Structure-guided steric hindrance engineering of Bacillus badius phenylalanine dehydrogenase for efficient l-homophenylalanine synthesis

Background Direct reductive amination of prochiral 2-oxo-4-phenylbutyric acid (2-OPBA) catalyzed by phenylalanine dehydrogenase (PheDH) is highly attractive in the synthesis of the pharmaceutical chiral building block l-homophenylalanine (l-HPA) given that its sole expense is ammonia and that water is the only byproduct. Current issues in this field include a poor catalytic efficiency and a low substrate loading. Results In this study, we report a structure-guided steric hindrance engineering of PheDH from Bacillus badius to create an enhanced biocatalyst for efficient l-HPA synthesis. Mutagenesis libraries based on molecular docking, double-proximity filtering, and a degenerate codon significantly increased catalytic efficiency. Seven superior mutants were acquired, and the optimal triple-site mutant, V309G/L306V/V144G, showed a 12.7-fold higher kcat value, and accordingly a 12.9-fold higher kcat/Km value, than that of the wild type. A paired reaction system comprising V309G/L306V/V144G and glucose dehydrogenase converted 1.08 M 2-OPBA to l-HPA in 210 min, and the specific space–time conversion was 30.9 mmol g−1 L−1 h−1. The substrate loading and specific space–time conversion are the highest values to date. Docking simulation revealed increases in substrate-binding volume and additional degrees of freedom of the substrate 2-OPBA in the pocket. Tunnel analysis suggested the formation of new enzyme tunnels and the expansion of existing ones. Conclusions Overall, the results show that the mutant V309G/L306V/V144G has the potential for the industrial synthesis of l-HPA. The modified steric hindrance engineering approach can be a valuable addition to the current enzyme engineering toolbox. Supplementary Information The online version contains supplementary material available at 10.1186/s13068-021-02055-0.

l-HPA synthesis are constrained by lengthy steps, high cost, environmental pollution, and poor enantioselectivity, rendering industrial processes ultimately unsustainable [7,8]. By contrast, biocatalysts are environmentally friendly, mild reaction conditions, and have excellent selectivity [9,10]. Thus, various enzymatic routes including kinetic resolution catalyzed by lipase [11], in situ racemization catalyzed by hydantoinase [12,13], and asymmetric transamination catalyzed by transaminase [14,15] were recently developed and applied to l-HPA synthesis. Nevertheless, the foregoing reactions also have limitations, such as a theoretical maximum 50% lipase yield, poor applicability because of the strict d-enantioselectivity of hydantoinase, and undesirable equilibrium shifts in the reversible transaminase transamination [16].
Direct asymmetric reductive amination of prochiral 2-oxo-4-phenylbutyric acid (2-OPBA) catalyzed by phenylalanine dehydrogenase (PheDH) is a promising approach in l-HPA synthesis. It consumes only free ammonia and generates only water as a byproduct ( Fig. 1). Asano et al. [17] reported the use of PheDH from Bacillus sphaericus (BsPheDH) to prepare l-HPA. Bradshaw et al. [18] investigated the application of recombinant E. coli cells harboring PheDH from Rhodococcus sp. M4 (RsPheDH) to produce l-HPA. They achieved 63% conversion using 60 mM substrate. Ahmad et al. [19] enhanced the applicability of RsPheDH in l-HPA production via cell immobilization. They achieved > 80% conversion using 67 mM substrate. Notably, Zhang et al. [20] stated that the combination of substrate fed-batch with continuous product removal improved substrate loading in l-HPA production catalyzed by PheDH from Thermoactinomyces intermedius (TiPheDH). A final substrate concentration of 510 mM was achieved after eight cycles of a fed-batch operation. An important constraint in PheDH-catalyzed l-HPA production is the poor catalytic efficiency toward non-native, bulky 2-OPBA compared with native phenylpyruvic acid (PPA). Consequently, the substrate loading is low and fails to meet the requirements of industrially and commercially viable biocatalysts and biocatalysis processes [21].
Directed evolution techniques are powerful tools for generating prolific sources of biocatalysts [22,23]. Steric hindrance of the substrate-binding pocket plays a significant role in modulating the catalytic properties of an enzyme, including enantioselectivities [24,25], substrate specificities [26][27][28], and catalytic activities [25,[28][29][30]. Recent enzyme engineering studies focused on increasing or conferring enzyme activity towards non-native bulky substrates by attenuating the steric hindrance of the substrate-binding pocket. The mutant constructed by Merck and Codexis [31] using transaminase ATA-117 as a mutagenic template is a remarkable example of such engineering, as the mutant successfully showed detectable reductive amination activity toward the bulky prositagliptin ketone that lacked any activity before, which was successfully used to catalyze the synthesis of the antidiabetic compound sitagliptin. Chen et al. [32] used rational design based on steric hindrance attenuation to induce asymmetric reductive amination activity of engineered amine dehydrogenases towards several previously unreactive long-chain α-aliphatic ketones. The hydroxynitrile lyase mutants engineered by Zheng et al. [33] showed remarkably increased asymmetric hydrocyanation activity toward the rigid pharmaco aldehydes, and successfully applied them to achieve the synthesis of the β 2 -adrenoreceptor agonist salmeterol. These successful advances demonstrated the feasibility of steric hindrance Fig. 1 Biosynthesis of l-HPA from 2-OPBA catalyzed by PheDH paired with a GluDH cofactor recycling system engineering of substrate-binding pockets to modulate the catalytic activity of enzymes towards non-native, bulky substrates.
In the present study, we report the structure-guided steric hindrance engineering of PheDH from Bacillus badius (BbPheDH) to improve its catalytic efficiency in l-HPA synthesis. The steric hindrance of the substratebinding pocket of BbPheDH was subjected to crucial residue identification and evaluation of the mutagenesis libraries constructed based on degenerate codons. Seven superior mutants were generated after four rounds of steric hindrance engineering. The optimal triple-site mutants significantly enhanced catalytic efficiency, substrate loading, and specific space-time conversion. Structure analysis elucidated the factors contributing to the enhanced catalytic efficiency observed in the enzymekinetics and transformation experiments.

Preliminary activity studies and molecular modeling
Initially, two PheDHs from B. badius (BbPheDH) and G. kaustophilus (GkPheDH) were selected as candidate enzymes in the development of an enhanced biocatalyst for l-HPA synthesis. As 2-OPBA had a larger side chain than the native substrate PPA, its catalysis by BbPheDH and GkPheDH was characterized by lower turnover frequency (k cat ) and less favorable catalytic efficiency (k cat /K m ) ( Table 1), which motivated us to consider enhancing catalytic efficiency toward 2-OPBA by the steric hindrance engineering. Biotransformation experiments were subsequently performed using BbPheDH and GkPheDH as biocatalysis, giving 67.5% and 53.6% conversions at 0.2 M substrate concentration within 8 h (Table 1 and Additional file 2: Figure S1). Due to its relatively higher conversion, BbPheDH was selected as the starting mutagenesis template for steric hindrance engineering.
To assess the feasibility of developing an enhanced biocatalyst for efficient l-HPA synthesis, we first generated a structural homology model of BbPheDH (Fig. 2a) based on the reported crystal structure of Rhodococcus sp. M4 PheDH [34] to develop hypotheses for mutagenesis library designs. The model revealed that the residues surrounding the catalytic active center of BbPheDH could be classified into two categories on the basis of their function in the substrate-binding process. Group one encompasses the catalytic residues that are appointed in the binding of the α-carboxylic and α-amino groups of the substrate, namely, K78, K90, D125, and N276, respectively, which were highly conserved among the PheDH family (Additional file 2: Figure S2). These residues create a hydrophilic environment that combines the hydrophilic α-amino acid moiety of the substrate (Fig. 2b). On the opposite side, a panel of hydrophobic residues creates a hindered hydrophobic environment forcing the substrate side chain into its ideal binding pose (Fig. 2c). Furthermore, limited structure-guided engineering was previously conducted on PheDHs and showed that most of the beneficial mutations introduced into the substratebinding pocket were hydrophobic amino acid residues [35][36][37]. Based on the foregoing information, we selected amino acids with strong hydrophobicity and weak steric hindrance as target substitutes in the subsequent mutagenesis experiments.

Mutagenesis library construction and evaluation based on attenuated steric hindrance
To identify mutagenesis candidates for initial library construction, 2-OPBA was docked into the catalytic active center of BbPheDH to investigate the unfavorable hindrance interactions. Based on the acquired binding pose, all amino acid residues except the smallest glycine and the conserved catalytic residues within 6 Å of the side-chain phenyl ring of 2-OPBA were selected for site-directed mutagenesis in the first round of steric hindrance engineering (Fig. 3a). Twenty-three single-site mutants were constructed at nine amino acid sites based on the designed degenerate codon (Additional file 1: Table S1) by whole-plasmid PCR [38]. All target mutants were isolated from the mixed libraries and confirmed by DNA sequencing. Three superior mutants L50V (M1-1), V309A (M1-2), and V309G (M1-3) were identified by initial activity screening of cell-free extracts (Additional file 2: Figure S3). Subsequent enzyme-kinetics experiments with the purified proteins demonstrated that the K m value of these mutants toward 2-OPBA did not change significantly (Fig. 3h), whereas the k cat values were 25, 61, and 89 s −1 , which were 2.3-, 5.7-, and 8.3-fold higher than that of the wild type (11 s −1 ), respectively (Fig. 3g).
As mutant M1-3 had the highest catalytic efficiency (74 s −1 mM −1 ) towards 2-OPBA, it was selected as the mutagenesis template in the second round of steric hindrance engineering to investigate the potential synergistic effects. The binding pose of M1-3 with 2-OPBA indicated eight amino acid residues within 6 Å of the side-chain phenyl ring of 2-OPBA. For the rapid detection of any synergy of a proximal steric hindrance with residue G309, the residues were simplified by doubleproximity filtering. We identified four mutagenesis candidates within 6 Å of both the phenyl ring of 2-OPBA and G309 (Fig. 3b). Thus, a mutagenesis library consisting of ten double-site mutants was constructed, and the superior mutants V309G/L50V (M2-1) and V309G/L306V (M2-2) were identified by the initial activity screening  Figure S4). The k cat /K m values of the purified mutant proteins, M2-1 and M2-2, toward 2-OPBA were determined to be 143 and 179 s −1 mM −1 , respectively, which were 8.4-and 10.5-fold higher than that of the wild type, respectively (Fig. 3f ). Notably, this change was also caused by the enhanced k cat value of the mutants (Fig. 3g).
With this result, the mutant M2-2 was selected as the template in the third round of steric hindrance engineering. Here, we investigated proximal steric hindrance synergy with V306. Three mutagenesis candidates were identified by double-proximity filtering using V306 as the key residue (Fig. 3c). The initial activity screening of the cell-free extracts of the ten triple-site mutants revealed two superior mutants V309G/L306V/V144A (M3-1) and V309G/L306V/V144G (M3-2) (Additional file 2: Figure  S5). The k cat values of the purified mutant proteins, M3-1 and M3-2, toward 2-OPBA reached 131 and 136 s −1 , respectively, which were 12.2-and 12.7-fold higher than that of the wild type, respectively (Fig. 3g); this change was accompanied by a 11.3-and 12.9-fold increase in the k cat /K m values, respectively (Fig. 3f ).
Inspired by the above results, we explored the possibility of increasing catalytic activity towards 2-OPBA using M3-2 as the template in the fourth round of steric hindrance engineering. Two mutagenesis candidates were identified (Fig. 3d) and four corresponding quadruplesite mutants were constructed. However, none of them Residues shown in gray bubbles, including in blue bubbles (a-d), were all close (within 6 Å) to the side-chain phenyl ring of 2-OPBA, while the residues in blue bubbles (b-d) are not only close to the side-chain phenyl ring of 2-OPBA, but also close to the focal residues (G309, V306, and G144, respectively). Activity was measured in NH 4 Cl/NH 4 OH buffer (2 M, pH 9.5) containing 1-20 mM 2-OPBA and 0.5 mM NADH at 30 °C and carried out at a 200-μL scale in 96-well microtiter plates by monitoring the initial decrease velocity of the absorbance at 340 nm (indicating NADH consumption). All determinations were performed in triplicate, and error bounds represent ± sd exhibited higher activity towards 2-OPBA than M3-2 (Additional file 2: Figure S6).
Together, seven superior mutants were successfully screened from the 47 elements in the site-directed mutagenesis library (Fig. 3e). The optimal triple-site mutant, M3-2 (V309G/L306V/V144G), showed a 12.7fold increase in the k cat value (Fig. 3g) compared with that of the wild-type BbPheDH, whereas no obvious change was observed in the K m value (Fig. 3h), which resulted in a 12.9-fold increase in the k cat /K m value (Fig. 3f ). This result was similar to that of other studies [25,32]. Subsequently, the kinetic parameters of BbPheDH and obtained mutants with superior activity toward NADH were assessed. The results showed that the catalytic efficiency of the mutants toward NADH was improved to varying degrees, which is beneficial for the practical application of these mutants to l-HPA synthesis. Moreover, the thermostabilities of BbPheDH and its superior mutants were assessed by measuring the T 50 30 value (the temperature at which 50% of enzyme activity is lost following a heat treatment for 30 min). It is not unexpected that some kind of stability-activity tradeoff occurred in all mutants, although this tradeoff is of no practical importance, since the T 50 30 value of all mutants still exceeded 50 ℃ (Additional file 1: Table S2).

Biocatalytic optimization analysis on the best mutant M3-2
The dependence of temperature and pH value were investigated for the reductive amination of 2-OPBA catalyzed by M3-2 paired with glucose dehydrogenase (GluDH) at 0.2 M substrate concentration. Figure 4a shows that 2-OPBA conversion was > 99% after 180 min when the reactions proceeded at 30-40 °C. Catalytic efficiency decreased slightly at 25 °C, with the reaction requiring 240 min to attain 99% conversion. At 50 °C, however, quantitative conversion was < 90% after 240 min. We then evaluated the effects of NH 4 Figure 4b shows that the fastest conversions occurred at pH 8.5 and the quantitative conversion was > 99% within 90 min. We subsequently performed the reductive amination of 2-OPBA in various concentrations of NH 4 OH/HCOONH 4 buffer (pH 8.5) and NAD + to determine the influences of NH 4 + and NAD + concentration. The 2-OPBA conversion was > 99% at 1-3 M NH 4 + . By contrast, 4 M and 5 M NH 4 + lowered the conversion rate (Fig. 4c). An activity assay indicated that low GluDH activity might explain the observed decrease in conversion rate in the presence of high NH 4 + concentrations (Additional file 2: Figure S7). Moreover, 2-OPBA could be fully converted when the NAD + concentration was 0.3 mM, 0.5 mM, or 1 mM; however, the conversion dropped to < 50% when the NAD + concentration was < 0.3 mM (Fig. 4d).

Preparative scale synthesis of l-HPA employing mutant M3-2
Using the optimized reaction system (30 °C; pH 8.5; 3 M NH 4 + ; 0.3 mM NAD + ), the preparative scale synthesis of l-HPA by reductive amination of 2-OPBA was carried out in 100 mL reaction volume. The pH value was maintained at 8.5 by titrating concentrated NH 3 ·H 2 O during the reaction course. Figure 4e shows that 0.2 M and 0.3 M 2-OPBA were converted by > 99% conversion within 20 min and 30 min, respectively. However, substrate inhibition was observed at high concentrations. The quantitative conversion rates were < 70% and < 30% after 240 min in the presence of 0.4 M and 0.5 M substrate, respectively (Additional file 1: Table S3). Similar results were reported for TiPheDH from T. intermedius [39]. We speculate that high substrate concentration may have an irreversible adverse effect on the enzyme [40].

Continuous substrate fed-batch l-HPA synthesis
Substrate concentration is a crucial factor in industrially feasible biocatalytic processes. Therefore, a combination of substrate fed-batch with product removal strategy was applied for continuous l-HPA synthesis. During the reaction course, 2-OPBA was repeatedly added to the initial NH 4 OH/HCOONH 4 reaction medium (pH 8.5) containing 0.3 mM NAD + , 3 M NH 4 + , 10 g L −1 cell-free extracts of M3-2, and 12 g L −1 cell-free extracts of GluDH. The substrate concentration was fixed at ~ 0.3 M to maintain high reaction efficiency. Meanwhile, l-HPA product was continuously precipitated from the reaction solution and isolated after filtration and drying. Consequently, 1.08 M 2-OPBA with a quantitative conversion of 90.2% was effectively transformed after four times fed-batch over 210 min (Fig. 4f ), and the specific space-time conversion reached 30.9 mmol g −1 L −1 h −1 . Comparisons with previously reported biocatalysis processes revealed that our M3-2 had the highest substrate loading and specific space-time conversion in l-HPA production (Additional file 1: Table S4). Furthermore, M3-2 displayed perfect stereoselectivity and yielded l-configuration HPA with up to 99% enantiomeric excess (ee). These results underscore the strong potential of M3-2 as a biocatalyst for continuous l-HPA synthesis.

Docking simulation and tunnel analysis of wild type and derived mutants
To better understand the enhancement in the turnover frequency (k cat ) observed in enzyme-kinetics experiments at the atomic and molecular levels, we performed docking simulations using the available structural information. The substrate 2-OPBA was docked into the catalytic active center of the wild-type BbPheDH and of the derived mutants M1-3, M2-2, and M3-2 (Fig. 5). The docking simulations revealed poor adaptability of 2-OPBA to the substrate-binding pocket of wild-type BbPheDH, compared with the binding pose of the native substrate PPA (Additional file 2: Figure S8). The relatively bulky hydrophobic phenyl group side chain of 2-OPBA could not fit into the substrate-binding pocket of wildtype BbPheDH in a relatively stretched configuration (Fig. 5a, b), which may prevent the hydride transfer of the substrate in proper orientation. The low degree of freedom of the phenyl group side chain may be the reason for the low k cat value, and thereby, for the low k cat /K m value of the wild-type BbPheDH toward 2-OPBA. This observation was similar to that in a previous study [41]. In contrast, the incorporation of three crucial mutations, V309G, L306V, and V144G, gradually enlarged the volume of the substrate-binding pocket compared with that of BbPheDH. With this change, the bulky phenyl group side chain of 2-OPBA could fit into the enlarged substrate-binding pocket with a relatively stretched configuration at a high degree of freedom ( Fig. 5c-h). Hence, the mutant M3-2 showed an enhanced k cat value toward 2-OPBA compared with that of the wild-type BbPheDH. Moreover, the observed enlargement of the substrate-binding pocket can be further supported by calculating the volume of the substrate-binding pocket, which revealed that the volume of the eventual mutant M3-2 was about 196.9 Å 3 larger than that of the wildtype BbPheDH (Additional file 1: Table S5). These results demonstrated that the enlargement of the substratebinding pocket in a reasonable range may be beneficial to enhance the k cat value of enzymes toward the substrates with bulky hydrophobic side chain [32].
Tunnels represent potential transport pathways for small molecules, water molecules, and ions, and play a significant role in the functioning of a large variety of proteins [42,43]. The accessibility of a tunnel depends largely on its shape, size, and amino acid composition. To a certain extent, it can be modified by protein engineering [44,45]. Therefore, we investigated whether steric hindrance engineering alters BbPheDH tunnels. Tunnels analyses based on the structure homology model of M3-2 revealed that, compared with BbPheDH, the  M3-2 (g, h). Structural homology models of the enzyme are shown in light blue cartoon or surfaces representations. Substrate 2-OPBA is shown as cyan sticks or balls. Initial residues V144, L306, and V309 are shown as yellow sticks. Mutagenized residues G144, V306, and G309 are shown as magenta sticks. Coenzyme NADH is shown as orange sticks incorporation of V144G, L306V, and V309G formed a new tunnel (shown in magenta) in the vicinity of the catalytic active center (Fig. 6), suggesting a new transport pathway for substrate access and product egress. Furthermore, V144G, L306V, and V309G mutations markedly altered the shapes of the existing tunnels (Fig. 6). The widened tunnels (shown in red and orange) could further accelerate the transport rates of substrate and product. Both pieces of information lend support to the enhancement of the k cat value toward 2-OPBA catalyzed by M3-2 observed in the enzyme-kinetics experiments. This observation demonstrated that the formation of new tunnels and widening of existing ones can modulate the catalytic activity of enzymes. It also resembled the findings reported for prior studies [46,47].

Conclusions
Here, we employed a modified steric hindrance engineering approach to create an enhanced biocatalyst with BbPheDH as a mutagenesis template for efficient l-HPA synthesis. Seven superior mutants were screened from the mutagenesis libraries constructed based on molecular docking, double-proximity filtering, and a degenerate codon. The k cat and k cat /K m values of the mutant M3-2 (V309G/L306V/V144G) were 12.7-and 12.9-fold higher than those of the wild type, respectively, it converted 1.08 M 2-OPBA to l-HPA in 210 min, and its specific space-time conversion was 30.9 mmol g −1 L −1 h −1 . The substrate loading and specific space-time conversion are the highest values reported so far, demonstrating that mutant M3-2 is a useful biocatalyst with potential for Fig. 6 Tunnels of wild-type BbPheDH (a, b) and optimal mutant M3-2 (c, d) calculated by CAVER plugin of PyMOL. Structural homology models of the enzyme are shown in light blue cartoon. Calculated tunnels are shown as balls. Initial residues V144, L306, and V309 are shown as yellow sticks. Mutagenized residues G144, V306, and G309 are shown as magenta sticks the industrial production of l-HPA. Structure analysis demonstrated that mutations altered the substrate-binding pocket and tunnel, and variations in steric hindrance modulate the catalytic activity of BbPheDH towards 2-OPBA. Moreover, our study indicates the combination of molecular docking with a double-proximity filter as a promising workflow for capturing the proximal steric hindrance synergistic effects. The modified steric hindrance engineering approach can be a valuable addition and extension of the current enzyme engineering toolbox.

Sequence alignment and site-directed mutagenesis
Amino acid sequence alignment was performed using the MUSCLE server (https:// www. ebi. ac. uk/ Tools/ msa/ muscle/) [48] and displayed using the Esprit server (http:// espri pt. ibcp. fr) [49]. Site-directed mutagenesis was introduced by whole-plasmid PCR [38] using primers containing the required degenerate codons and the pET-28a (+) plasmid inserts the pdh gene as the template. The PCR products were analyzed by agarose gel electrophoresis and transformed into E. coli BL21 (DE3) cells. All target mutants were isolated from the constructed mixed libraries and confirmed by DNA sequencing (Talen-Bio, Wuxi, China).

Protein expression and purification
Escherichia coli BL21 (DE3) cells containing recombinant pET-28a (+) plasmids were cultured in Luria-Bertani (LB) medium with kanamycin (50 mg L −1 ) at 37 °C and with shaking at 200 rpm for 2 h. Enzyme expression was induced with IPTG (final concentration = 0.2 mM) when the bacteria reached OD 600 = 0.6-0.8. The cells were incubated at 20 °C and with shaking at 200 rpm for 12 h. His-tagged enzymes were purified with an ÄKTA purifier system (GE Healthcare, Little Chalfont, UK). The cells were resuspended in buffer A (100 mM potassium phosphate, 150 mM NaCl, and 20 mM imidazole; pH 7.5) and disrupted with an ultrasonic cell disruption system (SCIENTZ-IID; Ningbo Scientz Biotechnology, Ningbo, Zhejiang, China) followed by centrifugation at 12,000 rpm and 4 °C for 30 min to remove cell debris and obtain cell-free extracts. The latter were passed through a 0.22-μm filter and loaded into a standard Ni-NTA affinity column (Thermo Fisher Scientific, Waltham, MA, USA) pre-equilibrated with buffer A. The column was then gradient-eluted with buffer B (100 mM potassium phosphate, 150 mM NaCl, and 500 mM imidazole; pH 7.5). Eluted fraction purity was determined by SDS-PAGE (Additional file 2: Figure  S9).

Activity and kinetic parameters analysis
Reductive amination activity was determined at a 200-μL scale in 96-well microtiter plates at 30 °C. The rate of initial decrease of the absorbance was measured at 340 nm (indicating NADH consumption) using a Multi-Skan GO UV-spectrometer (Thermo Fisher Scientific). The reaction mixture consisted of NH 4 Cl/NH 4 OH buffer (2 M; pH 9.5), 10 mM 2-OPBA, 0.2 mM NADH, and a certain amount of cell-free extract. One unit of enzyme activity (1 U) was defined as the amount of enzyme oxidizing 1 μM NADH per minute. Kinetic parameters with respect to 2-OPBA (PPA) were evaluated in NH 4 Cl/ NH 4 OH buffer (2 M; pH 9.5) at 30 °C in the presence of 1-20 mM 2-OPBA (PPA) and 0.5 mM NADH with the purified enzymes. Kinetic parameters with respect to NADH were evaluated in NH 4 Cl/NH 4 OH buffer (2 M; pH 9.5) at 30 °C in the presence of 0.1-0.5 mM NADH and 20 mM 2-OPBA with the purified enzymes (Additional file 1: Table S6). K m and V max were calculated in GraphPad Prism 8 (GraphPad Software, La Jolla, CA, USA) by nonlinear curve fitting of the initial rate versus substrate concentration data to the Michaelis-Menten equation.

Thermostability analysis
Thermostability assays of wild-type BbPheDH and its mutant were conducted by measuring T 50 30 , which is the temperature at which 50% of the enzyme activity is lost after 30 min heat treatment. The enzymes were incubated at various temperatures (30-70 °C) for 30 min, and their residual activity was measured. Specific activity before incubation was normalized as 100%. Enzyme activity was measured in NH 4 Cl/NH 4 OH buffer (2 M; pH 9.5) containing 0.2 mM NADH and 10 mM 2-OPBA.

Homology modeling and molecular docking
Structural homology models of BbPheDH and its mutants were constructed using the SWISS-MODEL server (https:// swiss model. expasy. org/) [50] based on the crystal structure of the PheDH from Rhodococcus sp. M4 (PDB: 1C1D) [34]. The 3D structure of the ligand was obtained from the PubChem website (https:// pubch em. ncbi. nlm. nih. gov/) [51]. Ligand energy was minimized with the OPTIMIZE plugin of PyMOL [52]. AutoDock v. 4.2 [53] was used for docking simulations. Prior to docking, each protein was protonated at pH 9.5 using the H++ server (http:// bioph ysics. cs. vt. edu/ H++) [54] to mimic the experimental condition. All water molecules were removed, and nonpolar hydrogen atoms were added with MGLTools v. 1.5.4. The components of the catalytic triad (K78, K90, and N276) were defined as flexible residues. A grid box of 50 × 50 × 50 Å with 0.375 Å spacing encompassed the active cavities of BbPheDH and its mutants and was set as the search space for suitable substratebinding poses. Energies of the interactions between the ligands and the receptors were calculated with MGLTools v. 1.5.4. The lowest energy docking pose was selected for the subsequent molecular dynamics simulations.

Tunnel analysis
The tunnels of BbPheDH and its mutant M3-2 were calculated with the CAVER v. 3.0 plugin of PyMOL [55]. The probe radius was set to 1.2 Å, the default settings were applied, and the center of the catalytic triad (K78, K90, D125, and N276) was the starting point. All water molecules were removed, nonpolar hydrogen atoms were added with MGLTools v. 1.5.4, and the volumes of the substrate-binding pockets of BbPheDH and its mutants were calculated with the ProteinsPlus tool (https:// prote ins. plus/) [56]. The PDB format files were uploaded for the calculations.
Additional file 1: Table S1. Primers used for site-directed mutagenesis. Table S2. Enzyme thermostabilities of BbPheDH and its superior mutants. Table S3. Effects of substrate concentration on the asymmetric reductive amination of 2-OPBA catalyzed by mutant M3-2. Table S4. Comparison between M3-2 and other reported PheDHs for the synthesis of l-HPA from 2-OPBA. Table S5. Calculated substrate-binding pocket volume and area of BbPheDH and its superior mutants. Table S6. Kinetic parameters of BbPheDH and its superior mutants toward NADH.
Additional file 2: Figure S1. Time course of asymmetric reductive amination of 2-OPBA catalyzed by BbPheDH and GkPheDH. Figure S2. Amino acid sequence alignment of the PheDHs from different sources. Figure S3. Relative activity of the single-site mutants constructed in the first round of steric hindrance engineering. Figure S4. Relative activity of the double-site mutants constructed in the second round of steric hindrance engineering. Figure S5. Relative activity of the triple-site mutants constructed in the third round of steric hindrance engineering. Figure  S6. Relative activity of the quadruple-site mutants constructed in the fourth round of steric hindrance engineering. Figure S7. Relative activity of BbPheDH and GluDH in different concentrations of NH 4 OH/HCOONH 4 buffer (pH 8.5). Figure S8. Binding poses of native substrate PPA (a) and bulky substrate 2-OPBA (b) with BbPheDH. Figure S9. SDS-PAGE analysis of BbPheDH and its superior mutants.