Structural and functional insights into the iminium ion intermediate in AsKslB-mediated Pictet-Spengler reaction

Abstract

Pictet-Spengler (PS) reactions are pivotal in the biosynthesis of bioactive alkaloids and pharmaceuticals, yet key structural details underlying their enzymatic catalysis remain insufficiently understood. We identified AsKslB from Actinosynnema sp. ALI-1.44 as a Pictet-Spenglerase with broad substrate scope that catalyzes the stereoselective condensation of l-tryptophan (l-Trp) and α-ketoglutarate (α-KG) to form kitasetalic acid (KA), a tetrahydro-β-carboline (THβC). High-resolution crystal structures of apo, substrate-, intermediate-, and product-bound forms elucidate the full catalytic trajectory and key residues. Crucially, the elusive iminium ion intermediate (IM-1) and a synchronously released water molecule are captured, providing direct structural evidence for the initiating cyclization step of Pictet-Spengler reaction. Glu276 undergoes conformational changes essential for catalysis. These findings offer detailed mechanistic insights into Pictet-Spenglerase function and establish AsKslB as a promising biocatalyst for stereoselective N-heterocycle synthesis.

1. Introduction

The Pictet-Spengler reaction (PSR) can efficiently generate alkaloids by condensing β-arylethylamine and carbonyl compounds (aldehydes or ketones), which is prolongedly and extensively applied in organic synthesis [[1], [2], [3]]. These intermolecular cyclizations can also be catalyzed by Pictet-Spenglerase (PSase) under mild environmentally benign condition [[1], [2], [3]]. Due to their catalytic efficiency and selectivity, PSases are progressively engineered as novel biocatalysts for constructing valuable N-heterocyclic compounds [4,5]. To date, various PSases have been identified across different living systems, including plants, animals, fungi, and bacteria [3,[6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21]]. However, only four PSases have been well characterized both biochemically and structurally (Fig. 1). They can be categorized into two main types: one is plant derived PSases, including strictosidine synthase (STR) and norcoclaurine synthase (NCS); the other is identified from rare actinomycetes including McbB and KslB (Fig. 1) [[6], [7], [8], [9], [10], [11], [12]]. Especially, the KslB can efficiently biosynthesize 1,1-disubstituted THβC with high diastereoselectivity by asymmetric construction using l-Trp derivatives and α-keto acid [10,11]. Furthermore, KslB exhibits broad substrate promiscuity and is compatible with cascade reactions, highlighting its potential in biosynthesis of THβC scaffolds [11]. These features indicate that rare actinomycetes are new emerging resource in PSases development. Therefore, continuous mining of new PSases using KslB as a unique screening marker could effectively expand the pool of available tool enzymes. More importantly, this discovery approach enables access to more intuitive and reliable enzymatic data including the essential natural transition state: formation and conversion of iminium ion intermediate (IM-1) (Fig. S1), which has remained elusive in the PSR and has never been directly observed.

Fig. 1.

PSR catalyzed by four kinds of PSases with biochemical and crystallographic characterized. (A) STR. (B) NCS. (C) McbB. (D) KslB.

Given the feasibility of the aforementioned strategy, we employ KslB as a probe to perform a genome-wide screening of rare actinomycetes, leading to the discovery of a new homologous PSase, AsKslB, from Actinosynnema sp. ALI-1.44. AsKslB catalyzes the stereoselective reaction between l-Trp and α-KG to form KA and exhibits broad substrate tolerance. Structural analysis of AsKslB in apo, l-Trp-, KA-, and IM-1-bound states reveals the full catalytic trajectory. These multistate co-crystal structures provide a comprehensive view of the active site architecture and catalytic configuration of AsKslB, highlighting the essential and dynamic role of Glu276 in AsKslB activity. Most notably, the successful capture of the elusive IM-1 offers the first direct structural snapshot of this key transition state in the PSR, enabling precise mechanistic dissection and paving the way for rational engineering and broader application of PSases.

2. Materials and methods

2.1. General experimental procedures and methodology

Escherichia coli DH5α (Invitrogen) was used as the host strain for general DNA cloning. For protein expression, E. coli BL21(DE3) (Novagen) was employed. The gene encoding AsKslB was derived from the genome of Actinosynnema sp. ALI-1.44 [22]. Tryptophan analogs, α-keto acids and aldehydes were obtained from Aladdin Scientific Corporation, Shanghai Yuanye Bio-Technology Co., Ltd. and Shanghai Macklin Biochemical Technology Co., Ltd. Analytical high performance liquid chromatography (HPLC) was operated on Waters Alliance e2695 Separations Module with a 2998 PDA detector using an Agilent Zorbax SB-C18 column (150 × 4.6 mm, 5 μm) under the following program: solvent A, 0.1 % AcOH in ddH₂O; solvent B, 0.1 % AcOH in CH₃CN; 5 % B to 95 % B (0−8 min), 95 % B (8−16 min), 95 % B to 5 % B (16−18 min), 5 % B (18−20 min); flow rate 0.4 mL/min; UV detection at 276 nm. HR-ESI-MS spectra were recorded on a ThermoFisher Q Exactive Plus (Thermo). Optical rotation was measured on a SGW-568A polarimeter (Shanghai INESA Physico-Optical Instrument Co., Ltd.). Nuclear magnetic resonance (NMR) spectra were recorded on an AVANCE III HD 600 spectrometer (Bruker) at 600 MHz for ¹H nucleus and 150 MHz for ¹³C nucleus.

2.2. Protein expression and purification of AsKslB

The gene encoding AsKslB was synthesized by Sangon Biotech (Shanghai, China). It was inserted between the NdeI and HindIII restriction sites of the pET-28a(+) vector, enabling expression of an N-terminal 6 × His-tagged fusion protein.

The recombinant plasmid pET-28a(+)/AsKslB was transformed into E. coli BL21(DE3) competent cells and cultured in 15 mL LB medium supplemented with 50 μg/mL kanamycin and incubated at 37 °C with shaking at 200 rpm for ∼12 h. The preculture was then transferred into 500 mL LB medium containing the same antibiotic and grown under identical conditions until the OD₆₀₀ reached 0.6–0.8. Protein expression was induced by the addition of isopropyl β-d-1-thiogalacto- pyranoside (IPTG) to a final concentration of 0.4 mM, followed by incubation at 16 °C with shaking at 160 rpm for 20 h. All subsequent steps were performed at 4 °C unless otherwise specified. Meanwhile, the KslB was purified as previously described [[10], [11], [12]].

Enzymatic assays of AsKslB were conducted following previously reported methods with minor modifications [[9], [10], [11], [12]]. Each reaction mixture (50 μL total volume) contained 50 mM HEPES buffer (pH 7.0), 1 mM l-Trp, 2 mM α-KG, and 6.0 μM AsKslB. A denatured AsKslB sample served as a negative control. Reactions were incubated at room temperature (∼23 °C) for 2 h and subsequently quenched by the addition of 100 μL methanol. After centrifugation (three rounds), a 30 μL aliquot of the supernatant was analyzed by HPLC.

To obtain the enzymatic product KA of AsKslB, a scale-up overnight reaction was performed. These overnight reactions (500 μL) containing 5 mM l-Trp, 5 mM α-KG and 93 μM AsKslB in 50 mM potassium phosphate (pH 8.0) were quenched when HPLC analysis indicated complete conversion. The mixtures were deproteinized by ultrafiltration, lyophilized, and the residues dissolved in ddH₂O for optical rotation measurement. After a second lyophilization, the AsKslB product was dissolved in D₂O for NMR and optical rotation identification. The KA, utilized as standard sample in optical rotation data comparison, was catalyzed by KslB as previously described [[10], [11], [12]].

For pH optimization, buffer systems ranging from pH 4.0 to 10.0 were screened, identifying pH 7.5 as optimal. Further refinement between pH 7.0−8.0 pinpointed pH 7.6 as the optimal condition. Temperature optimization was similarly performed, with initial screening from 15 to 50 °C indicating 25 °C as favorable. Fine-tuning between 20 and 30 °C confirmed 26 °C as the optimal reaction temperature. All temperature experiments were conducted in a 50 mM HEPES buffer system (pH 7.6) containing 1 mM l-Trp, 2 mM α-KG, and 6.0 μM AsKslB. Reactions were terminated by adding 100 μL methanol, followed by triple centrifugation, with 30 μL of supernatant collected for HPLC analysis.

To investigate the reaction kinetics, time-course assays were performed under isothermal conditions (26 °C), with sampling at 5, 10, 20, 30, 40, 50, 60, 90, and 120 min, as well as 16 h. All reactions were done in triplicate and quenched with methanol (100 μL) prior to centrifugation and HPLC analysis.

Substrate specificity was tested similarly. For l-Trp analogs, each reaction contained 50 mM HEPES buffer (pH 7.6), 1 mM analog, 2 mM α-KG, and 6.0 μM AsKslB (Fig. S6). For α-KG analogs, 1 mM l-Trp and 2 mM analog were used (Fig. S6). Reactions were incubated at 26 °C for 16 h, quenched, and centrifuged. Target products were collected, dried, and redissolved in 30 μL methanol for HR-ESI-MS analysis. A denatured AsKslB sample was used as a negative control.

Enzyme reactions and activity assays for AsKslB mutants were conducted under identical conditions as the wild-type.

2.3. Site-directed mutagenesis, expression and purification of the mutant proteins

Site-directed mutagenesis of AsKslB was conducted using the Quick Change Site-Directed Mutagenesis Kit or overlap extension-PCR method according to the manufacturer's protocol (with primer pairs listed in Table S1). A total of 13 AsKslB variants were constructed: F55A, I59A, R61A, F98A, M225A, V226A, S227A, Y232A, R258A, T263A, K266A, E276L, and T263A/K266A. The mutant plasmids were first transformed into E. coli DH5α for sequence verification. Confirmed constructs were subsequently transformed into E. coli BL21(DE3) for protein expression.

Protein expression and purification of each mutant followed the same protocol as that used for wild-type AsKslB. Ultimately, all mutants were successfully expressed and purified for enzymatic assays.

2.4. Crystallization and structure determination

Crystals of AsKslB were obtained after 5–6 days of incubation at 23 °C. Apo-AsKslB crystals were grown using the hanging-drop vapor diffusion method in 0.1 M sodium cacodylate (pH 6.5) and 20 % PEG 400. For the AsKslB/l-Trp complex, crystallization conditions comprised 27.2 % PEG 3350 and 0.1 M Tris (pH 7.5). In contrast, crystals of AsKslB/IM-1 and AsKslB/KA complexes were obtained via sitting-drop vapor diffusion with the following conditions: IM-1 complex—15 % PEG 8000, 0.1 M ADA (pH 6.6); KA complex—0.1 M Tris (pH 8.0), 20 % PEG 4000.

To capture the reaction intermediate and product, purified AsKslB (10 mg/mL) was co-incubated with l-Trp (5 mM) and α-KG (5 mM) at 8 °C for time intervals ranging from 1 to 6 h prior to crystallization screening. The IM-1-bound structure was determined from one of the crystals obtained with the 2 h incubation mixture, while the product-bound structure was derived from the 6 h incubation. The substrate-bound structure was obtained by incubating AsKslB with l-Trp alone. Crystals were cryoprotected by brief immersion in reservoir solution containing 25 % (v/v) glycerol and subsequently flash-cooled in liquid nitrogen.

X-ray diffraction data were collected at beamline BL18U of the Shanghai Synchrotron Radiation Facility (SSRF). The datasets were collected at resolutions of 2.6 Å for AsKslB, 2.4 Å for AsKslB/l-Trp, 1.7 Å for AsKslB/IM-1, and 2.1 Å for AsKslB/KA, respectively. Data processing and scaling were performed using HKL2000 for AsKslB, AsKslB/l-Trp and AsKslB/IM-1, DIALS/xia2 and Aimless (CCP4 suite) for AsKslB/KA. Initial phasing was achieved by molecular replacement with Phaser (CCP4), by using the McbB monomer (PDB ID: 3X27) as the search model. Model building and refinement were conducted using Coot and Refmac5, respectively. Crystallographic data and refinement statistics are summarized in Table S2.

Molecular graphics were prepared using PyMOL. Final coordinates and structure factors have been deposited in the Protein Data Bank under accession codes: PDB ID 9UWA (AsKslB), 9UWB (AsKslB/l-Trp), 9UWC (AsKslB/IM-1), and 9UWR (AsKslB/KA).

3. Results and discussion

3.1. Biochemical characterization of AsKslB

AsKslB (WP_076988106.1/ONI83199.1), derived from rare Actinosynnema sp. ALI-1.44, shares high sequence homology (68 % identity, 82 % similarity) with KslB [22]. To demonstrate its activity in vitro, the corresponding N-terminal His₆-tagged AsKslB was purified. Subsequent enzymatic analysis with substrates: l-Trp and α-KG was evaluated across various buffer systems (acetate, phosphate, glycine-NaOH, Tris, and HEPES) spanning a broad pH range (pH 4.0–10.0). As expected, all these reactions yielded the theoretical product KA, which in return further verified AsKslB is an actual new PSase (Fig. S2 and S3). Meanwhile, aforementioned reaction comparisons also synchronously confirmed the optimal catalysis condition of AsKslB is HEPES buffer with pH 7.6 at 26 °C (Fig. S2 and S3). Additionally, the presence of common cofactors (NAD, NADH, FAD, ATP) and metal ions (Ca²⁺, Mg²⁺, Cu²⁺, Fe²⁺, Co²⁺, Cd²⁺, Fe³⁺) had no significant effect on the catalytic process of AsKslB, which indicated the catalysis of AsKslB is economical for no need of any cofactors (Fig. S4). Time-course of AsKslB-catalyzed enzymatic conversion of l-Trp and α-KG to KA demonstrated no substantial increase in yield between 2 h and 16 h (Fig. S5).

To evaluate the substrate scope of AsKslB (Fig. S6), a series of enzymatic reactions were conducted. The results first revealed AsKslB can not recognize d-Trp, tryptamine, l-tryptophanol, or l-tryptophanamide, which suggest that l-stereochemistry and carboxyl group of Trp is essential in these PSRs (Fig. S7). In contrast, AsKslB can efficiently catalyze reactions with 4-fluoro(F)-, 5-F-, 6-F-, 5-chloro(Cl)-, 6-Cl-, 7-Cl-, 5-methyl(Me)-, 7-Me-, 5-methoxy(MeO) substituted l-Trps, exhibiting broad tolerance toward the substitutions on indole ring of l-Trp (Fig. 2A and C, Fig. S7A). However, AsKslB cannot utilize the α-Me-l-Trp, 5-hydroxyl-l-Trp, 5-Br- and 6-Br-l-Trp, which indicated that specific modification of these sites in l-Trp is limited during its catalysis (Fig. S7B).

Fig. 2.

Relative conversion efficiencies of AsKslB-catalyzed reaction toward various substrates. (A) Relative conversion efficiencies of AsKslB-catalyzed reaction with 1b and 1a−10a. (B) Relative conversion efficiencies of AsKslB-catalyzed reaction with 1a and 1b−8b. (C) Chemical structures of substrates could be recognized by AsKslB.

To further evaluate the substrate promiscuity of AsKslB, a panel of α-keto acids and aldehydes was screened (Fig. S6). AsKslB efficiently accepted 2-ketobutyric acid, 2-ketohexanoic acid, 2-oxovaleric acid, oxaloacetic acid, α-KG, 2-oxoadipic acid, succinic semialdehyde (SSA), and glyoxylic acid as counterparts of l-Trp to form THβCs (Fig. 2B and C, Fig. S7C). In contrast, no detectable products were observed when using pyruvic acid, 4-oxopentanoic acid, acetaldehyde, methylglyoxal, hydroxyacetic acid, diethyl ketomalonate, methyl 4-oxobutanoate, 1,4-benzopyrone, or 4-hydroxylphenylacetone in aforementioned PSR (Fig. S7D). All resulting products were subsequently confirmed with HR-ESI-MS analysis (Fig. S8−S25). Especially, the KA derived from AsKslB catalysis was finally identified by NMR and optical rotation comparison (Figs. S26 and S27, Table S3), which in return facilitated the stereoselectivity confirmation of all enzymatic products [11].

3.2. Overall structure of AsKslB

To elucidate the structural basis of AsKslB-mediated PSR, we determined the crystal structure of AsKslB at 2.6 Å resolution (PDB ID: 9UWA). Data collection and refinement statistics are summarized in Table S2. The asymmetric unit contains a single polypeptide chain (Fig. 3A). The biologically active dimer is formed through crystallographic symmetry operations (Fig. 3B and S28). The α3 and α4 helices from the neighboring monomer play a crucial role in stabilizing the dimer interface and contribute to the structural integrity of the catalytic site (Fig. 3A and B, Fig. S28). The overall structure of AsKslB shows significant similarity to other actinobacterial PSases. It aligns closely with KslB [The root-mean-square deviation (RMSD) = 0.670 Å) and McbB (RMSD = 1.112 Å] (Fig. S29). Each AsKslB monomer comprises two distinct domains: an N-terminal domain containing seven α-helices (α1–α7) and a short α-helices (H3₁₀-1), and a C-terminal domain composed of ten antiparallel β-strands (β1–β10) that form a β-barrel motif, flanked by two short α-helices (α8 and H3₁₀-2) and one long α9 helix (Fig. 3A). The absence of electron density for the N-terminal 6 × His tag and the Ala158–Ser179 segment indicates intrinsic flexibility in these regions.

Fig. 3.

Crystal structures of AsKslB, AsKslB/l-Trp, AsKslB/IM-1 and AsKslB/KA. (A) The ribbon diagram of the overall structure of apo-form AsKslB monomer. The α-helices and β-strands are colored blue and wheat, respectively. Secondary structure elements are labeled in black. NTD: N-terminal domain; CTD: C-terminal domain. (B) Comparison of the overall structures of AsKslB (green), AsKslB/l-Trp (organge), AsKslB/IM-1 (gray) and AsKslB/KA (cyan). The α3 and α4 helices are labeled in black. (C) Structural basis of l-Trp binding in the AsKslB active site. l-Trp and interacting residues are depicted as ball-and-stick models. Residues from the adjacent monomer (yellow Cα atoms) engage in binding are labeled with prime (′). The l-Trp omit map is shown as mesh contoured at 3.0σ and measured hydrogen-bond/hydrophobic interaction distances are labeled. (D) Location of IM-1 in the AsKslB active site. The omit maps of IM-1 and the water molecule (HOH) are shown as mesh contoured at 2.0σ and measured hydrogen-bond interaction distances are labeled. (E) KA binding mechanism. Amino acids involved in stabilizing the THβC of KA are shown as ball-and-stick models. The omit electron density map for KA is displayed as mesh and contoured at 3.0σ. (F) Confrontational changes of Glu276 and the binding ligands. Glu276 are colored green in AsKslB, wheat in AsKslB/l-Trp, gray in AsKslB/IM-1 and cyan in AsKslB/KA. Hydrogen bonds formed by Glu276 in AsKslB/l-Trp, AsKslB/IM-1 and AsKslB/KA are shown as wheat, gray and cyan dashes, respectively. Water molecule in AsKslB/IM-1 is colored and labeled pink.

To investigate reaction mechanism, we solved three co-crystal structures: AsKslB bound to l-Trp (AsKslB/l-Trp), to the IM-1 (AsKslB/IM-1) and to KA (AsKslB/KA), obtained by co-crystallization with 4 mM l-Trp, 2 mM α-KG, or both (Fig. 3, Fig. S30–S33, Table S2). The AsKslB/l-Trp, AsKslB/IM-1, and AsKslB/KA complexes were resolved at 2.4 Å, 1.7 Å, and 2.1 Å resolution, respectively (PDB ID: 9UWB, 9UWC, 9UWR). No substantial global conformational change was observed in the ligand-bound structures compared with the apo form (Fig. 3B). The RMSDs were 0.288 Å (529 Cα atoms) for AsKslB/l-Trp, 0.135 Å (522 Cα atoms) for AsKslB/IM-1, and 0.275 Å (508 Cα atoms) for AsKslB/KA, respectively.

3.3. Recognition and binding mechanism of the substrate l-Trp

As observed in McbB and KslB, the AsKslB homodimer forms a substrate-binding cavity at the monomer–monomer interface that accommodates l-Trp (Fig. S30A and B). Well defined electron density corresponding to l-Trp is observed within this pocket (Fig. 3C, Fig. S30C and D). The l-Trp backbone and side chain are stabilized through distinct interactions: the carboxylate group engages in hydrogen bonding with the backbone amide nitrogens of Met225, Val226, and Ser227, while hydrophobic contacts primarily anchor the indole side chain (Fig. 3C). Functional mutagenesis revealed that substitution of Met225 with alanine reduced catalytic activity by approximately 50 %, whereas the V226A mutant resulted in complete loss of activity.

In contrast, the S227A variant retained wild-type activity (Fig. 4). This is consistent with structural observations: the side chains of Met225 and Val226 project into the substrate pocket and likely participate in van der Waals interactions with the ligand, whereas the side chain of Ser227 is oriented away from the binding site, explaining its limited functional role. Additionally, the carboxylate group forms a fourth hydrogen bond with the hydroxyl group of Tyr232, and the main-chain amide nitrogen of l-Trp engages in a fifth hydrogen bond with the side chain of Glu276, the catalytic residue. These five cooperative hydrogen bonds effectively secure the main-chain conformation of l-Trp (Fig. 3C, Fig. S30C). The indole moiety of l-Trp is embedded in a hydrophobic cleft, where it is stabilized predominantly by shape and complementarity and van der Waals interactions. In the substrate-binding site, the indole ring of l-Trp is flanked by Arg61 and Thr263, forming a sandwich-like architecture. The methylene groups of the Arg61 side chain are positioned near the π-face of the indole, contributing CH–π interactions (∼3.5 Å), while the methyl group of Thr263 packs against the opposite face through hydrophobic van der Waals contacts (∼3.5 Å). This dual-face clamping restricts the rotational freedom of the indole ring, thereby enhancing substrate specificity and binding stability (Fig. 3C, Fig. S30D). A surrounding hydrophobic network composed of Phe55 (∼3.8 Å), Ile59 (∼4.3 Å), and Phe98 (∼3.8 Å)—all derived from the adjacent monomer—encloses the binding cavity and further stabilizes the indole ring (Fig. S30D). Alanine substitutions at these hydrophobic positions significantly reduced (I59A) or abolished (F55A, F98A) enzymatic activity, confirming their critical roles in substrate recognition and catalysis (Fig. 4).

Fig. 4.

HPLC profiles and relative activities of the AsKslB or its mutants-catalyzed reaction with α-KG and l-Trp. (A–B) (i) AsKslB, (ii−xiv) F55A, I59A, R61A, F98A, M225A, V226A, S227A, Y232A, R258A, T263A, K266A, E276L, and T263A/K266A mutants of AsKslB. (C) Relative activities of AsKslB and its mutants.

Comparison with the KslB/Trp complex (PDB: 9NSC) and McbB/Trp complex (PDB: 3X27) reveals conserved binding of the amino and carboxylate groups across all three enzymes (Fig. S31A and B). Unlike KslB and McbB, AsKslB adopts a unique binding mode for the indole ring of l-Trp (Fig. S31A and C). In AsKslB, the indole moiety is tightly clamped between Arg61 and Thr263 and orients nearly perpendicular to the aromatic side chain of Phe98 from the adjacent monomer, enabling a T-shaped interaction with Phe98. In contrast, in KslB and McbB, the indole rotates by ∼90°, adopting a parallel stacking mode with Phe98. On the opposite side of the indole ring, stabilization is provided by conserved hydrophobic residues—Thr261 in KslB and Leu85 in McbB—both of which occupy spatial positions analogous to Thr263 in AsKslB (Fig. S31B and C).

3.4. Structural insights into the binding of the iminium ion intermediate IM-1

The iminium ion intermediate IM-1 occupies the AsKslB active site through hydrogen bonding and hydrophobic interactions (Fig. 3D, Fig. S32). Its α-KG moiety forms four specific hydrogen bonds, with the α-carboxyl group interacting with Thr263 and Lys266 (2.7 Å and 2.5 Å, respectively), while the γ-carboxyl group establishes two hydrogen bonds with the side chain of Arg258 (3.0 and 3.1 Å) (Fig. 3D). These observations suggest that Arg258, Thr263, and Lys266 cooperatively contribute to α-KG coordination and stabilization within the active site. Although a crystal structure with α-KG alone was not obtained, the binding mode inferred from the IM-1 complex supports this hypothesis. Mutagenesis experiments further highlight the functional importance of these residues (Fig. 4). The complete loss of activity in the R258A and T263A/K266A double mutants indicates a critical role for Arg258 in γ-carboxyl anchoring, while the reduced activity in T263A and K266A single mutants implies that both residues are individually important but functionally complementary in stabilizing the α-carboxyl group during the catalytic process. The l-Trp-derived carboxyl group recapitulates the conserved binding mode of free l-Trp, forming four backbone-mediated hydrogen bonds to Met225-NH (2.8 Å), Val226-NH (2.7 Å), and Ser227-NH (2.9 Å) (Fig. S32C). The indole ring engages in multivalent van der Waals stabilization through Phe55, Ile59 and Phe98, mirroring the hydrophobic cage architecture observed in the l-Trp-bound state (Fig. S32D). Notably, crystallographic analysis revealed a structurally trapped water molecule bridging the catalytic residue Glu276 and the N atom of IM-1, stabilized by dual hydrogen-bonding interactions with Glu276-Oε (2.8 Å) and IM-1-Nα(3.0 Å) (Fig. 3D). The complete loss of enzymatic activity in Glu276L mutant, where this critical hydrogen-bonding network is disrupted, provides functional validation of the observed structural configuration (Fig. 4). Given that IM-1 is generated via Glu276-mediated dehydration of the l-Trp/α-KG adduct, this water species likely represents the post-reaction aqueous byproduct prior to its dissociation from the active site (Fig. 3, Fig. 5).

Fig. 5.

Proposed mechanism of AsKslB-catalyzed PSR.

3.5. Binding architecture for the final product KA

The product, KA, adopts a well-defined conformation in the active site, stabilized by a combination of hydrogen bonding and hydrophobic interactions with AsKslB (Fig. 3E, Fig. S33). The l-Trp derived and α-KG derived carboxyl moiety of KA forms the same hydrogen bonds interaction network as that in IM-1 bound structure (Fig. S33C and D). In the KA-bound structure, the indole moiety undergoes an ∼90° rotation, adopting a parallel stacking orientation with Phe98. Meanwhile, the methyl group of Thr263 continues to engage the opposite face via hydrophobic van der Waals interactions. Additionally, Phe55 and Ile59 from the adjacent monomer contribute hydrophobic contacts that further stabilize KA binding (Fig. 3E).

3.6. Structural dynamics of AsKslB during catalytic cycling

Structural snapshots of AsKslB in four distinct functional states—apo, l-Trp-bound, IM-1-bound, and KA-bound (Fig. 3)—reveal a coordinated reorganization of the active site during the PSR. Most active site residues maintain highly similar conformations throughout the catalytic cycle, with the notable exception of the catalytic residue Glu276, which undergoes marked positional shifts (Fig. 3F, Fig. S34). Uponl-Trp binding, the side chain of Glu276 reorients to form a hydrogen bond with the main chain amide nitrogen of l-Trp, positioning it for subsequent nucleophilic attack on α-KG (Fig. 3F). During the transition from l-Trp to IM-1, the backbone of l-Trp remains largely unchanged, while the indole ring undergoes a ∼12° rotation upon the reaction (Fig. 3F). Additionally, Glu276 adopts a distinct conformational change to stabilize a water molecule situated between its side chain and the IM-1 intermediate (Fig. 3F). This water molecule is most likely derived from the dehydration step that accompanies iminium ion formation.

During the conversion from IM-1 to the final product KA, cyclization of the l-Trp side chain leads to the formation of a THβC ring. This transformation is accompanied by an ∼90° rotation of the indole moiety, resulting in a markedly altered ligand conformation (Fig. S34). Despite this substantial rearrangement, the α-KG moiety remains anchored by a conserved hydrogen-bonding network involving Thr263 and Lys266, serving as a rigid scaffold throughout catalysis (Fig. 3, Fig. S32−S34). Concomitantly, Glu276 returns to a position closely resembling its orientation in the l-Trp bound state (Fig. 3F, Fig. S34). We propose that this dynamic repositioning of Glu276 during product formation is modulated by transient water generation and release (Fig. 3F), which temporarily alters the electrostatic environment to promote product stabilization and active site reset.

3.7. Structural comparison with KslB, STR and NCS

As previously described, the overall structures of AsKslB and KslB are highly conserved (Fig. S29), with nearly identical sequences and conformations of active site residues (Fig. S31). This structural conservation supports a shared catalytic mechanism, further corroborated by the superimposable conformations of l-Trp and the final product in both enzymes, as well as their comparable enzymatic activities (Fig. S31 and S35A).

Notably, in AsKslB complexes (AsKslB/IM-1 and AsKslB/KA), the α-KG moiety adopts a consistent conformation stabilized by hydrogen bonds with Arg258, Thr263 and Lys266 (Fig. 3D, Fig. S33D). In contrast, the KslB structure (PDB ID: 9NSU) reveals a distinct binding mode for SSA, a product-derived moiety (Fig. S35B). In comparison to KA, the γ-carboxyl group of SSA no longer engages in hydrogen bonding with the conserved Arg256 (corresponding to Arg258 in AsKslB). Instead, it undergoes an approximately 90° rotation toward the side of Glu274 (Glu276 in AsKslB). Notably, in this structure, the electron density for the loop containing Glu274 is not observable (Fig. S35B). The absence of defined electron density around Glu274 and the accompanying conformational change of the product moiety may represent structural dynamics involved in the substrate release process.

To date, aside from KslB and its homologs, only two kinds of plant-derived PSases have been structurally characterized: STR and NCS [3,6,7]. Structural comparison shows that these two plant enzymes are not structurally similar to each other and are entirely different from KslB (Fig. 3, Fig. S36). Sequence alignment indicates that AsKslB shares only 14 % and 13 % sequence identity with STR and NCS, respectively, with overall similarity below 30 % (data not shown). Therefore, these three types of enzymes are completely non-conserved in both sequence and structure and belong to distinct protein families. Interestingly, although AsKslB and STR share very low sequence identity and distinct overall structures, the chemical features of their respective substrates—l-Trp in AsKslB and tryptamine in STR—are similar. As a result, both enzymes employ analogous chemical strategies to stabilize substrate binding: hydrogen bonds and hydrophobic interactions orient the polar amino group and the indole ring of the substrate, respectively (Fig. 3, Fig. S30 and S36B). However, these similarities arise from convergent chemical logic rather than structural homology. In contrast, NCS catalyzes the condensation of dopamine and 4-hydroxyphenylacetaldehyde, resulting in a substrate-binding mode that is entirely different from that of AsKslB.

In summary, AsKslB and KslB are highly conserved in both structure and active site architecture, supporting a shared catalytic mechanism, whereas STR and NCS are structurally and evolutionarily distinct PSase families, with only convergent chemical strategies underlying substrate recognition and stabilization.

3.8. Site-directed mutagenesis of AsKslB

To investigate the catalytic roles of key residues, targeted mutagenesis was performed based on sequence alignment (Figs. S37 and S38) and structural insights, and 13 mutants of AsKslB were generated (Fig. 4, Table S1). Site-directed mutagenesis revealed distinct functional contributions of active site residues. The F55A, F98A, V226A, Y232A, R258A, and E276L mutants exhibited complete loss of enzymatic activity (Fig. 4). Structural insights demonstrated that Phe55, Phe98, Val226, and Tyr232 stabilize the tryptophan-binding pocket through robust hydrophobic interactions and specific hydrogen bonds (Fig. 3C–E). Arg258 anchored the γ-carboxyl group of α-KG via dual hydrogen bonds (Fig. 3D). Disruption of these interactions led to collapse of substrate binding. Glu276 acted as the catalytic base essential for the dehydration step, explaining the complete inactivation upon mutation to leucine (Fig. 4). The I59A, M225A, T263A, and K266A mutants showed markedly reduced activity (Fig. 4). Ile59 and Met225 contributed to Trp binding, whereas Thr263 and Lys266 facilitated α-KG coordination (Fig. 3C and D). The T263A/K266A double mutant (no activity) confirmed their essential role in α-KG binding. Partial activity retention was observed in R61A and S227A (Fig. 4). Arg61 maintained weak van der Waals contact with the indole ring, while Ser227 provided a backbone-mediated hydrogen bond to l-Trp, preserving limited substrate stabilization. These findings delineate a functional hierarchy among active-site residues, from indispensable catalytic and structural determinants to auxiliary modulators, shaped by the nature and magnitude of their molecular interactions.

3.9. Proposed mechanism

In classical Pictet-Spenglerases (PSases) including McbB, KslB, and STR, a conserved glutamate residue typically deprotonates the substrate amino group, thereby activating it for nucleophilic attack on the carbonyl moiety, resulting in Schiff base formation and subsequent cyclization [3,9,12,23,24]. In contrast, NCS exhibits a distinct catalytic mechanism [3,25]. Based on comprehensive structural and biochemical analyses, we propose a catalytic pathway for the AsKslB-mediated PSR (Fig. 5). Specifically, Glu276 facilitates deprotonation of the l-Trp amino group, initiating nucleophilic attack on the α-KG carbonyl carbon to form a carbinolamine intermediate (Fig. 5, state I to II). The substrate is simultaneously stabilized through hydrophobic interactions (Phe55, Ile59, Phe98) and hydrogen bonding (Met225, Val226, Ser227, and Tyr232), ensuring optimal spatial orientation for cyclization. Subsequent dehydration of the carbinolamine intermediate yields the IM-1 intermediate, whose structure has been resolved for the first time in this study, providing direct molecular evidence for the reaction's stereochemical pathway.

Crystallographic data and electron density maps reveal that IM-1 adopts a singular conformation that perfectly matches the experimental density, strongly supporting nucleophilic attack occurring on the si-face (Fig. 3D, Fig. S32). Key residues surrounding IM-1, including Lys266, Arg258, and Thr263, stabilize the si-face conformation through hydrogen bonding and spatial arrangement, creating an optimal environment for nucleophilic attack at the indole C2 position. Alternative conformations were investigated through modeling studies, which demonstrated that only the si-face conformation aligns with the experimental density, while the re-face conformation shows poor agreement (Fig. S39).

Structural analysis suggests that rotation of the indole ring around C3 from si-to re-face would likely result in steric clashes with Phe98, with additional steric hindrance from Thr263 and Lys266 disfavoring re-face attack at C2′ of α-KG (Fig. S39). Comparative analysis ofl-Trp, IM-1, and KA structures indicates that during Trp-to-KA conversion, the indole ring undergoes approximately 90° tilt towards the α-KG module (Fig. 3F), avoiding collision with Phe98, whereas full si-to-re inversion would require an energetically unfavorable 180° flip (Fig. S39). Significantly, all resolved structures demonstrate Trp exclusively in the si-conformation, and Phe98-to-Ala mutation results in complete loss of activity, confirming Phe98's critical role in stabilizing si-face binding.

The KslB crystal structure reported by Kim et al. [12] employed molecular docking simulations to predict both si- and re-face binding modes of IM-1 in the absence of resolved intermediates. Their computational results suggested that the re-face binding mode would experience steric interference from Phe98 during cyclization, preventing product formation. While these findings partially corroborate our structural observations, key distinctions exist: both studies confirm si-face nucleophilic attack, but our structural data exclusively captured the si-face binding mode of IM-1. Moreover, our analysis demonstrates that steric constraints imposed by Phe98 likely prevent indole ring flipping from si-to re-face orientation. We therefore propose that the indole ring maintains si-face orientation throughout the catalytic cycle. These structural insights collectively demonstrate that the amino acid composition and physicochemical properties of the AsKslB active site stabilize the indole ring in si-face orientation during Trp-to-KA conversion, facilitating efficient nucleophilic attack followed by an approximately 90° indole ring tilt during cyclization to yield KA. the indole C2 position executes.

During cyclization, a 6-endo-trig nucleophilic attack is carried out by the C2 C3 double bond of the indole ring on the imine (Fig. 5, state II to III), while the l-Trp backbone remains largely unchanged and the indole ring rotates by about 90°. Glu276 undergoes conformational rearrangement to stabilize a water molecule positioned between its side chain and the IM-1 intermediate, likely originating from dehydration during iminium ion formation. The cyclization product adopts (R)- and (S)-configurations at C2 and C2′, respectively, after which Glu276 functioning as a general base to remove the C2 proton (Fig. 5, state III to IV). This comprehensive mechanism integrates substrate activation, intermediate stabilization, and stereochemical control, all of which are essential for AsKslB catalytic efficiency. The resolution of the IM-1 intermediate structure provides compelling structural evidence supporting si-face selectivity, underscoring both the novelty and significance of this finding for understanding the enzyme's stereochemical control.

4. Conclusions

In this study, we characterized the function and structure of a novel PSase, AsKslB, through comprehensive biochemical and crystallographic analyses. AsKslB was demonstrated to condensate l-Trp and α-KG to form KA. Substrate scope assays revealed that AsKslB could utilize different F, Cl, Me, MeO substituted l-Trp, along with different α-keto acids as substrates. The X-ray structure of AsKslB indicated that it adopts a fold similar to those of KslB and McbB. Co-crystal structures with l-Trp and KA revealed the substrate binding mode. Importantly, a key intermediate IM-1 in the AsKslB-mediated PSR was captured, which provides direct evidence for the si-face attack of the indole C2 on the imine. Site-directed mutagenesis further demonstrated that residues Phe55, Ile59, Phe98, Arg258, Thr263, Lys266, and Glu276 play essential roles in substrate recognition and catalysis. These residues may serve as rational targets for enzyme engineering aimed at expanding the catalytic repertoire of AsKslB. Together, this work provides direct structural evidence for the formation of IM-1, reveals the stereochemical course of the reaction, and offers new mechanistic insights into PSase catalysis, which may accelerate the development of engineered PS biocatalysts.

CRediT authorship contribution statement

Yan-Bin Teng: Writing – original draft, Supervision, Software, Investigation, Funding acquisition. Zhi Qiao: Writing – original draft, Software, Investigation, Formal analysis. Chunya Xie: Writing – original draft, Validation, Investigation. Xiaona Yang: Writing – original draft, Validation, Data curation. Xinyu Liu: Validation, Formal analysis. Zhengrong Zou: Validation. Yunchang Xie: Writing – review & editing, Supervision, Funding acquisition. Xuan Zhang: Writing – review & editing, Supervision, Investigation. Qi Chen: Writing – review & editing, Supervision, Funding acquisition.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix. ASupplementary data

The following is the Supplementary data to this article: