Screening of HHDHs
To commence the study, we chose the dehalogenation of racemic 3-chloro-1-phenylpropan-1-ol (1a) and the azide-mediated ring-opening of racemic 2-phenyl-oxetane (1b) as the respective model reactions for dehalogenation and ring-opening processes. Dozens of recombinant E. coli BL21 (DE3) strains harboring HHDH enzymes (Supplementary Table 1), maintained in our laboratory collection32–36, were initially evaluated for their catalytic efficiency in the dehalogenation model reaction (Supplementary Table 2). Background dehalogenation reaction was not observed either in the absence of E. coli cells or with E. coli cells that do not express HHDH (entries 1–2, Fig. 2a). Although the majority of tested enzymes demonstrated negligible or very low catalytic activity, several HHDHs were found to show relatively good catalytic efficiency with the yields of 1b > 10% (entries 3–10, Fig. 2a). Among them, the enzymes HheA5, HheC, and HheD8 exhibited a moderate to good enantioselectivity (E = 13–76). Although non-enzymatic conversion of γ-haloalcohols to oxetanes can be achieved40, finding an enantioselective catalyst for this transformation remains elusive. Subsequently, the active HHDHs underwent further evaluation with the model ring-opening reaction (Supplementary Table 3). No background ring-opening reactions were detected (entries 1–2, Fig. 2b). Unexpectedly, nearly all exhibited very low catalytic efficiency for the oxetane ring-opening, except for the enzymes HheD8 and HheD15. It is noteworthy that HheD8, which originates from strain Thauera aminoaromatica S2, also demonstrated moderate R enantioselectivity (E = 8), resulting in the formation of γ-azidoalcohol (R)-1c with 74% e.e and 18% yield (entry 7, Fig. 2b). To obtain effective biocatalysts for establishing a biocatalytic platform for the enantioselective formation as well as ring-opening of oxetanes, HheD8 was chosen as the progenitor enzyme for further directed evolution study, as it has demonstrated relatively good catalytic performances in both dehalogenation and ring-opening model reactions.
Directed evolution of HheD8
Given that the protein structure of the wild-type HheD8 (HheD8-WT) was not resolved at that time, we acquired a predicted structure model of HheD8-WT from the AlphaFold Protein Structure Database (https://alphafold.ebi.ac.uk, AFDB code: AF-N6YXW4-F1). The molecular docking of the oxetane (R)-1b into the active site of the HheD8-WT/AF model was then performed to obtain a probable binding pose for subsequent protein engineering efforts. In an attempt to enhance both the catalytic efficiency and enantioselectivity for the formation and ring-opening reactions of oxetane, we strategically targeted eight amino acid residues (F19, A69, Y168, M124, R127, Q160, N161 and R182) lining the active site pocket (Fig. 2c). Sequential rounds of site-saturation mutagenesis (SSM) were carried out on these selected target residues, with the resulting variants being screened by whole-cell biotransformation for enhanced enantioselectivities and/or substrate conversions. In each round, variants exhibiting enhancements underwent further evaluation through separate model reactions: the dehalogenation of (rac)-1a and the ring-opening of (rac)-1b with azide.
In the dehalogenation reaction screening, four rounds of iterative SSM led to the identification of beneficial variants: HheD8-M1 (A69F), HheD8-M2 (A69F/R127G), HheD8-M3 (A69F/M124P/R127G), and HheD8-M4 (A69F/M124P/R127G/R182W). These variants exhibited remarkably improved enantioselectivity and/or catalytic activity compared to the HheD8-WT in the dehalogenation of (rac)-1a (Fig. 2d). Notably, the quadruple-mutant HheD8-M4 excelled, achieving 49% conversion of 20 mM (rac)-1a and formation of (R)-1b with 99% e.e. after 8 h. In the case of ring-opening reaction screening (Fig. 2e), the mutant HheD8-M1 (A69F) was also identified in the first round of SSM and exhibited exceptional enantioselectivity in stark contrast to the HheD8-WT (E > 200 vs E = 8). Using the HheD8-M1 variant as the parent enzyme, three additional rounds of iterative SSM were conducted to create mutants HheD8-M5 (A69F/R127L), HheD8-M6 (A69F/M124P/R127L), and HheD8-M7 (A69F/M124P/R127L/R182W), which displayed progressive improvements in catalytic activity while maintaining excellent enantioselectivity (Fig. 2e). The exceptionally performant quadruple-mutant HheD8-M7 delivered (R)-1c with > 99% e.e. and achieved 46% conversion of 20 mM (rac)-1b after 10 h. In a comparative assessment of the dehalogenation reaction, the mutants HheD8-M5, HheD8-M6, and HheD8-M7 were found to exhibit inferior catalytic performance relative to mutant HheD8-M4 (entries 7–9 vs entry 6, Supplementary Table 4). Concurrently, the mutants HheD8-M2, HheD8-M3, and HheD8-M4 were assessed with the ring-opening reaction. The results revealed that the HheD8-M4 outperformed HheD8-M7 in catalytic efficiency, achieved 50% conversion of 20 mM (rac)-1b and produced (R)-1c with 99% e.e. (entry 9 vs entry 6, Supplementary Table 5).
Structural analysis of mutant HheD8-M3
For possible understanding the impact of these mutagenesis on catalytic efficiency and enantioselectivity, we sought to determine the crystallographic structure of the HheD8-M4. However, this mutant tended to precipitate during the purification process. The mutant HheD8-M3 (A69F/M124P/R127G), which lacks the R182W mutation, showed improved solubility that favored the crystallization process. We successfully resolved the X-ray crystal structure of the HheD8-M3 complex with chloride as a ligand in the halide-binding site to a resolution of 2.40 Å (PDB code: 8XXB, Supplementary Fig. 1). Our further attempts to obtain the crystals of the HheD8-M3 complexed with oxetane or azide were not successful. Overlap analysis revealed that the mutant HheD8-M3 adopts an overall architecture similar to the HheD8-WT/AF (Supplementary Fig. 2), with a highly matched catalytic triad composed of residues (S117-Y130-R134). Through detailed comparison of these mutated residues (Extended Data Fig. 1a), we speculate that the A69F mutation, which replaces the nonpolar amino acid alanine (A) with the larger aromatic amino acid phenylalanine (F), leads to enhanced catalytic enantioselectivity through aromatic π-π stacking interactions, while it slightly decreases catalytic efficiency by reducing the space size of the active site pocket. Our inference is consistent with the experimental results comparing the catalytic performances of HheD8-WT and HheD8-M1 (Fig. 2c). Mutations M124P and R127G introduce smaller amino acid residues, enlarging the active site pocket, which is thought to enhance catalytic efficiency (Extended Data Fig. 1b). This speculation is supported by the observed stepwise increase in activity for the mutants HheD8-M2 and HheD8-M3, following the sequential introduction of these mutations (Fig. 2c).
Scope for enantioselective formation of oxetanes
Subsequently, the substrate scope of the biocatalytic platform for the enantioselective dehalogenation of γ-haloalcohols was explored on a preparative-scale (Fig. 3, Supplementary Table 6). Various aryl γ-chloroalcohols bearing electron-donating or electron-withdrawing substituents on the phenyl ring (1-15a) were accepted to furnish both chiral (R)-oxetanes 1-15b (32–46% yield, 93->99% e.e.) and (S)-γ-chloroalcohols 1-15a (47–53% yield, 86->99% e.e.) with good yields and high optical purities. It is worth noting that the presence of steric hindrance at the ortho position of the aromatic ring (2-5a) was found to be compatible with the biotransformation. Substitutions were well tolerated at both the para and meta positions (6-14a). Smooth conversion of substrate 14a, featuring a strong electron-withdrawing trifluoromethyl group, to chiral oxetane (R)-14b was achieved, albeit with a slight decrease in enantioselectivity. Interestingly, introducing two fluoro substituents on the phenyl ring (15a) did not hamper the reaction; rather, it successfully delivered both chiral (R)-15b and (S)-15a with excellent enantioselectivity (E > 200). The substrate containing a bulky naphthalene moiety (16a) underwent effective and enantioselective dehalogenation using the mutant HheD-M3, yielding chiral oxetane (R)-16b and γ-chloroalcohol (S)-16a with 37% yield, > 99% e.e. and 50% yield, 96% e.e., respectively. Additionally, enantioselective dehalogenation of α-,α-disittuted γ-chloroalcohol 17a also proceeded smoothly to furnish the (R)-17b (36% yield, > 99% e.e) and (S)-17a (49% yield, > 99% e.e). Remarkably, the reaction also tolerated structural perturbations, accommodating the substitution of the aryl ring with several heterocyclic moieties such as pyrimidine (18a), quinoline (19a), and thiophene (20a). Chiral heterocyclic γ-chloroalcohols (18-20a, 48–50% yields and 89->99% e.e.) and oxetanes (18-19b, 42–44% yields and 96–97% e.e.) were successfully isolated with the exception of the unstable oxetane 20b. Moreover, the reaction exhibited tolerance toward the alkyl-substituted substrate 21a, affoding chiral alkyl oxetane (R)-21b (55% yield, > 99% e.e.) and alkyl γ-chloroalcohol (S)-21a (48% yield, > 99% e.e.). Furthermore, the biocatalytic system was applied to the enanioselective generation of chiral γ-bromoalcohol (S)-22a (50% yield, > 99% e.e) and γ-iodoalcohol (S)-23a (48% yield, > 99% e.e).
Scope for enantioselective ring-opening of oxetanes
We next turned our attention to the substrate generality for the enantioselective ring-opening of oxetanes (Fig. 4, Supplementary Table 7). As expected, aromatic substituted oxetanes (1-14b) with mono-substituent at the para-, meta-, and ortho positions were well tolerated, affording the corresponding chiral (R)-γ-azidoalcohols (41–50% yield, 91->99% e.e) and (S)-oxetanes (30–49% yield, 86->99% e.e) with good to excellent enantioselectivities (E = 72->200). The electronic characteristics of the substituents show a modest effect on both the efficiency and enantioselectivity of the ring-opening reaction. Oxetane 15b, bearing two fluoro meta-substituents, was also successfully accommodated to yield the chiral (R)-15c and (S)-15b with > 99% e.e. and high yields. Additionally, efficient ring-opening of sterically hindered oxetanes 16b and 17b was achieved using the mutants HheD8-M3 and HheD8-M7, respectively. Moreover, heterocyclic oxetanes contaning pyrimidine (18b) or quinoline (19b) served as competent substrates for the mutant HheD8-M4. Alkyl oxetane 21b was also smoothly converted to deliver the corresponding chiral alkyl γ-azidoalcohol (R)-21c and oxetane (S)-21b. Furthermore, we challenged the ring-opening reaction with other anionic nucleophiles. The reaction accepted the nucleophilic cyanide, generated in situ from mandelonitrile36, yielding the chiral γ-cyanoalcohol (R)-1d with 46% yield and > 99% e.e.. The nitrite demonstrated as an ambident nucleophile, leading to the formation of chiral γ-nitroalcohol (R)-1e and γ-diol (R)-1f through attack by its nitrogen and oxygen atoms, respectively. Cyanate and thiocyanate were also examined but exhibited very low reactivity and enantioselectivity in the ring-opening reactions (Supplementary Table 8). Further efforts in enzyme screening and protein engineering are required to facilitate oxetane ring-opening reactions utilizing these nucleophilic agents.
Large-scale reactions
Substrate tolerance is a critical metric for assessing the effectiveness of biocatalytic approaches. To demonstrate its practical applicability, the biocatalytic platform was then evaluated at higher substrate concentrations within a biphasic system (PB buffer: n-hexane = 5:1), employing the variant E. coli (HheD8-M4) whole cells as biocatalysts. The enantioselective dehalogenation reactions of 40–140 mM (rac)-1a proceeded smoothly to completion with 47–50% conversions over a period of 3–48 h (Extended Data Fig. 2, Supplementary Table 9). The decrease in reaction pH, caused by proton release at higher substrate concentrations, may diminish the reaction efficiency. By maintaining the reaction pH at 8.5 ± 0.1 through the addition of aqueous sodium hydroxide solution, a large-scale reaction to convert 200 mM of (rac)-1a (20 mmol, 34 g/L) was performed and completed smoothly within 33 h (Fig. 5a). For the enantioselective ring-opening of 40–200 mM oxetane (rac)-1b, reactions consistently reached 50% conversions within 12–36 h (Extended Data Fig. 3, Supplementary Table 10). A large-scale reaction of 200 mM (rac)-1b (20 mmol, 27 g/L) was successfully conducted to completion after 39 h (Fig. 5a). Notably, in both the large-scale dehalogenation and ring-opening reactions, all chiral compounds were obtained in gram quantities, achieving high isolated yields (39–49%) and excellent enantiopurities (97->99% e.e.). These results substantiate that the biocatalytic platform accommodates large-scale synthesis at high substrate concentrations while maintaining catalytic efficiency and enantioselectivity, thereby emphasizing its synthetic potential for industrial applications.
Biocatalytic cascade reactions
The favorable compatibility of enzymatic reactions facilitates their application in biocatalytic cascade processes, thus omitting the need for intermediate purification and isolation steps41. We subsequently explored the feasibility of integrating and executing both the formation and ring-opening reactions of oxetanes in a one-pot, one-catalyst cascade system. This setup was envisaged to enable the enantioselective transformation of γ-haloalcohols into γ-azidoalcohols via the unisolated oxetane intermediates. As depicted in Fig. 5b, three γ-haloalcohols (1a, 6-7a) underwent evaluation using E. coli (HheD8-M4) cells. The results indicated that all reactions proceeded efficiently, resulting in the formation of both chiral γ-azidoalcohols and γ-haloalcohols with excellent isolated yields (47–50%) and as single enantiomers (> 99% e.e.). The successful integration of the biocatalytic cascade reactions further highlighted the operational flexibility of the biocatalytic platform.
Transfomations of chiral products
The developed biocatalytic platform enables the synthesis of both (R)- and (S)-enantiomers of oxetanes as well as various chiral γ-substituted alcohols. These provide a versatile platform for synthesizing many valuable compounds, particularly serving as key intermediates and functional groups in bioactive molecules. Representative transformations were subsequently carried out using the enzymatically synthesized chiral products (Fig. 5c). The γ-chloroalcohol (S)-1a can be transformed into the chiral (R)-3-phenylisoxazolidine (1aa), an important heterocyclic motif found in anticancer agents42. Smooth conversion of the chiral (S)-1a to furnish (S)-1ab was also realized, providing a key precursor for the synthesis of (R)-dapoxetine43. Additionally, the chiral oxetane (R)-1b was successfully converted into (R)-2-phenyltetrahydrofuran (1ba) through a straightforward ring-expansion reaction44. We also performed the conversion of (R)-1b into the chiral (R)-1bb, a crucial precursor in the synthesis of (R)-tomoxetine45. Moreover, the presence of an azide group enables copper-catalyzed click chemistry46, allowing for the modification of γ-azidoalcohol (R)-1c with a simple alkyne to afford the chiral γ-hydroxytriazole (R)-1ca. The reaction of (R)-1c with a symmetrical ketone was aslo performed to deliver medium-sized ring lactam (R)-1cb, a commonly important intermediate in the synthesis of nitrogen-containing compounds47. The absolute stereochemistry of (R)-1cb was ascertained by single-crystal X-ray diffraction analysis (Supplementary Table 11), which also demonstrated R-enantioselectivity during the enzymatic dehalogenation and ring-opening reactions. Furthermore, the transformations of chiral γ-cyanoalcohol (R)-1d yielded the chiral lactone (R)-1da and the γ-hydroxyamide (R)-1db, both serving as building blocks in the synthesis of (R)-fluoxetine and (R)-norfluoxetine48. For the nitrite-mediated ring-opening products, the γ-nitroalcohol (R)-1e can be readily reduced to generate the chiral γ-aminoalcohol (R)-1ea49. On the other hand, esterification of chiral γ-diol (R)-1f yielded a key intermediate, (R)-1fa, for the systhesis of natural piperidine alkaloids50. Taken together, these representative transformations resulting in the generation of useful derivatives showcased the synthetic scalability of the biocatalytic platform.