically encountered in enzyme evolution efforts. To enhance the efficiency of evolution, we first targeted potential stability-activity hotspots in the WecE active site prior to constructing the CAST library, which was based on various assessments including the evolutionary conservation and relative folding energy of mutations. The two best-performing mutants, M3 and M4, exhibited an 83.04- and 641.49-fold increase, respectively, in the half-life at 40 °C. Additionally, M3 and M4 exhibited a 37.77- and 31.37-fold increase in catalytic activity for valienamine synthesis, respectively. Our study presents a target-oriented approach to address the challenging problem of simultaneously improving both stability and activity and aims at achieving stable and highly active SAT enzymes for the production of valuable amino sugars.
Acknowledgments
This work was supported by the National Key Research and Development Program of China (2018YFE0200501 and 2020YFA0907700) and the National Natural Science Foundation of China (32271306 and 21977067).
Compliance with ethics guidelines
Runxi Wang, Lu Qiao, Mufei Liu, Yanpeng Ran, Jun Wang, Wupeng Yan, Yan Feng, and Li Cui declare that they have no conflict of interest or financial conflicts to disclose.
Appendix A. Supplementary data
Supplementary data to this article can be found online at
https://doi.org/10.1016/j.eng.2024.04.026.
PDF
(3491KB)
