Molecular and Structural Basis for Cγ-C Bond Formation by PLP-Dependent Enzyme Fub7

Abstract

Pyridoxal 5’-phosphate (PLP)-dependent enzymes that catalyze γ-replacement reactions are prevalent, yet their utilization of carbon nucleophile substrates is rare. The recent discovery of two PLP-dependent enzymes, CndF and Fub7, has unveiled unique C-C bond forming capabilities, enabling the biocatalytic synthesis of alkyl- substituted pipecolic acids from O-acetyl-L-homoserine and β-keto acid or aldehyde derived enolates. This breakthrough presents fresh avenues for the biosynthesis of pipecolic acid derivatives. However, the catalytic mechanisms of these enzymes remain elusive, and a dearth of structural information hampers their extensive application. Here, we have broadened the catalytic scope of Fub7 by employing ketone-derived enolates as carbon nucleophiles, revealing Fub7’s capacity for substrate-dependent regioselective α-alkylation of unsymmetrical ketones. Through an integrated approach combining X-ray crystallography, spectroscopy, mutagenesis, and computational docking studies, we offer a detailed mechanistic insight into Fub7 catalysis. Our findings elucidate the structural basis for its substrate specificity, stereoselectivity, and regioselectivity. Our work sets the stage ready for subsequent protein engineering effort aimed at expanding the synthetic utility of Fub7, potentially unlocking novel methods to access a broader array of noncanonical amino acids.

Graphical Abstract

Fub7, a pyridoxal 5’-phosphate (PLP)-dependent enzyme, catalyzes stereoselective and substrate-dependent regioselective Cγ-C bond formations. Now a structural and enzymological study have elucidated the structural and molecular underpinnings of its catalytic specificity and selectivity, paving the way for harnessing its activity in the synthesis of cyclic noncanonical amino acids.

Introduction

Noncanonical amino acids (ncAAs) are essential building blocks in natural products,^[1] peptide and small molecule therapeutics,^[2] and designer proteins.^[3] Of particular note are the conformationally constrained cyclic ncAAs. Incorporating these ncAAs into peptidomimetic drugs can enhance their conformational rigidity leading to higher potency and better efficacy. For instance, bicyclo[3.1.0]proline is a key structural motif found in the protease inhibitors Boceprevir^[4] and Nirmatrelvir^[5], which are clinically used to treat hepatitis C and COVID-19, respectively (Figure 1). Similarly, cyclic ncAAs such as bicyclo pipecolic acid and 3-ethyl-pipecolic acid are important moieties in the anti-HIV drug Nelfinavir and the antibiotic Pirlimycin, respectively (Figure 1). Despite their growing significance in chemistry, biology, and medicine,^[6] developing robust methods for synthesizing ncAAs remains challenging.^[7] This is particularly true for cyclic ncAAs, such as substituted pipecolic acids, where multi-step chemical syntheses are often required and overall synthetic efficiency is low.^[8]

Figure 1.

Examples of peptide therapeutics containing cyclic ncAAs.

Enzymes offer unparalleled selectivity, sustainability, and evolvability, making them an attractive alternative for the preparation of ncAAs that can shortcut traditional synthetic routes.^[9] However, most biocatalysts developed thus far are applied to the synthesis of acyclic ncAAs,^[10] such as ammonia lyases for the synthesis of β-amino acids, tyrosine phenol lyase (TPL)^[11] and tryptophan synthase (TrpS)^[12] for the preparation of tyrosine and tryptophan analogues, respectively. In contrast, biocatalytic tools for the synthesis of cyclic ncAAs are scarce.^[13]

Recently, CndF and Fub7, two unusual pyridoxal 5’-phosphate (PLP)-dependent enzymes were discovered in the biosynthetic pathways of citrinadin and fusaric acid, respectively.^[14] Both enzymes have been investigated for the biocatalytic synthesis of various alkyl-substituted pipecolic acids (Figure 2a). Despite a low sequence identity (~23%) shared between these two enzymes, similar catalytic mechanisms have been proposed: the common substrate O-acetyl-L-homoserine (OAH) first undergoes γ-elimination to generate vinylglycine ketimine (VGK), a Michael acceptor intermediate that is crucial for the subsequent 1,4-conjugate addition with different carbon nucleophiles. The amino acid products released from this process readily cyclize to form dehydropipecolic acids, which can then undergo either enzymatic or chemical imine reduction to yield a variety of alkyl-substituted pipecolic acids.

Figure 2.

PLP-dependent enzymes catalyzing C-C bond formation. a) Previous studies showed that CndF uses β-keto acid derived enolate while Fub7 chooses aldehyde-derived enolate and catalyzes stereoselective C-C bond formation. Both enzymes were proposed to operate through an elimination-addition sequence via a shared vinylglycine ketimine (VGK) Michael acceptor.^[14] b) This work explores the possibility of using ketone-derived enolate as carbon nucleophiles and whether Fub7 exerts any regioselective control with unsymmetrical ketones. Regioisomeric ratio, r.r.; diastereoisomeric ratio, d.r.

The discovery of CndF and Fub7 has opened new avenues for the biosynthesis and chemoenzymatic synthesis of pipecolic acid derivatives, yet many mechanistic questions remain to be answered. First of all, while PLP-dependent enzymes catalyzing γ-replacement reactions are well-known, those that utilize carbon nucleophiles for the formation of Cγ-C bonds are rare. The mechanism how CndF and Fub7 select and activate carbon nucleophiles is still not well understood. Additionally, CndF and Fub7 exhibit distinct preferences for nucleophile substrates: CndF typically prefers β-keto acid-derived enolates, while Fub7 predominantly uses aldehyde-derived enolates. The basis for their complementary substrate specificity has yet to be determined. Moreover, while Fub7 catalyzes stereoselective C-C bond formations, as demonstrated by the production of pipecolic acid derivative 1 (Figure 2a), the mechanism underlying this stereoselectivity are currently unknown. In summary, the absence of structural insights and comprehensive mechanistic understanding of these emerging Cγ-C bond forming PLP-dependent enzymes hinders further protein engineering efforts and limits their wider applications.

In this study, we expanded the catalytic repertoire of Fub7, employing ketone-derived enolates as nucleophiles (see Figure 2b). Notably, we discovered that Fub7 catalyzes substrate-dependent regioselective α-alkylation of unsymmetrical ketones. We determined the X-ray crystal structures of Fub7 in complex with different reaction intermediates, and through a combination of spectroscopic characterization, mutagenesis, and computational docking studies, we now provide a comprehensive understanding of Fub7 catalysis.

Results and Discussion

At the outset of this study, we chose to use ketone substrates to investigate the Cγ-C bond formation catalyzed by Fub7. Compared to aldehydes and β-keto acids, ketones are less prone to enolization due to the higher pKa for their α-hydrogens. Therefore, we hypothesized that employing chemically less reactive ketone substrates could better represent Fub7’s true catalytic capabilities in recognizing and activating carbonyl substrates. In addition, unsymmetrical ketones, which generate two enolate regioisomers, provide an opportunity to determine whether Fub7 can impose regioselective control during C-C bond formation. This aspect was not considered in previous studies with either CndF and Fub7, as the carbon nucleophiles tested (i.e. β-keto acids or aldehydes) have only one α-site for deprotonation. In a similar vein, asymmetric alkylation of ketones was recently demonstrated with an engineered variant of TrpB,^[15] but the ketone substrates used therein also possessed only a single deprotonation site. Consequently, regioselective α-alkylation of ketones remains an unexplored area for PLP-dependent enzymes.

We commenced our investigation with 2-hexanone 2a, a constitutional isomer of Fub7’s native substrate, n-hexanal. Given their structural similarity, we hypothesized that the activation of 2a by Fub7 would be analogous to that of n-hexanal. Consequently, we anticipated the formation of 5-propyl-6-methyl dehydropipecolic acid (3a, see Figure 3a) when incubating 2a with Fub7 and OAH. Recognizing that 3a epimerizes spontaneously at C5 in aqueous solution, which complicates isolation and structural characterization, we incorporated the oxidase Fub9 into a coupled assay with Fub7, as previously described.^[14b] This cascade was intended to further oxidize and aromatize 3a into its corresponding picolinic acid derivative for easy detection and isolation. This approach proved successful, resulting in the isolation of 5-propyl-6-methyl-picolinic acid (4a) with its identity confirmed by NMR and mass spectrometry (data provided in Table S1). However, a careful analysis of the reaction mixture unveiled an additional enzymatic product with the same molecular weight as 3a. Since this species was resistant to Fub9-catalyzed aromatization, we derivatized it by chemical reduction using NaBH₃CN. Isolation and structural characterization of the resulting product revealed its identity as cis-6-butyl-pipeoclic acid (6a), suggesting that its precursor was likely 6-butyl-dehydropicolinic acid (5a), as depicted in Figure 3a.

Figure 3.

Exploring the capacity of using ketone-derived enolate as carbon nucleophiles in Fub7-catalyzed reactions. a) Reaction schemes with ketone substrates 2a-2d. The two possible alkylation sites C₁ and C₃ are colored in yellow and blue, respectively b) Substrate-dependent regioselective α-alkylation of unsymmetrical ketones. The analytical yield and regioisomeric ratio are determined by LC-MS analysis. Extracted ion chromatogram traces are shown.

The nearly 1:1 ratio of products 4a and 6a closely matches the literature value of enolate mixture distribution (58:42) derived from ketone 2a under the thermodynamic control.^[16] This correlation suggests that both the thermodynamic enolate and kinetic enolate can participate in the C-C bond formation catalyzed by Fub7. Intriguingly, we noted an increase in transformation efficiency with the use of longer ketone substrates (Figure 3b). As the substrate chain length grows, there was also a progressive preference for alkylation at the less-substituted C1 position over the more substituted C3 position in unsymmetrical ketones. This trend was particularly pronounced with substrate 2d, where C-C bond formation occurred almost exclusively at the C1 position. The high degree of regioselectivity enabled the isolation of 6-heptyl-dehydropipecolic acid (5d) and its reduced form (6d) as single stereoisomers (Table S3–S4). This straightforward product distribution also facilitated the characterization of steady-state kinetics for the ketone alkylation reaction. Kinetic analysis revealed an apparent K_M of 0.62 mM for 2d and a k_cat of 7.7 min⁻¹ (Figure S1), suggesting that ketones are competent carbon nucleophiles in Fub7-catalyzed reaction. The turnover number k_cat is comparable to that observed for Fub7’s native substrate n-hexanal (k_cat = 4.0 min⁻¹), while the Michalis constant K_M is much lower than that of n-hexanal (K_M = 11 mM, Figure S2). The comparable turnover numbers observed for both nucleophiles indicate that the enolization of these different carbonyl substrates is probably not the rate-limiting step in the overall transformation process. Additionally, the lower K_M value for longer ketone substrate aligns with previous observation that Fub7 generally favors nucleophiles with longer carbon chains.^[14b]

To better understand the substrate-dependent regioselectivity, we sought to determine the crystal structure of Fub7. We successfully crystallized Fub7 in its holo-form and solved its crystal structure at a resolution of 1.76 Å (Table S5). Fub7 crystallized as a dimer of dimers (Figure S3), with each monomer comprising an N-terminal oligomerization domain (residues 6–60), a large PLP-binding sub-domain (residues 61–294), and a C-terminal sub-domain (residues 295–432). The overall fold of Fub7 is similar to that of other Class I fold-type PLP-dependent enzymes.^[17] Structural comparison using the Dali server^[18] reveals that Fub7 bears the closest resemblance to OAH sulfhydrylase MetY,^[19] which is involved in methionine biosynthesis (PDB entry 7KB0), hinting an evolutionary link between the Cγ-C bond-forming Fub7 and Cγ-S bond-forming PLP-dependent enzymes. The active site of Fub7 is sandwiched between the large and small sub-domains, right at the dimer interface. The N-terminal α2-helix and its flanking loops (residues 34–59) help to define the active site. B-factor analysis indicates that these loops and α-helices have high flexibility, suggesting that the holo-Fub7 exists in an open conformation (Figure S3). Within the active site pocket, clear electron density defines a PLP cofactor that is in the internal aldimine form covalently linked to lysine residue K211 (Figure 4a).

Figure 4.

X-ray crystal structures of Fub7 in complex with various reaction intermediates. a) Holoenzyme (PDB Entry 8EQW). b) Vinylglycine ketimine intermediate (PDB Entry 8ERB). c) 2-Aminocrotonate intermediate (PDB Entry 8ERJ). Left: Single-crystal microspectrophotometer UV-Vis spectra. Middle: close-up view of the active site and F₀-F_c polder omit map of the PLP or PLP-bound intermediate (grey mesh, contoured at 3.0 s). Hydrogen bond interactions are indicated by dashed black lines. Residues from monomer A are coloured in cyan, whereas residues from monomer B are coloured in salmon. PLP and PLP-bound intermediates were shown in stick-model and coloured in orange. Right: Structures and interactions of assigned intermediates based on electron density and UV-Vis spectra.

To gain mechanistic insights, we soaked holo-Fub7 crystals with its native substrate, OAH, and its substrate analogue L-vinylglycine. The latter was chosen to bypass the γ-elimination step. Within minutes after OAH was introduced, the holo-Fub7 crystals completely transitioned from yellow to colorless (Figure S4). Single crystal micro-spectrophotometer analysis confirmed the observation, showing the disappearance of the characteristic absorption peak at 425 nm (Figure 4b). In contrast, upon soaking with L-vinylglycine, the crystals shifted in color from yellow to orange (Figure S5). A corresponding microspectrophotometer analysis revealed a new absorption peak at 485 nm, and increase absorbance at 350 nm (Figure 4c). These changes suggest that distinct PLP-bound intermediates were trapped and stabilized in crystallo. Subsequent X-ray crystallographic analysis solved the structures of these intermediates at 1.98 Å and 2.16 Å resolution, respectively (Table S5). The electron density in the OAH-soaked crystal clearly demonstrated that the substrate had undergone the γ-elimination reaction, with the remaining amino acid carbon and nitrogen atoms aligning in a coplanar geometry (Figure 4b). Interestingly, in the L-vinylglycine-soaked crystal, the electron density at the active site closely resembled that of the OAH-soaked crystal (Figure 4c). We interpreted these electron densities to represent an s-trans configured VGK intermediate and an (E)-2-aminocrotonate intermediate, respectively.

Although it is not possible to differentiate between these two intermediates with nearly identical geometries at the current resolutions, they possess vastly different conjugation systems and our structural assignments are supported by their markedly different UV-Vis absorption spectra collected in crystallo. Furthermore, analogous spectral shifts observed in solution assays reinforce our structural interpretations (Figure 5). The formation of the 2-aminocrotonate intermediate from L-vinylglycine, rather than the VGK intermediate, is consistent with its incompetence to engage in the C-C bond formation: L-vinylglycine failed to substitute OAH in any C-C bond-forming reaction; Instead, it exclusively underwent deamination reaction to produce α-ketobutyrate. Thus, while vinylglycine could act as an analog in the γ-lyase reaction of OAH, it is not an equivalent for the γ-replacement reaction. A proposed mechanism for the formation of the 2-amoinocrotonate intermediate during the Fub7-catalyzed deamination of L-vinylglycine is presented in Figure S5.

Figure 5.

Spectroscopic characterization of Fub7 in solution. A) Fub7 mixed with OAH. B) Fub7 mixed with L-vinylglycine.

It is also noteworthy to mention that the s-trans configuration of the VGK intermediate unambiguously determined by our high-resolution crystal structures, contrasts with the widely accepted s-cis configuration (as depicted in Figure 2a), that was first established by Chang and Walsh through isotope labeling studies four decades ago.^[20] Such discrepancy suggests that the conclusion drawn from previous stereochemical analysis of cystathionine γ-synthase in Salmonella typhimurium may not be universally applicable to all γ-substituting PLP-dependent enzymes. Specifically, it seems that these earlier generalizations do not hold for Fub7.

A structural comparison between the intermediate-bound state (i.e. the Fub7-VGK complex) and the holo-structure of Fub7 unveiled significant conformational changes upon ligand binding, including a substantial shift over 5 Å for helix α16. B-factor analysis suggests that this closed-conformation is considerably more rigid than the open state observed in the holo-enzyme (Figure S6). Such conformational changes are consistent with an induced fit mechanism, presumably serving to shield the VGK intermediate from unwanted nucleophilic attacks (e.g. by water). Nevertheless, the VGK intermediate is not completely occluded. A hydrophobic tunnel, approximately 21 Å in length, was identified at the dimer interface in this closed state (Figure 6a). We propose that this tunnel allows nucleophilic co-substrates to reach the active site pocket. The tunnel’s hydrophobic nature and substantial length likely account for Fub7’s preference for long-chain hydrophobic carbon nucleophiles. Comparing the crystal structure of Fub7 with the AlphaFold2 model of CndF revealed that CndF has a much wider and shallower active site pocket, which may explain why CndF prefers small and hydrophilic β-keto acid substrates (Figure S7). Detailed analysis would require a crystal structure of CndF in its closed-state.

Figure 6.

Structural basis for Fub7-catalyzed C-C bond formation. a) Substrate entrance tunnel revealed from the crystal structure of the Fub7-VGK complex. b) Docking of enamine intermediate post C-C bond formation with the kinetic enolate derived from ketone 2d. The trans-configuration of enamine is based on the s-trans configuration unambiguously established for its precursor VGK. c) Docking of enamine intermediate post C-C bond formation with the enolate derived from 2-methyl-pentanal. Hydrogen bonds are shown as black dash lines. d) Based on the docking model shown in c), it is possible to propose a likely transition-state for the enolate attack, and the stereochemical outcome is consistent with the stereo-configuration of final product 1

The orientation of Fub7’s substrate entrance tunnel delineates the trajectory for nucleophilic attack on the VGK intermediate, as the nucleophile can only approach this Michael acceptor through the re-face of the PLP. Yet, mechanistic details regarding the C-C bond formations are still lacking. For instance, the precise mechanism by which a carbonyl substrate is converted into an enolate, as well as the means by which regioselectivity and stereoselectivity are controlled, are still unknown. In pursuit of these insights, we aimed to capture a crystal structure of the Fub7-VGK-nucleophile ternary complex by exposing the OAH-soaked colorless crystals to a non-enolizable aldehyde/ketone (e.g. 2,2-dimethylpentanal). However, repeated efforts were met with challenges: soaking the crystals with aldehyde or ketone ligands consistently resulted in crystal deformation or poor diffraction quality. This is presumably due to nonenzymatic reactions at high concentrations of carbonyl ligands with lysine residues on the surface of Fub7, which disrupted the integrity of the protein crystal lattice.

While we were unable to obtain a ternary complex structure, the closed-conformation of Fub7-VGK binary complex structure served as a reliable model, enabling us to dock various post C-C bond-forming reaction intermediates into the active site. As depicted in Figure 6b, when we docked the hypothesized E-enamine intermediate en route to product 5d, we found that the alkyl chain of the ketone substrate fits well within the hydrophobic tunnel. Additionally, the ketone’s carbonyl group accepts a hydrogen bond donated from residue Y113. We also performed similar docking procedures for enamine intermediates en route to products 3a and 5a, respectively (Figure S8). These docked models suggest that Fub7 can accommodate both the thermodynamic and kinetic enolates derived from substrate 2a. Moreover, the sole polar contact is a hydrogen bond between the enol(late) oxygen and Y113. However, in the case of C3-alkylation, the methyl group at C1 is positioned towards residue R273(B), potentially leading to unfavorable interactions if the rotation of the R273(B) side chain is restricted. Such scenario is likely when the ketone’s alkyl chain lengthens. Consequently, substrates with longer alkyl chains tend to be less favorable for C3-alkylation. This tendency explains the observed progressive preference for C1-alkylation with unsymmetrical ketones.

With the substrate-dependent regioselectivity explained, we next performed similar docking studies to investigate Fub7’s stereoselectivity. As illustrated in Figure 6c, docking of the enamine intermediate derived from α-methyl pentanal revealed how Fub7 determines the stereochemical outcome observed in product 1. The docked model shows that the aldehyde carbonyl oxygen forms a hydrogen bond with Y113, echoing the interactions noted in the aforementioned models, while the α-methyl group snugly fits into a compact side pocket near T388. The limited size of this pocket imply that only aldehydes with small functional groups at the α-carbon are accommodated, corroborating previous findings that Fub7 does not tolerate bulky aldehydes with ethyl or larger substituents at Cα.^[14b] Furthermore, according to the docked model, Fub7 favors aldehyde enolate attack through its re-face, leading to R-configured stereocenters at C5 as the major diastereomer product 1 (Figure 6d). On the contrary, if the enolate attacks through its Cα si-face, its oxyanion will shift to the side pocket, and its α-methyl group will point to R60(B) and Y113, resulting in unfavorable interactions (Figure S9). Therefore, the stereoselective control likely originates from the fixed recognition of the enolate oxyanion through Y113 and preferred orientation of the α-methyl group and the aliphatic backbone.

Based on our analysis, we postulated that 1) enlarging the T388 side pocket might allow Fub7 to recognize aldehydes with bulkier functional groups at the Cα position; and 2) mitigating the adverse interactions caused by R60(B) may potentially alter Fub7’s selectivity. To test this hypothesis, we created T388S and R60A mutants and evaluated their activities. As shown in Figure 7, the R60A mutant is catalytically inactive, presumably because R60(B) is essential for anchoring the phosphate group of the PLP cofactor via a salt bridge, a crucial interaction observed in all Fub7 crystal structures described above. An alanine substitution at this site disrupts this important interaction, thereby impeding catalysis. Nevertheless, the T388S mutant successfully synthesized the expected dehydropipecolic acid derivative when 2-ethyl-hexanal was used as the nucleophile (Figure S10). In contrast, the wild-type (WT) protein showed virtually no activity with this bulky substrate. Collectively, these results confirmed the hypothesized steric hinderance caused by T388 when interacting with the ethyl substituent and validated our docking analysis.

Figure 7.

Activity of Fub7 mutants. Asterisks indicate mutants with no measurable activity under assay condition. K211A mutant was used as a negative control.

In every docked model, Y113 represents the sole direct polar interaction with nucleophile substrate, engaging with the enolate oxyanion during the C-C bond formation. Another polar residue, R273(B), although in proximity, remains an enigma in terms of its catalytic function due to its considerable distance (>5 Å) from the alkylation site. The absence of any residue that could evidently serve as a general base for the deprotonation at the substrate’s α-carbon leads us to surmise that Fub7 may preferentially bind to the enol tautomer of carbonyl substrates, despite their low population in solution. We propose that Y113 could act as a general base, facilitating deprotonation of the enol and directing the resulting enolate towards the VGK intermediate. This hypothesis also provides a plausible explanation for the observed product ratio between 4a and 6a: should Fub7 bind nonenzymatically formed enols – both thermodynamic and kinetic – equally well, Y113 would activate each for the subsequent nucleophilic attack.

To investigate the catalytic roles of residues Y113 and R273(B), we performed site-directed mutagenesis at these sites and assessed the activity of each variant in vitro. As shown in Figure 7, Y113 is indispensable for C-C bond formation. The isosteric substitution of Y113 with phenylalanine, which lacks a hydroxyl group, completely abolished Cγ-C bond-forming activity with both aldehyde and ketone substrates. Interestingly, this substitution led to a 2.5-fold enhancement in the γ-lyase activity for OAH. These findings underscore the essential role of the phenolic hydroxyl group of Y113 in mediating the nucleophilic attack at the VGK intermediate. These observations support the notion that Y113 functions as a general base in the deprotonation of enol substrates. In line with this mechanistic view, the general base Y113 is also predicted to catalyze the oxa-Michael addition reaction with alcohol nucleophiles. This prediction was confirmed as the Y113F mutant was also deficient in forming the Cγ-O bond. Replacing Y113 with histidine significantly impaired both Cγ-C and Cγ-O bond forming activities, yet the γ-lyase activity remained unaltered. The inability of the Y113H mutant to serve as a general base in nucleophilic addition reactions may be attributed to geometric constraints and the suboptimal positioning of the histidine’s side chain. Nonetheless, these results highlighted the importance of Y113’s phenolic hydroxyl group in the nucleophilic addition reaction, while it appears to be dispensable in the γ-elimination reaction. The non-involvement of Y113 in the γ-elimination reaction catalyzed by Fub7 starkly contrasts with the findings from previous studies on the methionine γ-lyase, where a similar phenylalanine mutation at the corresponding tyrosine residue significantly reduced the γ-lyase activity (i.e. a 910-fold decrease in k_cat).^[21] This disparity suggests that even seemingly conserved residues may fulfill distinct roles in evolutionarily related enzymes, such as γ-replacement enzymes in comparison to γ-lyases.

On the other hand, R273(B) has been identified as essential for the γ-elimination reaction, with three mutations at this residue (namely R273A, R273Q, and R273H) completely abolishing the γ-lyase activity (Figure 7). The failure of the initial γ-elimination inevitably interrupts the entire γ-replacement process, hence the cessation of all Cγ-X bond-forming activities in these mutants was anticipated. Remarkably, replacing R273 with lysine not only fully restored the γ-lyase activity but also partially recovered the Cγ-O and Cγ-C bond-forming activities to 40% and 20%, respectively. These results suggest that R273 is pivotal for the γ-elimination reaction, presumably by providing favorable electrostatic interaction to the negatively charged acetate leaving group, a role can be mimicked by a similarly positively charged lysine substitution. Our hypothesis is supported by docking analysis with substrate OAH (Figure S11). When modeled as the PLP-bound external aldimine, R273 is found to be in close proximity (4.2 Å) to the substrate acetyl group. It is conceivable that following the elimination step, the departing acetate byproduct is likely drawn to R273 via electrostatic stabilization.

In summary, this study elucidates the structure-function relationship and mechanistic intricacies of Fub7 catalysis (Figure 8). Fub7 initiates the reaction by catalyzing the γ-elimination of OAH and generates an s-trans-configured VGK intermediate. K211 likely serves as the general base/acid catalyst for all requisite proton transfers, while R273(B) assists in the release of the acetate byproduct. Although Y113 does not participate in the initial stage, it becomes critical in the activation of incoming nucleophiles, facilitating deprotonation of both enol and alcohol substrates. The enzyme’s hydrophobic tunnel is selective for nucleophile substrates with lengthy carbon chains and hydrophobic substituents. The selectivity in substrate binding predisposes the reaction towards C1 alkylation in unsymmetrical ketones and guides the enolate to attack via the re-face, culminating in stereoselective Cγ-C bond formation.

Figure 8.

Proposed catalytic mechanism for Fub7. a) Highlighted role of R273(B) in the γ-elimination stage. b) Proposed nucleophilic addition transition state with kinetic enol derived from unsymmetrical ketones. c) Molecular mechanism for the stereoselective control of enolate attack. d) Similar activation mode proposed for oxo-Michael addition reaction with alcohol nucleophiles. Many steps are omitted for clarity due to similarity with other common PLP-dependent enzymes catalyzing γ-replacement reactions.

Conclusion

PLP-dependent enzymes catalyzing γ-replacement reactions are diverse, yet the majority of those thoroughly investigated are Cγ-S bond-forming enzymes involved in methionine or cystathionine biosynthesis.^[22] To date, only less than a handful of examples were reported to catalyze C-C bond formation at the γ-position.^[23] Our studies have provided the first crystal structure of a Cγ-C bond forming PLP-dependent enzyme, shedding light on the structural underpinnings of Fub7 catalysis. This includes delineating the structural determinants for its substrate specificity, regioselectivity, and stereoselectivity. Our findings now paves the way for future structure-guided protein engineering efforts with Fub7, aiming to expand its synthetic utility, such as orchestrating photocatalytic radical reactions^[24] and [3+2] cycloaddition reactions.^[23a] We expect that unlocking the full catalytic prowess of Fub7 could forge new methods for the synthesis of noncanonical amino acids.