Efficient chemoenzymatic synthesis of α-aryl aldehydes as intermediates in C–C bond forming biocatalytic cascades

Abstract

Multi-enzyme biocatalytic cascades are emerging as practical routes for the synthesis of complex bioactive molecules. However, the relative sparsity of water-stable carbon electrophiles limits the synthetic complexity of molecules made from such cascades. Here, we develop a chemoenzymatic platform that leverages styrene oxide isomerase (SOI) to covert readily accessible aryl epoxides into α-aryl aldehydes through a Meinwald rearrangement. These unstable aldehyde intermediates are then intercepted with a C–C bond forming enzyme, ObiH, that catalyzes a transaldolase reaction with l-threonine to yield synthetically challenging β-hydroxy-α-amino acids. Co-expression of both enzymes in E. coli yields a whole cell biocatalyst capable of synthesizing a variety of stereopure non-standard amino acids (nsAA) and can be produced on gram-scale. We used isotopically labelled substrates to probe the mechanism of SOI, which we show catalyzes a concerted isomerization featuring a stereospecific 1,2-hydride shift. The viability of in situ generated α-aryl aldehydes was further established by intercepting them with a recently engineered decarboxylative aldolase to yield γ-hydroxy nsAAs. Together, these data establish a versatile method of producing α-aryl aldehydes in simple, whole cell conditions and show that these intermediates are useful synthons in C‒C bond forming cascades.

Graphical Abstract

Introduction

Enzymatic cascades are highly sought for their potential to rapidly build molecular complexity and circumvent the need to isolate intermediates.^1,2 The opportunities enabled by multi-enzyme cascades are exemplified by the recent process-scale syntheses of islatravir and molnupiravir, in which enzymes from multiple species were engineered to produce high value target molecules.^3,4 While some enzymatic cascades have been leveraged for preparative-scale synthesis, the scope of currently accessible products is dwarfed in comparison to the vast molecular diversity achieved in Nature.¹ This natural diversity is accomplished by a constellation of enzymes that are often tuned to perform specific chemical transformations on single substrate.¹ In contrast, broadly useful synthetic methodologies are characterized by their ability to perform well-defined transformations on large numbers of substrates. To meet this synthetic ideal with enzymatic cascades, each enzyme in the cascade must have a complementary and broad substrate scope. There have been phenomenal advances in the development of such promiscuous biocatalytic cascades that perform sophisticated functional group interconversions, such as the conversion of racemic alcohols into chiral amines.^5,6 However, there are comparatively few biocatalytic cascades that catalyze a well-controlled C‒C bond forming reaction on diverse substrates, in part due to the limited availability of water-stable reactive carbon precursors. Hence, biocatalytic strategies to produce reactive carbon electrophiles are poised to enable a plethora of enzymatic cascades capable of rapidly and efficiently constructing complex molecular scaffolds.

Aldehydes are preeminent substrates for C–C bond forming enzymes. A key characteristic of their reactivity (and of other carbonyl containing molecules), is their ability to tautomerize and form both electrophilic and nucleophilic carbon species. The exploitation of this dual reactivity is widespread in Nature and a key theme in central metabolism.⁷ Correspondingly, multiple enzymes that feature aldol chemistry have independently evolved to catalyze these valuable transformations. Their ability to catalyze enantioselective transformations with simple starting materials has motivated many engineering and synthetic studies. For example, aldolases have been leveraged to perform key C–C bond forming steps in biocatalytic systems that produce complex carbohydrates^8,9, functionalized α-ketoacids^10,11, and amino acid analogs.^12–15 Independent of the particular enzyme employed, aldol chemistry is fundamentally constrained by substrate availability. Although thousands of aromatic and aliphatic aldehydes are commercially available, the simple α-aryl aldehydes are notably scarce (Figure 1b). This sparsity is likely due to the intrinsic instability of α-aryl aldehydes as conjugation with the arene stabilizes the enol tautomer, leading to rapid depletion of the reagent. Given the ubiquity of aldehydes in biocatalytic cascades, the development of a robust method to synthesize and capture α-aryl aldehydes in C–C bond forming reactions would provide access to valuable reagents for diverse biocatalytic transformations.

Figure 1.

α-Aryl aldehydes motifs in biosynthesis and in situ generation for biocatalysis. A) α-Aryl aldehydes are intermediates in biosynthesis. The carbons derived from α-aryl aldehydes are shown in red. B) Styrene oxide analogs are the ideal entry point for the generation of α-aryl aldehyde intermediates. Previous work utilizing styrene, styrene monooxygenase (SMO), and styrene oxide isomerase (SOI) successfully produced α-aryl aldehydes. Chemoenzymatic synthesis of styrene oxides from aryl carbonyl compound expands the potential substrate pool for SOI. C) This work: mechanism of SOI and C–C bond forming cascade catalysis to produce non-standard amino acids (nsAAs). α-Aryl aldehydes generated by SOI are intercepted by PLP-dependent, C–C bond forming enzymes yielding structurally diverse amino acids.

In Nature, α-aryl aldehydes are typically generated through the Ehrlich pathway (Figure S1),¹⁶ which is involved in aromatic amino acid catabolism and provides key intermediates in the metabolic synthesis of bioactive natural products (Figure 1a).¹⁷ Wang and coworkers utilized enzymes from the Ehrlich pathway to produce an α-aryl aldehyde substrate for norcoclaurine synthase en route to benzylisoquinoline alkaloid analogs.¹⁸ However, to access the α-aryl aldehyde synthon through the Ehrlich pathway, multiple enzymes must operate on a corresponding non-standard aromatic amino acids (nsAAs), hampering the viability of this method in a preparative synthetic context. A potentially competitive approach to generating α-aryl aldehydes involves redox-mediated transformations of carboxylic acids or primary alcohols.¹⁹ Such enzymes are generally quite effective, but the requirement of accessory enzymes for cofactor regeneration and redox balance adds to the complexity of such an approach. The direct, biocatalytic anti-Markovnikov oxidation of styrene was previously demonstrated by Hammer and coworkers.²⁰ While the transformation is compelling, the scope of the engineered P450 oxidase is limited at preparative scales. In comparison, purely synthetic approaches involve Wittig-like reactions that require purification of the unstable α-aryl aldehydes prior to utilization in downstream reactions.²¹

We were drawn to recent reports from the Zhi Li lab who produced α-aryl aldehydes using a two-enzyme cascade from Pseudomonas sp. VLB120.^22–25 In this system, a flavin-dependent styrene monooxygenase (SMO) catalyzes epoxidation of styrene to produce styrene oxide. This molecule is the substrate for styrene oxide isomerase (SOI), an integral membrane enzyme. SOI catalyzes a redox neutral Meinwald rearrangement²⁶ to produce α-aryl aldehydes, which were shown to be amenable to a variety of functional group interconversions in cascade reactions with transaminases and dehydrogenases (Figure S2).^22–24,27 However, the SMO-SOI cascade has limitations. Chiefly, efficient formation of the α-aryl aldehyde intermediate requires both enzymes to react with non-natural substrate analogs. The challenge of using two enzymes to generate the α-aryl aldehyde intermediate is compounded when considering the functional diversity of commercially available styrenes is relatively low (Figure 1b).

Here, we design and implement a chemoenzymatic route to access a wide array of α-aryl aldehydes for subsequent biocatalytic elaboration through C–C bond forming reactions (Figure 1c). To circumvent SMO, most aryl epoxides were made in a single step via the Corey-Chaykovsky reaction from the vast pool of inexpensive aryl aldehydes or through bromination of acetophenones and subsequent reductive epoxidation (Figure S3). These substrates were reacted with SOI in whole cell fashion, and the resulting aldehydes were intercepted in situ by C–C bond forming enzymes. The l-threonine (Thr) transaldolase ObiH was shown to react with an exceptionally broad scope of SOI-generated aldehydes, yielding non-standard β-hydroxy amino acids with excellent d.r. and ee. We used isotopically labeled substrates to clarify the mechanism of SOI and show that SOI catalyzes a concerted isomerization with a stereospecific 1,2-hydride-shift. We further showcase the utility of SOI by combining it with an additional C–C bond forming enzyme, UstD. UstD is a pyridoxal phosphate (PLP)-dependent enzyme that catalyzes a formal decarboxylative aldol reaction of l-aspartate into diverse aldehydes, generating γ-hydroxy-α-amino acids.^28,29 We show that UstD can intercept the α-Aryl aldehydes intermediates to generated structurally diverse nsAAs. Together, these data demonstrate the use of SOI in tandem with C–C bond forming enzymes; adding a robust method for accessing metastable C-electrophiles in aqueous conditions.

Results and Discussion

α-Aryl aldehydes produced from SOI can be intercepted with purified ObiH.

We began by testing the ability of purified ObiH to intercept phenylacetaldehyde produced by heterologously-expressed SOI and commercially-available styrene oxide (2a). Following the report from Wu et al., we expressed the membrane-integrated SOI enzyme as a C-6xHis construct in E. coli BL21(DE3).^22,23 Rather than purify the catalyst, which is a notoriously cumbersome process for membrane-integrated proteins, we tested the activity of the cascade in reactions with whole cells expressing SOI, purified ObiH and its substrate Thr (1), and styrene oxide (2a), to generate a stable amino acid product in a one-pot reaction (Figure 2a). Gratifyingly, we observed significant accumulation of β-hydroxyhomophenylalanine (2b) by UPLC-MS and confirmed that product formation requires the combination of SOI and ObiH (Figure S4). By changing the concentration of ObiH, we were able to study the efficiency of the two-enzyme system (Figure 2b). Increasing the concentration of ObiH in these reactions resulted in a commensurate increase in product formation when reactions were run to incomplete conversion. These data indicate that, under these conditions, SOI rapidly converts styrene oxide into phenylacetaldehyde and ObiH catalyzes the flux-limiting step.

Figure 2.

One-pot biocatalytic cascade with SOI and ObiH. A) BL21 (DE3) cells expressing SOI and purified ObiH generate a stable amino acid from 25 mM styrene oxide (2a or 3a), 100 mM Thr (1), 100 mM Tris•HCl pH 8.5, 0.1% dry cell weight (dcw) and variable amounts of purified ObiH. B) Fold increase in total product observed from styrene oxide (2a) and p-bromostyrene oxide (3a) is shown as a function of ObiH concentration. Data indicate how the flux-limiting step shifts as a function of substrate.

Based on our observation that ObiH limits flux through the cascade with styrene oxide, we hypothesized substrates that react more sluggishly with SOI would be less dependent on ObiH concentration. In particular, styrene oxide analogs substituted with electron withdrawing groups would destabilize the positive charge that is proposed to accumulate at the benzylic carbon during isomerization (Figure S5).²² We repeated the ObiH titration with p-bromostyrene oxide (3a), which revealed a shallower dependence on ObiH concentration; a 16-fold increase in ObiH concentration only produced a ~2.5-fold increase in product formation (Figure 2b). Nevertheless, these data showed that α-aryl aldehydes generated via SOI in whole cell catalysis could be intercepted by ObiH.

Co-expression of SOI and ObiH results in a whole cell biocatalyst with exceptional activity.

Operational simplicity is a critical element for the adoption of new synthetic methods. We therefore sought to develop a whole cell catalyst in which SOI and ObiH are co-expressed in E. coli BL21 (DE3) on separate expression vectors. Such approaches are well precedented and greatly simplify the work required by preparing both catalysts in parallel as a single unit (Figure 3a).²² We used two common methods for preparing and deploying the whole cell catalysts. Cell pellets harvested from expression cultures were either extruded through a syringe and flash-frozen in liquid nitrogen or simply frozen and lyophilized. Activity comparison between the two methods showed that each produced highly active catalysts that can be stored for months while maintaining activity (Figure S6). Lyophilized cells were utilized during optimization and initial exploration of the substrate scope for accurate accounting of catalyst mass and reproducible handling. Frozen wet cells were used for subsequent preparative-scale reactions as lyophilization requires additional time and specialized equipment that may not be accessible in all laboratory settings. Importantly, both preparations yielded catalysts that can easily be deployed in large scale reactions.

Figure 3.

E. coli coexpressing SOI and ObiH efficiently produces β-hydroxy-α-amino acids from epoxides and Thr. A) Cascade catalysis with cells co-expressing SOI and ObiH. B) Product formation of β-hydroxyhomophenylalanine (2b, dark red) and p-bromo-β-hydroxyhomophenylalanine (3b, blue) with SOI-ObiH cells in overnight reactions. Relative product formation is scaled to the maximum observed product for each compound. C) Product formation over time for 2b (dark red) and 3b (blue) with 0.1% dcw SOI-ObiH cells. D) Preparation of catalyst as either frozen wet cell or dry cells; both can be deployed for preparative scale reactions. Gram-scale synthesis and isolation of 2b at 10 mmol scale. Synthesis of 3b, which has low aqueous solubility, was improved by either increasing catalyst loading or decreasing substrate concentration.

We next sought to understand the impact of catalyst loading on product formation (Figure 3b). The product concentration reached a maximum with 0.05% dry cell weight (dcw), with yields corresponding to ~80% of the theoretical titer with 2a as the substrate. When we added more cells, instead of plateauing, we observed that reaction yields decreased. We considered whether endogenous reducing enzymes could consume the phenylacetaldehyde intermediate. A strain of E. coli in which multiple reducing enzymes were knocked out, E. coli K-12 MG1655 (DE3) RARE, was engineered by the Prather Lab and has been used in biocatalytic applications that feature aromatic aldehydes.³⁰ We hypothesized that use of this strain might ameliorate depletion of the aldehyde in whole-cell reactions. However, when we expressed the SOI-ObiH cascade in this strain of E. coli we observed no change in product formation in comparison to the more commonly used BL21 DE3 strain, which was used for the rest of this study (Figure S7). While the complex cellular milieu prevents the determination of the specific pathway that is consuming the intermediate, aldehydes as well as the epoxide starting material are generally prone to reactions with cellular nucleophiles, which may be sequestering the substrate and intermediate at higher cell loadings. Cell loading experiments were repeated with 3a. In these reactions, 10-fold higher cell loadings were required to achieve maximal product formation relative to styrene oxide (Figure 3b), which we attribute to the lower activity of SOI on the p-bromo substrate (Figure 2b).

To test the scalability of this two-enzyme cascade, we measured reaction progress and conducted preparative scale reactions. With 2a at 25 mM and catalyst loading of 0.1% dcw, reaction yields reached 50% in ~30 minutes and slowly increased to 83% overnight (Figure 3c). While we were interested in tracking the abundance of phenylacetaldehyde during this process, such studies were precluded by its instability. Isolated yields for 2b changed little as substrate loading, catalyst preparation, and catalyst loading were varied; with yields between 52–69% (Figure 4 and Table S1). Inspired by this success, reactions with 2a were scaled to 10 mmol, from which 1.21 g of 2b was isolated, corresponding to a 62% yield (Figure 3d). Reactions with 25 mM 3a, which reacts slower than 2a, reached a 37% yield overnight on analytic scale. Initial preparative-scale reaction conditions with 3a produced 3b with just 12% isolated yield. Increasing the catalyst loading 10-fold to 1.0% wcw (wet cell weight) resulted in an increase in 3b to 32%. We observed a large fraction of insoluble 3a remained after these overnight reactions. Therefore, substrate loading was reduced 10-fold to 2.5 mM 3a while maintaining 100 mM Thr, which increased the isolated yield to 51% at 0.1% wcw. We attempted reactions with 2.5 mM 3a and 1.0% wcw, but isolation of the relatively dilute product was stymied by the large amount of cellular material and unreacted Thr.

Figure 4.

Preparative scale reactions with SOI-ObiH whole cells. Reactions were conducted with 1 mmol epoxide unless otherwise stated. General reaction conditions used 25 mM epoxide, 100 mM l-Thr, 100 uM PLP, 100 mM Tris•HCl pH 8.5, 5% EtOH, and variable concentrations of SOI-ObiH whole cells: 0.1% wcw^a, 1.0% wcw^b, or 0.1% dcw^d; 3b^c (1 mmol epoxide, 2.5 mM, 0.1% wcw), 2b^d and 5b^d (2.5 mmol epoxide, 25 mM, 0.1% dcw). Products 8b and 8c were isolated from a single reaction with 8a. Product purity was assessed via ¹H NMR. All products were derivatized with Marfey’s reagent to determine dr indicating that all products were isolated with dr >19:1 and ee >99:1 with clean 2R, 3S stereochemistry unless otherwise noted. We estimate reactions at 1.0% wcw to have 5 μM ObiH present.

Previous studies by us and others have shown that ObiH has excellent stereoselectivity.^14,31,32 The S-isomer at the 2-position (C_α) is formed due to the complete facial selectivity of the ObiH quinonoid nucleophile. ObiH forms the 3-(R)-isomer with high, >20:1 selectivity under initial velocity conditions for most substrates tested, which matches the absolute configuration of obafluorin, the natural product produced in the biosynthetic pathway encoding ObiH.³¹ NMR analysis and derivatization with the chiral shift reagent l-FDAA (Marfey’s Reagent)³³ followed by UPLC-MS analysis both showed that 2b and 3b are formed as single diastereomer, which we assign as (2S, 3R). Notably, this high stereochemical purity is maintained even as reactions proceed to high yield, which stands in contrast to the well-studied Thr aldolases³⁴ and the reaction of ObiH with aryl aldehydes.¹⁴

SOI-ObiH form a promiscuous biocatalytic cascade.

We next sought to explore the substrate scope of the SOI-ObiH cascade. Commercial epoxide substrates were used when available, but the majority of substrates assayed were synthesized through the Corey-Chaykovsky reaction. This one-step C–C bond forming reaction readily converts aldehydes and ketones into terminal epoxides using sulfonium iodide and NaH (Figure S3). These starting materials provide one of the most structurally diverse entry points for the synthesis of α-aryl acetaldehyde analogs (Figure 1b). Furthermore, the multi-enzymatic transformation was tolerant of contaminants present in crude mixtures of epoxide substrates, allowing us to utilize epoxides that are otherwise sensitive to chromatographic purification with silica gel. With this approach, we were able to generate and intercept diverse α-aryl aldehydes with ObiH and demonstrate the preparative-scale utility of this cascade (Figure 4).

We assayed the efficiency of the SOI-ObiH cascade with a series of styrene oxides. We began with an isotopically labeled styrene oxide in which the oxirane is fully substituted with deuterium (2a-2,3,3-d₃) which gave an amino acid product (2b-3,4,4-d₃) with deuterium at C_β and C_γ in 48% yield with 1.0% wcw, indicating the intramolecular transfer of a hydride during isomerization. We discuss the mechanistic implications associated with deuterated styrene oxides below and transitioned to epoxides with substitutions on the arene ring. Under standard reaction conditions, p-bromostyrene oxide (3a) and p-chlorostyrene oxide (4a) produced modest isolated yields of 16% and 36% respectively with equivalent catalyst loading (0.1% wcw). Increasing catalyst loading to 1.0% wcw with 25 mM 4a enabled product isolation in 59% yield. In contrast, p-fluorostyrene oxide (5a) was isolated in 61% yield at 0.1% wcw catalyst loading. Additional halogen substitutions were tolerated as the product from a reaction with 2,4-difluorostyrene oxide (6a) resulted in an isolated yield of 53%. When considering other functional groups, we were excited about the possibility of entering the cascade with p-nitrostyrene oxide substrate, which would generate the native substrate for ObiH (4-nitrophenylacetaldehyde). However, we were unable to detect amino acid product from this substrate. Given that 4-nitrophenylacetaldehyde is the native substrate of ObiH, we attribute the lack of activity in the cascade to SOI and hypothesize that the electron withdrawing properties of this substrate destabilizes the partial positive charge build up on the benzyl carbon during isomerization (Figure S5). Other styrene oxide analogs functionalized with electron withdrawing groups also failed to generate product through the cascade (o-NO₂, m-NO₂, p-CN, and p-CF₃). In contrast, styrene oxides functionalized with electron rich substituents showed high activity in the SOI-ObiH cascade. p-Methoxy styene oxide (7a) reacted in the cascade with a yield of 51% at 1.0% wcw. We were interested accessing β-hydroxyhomotyrosine (8c) from the SOI-ObiH cascade as it is an important motif in pharmaceutically relevant natural products.^35–37 The corresponding p-hydroxy stryene oxide proved difficult to access, but we successfully transformed the acetylated epoxide (8a) into 8b and 8c with isolated yields of 18% and 17% from a single reaction. The benzodioxole containing epoxide derived from piperonal (9a) into an amino acid product with a 41% yield at 1.0% wcw. We then transitioned to styrene oxides substituted with aliphatic functional groups (10a-12a). Styrene oxides substituted with methyl groups at various positions on the phenyl ring all produced β-hydroxy-α-amino acid products, with 37–48% yield at 0.1% wcw catalyst loading. Standard reaction conditions with 10a gave 10b in 37% yield. We increased the catalyst loading to 1.0% wcw, which afford 10b in 58% yield. p-Isopropyl styrene oxide (13a) reacted less efficiently in the SOI-ObiH cascade, resulting in an isolated yield of 26% at 1.0% wcw. The isomeric 2,4,6-trimethylstryene oxide (14a) was successfully transformed through the cascade, albeit with a low yield of 11%. The reactivity of the trimethyl substrate suggested that other bulky substrates might also react in the SOI-ObiH cascade. Naphthyl-2-aldehyde was previously observed to react with ObiH.¹⁴ Here, we observed that the naphthyl-1-epoxide 15a reacted in the cascade to produce the corresponding amino acid 15b with a 25% yield. To assess whether SOI could convert epoxide functionalized aromatic heterocycles into the corresponding α-aryl aldehyde, we synthesized oxiranyl pyridine substrates from 2-, 3-, and 4-pyridine carboxaldehydes. However, no product was observed from the SOI-ObiH cascade. We speculated the electron-deficient nature of the pyridine heterocycle prohibited the Meinwald-like rearrangement, which correlates well to the inactivity on electron deficient styrene oxides. We therefore screened reactions with the electron rich furan and thiophene-containing epoxides. Both epoxides were successfully transformed into the corresponding amino acids 16b and 17b with an 18% yield.

Di-substitution at the α-position in the epoxide starting material presents a distinct challenge for SOI. Activity with α-methyl styrene oxide (18a) had been reported previously by Wu et al. We observed that racemic 18a reacted in the SOI-ObiH cascade to give γ-methyl-β-hydroxyhomophenylalanine (18b) in 27% yield as a 2:1 mixture of diastereomers at C_γ which were readily resolved via UPLC-MS (Figure 5a). ¹H-NMR analysis indicated the major isomer has an anti-relationship between the hydroxy and methyl groups at C3 and C4, respectively, implying a 4R configuration for the major isomer. We synthesized the epoxide from 1-indanone (19a) which reacted in the cascade to give the bicyclic amino acid product (19b) in 25% yield and 10:1 d.r. Additional α-branched styrene oxide analogs including 1-tetralone and α-methyl styrene oxide analogs were also tested and appeared to form stable amino acids products as determined by UPLC-MS analysis and ¹H-NMR data from preparative scale reactions. However, low yields and convoluted structural data precluded isolation and full characterization (Table S1).

Figure 5.

Stereochemical analysis of the SOI reaction. A) Reactions were run with 25 mM 18a and 0.1% wcw SOI-ObiH catalyst. Products were analyzed via UPLC-MS, which resolves the diastereomers formed at the 4-position of 18b. Isolated material from preparative scale reactions contains two diastereomers at C-4 with product ratios of 65:35. 1H-NMR analysis indicates that the 4R configuration is the major product. B) Reactions with 20 μM purified ObiH and 20 mM 2-phenylpropanal (22) were quenched after 1 h and yield 18b with a 60:40 mixture of diastereomers. C) Reactions with 25 mM rac-18a and 0.1% wcw SOI-ObiH catalyst were quenched after 3 min and yield 18b with a 70:30 mixture of diastereomers. D) Reactions with 25 mM (S)-18a and 0.1% wcw SOI-ObiH catalyst were quenched after 10 min and yield 18b with a 20:80 mixture of diastereomers. E) Same as D), except (R)-18a is used and yields a 95:5 ratio of diastereomers.

The promiscuous nature of SOI encouraged us to probe epoxides that were not α-aryl epoxides. We observed the cinnamaldehyde-derived epoxide reacted to generate β-hydroxy nsAA 20b, albeit with low yield. Substitution with an additional methyl group increased the reactivity of the substrate, producing 21b in 17% yield. We assayed other seemingly suitable non-aryl epoxides but failed to detect SOI activity with any epoxides that are not adjacent to an extended pi system (Table S2). Nevertheless, these are the first data demonstrating the SOI can catalyze a reaction with non-aryl epoxides. The overall scope presented here demonstrates the compatibility of SOI and ObiH to efficiently synthesize a variety of complex β-hydroxy nsAAs.

SOI catalyzes a concerted and stereospecific isomerization.

We rationalized that diastereomeric enrichment observed in 18b was a result of either SOI or ObiH exerting a stereopreference for one of the α-methyl-styrene oxide enantiomers or the disubstituted aldehyde intermediates, respectively. The intermediate produced by SOI in this reaction, 2-phenylpropanal (22), is commercially available and was subjected to reactions with purified ObiH. Analysis via UPLC-MS revealed that ObiH forms the two diastereomers of 18b in nearly equal populations, indicating ObiH exerts negligible stereoselectivity with this substrate (Figure 5b). We next subjected rac-18a to the SOI-ObiH cascade and quenched the reaction at early time points, which revealed a 70:30 mixture of diastereomers, which is consistent with the known stereochemical preference of SOI (Figure 5c).²² To study the stereochemical outcome of the hydride transfer, we next prepared both enantiomers of α-methyl styrene oxide ((S)-18a and (R)-18a) and subjected them to the SOI-ObiH cascade. A change in the Cahn-Ingold-Prelog priorities during the transformation of 18a implies that inversion of configuration maintains the S/R designation in the product. Analysis via UPLC-MS clearly showed that each substrate generated a distinct stereoisomer when reactions were analyzed at early timepoints, 4-(S)-18b (20:80) (Figure 5d) and 4-(R)-18b (95:5) (Figure 5e), precluding a carbocation intermediate. Previously, Wu et al. noted that the SOI reaction is enantiorententive for substrates with substituents at the benzyl carbon (C-2) and has a slight preference for (R)-18a from kinetic resolution studies with rac-18a.²² It was hypothesized SOI forms a carbocation intermediate after C–O cleavage and that the active site environment directs a stereospecific hydride shift to produces a single enantiomer of α-methyl-phenylacetaldehyde (Figure S5).^22,23,25 However, reactions with rac-18a produced both enantiomers of the corresponding aldehyde. If the mechanism were to proceed through a sp²-hybridized, carbocation intermediate, a stereospecfic hydride shift would only produce a single enantiomer of the chiral aldehyde. Instead, these data indicate that SOI catalyzes isomerization with inversion of configuration, which is indicative of a concerted Meinwald rearrangement.

Cascade reactions with (S)-18a and (R)-18a were also allowed to progress past the initial velocity regime and we observed that the diastereomeric ratio decreased overnight to 53:47 and 65:35 for (S)-18a and (R)-18a, respectively (Figure S8). We hypothesized that decrease in d.r. was a result of accumulation and enolization of 2-phenylpropanal prior to being intercepted by ObiH. To probe enolization of α-aryl aldehydes under these conditions, the SOI-ObiH cascade was performed in D₂O with 2a as a substrate. The amino acid product isolated from this reaction showed 85% deuterium incorporation at C_γ (Figure S9), providing strong evidence that the phenylacetaldehyde intermediate readily undergoes enolization under these reaction conditions.

We were intrigued by the concerted mechanism of SOI, which we sought to further substantiate through deuterium labelling experiments. As previously noted, reactions with 2a-2,3,3-d₃ gave an amino acid product (2b-3,4,4-d₃) with deuterium at C_β and C_γ, consistent with isomerization via a hydride shift (Figure 6a). However, it was unclear which of the two terminal hydrogen atoms were being transferred from C-3 to C-2 during isomerization. To this end, we synthesized stereo-enriched cis and trans β-monodeutero styrene oxides ((±)-cis-2a-3-d₁ and (±)-trans-2a-3-d₁) (Figure 6b and Figure S10). Each substrate was reacted with the SOI-ObiH cascade and the location of deuterium atoms in the final α-amino acid product was determined by ¹H-NMR after product isolation. In the case with the deuterium cis to the phenyl ring ((±)-cis-2a-3-d₁), the amino acid product contained deuterium only at C_β, indicative of migration of the trans hydrogen (Figure 6b). Concurrently, work from the Zhi Li lab also demonstrated that SOI is stereospecific for the trans migration by observing transfer of a trans-methyl group over the cis-hydrogen.²⁵ Consistent with this observation, reactions with the (±)-trans-2a-3-d₁ showed no deuterium at C_β and 25% deuterium at C_γ. The theoretical maximum labelling at C_γ would be 50%, indicating efficient transfer by SOI followed by hydrogen exchange due to enolization in protic solvent. Together, these experiments clarify the mechanism of SOI, which catalyzes a concerted isomerization with stereospecific 1,2-hydride-shift

Figure 6.

SOI catalyzes a concerted, stereoselective hydride transfer. A) Proposed mechanism for SOI catalyzed isomerization involving concerted ring-opening concomitant with a 1,2-H shift. B) Reactions with 25 mM 2a and 0.1% wcw SOI-ObiH catalyst were allowed to proceed overnight followed by isolation of the products. ¹H-NMR spectra for isolated β-hydroxyhomophenylalanine and isotopologs. The presence of ²H is inferred by diagnostic changes in the splitting and loss of a signal relative to the fully protic product (purple). Residual proton at C_β with (±)-cis-2a-3-d₁ is due to incomplete deuteration of the starting material (green), see SI for details. Reaction with (±)-trans-2a-3-d₁ results in ²H-transfer followed by exchange with the protic solvent, which increases the proton integration (red).

α-Aryl acetaldehydes can be intercepted by an additional C–C bond forming enzyme.

We were originally attracted to ObiH because its native substrate is an α-aryl aldehyde.^31,32 To further establish the synthetic utility of these reactive intermediates, we sought to couple SOI with a different C–C bond forming enzyme. UstD is a PLP-dependent enzyme originally identified in the biosynthesis of Ustiloxin that catalyzes a stereoselective, decarboxylative aldol addition into diverse aldehydes to generate γ-hydroxy amino acids.²⁹ We recently engineered UstD for increased activity, which yielded a variant containing seven mutations from wild-type, named UstD^v2.0, that can react with a variety of aldehydes.²⁸ To test whether the α-aryl aldehydes generated by SOI can be intercepted by UstD^v2.0, we co-expressed each enzyme in E. coli BL21 (DE3). We observed evidence of product formation via UPLC-MS from styrene oxide (2a). We conducted an analytical substrate scope for the SOI-UstD cascade which showed that UstD was also capable of intercepting numerous aldehyde intermediates produced by SOI (Figure S11), albeit with low apparent activity.

To increase yields for preparative reactions with 2a, cell loading was raised from 0.03–3% wcw but were still low-yielding (<30% yield) and the excess cell mass stymied product isolation (Figure S12). Given our experiences with the SOI-ObiH cascade, we hypothesized that UstD was flux-limiting for the reaction. Therefore, we tested reactions with just 0.1% wcw SOI-only cells and high concentrations, 2.5–5 μM, of purified UstD^v2.0, which would reduce the introduction of cell mass that impedes purification. Use of this reaction format significantly increased yield of 24 to 74%, which was subsequently shown via Marfey’s analysis to be a 5:2 mixture of diastereomers (Figure 7). This modest diastereoselectivity at C_γ is in contrast with previous studies, which showed that UstD forms the 4S isomer with high selectivity with aryl aldehydes.²⁸ The stereocenter at the 2 position is set in a separate step in the proposed mechanism, and generally occurs with pristine S-selectivity with fold-type I PLP-dependent enzymes. A deuterated analogue of 24 was synthesized by performing the cascade reaction in D₂O. The product (24–2,5,5-d₃) was isolated with a 44% yield and a 2:1 mixture of diastereomers. We observed deuterium incorporation at C_δ (75%) and C_α (90%). This labeling is consistent with the proposed mechanism of UstD by Ye, which proceeds through a proton exchange at C_α.²⁹ Reactions with p-fluoro styrene oxide (5a) were also successful, and 25 was isolated in 53% yield and 7:2 d.r. We also noticed significant accumulation of a white precipitate in this reaction. We isolated this material and assigned the structure as a trimer of the α-aryl aldehyde formed by SOI (26). The formation of such species highlights the challenge of intercepting α-aryl aldehydes more broadly, as non-catalytic reaction pathways can be a significant hurdle. We tested reactions with the epoxide derived from cinnamaldehyde (20a) and although we observed evidence of product formation by analytical Marfey’s derivatizations and NMR, the yield was too low for complete characterization. The yield and diastereoselectivity could plausibly be improved by directed evolution of UstD. For such studies, observation of a small amount of initial activity is often the largest uncertainty to developing a robust reaction. Together, these experiments demonstrate that the α-aryl aldehyde can be generated in situ by SOI under whole cell conditions and be intercepted with a C–C bond forming enzyme that does not natively react with this class of substrate, thereby rapidly building molecular complexity.

Figure 7.

Synthesis of γ-hydroxy amino acids through a cascade reaction with SOI and UstD^v2.0. Reactions were conducted with 1 mmol epoxide. General reaction conditions used 25 mM epoxide, 50 mM l-Asp (23), 100 μM PLP, 100 mM KPi pH 7.0, 100 mM NaCl, 5% EtOH, 2.5–5 μM UstD^v2.0, and SOI whole cells at 0.1% wcw. Product purity and d.r. was assessed via ¹H NMR and derivatization with Marfey’s reagent. The aldehyde trimer 26 was isolated as a side product from a reaction of 5a under the standard reaction conditions.

Conclusions

Here we demonstrated a chemoenzymatic approach to generating diverse α-aryl aldehydes, which are versatile synthons for the synthesis of structurally diverse molecules. We show that the SOI-ObiH cascade is highly promiscuous and can generate a large panel of functionally rich, β-hydroxy-α-amino acids, which are highly sought molecules due to their prevalence in bioactive compounds. ObiH can deliver these nsAAs in both good yield and exceptional diastereoselectivity, which has been a challenge with the related Thr aldolase enzymes and traditional synthetic methodologies. Prompted by the utility of these α-aryl aldehydes, we sought to rectify conflicting mechanistic information about the Meinwald rearrangement catalyzed by SOI. A series of labelling studies indicate that SOI catalyzes a stereospecific, concerted 1,2-hydride shift without the intermediacy of a benzylic carbocation. While α-aryl aldehydes have often been avoided in past studies due to their intrinsic instability, this work establishes robust chemoenzymatic methodology to access these intermediates and provides a foundation for their elaboration in C–C bond forming cascades.

ABBREVIATIONS

SOI

styrene oxide isomerase

SMO

styrene monooxygenase

PLP

pyridoxal phosphate

Thr

l-threonine

nsAA

non-standard amino acid

wcw

wet cell weight

dcw

dry cell weight

L-FDAA

Marfey’s Reagent