SPIN-guided engineering of a novel (R)-amine transaminase from Fusarium albosuccineum for enantioselective synthesis of chiral piperidyl amines

Abstract

Aminotransferases are promising green biocatalysts for the synthesis of chiral amines, yet their limited catalytic efficiencies restrict broader industrial applications. In this study, a novel (R)-amine transaminase, FalAT, was identified from Fusarium albosuccineum through genome mining. FalAT exhibited optimal activity at 30 °C and pH 7.0 and catalyzed the conversion of 1-Boc-3-piperidone to (R)-1-Boc-3-aminopiperidine with >99 % enantiomeric excess. To efficiently enhance its catalytic performance, a Substrate–Protein Interaction Network (SPIN) strategy was implemented, integrating structure-guided analysis, molecular docking, virtual saturation mutagenesis, and dual energy–distance filtering. SPIN first screened and constructed a mutational library covering 70 amino acid positions, which was subsequently narrowed to 9 key residues, ultimately yielding 15 candidate mutants for experimental validation. Experimental results showed that five mutants exhibited higher catalytic activity than the wild-type enzyme, among which R126A was the most effective, displaying approximately a 4-fold increase in catalytic activity and a 13-fold enhancement in catalytic efficiency (k_cat/K_m = 2.05 s⁻¹ mM⁻¹). Molecular dynamics simulations revealed that the R126A mutation expanded the active-site cavity, alleviated steric hindrance, and strengthened hydrophobic interactions, thereby improving substrate binding and catalytic turnover. Furthermore, substrate profiling demonstrated that FalAT possesses moderate substrate promiscuity. Overall, the SPIN strategy significantly improved the catalytic performance of FalAT while markedly reducing experimental workload, providing an efficient and generalizable approach for the directed evolution of (R)-amine transaminases for the green synthesis of chiral amines.

1. Introduction

Chiral amines are key structural motifs in numerous pharmaceuticals and play a pivotal role in the industrial synthesis of antidiabetic agents such as linagliptin and alogliptin [[1], [2], [3]]. (R)-1-Boc-3-aminopiperidine is an important intermediate commonly used for the preparation of dipeptidyl peptidase-IV (DPP-4) inhibitors and in the synthesis of gliptin drugs. Conventional chemical routes to chiral amines typically rely on noble-metal catalysis or chiral auxiliaries and often suffer from lengthy procedures, high environmental burden, and elevated costs [4,5]. By contrast, aminotransferases (ATs), which employ pyridoxal-5′-phosphate (PLP) as a cofactor, catalyze reversible amino-transfer between ketones and amine donors under mild conditions with high stereoselectivity and regioselectivity, and thus represent an attractive green approach for chiral amine synthesis [1,6].

Among reported aminotransferases, (R)-amine transaminases (R-ATAs) are particularly valuable for the preparation of non-natural R-configured chiral amines and have shown unique advantages in the synthesis of pharmaceutical intermediates [[7], [8], [9]]. However, most characterized R-ATAs derive from bacteria and possess narrow native substrate scopes with limited activity toward sterically demanding or structurally complex substrates, which restricts their industrial applicability [10]. For example, the key chiral intermediate of linagliptin and alogliptin, (R)-1-Boc-3-aminopiperidine, is a challenging substrate for transaminase catalysis because of the high rigidity of its piperidine ring and the bulky substituents it carries.

To overcome limitations of existing enzyme resources and catalytic performance, A previously uncharacterized fungal R-ATA was mined from Fusarium albosuccineum and designated FalAT. This enzyme exhibited high stereoselectivity (>99 % ee) toward the target intermediate. However, its catalytic efficiency required improvement for industrial application. For the first time in the engineering of (R)- amine transaminases, a Substrate–Protein Interaction Network (SPIN) strategy is introduced. By integrating AlphaFold structure prediction, molecular docking, virtual saturation mutagenesis, and the energy-distance dual filtering scheme, SPIN enabled precise identification of key active-site residues and the construction of a compact, high-value mutation library. This approach rapidly yielded mutants with markedly improved performance. Our study thus expands the pool of fungal R-ATAs and proposes a rational, efficient strategy for enzyme engineering toward industrial use.

2. Materials and methods

2.1. Materials

Plasmids were synthesized by Anhui General Biotech Co., Ltd.; PLP and amine donors were obtained from Aladdin; acetonitrile, methanol and Breath-Easy membranes were purchased from Sigma; other reagents were supplied by Sinopharm. An Avanti J-26s XP centrifuge (Beckman) and a JN-02C high-pressure cell disruptor (Guangzhou Juneng Bio) were used. A UV–vis spectrophotometer from Shanghai Yidian Analytical Instruments and a Shimadzu LC-20 HPLC system were used for analyses.

2.2. Methods

2.2.1. Mining of FalAT

A phylogenetic tree was constructed using MEGA 12 to identify candidate sequences showing significant evolutionary relatedness to representative R-ATAs [11]. For clarity, only the tree topology is presented in the manuscript. Multiple sequence alignments between the target protein and representative branched-chain aminotransferases (BCATs) were performed in GeneDoc to annotate conserved functional residues that informed subsequent functional analyses [12].

2.2.2. Construction, expression and purification of FalAT

The amino-acid sequence of FalAT was retrieved from UniProt (UniProt ID: A0A8H4KFE3). Considering codon-usage differences between species, the gene was codon-optimized for E. coli expression. The coding sequence was cloned into pET-15b using XhoI and BamHI sites. Plasmid DNA (50–100 ng/μL) was mixed with chemically competent BL21(DE3) cells on ice for 20 min, heat-shocked at 42 °C for 60–90 s, returned to ice for 5 min, and then recovered in 500 μL LB or SOC medium for 1 h before plating on antibiotic-containing LB agar. Positive clones were induced at 16 °C overnight. Cells were harvested and lysed on ice by sonication (30 min), followed by centrifugation at 12,000 rpm for 30 min at 4 °C. The supernatant was filtered through a 0.45 μm membrane and loaded onto a pre-equilibrated Ni²⁺ affinity column (equilibration buffer: 1 × PBS with 10–20 mmol L⁻¹ imidazole). After allowing protein binding on ice for 2 h, the column was washed and eluted. Elution fractions were analyzed by SDS-PAGE. Purified protein was dialyzed to remove imidazole, glycerol was added as a cryoprotectant, aliquoted and stored at −80 °C.

2.2.3. Enzyme activity assay

The enzyme activity assay was performed in a reaction mixture containing 20 mmol L⁻¹ (R)-1-Phenylethan-1-amine (the amine donor), 20 mmol L⁻¹ 1-Boc-3-piperidone (the ketone acceptor), 2 mmol L⁻¹ PLP, and 100 mmol L⁻¹ triethanolamine buffer (pH 7.0). Reactions were incubated at 30 °C for 30 min and then quenched with HCl prior to derivatization. The product (R)-1-Boc-3-aminopiperidine was derivatized with o-phthalaldehyde (OPA) prior to HPLC analysis. Briefly, 5 μL borate buffer and 2 μL reaction sample were mixed, followed by addition of 2 μL OPA reagent and vortexing; then 11 μL buffer was added. After 1 min the derivatized sample was injected for HPLC analysis. HPLC conditions: Shimadzu LC-20 instrument with a ZORBAX SB-C18 reversed-phase column. Mobile phase A: B = 6:4, flow rate 1.0 mL min⁻¹, detection wavelength 330 nm, column temperature 30 °C.

2.2.4. Determination of optimum temperature and pH

Standard aminotransferase assays were assembled at pH 7.0 and contained 20 mmol L⁻¹ (R)-1-Phenylethan-1-amine (the amine donor), 20 mmol L⁻¹ 1-Boc-3-piperidone (the ketone acceptor), 2 mmol L⁻¹ pyridoxal-5′-phosphate (PLP) and buffer. FalAT was added to a final concentration of 10 μg mL⁻¹. Reactions were performed for 30 min in a water bath at temperatures ranging from 20 to 65 °C (increments of 5 °C). Reactions were terminated by addition of HCl and product formation was quantified by HPLC to determine specific activity at each temperature. pH profiles were obtained using the same reaction composition across buffers covering pH 5.0–10.0: sodium phosphate buffers (pH 5.0–8.0), triethanolamine–HCl (pH 8.0–9.0), and glycine–NaOH (pH 9.0–10.0). Reactions were carried out at 30 °C for 30 min and terminated by HCl prior to HPLC quantification. All assays were performed in triplicate. The maximum activity observed across temperature or pH values was defined as 100 %, and relative activities were reported accordingly.

2.2.5. Substrate–Protein Interaction Network (SPIN) analysis

Three-dimensional models of FalAT were predicted using the AlphaFold Server (https://alphafoldserver.com/)and evaluated via the SAVES server for model quality. Using the structural model, FalAT covalently complexed with PLP and 1-Boc-3-piperidone was subjected to molecular docking in Discovery Studio 2019. Protein preparation included addition of hydrogens, repair of missing residues and charge assignment. Ligands were energy-minimized and charge-balanced prior to docking. Flexible docking was performed with the CDOCKER algorithm within the active-site region, generating 100 conformations that were ranked by interaction energy [13,14]. Residues within an 8 Å sphere centered on the ligand were extracted in PyMOL as candidate positions. Virtual saturation mutagenesis at each candidate position was performed using FOLDX 5.0, and average ΔΔG values were computed; residues with average ΔΔG <0 kcal mol⁻¹ were retained as thermodynamically favorable mutation sites. A further spatial filter was applied, requiring the minimum distance between the residue C-β atom and any ligand atom to be less than 8 Å, in order to focus on residues in direct proximity to the substrate [[15], [16], [17], [18]].

2.2.6. Substrate specificity assays

Substrate specificity was evaluated in standard reactions containing 20 mmol L⁻¹ (R)-1-Phenylethan-1-amine (the amine donor), 20 mmol L⁻¹ of the respective ketone substrate, 2 mmol L⁻¹ PLP and 100 mmol L⁻¹ triethanolamine buffer (pH 7.0). Reactions were performed at 30 °C for 30 min, terminated with HCl, and product formation was quantified by HPLC. Tested substrates included: 3-hydroxyacetophenone (S1), 4-hydroxyacetophenone (S2), 3-methoxyacetophenone (S3), 4-methoxyacetophenone (S4), 4-hydroxybutyrophenone (S5), 4-(4-methylphenyl)-2-butanone (S6), 3-piperidone (S7), 4-tert-butylcyclohexanone (S8), 4-methylcyclohexanone (S9), Cyclohexanone (S10), 1-acetylnaphthalene (S11), and 3-acetyl-2-fluoropyridine (S12). For substrates S7–S10, the corresponding amine products were derivatized with OPA prior to HPLC analysis following the procedure described in Section 2.2.3.

2.2.7. Kinetic parameter determination

Kinetic assays were performed with a fixed concentration of (R)-1-Phenylethan-1-amine (the amine donor) at 20 mmol L⁻¹ and PLP at 2 mmol L⁻¹, pH adjusted to 7.0. The concentration of the ketone substrate 1-Boc-3-piperidone was varied from 0.1 to 80 mmol L⁻¹. Enzyme concentration was 10 μg mL⁻¹. Reactions were incubated at 30 °C for 60 min. Product formation at each substrate concentration was measured to calculate initial rates. One unit (U) was defined as the amount of enzyme catalyzing the formation of 1 μmol product per minute at 30 °C. Kinetic parameters (K_m, k_cat and k_cat/K_m) were obtained by fitting the Michaelis–Menten equation using nonlinear regression.

2.2.8. Structural and functional analyses

To elucidate how mutations affect enzyme structure and substrate binding, molecular docking of wild-type and mutant enzyme–substrate complexes was performed using Discovery Studio 2019, and interaction types (hydrogen bonds, electrostatic interactions, hydrophobic contacts and unfavorable contacts) were analyzed. Active-site volumes and surface areas were computed using the CASTp and DoGSiteScorer web tools [[19], [20], [21]]. MD simulations (100 ns) of the wild-type and mutant enzyme–substrate complexes were carried out with GROMACS 2023 using the AMBER99SB-ILDN force field and the TIP3P water model; counterions (Na⁺/Cl⁻) were added to neutralize the system. After energy minimization, equilibration under NVT/NPT ensembles and production runs were performed. RMSD and RMSF profiles were extracted using GROMACS utilities to assess structural stability and pocket dynamics [[22], [23], [24], [25]].

3. Results

3.1. Mining and enzymatic characterization of FalAT

To address bottlenecks in industrial application of R-ATAs, Fungal sources were targeted to identify enzymes possessing distinct structural features and enhanced substrate adaptability. Using the well-characterized actinobacterial R-ATA (PDB ID: 3WWH) as a template, which has been successfully applied to the industrial production of sitagliptin intermediates after engineering [26,27], Sequence alignments and phylogenetic analyses were performed to identify candidate R-ATAs. From 24 type-IV transaminase sequences with >40 % identity, a previously uncharacterized candidate from Fusarium albosuccineum (UniProt ID: A0A8H4KFE3) was selected and designated FalAT (Fig. 1a) [10,28,29].

Fig. 1.

Mining and biochemical characterization of the novel (R)-amine transaminase FalAT from Fusarium albosuccineum. (a) Phylogenetic analysis of FalAT and representative (R)-ATAs; (b) Sequence alignment highlighting conserved residues; (c) Catalytic conversion of 1-Boc-3-piperidone to (R)-1-Boc-3-aminopiperidine; (d) Temperature dependence of FalAT activity measured in 100 mmol L⁻¹ triethanolamine buffer (pH 7.0) with 20 mmol L⁻¹ 1-Boc-3-piperidone and 20 mmol L⁻¹ (R)-1-phenylethan-1-amine (the amine donor); (e) pH profile of FalAT activity determined at 30 °C under the same substrate conditions.

FalAT shares approximately 46 % overall sequence identity with previously reported bacterial R-ATAs (such as the actinobacterial 3WWH) and with our lab's earlier M16AT [30], but displays marked differences in the O-pocket region of the active site (Fig. 1b). These distinctions may underlie FalAT's distinct capacity to accommodate bulky substrates. Consistent with this, both FalAT and selected mutants converted 1-Boc-3-piperidone to (R)-1-Boc-3-aminopiperidine (Fig. 1c) with high enantioselectivity (ee > 99 %; Fig. S1–2), establishing a link between structural divergence and functional novelty and motivating the subsequent targeted mutation design.

FalAT was cloned into pET-15b with an N-terminal 6 × His tag, expressed in BL21(DE3) and purified to give soluble protein (Fig. S3). The enzyme's temperature and pH optima were determined systematically. As shown in Fig. 1d, FalAT exhibits maximal activity at 30 °C; activity declines rapidly at higher temperatures (approximately 50 % activity at 40 °C and 20 % at 65 °C), indicating heat sensitivity likely due to thermal denaturation. At low temperatures, although the protein remains structurally stable, reduced molecular motions limit substrate binding and catalysis. The pH profile (Fig. 1e) indicates a maximum at pH 7.0, suggesting that neutral conditions best preserve structural integrity and catalytic residue protonation states. Activity decreased markedly at extreme pH (<5.0 or >9.0), likely reflecting disturbed charge distribution or altered protonation of catalytic residues. Accordingly, subsequent experiments were conducted at pH 7.0 using triethanolamine buffer.

3.2. Targeted mutagenesis based on Substrate–Protein Interaction Network (SPIN)

To systematically improve FalAT activity toward 1-Boc-3-piperidone, a Substrate–Protein Interaction Network (SPIN) strategy was developed. SPIN is a structure-guided pipeline that iteratively screens potential mutation sites and designs substitutions based on substrate physicochemical properties to rationally enhance enzyme activity. As transaminases commonly form dimers, the FalAT structure was predicted using AlphaFold Server. The predicted model displayed well-defined α-helices and β-sheets and an overall stable fold (ipTM = 0.83, pTM = 0.88). Moreover, visualization of the predicted model indicated that most residues were located in the blue region (pLDDT >90), and the PAE plot exhibited an overall low error distribution (Fig. S4). Model quality assessment via the SAVES server reported that 92.4 % of residues were located in allowed regions, which is above the 90 % confidence threshold and supports the reliability of the model for downstream docking, site identification, and rational mutagenesis (Fig. S5). Using Discovery Studio 2019, FalAT complexed with PLP and 1-Boc-3-piperidone was docked; the lowest energy pose (CDocker interaction energy = −73.7731 kcal mol⁻¹) was selected. Residues within an 8 Å sphere of the ligand center were collected (70 residues) (Fig. 2a). Virtual saturation mutagenesis at each position using FOLDX 5.0 (all 20 amino acids per site) produced average ΔΔG values, and residues with average ΔΔG <0 kcal mol⁻¹ were retained as thermodynamically favorable (20 sites in total) (Table S3).

Fig. 2.

Substrate–Protein Interaction Network (SPIN)-guided mutagenesis strategy for FalAT engineering. (a) Molecular docking and virtual saturation mutagenesis map showing residues within 8 Å of the substrate: red residues have average ΔΔG <0 kcal mol⁻¹ (thermodynamically favorable), blue residues have average ΔΔG ≥0 kcal mol⁻¹ (unfavorable); (b) Spatial filtering based on minimum C-β–ligand atom distances; (c) Functional-group-guided design of active-site mutations to enhance substrate binding and catalytic efficiency.

A further spatial filter was applied by requiring the minimum distance between the residue C-β and any ligand atom to be less than 8 Å, in order to focus on residues directly interacting with the substrate. After this secondary selection, nine candidate positions in the binding pocket with average ΔΔG less than 0 and appropriate proximity were retained: L52, H53, Y58, F113, Q115, R126, V145, H180, and W182 (Fig. 2b). Thus, the mutation search space was efficiently reduced to nine residues. Given that substrate specificity is determined by residues mediating binding in the active site, Substitutions were designed based on the substrate's functional groups to enhance complementary interactions.

1-Boc-3-piperidone contains two key functional groups: a hydrophobic tert butyl moiety and a polar carbonyl, while the piperidine ring imposes rigidity and steric demand. The tert butyl group cannot form hydrogen bonds and instead favors hydrophobic contacts with residues such as Ala and Val; therefore, residues in proximity to this group, including L52, H53, and H180, were considered for mutation to more hydrophobic residues such as Ala or Val to enhance hydrophobic complementarity and lower binding free energy [31]. However, Structural analysis indicated that the region surrounding the substrate tert-butyl group is spatially constrained, and the branched side chain of valine could introduce additional steric clashes with neighboring residues. Such local interference might distort the hydrophobic pocket and reduce conformational stability. In contrast, Ala, with its smaller and nonpolar side chain, maintains pocket compactness while strengthening hydrophobic interactions. Therefore, Ala was regarded as a more conservative and spatially compatible choice for this region. The carbonyl oxygen acts as a hydrogen bond acceptor and motivated mutations of nearby residues including Q115, W182, and Y58 to Ser or Ala in order to potentially introduce new hydrogen bond donors and improve substrate orientation and stabilization. Ser was chosen instead of threonine despite their similar hydroxyl functionality because Ser has a shorter and more flexible side chain that better fits the compact active-site cavity of FalAT. In contrast, the additional methyl group of Thr would increase steric hindrance and alter the hydrogen-bond geometry toward the substrate's carbonyl oxygen. As Q115, W182, and Y58 are located in a polar microenvironment close to the substrate carbonyl, spatial efficiency is critical. The smaller size of Ser allows favorable hydrogen-bond orientation without disrupting local residue packing. Although Thr remains a potential alternative, a conservative Ser mutation was adopted in this round to minimize conformational interference. For the rigid and bulky piperidine ring located near F113, V145, and R126, substitutions to residues with smaller or less bulky side chains such as Ala, Val, Ser or Phe were proposed to expand the local cavity volume or adjust the hydrophobic environment. However, considering that the side chains of Val and Phe are relatively bulky, they may cause steric clashes with neighboring residues and disturb the local packing within the confined catalytic pocket. In contrast, Ala can effectively alleviate steric crowding, while Ser, with its small size and hydroxyl group, is capable of forming stabilizing hydrogen bonds, thereby fine-tuning the pocket geometry and interaction strength while maintaining structural integrity. Therefore, Ala and Ser were regarded as the most suitable substitutions, balancing spatial adaptability and structural stability. These designs collectively formed a chemically and structurally rational mutation plan [[32], [33], [34]], as illustrated in Fig. 2c.

Following SPIN design, 15 mutants were constructed and screened. The single substitution R126A exhibited the most pronounced enhancement, approximately fourfold relative to wild-type. This improvement substantially exceeded that observed for other single mutants such as Q115S, which showed an approximately twofold increase in activity, indicating that R126 likely plays a pivotal role in substrate binding and transition state stabilization. Given its marked activity increase, R126A was selected for detailed kinetic, structural, and molecular dynamics analyses.

In contrast, certain combinatorial mutants (V145S/R126A) showed markedly reduced activity (approximately 50 % of wild-type), indicating that beneficial single mutations do not necessarily combine additively and may exhibit negative epistasis (Fig. 3b). Docking results suggest two possible causes: (1) the combined substitutions introduce new steric clashes that hinder substrate entry into the active site; (2) the double mutation perturbs hydrogen-bonding networks critical for substrate positioning and intermediate stabilization [35]. Excessive local flexibility from combined mutations may also destabilize pocket dynamics, shortening enzyme–substrate residence time and diminishing catalysis [36]. These findings underscore the importance of careful site selection and prioritization within the SPIN framework.

Fig. 3.

Enzymatic activity screening and substrate specificity analyses of wild-type FalAT and mutants. (a) Relative catalytic activities of single-site mutants toward 1-Boc-3-piperidone compared with wild-type; (b) Catalytic activities of combinatorial mutants containing R126A and other substitutions; (c) Substrate specificity profiles showing relative activities of wild-type FalAT and the R126A mutant toward each substrate under standard assay conditions (20 mmol L⁻¹ substrate, 20 mmol L⁻¹ (R)-1-Phenylethan-1-amine (the amine donor), 2 mmol L⁻¹ PLP, pH 7.0, 30 °C); (d) Chemical structures of the tested ketone substrate panel, including aromatic, aliphatic, and cyclic ketones.

3.3. Kinetic parameters, substrate scope, and semi-preparative application of R126A

Kinetic assays comparing FalAT and R126A across varying concentrations of 1-Boc-3-piperidone are summarized in Table 1. R126A displays a k_cat/K_m of 2.05 s⁻¹ mM⁻¹,approximately 13-fold higher than wild-type. The K_m of R126A is significantly reduced compared to the wild-type, indicating enhanced substrate affinity. Notably, the k_cat of R126A is slightly increased, suggesting accelerated turnover or product release. Structural analyses indicate that R126A enlarges the active-site cavity and reduces steric hindrance, thereby optimizing substrate accommodation and catalysis.

Table 1.

Kinetic parameters.

Enzyme	Km (mM)	kcat (s-1)	kcat/Km (s-1·mM-1)
WT	0.67 ± 0.15	0.11 ± 0.01	0.16 ± 0.04
R126A	0.10 ± 0.01	0.21 ± 0.01	2.05 ± 0.22

To further demonstrate the catalytic performance of FalAT and its mutant, a comparison of their kinetic parameters with representative reported transaminases was conducted (Table S1). The results show that R126A outperforms the wild-type in terms of kinetic parameters, indicating substantial improvements in both substrate affinity and catalytic turnover. When compared with previously reported transaminases, the catalytic efficiency of R126A was approximately 2.5-fold higher than that of MwoAT (originating from Mycobacterium sp. KWX21888.1), but lower than that of M16AT (originating from Mycobacterium sp. ACS1612), which catalyzes the small-molecule substrate pyruvate. Nevertheless, R126A exhibited a markedly lower K_m than M16AT, indicating a stronger substrate affinity despite its lower k_cat/K_m. This difference in overall catalytic efficiency is likely attributable to the substantial structural differences between the respective substrates. Moreover, even when acting on the sterically bulky substrate 1-Boc-3-piperidone, R126A displayed a catalytic efficiency comparable to that of the reference enzyme 3WWH (originating from Arthrobacter sp. KNK168), demonstrating its favorable substrate adaptability and promising biocatalytic potential [26,30,37].

Activity toward a panel of ketone substrates that included simple acetophenone derivatives, extended chain ketones and cyclic ketones was also assessed (Fig. 3c). As shown in Fig. 3d, for 4-methoxyacetophenone (S4) both R126A and wild-type exhibited high catalytic efficiencies relative to other substrates. For substrates such as 3-hydroxyacetophenone(S1) and 3-acetyl-2-fluoropyridine (S12), activity differences between R126A and the wild-type were small. However, for bulkier, longer-chain substrates such as 4-hydroxybutyrophenone (S5) and 4-(4-methylphenyl)-2-butanone (S6), R126A showed clearly superior activity compared with the wild-type, likely owing to active-site conformational changes that reduce steric clash and improve accommodation of larger molecules. For cyclic ketones such as 3-piperidone (S7), 4-tert-butylcyclohexanone (S8), 4-methylcyclohexanone (S9), and cyclohexanone (S10), R126A did not show improvements and in some cases was inferior to the wild-type, possibly due to ring-induced steric effects and the enzyme's specific recognition mechanisms for cyclic substrates.

To further evaluate the industrial applicability of the engineered enzyme, a semi-preparative scale reaction catalyzed by the R126A mutant was carried out under optimized conditions (substrate concentration: 20 mmol L⁻¹ 1-Boc-3-piperidone and 20 mmol L⁻¹ (R)-1-Phenylethan-1-amine (the amine donor), pH 7.0, 30 °C). The yield reached its maximum after reacting for 24 h, the desired product (R)-1-Boc-3-aminopiperidine was obtained with a yield of 12.1 % (Fig. S6). These results demonstrate that the R126A mutant can efficiently catalyze the target transformation under enlarged reaction conditions while maintaining excellent stereoselectivity. Nevertheless, the moderate yield indicates that the current catalytic efficiency still requires further enhancement through iterative saturation mutagenesis or active-site pocket engineering to meet the demands of industrial application.

3.4. Structural and functional analysis

To elucidate the structural basis for R126A's improved catalytic performance, molecular docking, active-site geometry analyses, and molecular dynamics simulations were combined to compare the wild-type and R126A structures and their substrate interactions.

Docking of the substrate complex into the AlphaFold FalAT model revealed a relatively dense interaction network in the wild-type complex. In the wild-type FalAT–substrate complex, residues H180, W182, Y58, F113, A274, V60 and H53 form the principal binding interface. These residues maintain a stable hydrophobic–hydrogen bond network via conventional hydrogen bonds, pi–alkyl and alkyl contacts, facilitating precise substrate recognition and positioning (Fig. 4a–b). In contrast, the R126A binding mode underwent significant rearrangement: while interactions with A274, W182 and V147 are retained, the key hydrogen bond involving Y58 is lost in the mutant. New C–H bond, pi–alkyl and alkyl contacts involving G214, A126 and H53 emerged in R126A (Fig. 4c–d). This reorganization suggests an increased contribution of hydrophobic and π-stacking interactions and a shift from polarity-dominated to hydrophobic-dominated binding, which may underlie the observed differences between binding stability and catalytic efficiency.

Fig. 4.

Structural and molecular dynamics analyses of wild-type FalAT and the R126A mutant. (a) Active-site pocket volume and surface area of the wild-type complex; (b) Binding interactions in the wild-type complex; (c) Active-site pocket volume and surface area of the R126A complex; (d) Binding interactions in the R126A complex; (e) Root mean square deviation (RMSD) profiles; (f) Root mean square fluctuation (RMSF) profiles highlighting reduced flexibility in residues 120–140.

Cavity analyses using CASTp and DoGSiteScorer show that the active-site pocket volume increases from 366.08 Å³ in wild-type to 434.18 Å³ in R126A, and the surface area increases from 425.24 Å² to 450.36 Å². The enlarged pocket volume and surface area facilitate accommodation of structurally diverse substrates, particularly those with longer-chains or larger bulk [38,39], consistent with the improved activity observed experimentally for larger substrates.

MD simulations (100 ns) were performed with GROMACS 2023 on both wild-type and R126A enzyme–substrate complexes to evaluate backbone stability, local flexibility, and pocket dynamics. Simulations were conducted under neutralized conditions with NPT ensemble settings.

RMSD analysis revealed that both protein backbones reached convergence without large-scale conformational drift, indicating that the overall structures remained stable throughout the 100 ns simulation. Nevertheless, the fluctuation amplitudes differed markedly: the WT exhibited an average RMSD of approximately 2.2 Å with periodic excursions up to 2.5 Å, whereas the R126A mutant displayed an average RMSD of approximately 1.8 Å and a substantially reduced overall amplitude. These data demonstrate that R126A possesses enhanced global conformational stability and a more compact backbone during the molecular dynamics trajectory. We hypothesize that this stabilization originates from local re-arrangements near the active-site pocket; replacement of the bulky, charged arginine with the smaller alanine alleviates steric hindrance and thereby consolidates the overall fold. The improved structural rigidity correlates well with the experimentally observed increase in catalytic efficiency. (Fig. 4e).

RMSF analysis further revealed that the R126A mutation significantly reduces local flexibility around the active-site vicinity(The equilibrium phase from 90 to 100 ns is selected for calculation). In residues 120–140, the wild-type exhibited an RMSF peak of 2.76 Å, whereas R126A decreased to 1.25 Å, representing a 55 percent reduction, indicating suppressed conformational fluctuations and enhanced local rigidity. Such local stabilization likely helps maintain consistent substrate recognition geometries, thereby enhancing catalytic efficiency (Fig. 4f).

In summary, the R126A mutant improves FalAT catalysis through multiple, synergistic mechanisms: (1) enlarging pocket volume and reducing steric hindrance to facilitate substrate access and accommodation; (2) enhancing hydrophobic character of the pocket to stabilize hydrophobic substrate moieties; (3) increasing local rigidity to promote stable enzyme–substrate interactions and productive conformations; and (4) stabilizing binding with reduced fluctuations to increase substrate residence time and conversion efficiency. These structural and dynamic changes underpin the observed enhancement in catalytic performance for R126A.

4. Discussion

The SPIN (Substrate–Protein Interaction Network) strategy established in this study represents an integrated semi-rational design framework that organically combines substrate docking, virtual mutagenesis, and dual energy–distance screening. This workflow is substrate-centered and driven by structural–energetic coupling principles. The novelty of SPIN lies not in the innovation of a single computational algorithm, but in the integration of multiple computational modules into a unified, mechanism-oriented mutation screening system. This enables an efficient transition from computational prediction to experimental validation, significantly improving the accuracy of identifying catalytically relevant residues while markedly reducing experimental workload.

Unlike conventional alanine-scanning or energy-based mutagenesis methods, SPIN follows a substrate-guided logic for residue selection. Alanine scanning systematically replaces residues with alanine to probe functional importance, yet neglects substrate conformations and interaction contexts, making it difficult to capture synergistic or allosteric effects. Energy-based mutation strategies primarily rely on global ΔΔG or stability predictions, which do not necessarily correlate with catalytic activity. In contrast, SPIN jointly considers residue–substrate distance (≤8 Å), FoldX calculated ΔΔG energy changes, and the compatibility between substrate functional groups and active-site residues, ensuring that the selected positions possess catalytic relevance both structurally and energetically.

Compared with existing enzyme-design platforms, SPIN exhibits distinct advantages in efficiency, hit rate, and practical applicability. Traditional semi-rational design tools (CAST and HotSpot Wizard) typically require constructing hundreds of mutants per round, with the hit rate for activity-enhancing mutations is generally low [40,41]. FireProt, on the other hand, mainly targets thermostability optimization and achieves limited success in catalytic improvement [42]. Although deep mutational scanning (DMS) can assess 10³–10⁴ variants simultaneously, it entails high experimental cost and yields a low proportion of beneficial mutations [43]. Rosetta provides high-precision conformational sampling but has limited predictive accuracy for enzyme design, with low experimental success rates for active-site redesign and typically requiring over 100 CPU hours per round; FoldX is computationally fast but lacks substrate information and spatial constraints, resulting in lower success rates for catalytic optimization, with predictions focusing more on stability than catalytic performance. By integrating substrate-guided spatial filtering (residue–ligand distance ≤8 Å), FoldX-based ΔΔG evaluation, and functional group matching, SPIN achieves synergistic selection for both structural tolerance and catalytic relevance. This energy–spatial coupling strategy retains the computational efficiency of FoldX (approximately 6 CPU hours per screening round) while enhancing functional correlation, requiring less than one-tenth of the computational cost of Rosetta [44,45].

In this study, SPIN successfully reduced the mutational search space of FalAT from hundreds of possible sites to 15 rationally designed variants, among which five (33 %) displayed higher catalytic activity than the wild-type enzyme. The optimal mutant, R126A, exhibited approximately 4-fold increase in catalytic activity and approximately 13-fold improvement in catalytic efficiency. These results demonstrate that SPIN can effectively enrich functional variants within a limited experimental scope and accurately guide the identification of key residues. Although R126A showed the best catalytic performance, this residue still holds potential for further optimization in terms of active-site geometry and catalytic dynamics. Future studies could employ iterative saturation mutagenesis at R126 to fine-tune hydrophobicity, polarity, and cavity volume while maintaining structural stability. Integrating such targeted saturation mutagenesis within the SPIN framework offers a promising strategy to balance enzyme stability and catalytic flexibility.

5. Conclusion

A novel fungal R-ATA, FalAT, was mined from Fusarium albosuccineum and its enzymatic properties together with catalytic behavior were comprehensively characterized. FalAT exhibits optimal activity at pH 7.0 and 30 °C and catalyzes 1-Boc-3-piperidone to (R)-1-Boc-3-aminopiperidine with high enantioselectivity (ee > 99 %). To enhance its catalytic performance, a Substrate–Protein Interaction Network (SPIN) strategy was applied, integrating structural analysis, molecular docking, virtual saturation mutagenesis, and energy–distance filtering. Experimental validation of candidate mutants identified R126A as the most effective variant, showing approximately a 4-fold increase in catalytic activity and a 13-fold improvement in catalytic efficiency compared with the wild type. MD simulations and pocket analyses indicate that R126A improves catalysis by enlarging the binding cavity, reducing steric hindrance, and enhancing local stability, thereby facilitating substrate binding and turnover.

Future work should focus on further engineering of FalAT for industrial robustness. Key targets include increased thermostability and organic-solvent tolerance because many pharmaceutical reactions operate at elevated temperatures, long durations, or in organic media. Potential strategies include introduction of disulfide bonds to enhance global stability, multipoint cooperative mutations to optimize flexible regions, and directed evolution under high-throughput selection to systematically improve thermostability and solvent tolerance. Combining these approaches could broaden FalAT's applicability in industrial green synthesis.

Moreover, FalAT and its engineered variants may find broader applications beyond chiral amine synthesis. The improved activity of R126A toward longer-chain ketones suggests potential utility in complex molecule synthesis. Future exploration of these enzymes in fine chemical transformations, natural product derivatization, and pharmaceutical intermediate synthesis could further promote industrial deployment of R-ATAs.

CRediT authorship contribution statement

Ruizhou Tang: Writing – review & editing, Writing – original draft, Visualization, Methodology, Formal analysis, Data curation. Jiahuan Li: Validation, Formal analysis, Data curation. Xiaole Yang: Visualization, Software, Resources, Data curation. Xia Tian: Visualization, Validation, Resources. Zehui Wang: Resources, Investigation, Formal analysis. Xuning Zhang: Visualization, Formal analysis, Data curation. Tingting Li: Writing – original draft, Validation, Supervision, Project administration, Methodology, Funding acquisition, Conceptualization.

Funding

We are grateful for financial support from for Jiangsu Ocean University for start-up funds (KQ19026).

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Xuning Zhang is currently employed by Jiangsu BestEnzymes Biotech Co. Ltd.

Appendix A. Supplementary data

The following is the Supplementary data to this article: