Redirecting a Native Ene‐Reductase Toward Desaturation With Reverse Enantioselectivity

ABSTRACT

Chiral enones are valuable motifs in synthetic intermediates and bioactive molecules, driving significant interest in biocatalysis. Although recent enzymatic desaturation strategies for substituted cyclohexenones provide efficient and highly enantioselective synthetic routes, none offer complementary stereoselectivity. To address this gap, we introduce a stereocomplementary biocatalytic system based on an old yellow enzyme (OYE), XenA from Pseudomonas putida. Although XenA natively catalyzes the reduction of electron‐deficient alkenes and exhibits negligible desaturation activity, protein engineering redirected its catalytic function toward desaturation, ultimately yielding a variant that accommodates a range of cyclohexanones with 85%–99% ee and 32%–98% yield. Remarkably, the optimal variant (XenA_4) possesses 46 mutations and exhibits an 11°C increase in melting temperature over the wild type. Mechanistic studies revealed that the unique dimeric structure of the enzyme is pivotal in controlling stereoselectivity by modulating the substrate‐binding orientation.

A native ene‐reductase, XenA, was repurposed to catalyze the reverse‐enantioselective desaturation of cyclohexanones. The final variant was obtained through extensive protein engineering, combining PROSS‐guided computational design with mutagenesis and screening. Mechanistic experiments and computational modelling further revealed the unique mechanistic features underlying this enzymatic desaturation process.

1. Introduction

Chiral enone motifs are widely present in bioactive molecules [1, 2, 3]—including pharmaceuticals and natural products—and serve as versatile building blocks for diverse downstream transformations [4]. Their pharmacological relevance and synthetic versatility have attracted considerable attention in synthetic chemistry [5, 6, 7] (Figure 1a). Among the various synthetic strategies for these compounds, traditional chemical synthesis typically relies on the use of transition‐metal catalysts and strong oxidants [7, 8, 9, 10, 11, 12, 13], which contrasts with the efficient and sustainable nature of biocatalytic approaches [14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25]. Recent advances in enzymatic approaches for preparing chiral cyclohexenones generally follow two mechanistic paradigms: the biocatalytic reduction of cyclohexadienones [14, 26, 27] or the desaturation of cyclohexanones [14, 21, 22] (Figure 1b). Although the reductive desymmetrization strategy has proven to be an efficient and stereoselective method, it has inherent limitations compared to oxidative desaturation. These include reliance on less stable and less commonly accessible substrates, as well as the requirement for expensive NADPH cofactors or complex cofactor‐regeneration systems. Furthermore, this strategy is not suitable for synthesizing mono‐substituted cyclohexenones via the reduction of phenol derivatives, owing to the stability of the aromatic rings. In contrast, enzymatic desaturation systems circumvent these limitations, highlighting their broader synthetic potential. Despite these advantages, previous reports on biocatalytic oxidative desaturation have focused on preparing only one enantiomer of the cyclohexenone product, with the complementary selectivity achievable only through the enzymatic reduction process [14, 21, 22]. This limitation poses a challenge for practical applications of enzymatic desaturation in organic synthesis and medicinal chemistry, where a pair of enantiomers can display distinct pharmacological properties, necessitating access to both of them for comprehensive evaluation and therapeutic development [28].

FIGURE 1.

Discovery and engineering of an old yellow enzyme (OYE) for enantiocomplementary desaturation. (a) Chiral cyclohexenones with a remote quaternary center are prevalent structural motifs in bioactive molecules. (b) Established biocatalytic platforms for preparing chiral cyclohexenones include oxidative desaturation systems using cyclohexanones as substrates and O₂ as the terminal oxidant (left), as well as biocatalytic reduction of cyclohexadienones using ene‐reductases along with NADPH regeneration systems (right). (c) An enzymatic desaturation system catalyzed by XenA variants achieves reversed enantioselectivity relative to the established desaturation systems. XenA variants exhibit unique dimeric structures, which play a crucial role in modulating substrate‐binding orientation to achieve different enantioselectivity.

To address this limitation, we screened an array of flavin‐dependent enzymes from different branches of the old yellow enzyme (OYE) family [29, 30], and identified XenA [31]—a thermophilic OYE from Pseudomonas putida—which displayed notable opposite enantioselectivity but initially exhibited very low activity. Although XenA was widely recognized as a versatile and efficient ene‐reductase for enone reduction [31, 32, 33, 34], its desaturation ability had not been previously reported. Through iterative rounds of semi‐rational enzyme engineering combined with computational protein engineering and condition optimization, we developed a variant, XenA_4, with significantly improved desaturation activity and substrate generality (up to 98% yield and >99% ee). Notably, a large portion of the introduced mutations were computationally designed using the Protein Repair One‐Stop Shop (PROSS) [35], an algorithm developed to enhance protein stability and expression in bacteria [36, 37]. Mechanistic studies further revealed key active site residues involved in catalysis, suggesting a flipped substrate‐binding conformation relative to other known desaturases (Figure 1c).

2. Results and Discussion

To identify enzyme candidates capable of catalyzing the desaturation of substrate 1a with the opposite enantioselectivity to known desaturases, we screened a panel of selected flavin‐dependent enzymes from different subclasses of the OYE family [29, 30]. This collection included GluER (from Gluconobacter oxydans), GsOYE (from Galdieria sulphuraria), CaOYE (from Chloroflexus aggregans), CtOYE (from Chroococcidiopsis thermalis), LeOPR1 (from Lycopersicon esculentum), OYE2.6 (from Scheffersomyces stipitis), OYE3 (from Saccharomyces cerevisiae), XenA (from Pseudomonas putida), TtER (from Thermus thermophilus HB8), and RmER (from Ralstonia metallidurans). The enzymatic reactions were first evaluated using cell lysates, and the results were subsequently validated with purified proteins (Table S2 for details). During the screening, only XenA displayed the desired stereoselectivity with measurable activity (approximately 2% yield), generating the chiral enone product (S)‐2a in 83% ee, while all the other enzymes favored the formation of (R)‐2a. Therefore, wild‐type XenA was selected as the starting point for improving its catalytic activity and enantioselectivity control through protein engineering.

Site‐saturation mutagenesis (SSM) libraries were initially generated and screened by targeting six active site residues (C25, Y27, I66, C70, K106, and Y183) [38], located in close proximity to the FMN cofactor (Figure S2a). This effort led to the discovery of a variant XenA_1 (wild‐type XenA + Y183F), which exhibited an over 18‐fold increase in yield relative to the wild type along with improved enantioselectivity (96% ee). This mutation was surprising, as the homologous position of Y183 in PtOYE (Y169) [14] had previously been identified as a key determinant of enzyme activity and selectivity. Subsequent mutagenesis—both at twelve additional sites (Figure S2b) and via random mutagenesis using XenA_1 as the parent—failed to identify any additional beneficial mutations. We reasoned that a significant hurdle in engineering XenA variants was their reduced thermostability compared to PtOYE or GkOYE enzymes [14, 22], a crucial attribute for achieving high conversions in biocatalytic desaturation reactions with the latter two. Based on this rationale, we aimed to enhance the thermostability of XenA_1 in the hope that a more stable variant would function at elevated temperatures, potentially improving its catalytic performance under these conditions. Moreover, this effort would yield a more robust scaffold for enzyme engineering, capable of tolerating additional beneficial yet destabilizing mutations [39].

To this end, we employed a computational protein design strategy for enhancing enzyme thermostability named PROSS [35]. PROSS is a fully automated algorithm that leverages atomistic Rosetta modeling and phylogenetic sequence constraints to optimize core packing, surface polarity, and backbone rigidity while maintaining active‐site integrity. These features enable PROSS to enhance protein thermostability and soluble bacterial expression while preserving function [34, 35]. Considering cost and screening efficiency, six of the nine mutants predicted by the one‐shot PROSS design cycle were selected for initial evaluation. Most variants exhibited comparable or improved desaturation activity while maintaining the desired selectivity (see Supporting Information for details). Among them, variant D8 (designated as XenA_2 in the evolutionary lineage, Figure 2a) catalyzed the desaturation of cyclohexanone 1a with 82% yield and 98% ee, a significant improvement compared to XenA_1. Notably, this PROSS‐designed variant, XenA_2, carried 41 mutations primarily scattered along the surface of XenA_1 (Figure 2b) and exhibited a nearly 6.9°C increase in melting temperature (T _m) compared to the wild type (61.71°C vs. 54.84°C, Figure 2c). The enzyme performance was further improved through an additional round of random mutagenesis, leading to the development of XenA_3 (XenA_2 + F304L). By introducing F304L as a beneficial mutation, the enzyme T _m value was increased to 64.56°C. This variant catalyzed the desaturation of 1a in 94% yield and 96% ee. Preliminary assessment of the substrate scope indicated that XenA_3 achieved only moderate yields when substrates with larger substituents were used (e.g., 42% yield for 1e bearing a trifluoromethyl group on the aryl ring). To address this issue, further protein engineering on XenA_3 through random mutagenesis and screening was performed with 1e as the model substrate, resulting in a superior catalyst, XenA_4 (XenA_3 + F232L + K346R + R363W), which is slightly more thermostable than XenA_3 (65.82°C vs. 64.56°C). Although the activity of both variants was similar at 45°C, the more stable XenA_4 was found to tolerate an increase in reaction temperature, improving the yield to 66% for substrate 1e (Figure 2a).

FIGURE 2.

Engineering of XenA for enzymatic desaturation with opposite enantioselectivity. (a) Evolution of XenA for the desaturation of 1a and 1e. Reactions with each variant were performed in triplicate using 1 mol% enzyme and 5 mM substrate under O₂ atmosphere (1 atm) for 48 h. Yields were determined by LC‐MS using 4‐methoxyphenol as the internal standard, and enantiomeric excess (ee) was measured using HPLC. All data points represent the average of triplicate runs (n = 3). Error bars, mean ± s.d. XenA_1: wild‐type XenA + Y183F; XenA_2: XenA_1 + 41 additional PROSS‐designed mutations (V83A, S94A, L53V, G284A, E191P, A124P, T299A, M45L, K295G, G258H, F337H, I175L, A316N, A252I, A139G, N143K, T350V, S348A, L296I, A121E, L33F, E273D, A258H, Q80E, Q243E, N317G, K157Q, A77E, A136P, Y237F, V85I, T306E, T274V, S208T, Q87R, Q315R, L262M, K195Q, I34V, A61S, and A199E); XenA_3: XenA_2 + F304L; XenA_4: XenA_3 + F232L + K346R + R363W. (b) Structure of XenA_1 (PDB ID 8AUG [33]). This variant was used as the input sequence for PROSS design. The 41 mutation sites introduced for XenA_2 (D8) are highlighted as salmon spheres. (c) Protein melting temperature (T _m) analysis. T _m values for XenA variants were determined by differential scanning calorimetry (DSC) in Kpi buffer (100 mM, pH 8) from 20°C to 90°C at a rate of 1°C/min. Protein samples were measured at a concentration of 2 mg/mL.

To demonstrate the generality of the evolved variant, we investigated the substrate scope of XenA_4 under the optimized reaction conditions (Figure 3). Gratifyingly, a broad range of prochiral cyclohexanones were suitable for this enzymatic system, including 4‐alkyl‐4‐aryl‐, 4,4‐dialkyl, and mono‐substituted substrates (2a–u). In particular, aryl‐substituted cyclohexanones bearing both electron‐withdrawing (2c–e, and 2h) and electron‐donating groups (2b, 2f, 2g, and 2j) were smoothly converted to the corresponding cyclohexenones in good yields (57%–98%) and excellent enantioselectivities (93%–99% ee). Naphthyl‐ (2l and 2m) and heteroaryl‐ (2i and 2k) substituted substrates were successfully desaturated in this reaction as well. Moreover, cyclohexanones with spirocyclic scaffolds (2n–q, bearing five to seven membered rings) were compatible with this enzymatic system, yielding products in moderate to good yields with excellent enantioselectivity. Notably, a couple of mono‐substituted cyclohexanones (2t and 2u) were also tested. Although the conversions for these substrates could be further improved, the enantioselectivity of the products remained excellent. Mono‐substituted chiral cyclohexenones are valuable synthetic intermediates [40, 41], which cannot be accessed through the enzymatic reduction approaches. To further demonstrate the preparative potential of this biocatalytic desaturation system, we scaled up the enzymatic transformations using 1a and 1n for synthesizing chiral enones at 0.5‐mmol and 0.2‐mmol scales, respectively (Figure 3). The final products, (S)‐2a and (S)‐2n, were isolated in excellent enantiomeric purity and good yields, showcasing the synthetic potential of this enzymatic system.

FIGURE 3.

Enzymatic desaturation of cyclohexanones using XenA_4. The experiments were performed using XenA_4 (1.0 mol%) with 5 mM cyclohexanone 1 in Kpi buffer (100 mM, pH 8) at 50°C under O₂ atmosphere (1 atm) for 48 h. Unless otherwise noted, all yields were determined via LC‐MS analysis using 4‐methoxyphenol as the internal standard. XenA_3: wild type XenA + Y183F + F304L + 41 PROSS‐designed mutations (V83A, S94A, L53V, G284A, E191P, A124P, T299A, M45L, K295G, G258H, F337H, I175L, A316N, A252I, A139G, N143K, T350V, S348A, L296I, A121E, L33F, E273D, A258H, Q80E, Q243E, N317G, K157Q, A77E, A136P, Y237F, V85I, T306E, T274V, S208T, Q87R, Q315R, L262M, K195Q, I34V, A61S, and A199E); XenA_4: XenA_3 + F232L + K346R + R363W. See Supporting Information for details. ^a)The reactions were performed using XenA_3 (1.0 mol%) at 45°C. ^b)Kpi buffer (100 mM, pH 9). ^c)72 h. ^d)The yields were determined via GC analysis.

To elucidate the origin of the reversed enantioselectivity and catalytic efficiency of the engineered XenA variants, we performed a comprehensive mechanistic investigation combining computational modeling with experiments. In related desaturation systems, the reaction is proposed to proceed via enolate formation by deprotonation at the C _α ‐position of the carbonyl group, followed by hydride transfer from the C _β to the FMN cofactor [14, 21, 22]. Selectivity in these systems is governed by substrate orientation in the active site, where specific hydrogen‐bonding interactions involving conserved histidines and π–π interactions between FMN and the aromatic substituent on the substrate determine which face of the enolate and which prochiral C _β –H approaches flavin N5 for hydride abstraction. In the XenA system, the reversed enantioselectivity is instead dictated by its unique active site environment.

To establish a structural basis for the mechanistic features of XenA, we first focused on determining the active oligomeric state of the enzyme. Biophysical characterization by size‐exclusion chromatography coupled with multi‐angle light scattering (SEC‐MALS) demonstrated that XenA_3 is a homodimer in the reaction buffer [42] (Figure 4a). Notably, available x‐ray structures of wild‐type XenA and its mutants [31, 33, 43], along with computational structural models of the XenA_3 variant, revealed that the C‐terminal α‐helix of one monomer intrudes into the active site of the adjacent monomer, creating a confined cavity (Figure 4b). Molecular dynamics (MD) simulations of the substrate‐bound complex revealed that W358 from the neighboring monomer sterically repels the aromatic ring of the substrate, orienting it opposite to the FMN cofactor and enforcing a flipped binding mode compared to other desaturases (Figure 4b). This inversion of facial orientation relative to the flavin cofactor leads to formation of the opposite product enantiomer compared to PtOYE [14]. This predicted binding mode was further supported by mutagenesis experiments and mechanistic probing using a stereospecifically β‐deuterated substrate (Figures 4c–e). Specifically, mutation of W358 to smaller residues (F and A, see Figure 4c, entries 2 and 3) led to a pronounced loss of stereoselectivity, underscoring the essential role of this bulky residue. Moreover, deletion of the C‐terminal helix—removing 8 or 18 residues, both encompassing W358—dramatically impaired enzyme selectivity (entries 4 and 5), highlighting the structural and functional significance of this unique active site located next to the dimer interface. Consistent with these observations, Michaelis–Menten kinetic analysis revealed a marked increase in the K _M value for the W358A variant (Figure 4d, entry 2), indicating that the bulky tryptophan residue is essential for maintaining substrate confinement required for high stereoselectivity. To further validate the catalytically relevant binding pose during the hydride transfer step, we prepared a stereospecifically β‐deuterated substrate, d _β‐1a, in which the labeled position is cis to the methyl group (see Section V of the Supporting Information). Desaturation of d _β‐1a catalyzed by XenA_3 resulted in exclusive formation of the non‐deuterated enone, (S)‐2a (Figure 4e). The stereospecific removal of the cis‐methyl deuterium further supports a binding mode in which the methyl group is oriented toward the FMN cofactor.

FIGURE 4.

Structural and mechanistic investigations. (a) SEC‐MALS analysis. The blue solid line represents the Rayleigh ratio, and the purple dashed line shows the calculated molecular weight (Da) profile across the elution volume (mL). (b) Catalytically relevant binding mode of 1a in XenA_3 active site as characterized from computational modeling. Substrate docking predictions were refined by performing restrained‐MD simulations, which included a harmonic potential restraint to keep the substrate in the active site next to the cofactor and prevent unbinding events (see Section XII and Supporting Information Figures S6, S7 for details). (c) Site‐directed mutagenesis experiments. All reactions were performed using XenA variants (1 mol%) with 5 mM 1a at 45°C under air for 3 h. (d) Kinetic parameters for the selected XenA variants. (e) Mechanistic investigation of substrate binding and enzyme stereospecificity using d _β‐1a. (f) H/D exchange experiments. All experiments were performed in deuterated Kpi buffer (pH 8) under N₂ atmosphere for 3 h. (g) Effects of pH on XenA_3 catalyzed desaturation.

Site‐directed mutagenesis of active site residues K106, H178, and H181 highlighted their impact on enzyme catalytic efficiency (Figure 4c, entries 6–8). As evidenced by the significant decrease in k _cat and increase in K _M (Figure 4d, entries 4 and 5), these residues are essential for the formation and stabilization of the enolate during the deprotonation step. Considering that the evolved XenA variant does not possess a conserved tyrosine residue adjacent to the α‐position of the ketone, we propose that water molecules within the active pocket, together with polar residues in the nearby top‐face region (Figure 4b), facilitate the deprotonation event. To investigate this hypothesis, we performed a series of H/D exchange experiments using XenA variants and a previously reported desaturase, PtOYE (Figure 4f). Interestingly, both XenA_3 and XenA_3‐F183Y variants exhibited non‐selective H/D exchange, whereas PtOYE displayed an approximately 10:1 ratio of deuterium incorporation favoring exchange at H_b (cis to the methyl group). This observation is consistent with our hypothesis and previous reports that, in PtOYE and related systems, a conserved tyrosine residue serves as the catalytic base for α‐deprotonation; however, this residue is not required in the present system. Furthermore, the reaction outcome is highly sensitive to buffer pH (Figure 4g), as demonstrated by pH‐dependent assays, corroborating that a basic environment facilitates α‐deprotonation and stabilizes the enolate intermediate.

3. Conclusion

In summary, we have developed an enantioselective enzymatic strategy for preparing chiral cyclohexenones with reverse enantioselectivity compared to established systems. A PROSS‐guided engineering approach was adopted during protein engineering process, greatly improving enzyme thermostability. After extensive protein engineering efforts, the final variant XenA_4 was obtained, which carried 46 mutations (around 12.7% of the protein sequence), including changes to both active site residues and residues distant from the catalytic pocket. XenA_4 exhibits an 11°C increase in T _m relative to the wild type and accommodates a wide range of cyclohexanones for enzymatic desaturation. Combined experimental and computational studies uncovered a unique binding mode and deprotonation mechanism for this enzymatic system that is distinct from those of known desaturases. Our work highlights that exploring naturally occurring enzymes with unique structural features to unlock their non‐native functions and selectivity patterns is a promising strategy for overcoming the limitations of existing biocatalytic systems. Together with previous desaturation systems, we now have an enantiocomplementary system for synthesizing chiral enones through enzymatic desaturation. Future studies will focus on improving the catalytic efficiency of the reverse selective desaturases to address current limitations, including low substrate concentrations, high catalyst loading, and slow reaction kinetics, thereby enhancing the synthetic utility of the system. In parallel, further mechanistic insights will be pursued through a combinatorial approach integrating computational modeling, spectroscopic analysis, and crystallography.

Author Contributions

Qing‐Qing Zeng: investigation, validation, data curation, writing – original draft. Cristina Berga: investigation, validation. Carla Calvó‐tusell: investigation, validation. Marc Garcia‐borràs: investigation, supervision, writing – original draft, writing – review and editing, funding acquisition. Zhen Liu: conceptualization, supervision, writing – original draft, writing – review and editing, funding acquisition, project administration, investigation.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File: anie73120‐sup‐0001‐SuppMat.pdf.

Zeng Q. Q., Berga C., Calvó‐Tusell C., Garcia‐Borràs M., and and Liu Z., Angewandte Chemie International Edition (65, no. 31 (2026): e3155545, 10.1002/anie.3155545

Data Availability Statement

The data that supports the findings of this study are available in the Supporting Information of this article.