Stereoselective Photoenzymatic Hydroarylation for the Construction of Quaternary Stereocenters

Abstract

Quaternary carbon stereocenters are a crucial component of many bioactive molecules but can be challenging to prepare stereoselectively. Olefin hydroarylations are an attractive means for preparing this motif, however, existing methods struggle to set stereocenters on substrates lacking traditional catalyst binding handles. Here, we report a stereoselective photoenzymatic olefin hydroarylation using a repurposed Baeyer-Villiger Monooxygenase. Three rounds of iterative site-saturation mutagenesis yielded a photoenzyme capable of preparing valuable tetrahydroquinolines in high yield with excellent enantioselectivity. The engineered variant accepts various arene substituents, highlighting the synthetic utility of this methodology. DFT calculations and control experiments suggest that the protein templates a through-space interaction between the tertiary radical and aromatic group, which attenuates the oxidation potential of the radical enabling C–C bond formation.

Graphical Abstract

Introduction:

Quaternary carbon stereocenters are a crucial structural motif present in numerous natural products and pharmaceuticals.¹ While their incorporation offers beneficial control over drug conformations and metabolic properties, their catalytic asymmetric synthesis is traditionally difficult because of the kinetic challenges associated with forming a sterically congested center.^{2, 3} Enolate/enamine alkylations and conjugate additions have been extensively developed for setting centers adjacent to carbonyls using chiral Lewis acids,⁴ quaternary ammoniums,⁵ transition metals,^{6, 7} and secondary amine catalysts (Figure 1A).^8–11 However, strategies for setting quaternary centers in substrates lacking carbonyls are far rarer, with asymmetric Heck reactions being the most well developed (Figure 1B).^{2, 12, 13}

Figure 1:

Common strategies for stereoselective assembly of quaternary carbons and outline of the developed photoenzymatic hydroarylation. AcHYAM= AcCHMO-M10-R327K-R490E-C326H-I491T-A479H.

Reactions involving carbocations and radical intermediates are ideal for synthesizing quaternary centers because of these intermediate’s high reactivity, near planar geometry, and early transition states for C–C bond formation enable them to overcome the kinetic challenges associated with C–C bond formation.^14–17 These features make it difficult for common small molecule catalysts to control the stereochemical outcome of these reactions. Nevertheless, some catalysts can set quaternary centers in select reactions, such as polyene cyclizations (Figure 1B).^18–20 These catalysts mimic terpene cyclase enzymes by preorganizing the substrate for concerted C–C bond formation with functional groups on the catalyst stabilizing the cationic character of a particular transition state.^{21, 22} Building on this general strategy, we hypothesized that a similar approach could be used to control the selectivity of radical-mediated reactions. Recently, our group reported that a flavin-dependent cyclohexane monooxygenase (CHMO)²³ can catalyze an asymmetric hydroamination to afford pyrrolidine products containing an α-tertiary stereocenter via radical intermediates (Figure 1C).²⁴ To facilitate C–N bond formation, the enzyme templated an interaction between a tertiary radical and the amine to generate an intermediate that can be oxidized by the flavin cofactor. Building on this mechanism, we investigated whether preorganization could be used to achieve C–C bond formation. Generalizing this preorganized mechanism for redox-neutral radical cyclizations would enable CHMOs to set fully substituted carbon stereocenters on substrates that are incompatible with existing catalyst architectures.

Results and Discussion:

We began our investigations with aniline 1 bearing a pendant 1,1-disubstituted alkene and explored a range of Acinetobacter calcoaceticus CHMO-M10 variants developed during the hydroamination directed evolution campaign (Figure 1D). While the parent enzyme and optimal variants from the first two rounds of engineering were low yielding (Figure S5.2), a variant from the third round of engineering (AcCHMO-M10-R327K-R490E-I491E) afforded the desired product 2 in 20% yield with 82:18 er with the remaining mass balance being unreacted starting material and < 1% yield of alkene reduction and hydration. To optimize the yield and enantioselectivity of the reaction, we began engineering the enzyme active site using an iterative site saturation mutagenesis strategy with a reduced codon trick.^25–28 To streamline the evolution campaign and enable product analysis by mass spectrometry, we prepared a deuterated analogue of our model substrate (3). The loss of deuterium upon hydroarylation to form deuterated product 4 was enabled differentiation of 3 and 4 by RapidFire Mass Spectrometry (Figure 2).²⁹ The directed evolution campaign focused on sites in proximity to FAD, assuming these residues would have a significant impact on the enzymes activity and selectivity.³⁰ In a first round, we identified residue Y246F as a beneficial mutation to improve both yield (47%) and enantioselectivity (89:11 e.r.). A similar rationale governed the sites chosen in the second round, and a subsequent mutation of S186W almost doubled the observed yield of the reaction (90%) and showed some increase in enantioselectivity (91:9 e.r.). In the third round, sites were chosen based on their adjacency to previously selected mutants, assuming mutations at these sites could attenuate the enantioselectivity of the enzyme. A final mutation of T187V close to the prior tryptophan mutation gave the final enzyme AcCHMO-M10-R327K-R490E-I491E-Y246F-S186W-T187V (AcHYAR, HYdro ARylase), which catalyzed the hydroarylation with excellent yield (96%) and excellent enantioselectivity (95:5 e.r.). Control experiments confirm that light is necessary for the reaction (Supplemental Figure 5.2). Moreover, FAD and N-phenylphenathiazine photocatalysts alone cannot catalyze the cyclization, highlighting the importance of the protein scaffold for achieving C–C bond formation (Supplemental Figure 11.2).

Figure 2:

Optimization performed with deuterated substrate to allow for mass analysis. Evolution campaign performed in iterative site saturation cycles using a reduced codon trick. Active site of crystal structure for final mutant AcHYAR (1.75 Å resolution, PDB 9ORN). Loop 1 in blue, loop 2 in pink.

To understand the structural changes to the protein brought about by directed evolution, we obtained a crystal structure of the enzyme in the absence of substrate. The structure was solved at 1.75 Å and shows distinct changes. The introduced tryptophan mutation (S186W) is located close to the flavin cofactor (3.5 Å), indicative of CH-π interactions.³¹ Furthermore, a disulfide bond is formed between residues at C326 and C489, which links the two loops that sit on top of the active site. This is in contrast to the hydroaminase AcHYAM, which has cysteine 326 mutated to histidine (C326H), preventing disulfide formation. In AcHYAR, both loops contain mutations important for the starting enzyme AcCHMO-M10-R327K-R490E-I491E (loop 1: R327K; loop 2: R490E, I491E). To probe the role of the disulfide bond, point mutations to alanine were expressed at both sites. In C326A, the yield was reduced to 20% and enantioselectivity to 85:15. In C489A, the yield was reduced to 29% and enantioselectivity to 88:12 (Supplementary Figure 5.3).

With a more active and selective variant in hand, we turned our attention to exploring the scope and limitations of this enzyme (Figure 3). In general, the enantioselectivity of AcHYAR is sensitive to substitution on the substrate. The enzyme tolerates a range of electron-rich substituents on the aniline motif, providing high yields with good enantioselectivities (Figure 3, 5–8). Bulkier functional groups were well tolerated, such as the 6-NHBoc substitution (9, 35%, 59:41 e.r.), albeit with moderate enantiomeric excess. Other bulky substituents, such as naphthalenes, were well tolerated and formed the anticipated product 20 in good yield (64%) but with no enantioselectivity. The related naphthalene 10 product was formed with low enantiomeric excess (65:35 e.r.), albeit in a modest yield of 37%. Excitingly, pyridine anilines are compatible with this method (Figure 3, 11), a substrate class that is incompatible with concurrent hydroarylation methodologies. However, the reaction was affected by the relative position of the pyridyl nitrogen (c.f. Supplement; section limitations), with most of the isomers yielding the olefin reduction product. Meta-substituted arenes are tolerated by this enzyme with the regioselectivity of the transformation correlating with the B5 sterimol value for the substituent (Supplemental Figure 11.5). For instance, small substituents like methyl and methoxy groups afford product with low regioselectivity (2.3:1 (12:13) and 1:1 (14:15), respectively) favoring functionalization distal to the substituent. More sterically demanding isopropyl and dimethylamino substituents afford high levels of regioselectivity (4:1 (16:17) and >10:1 (18:19), respectively).

Figure 3:

Scope of the explored C–C forming reaction.

On the styrenyl side, a range of substituents were well tolerated ranging from 2-methyl (21, 52%, 69:31 e.r.), 3-methyl (22, 82% yield, 86:14 e.r.), and 4-methyl (23, 49%, 54:46 e.r.) to the corresponding 2-methoxy (24, 90%, 78:22 e.r.), 3-methoxy (25, 64%, 78:22 e.r.), and 4-methoxy (26, 62%, 59:41 e.r.). Electron-deficient substituents such as trifluoromethyl groups were not tolerated and putatively lead to the alkylation of the cofactor (c.f. Supplement; section limitations). Nevertheless, the 4-fluoro-substituted product 27 was formed in modest yield (13%) and with good enantioselectivity (85:15 e.r.). Trisubstituted alkenes are accepted by the enzyme (Figure 3, 28), providing means of introducing alternative alkyl substituents at the quaternary center. Bulky substituents such as 2-naphthyl 29 afford the product in 63% yield, albeit with no significant enantiomeric excess (52:48 e.r.). The corresponding N-methyl product 30 was formed in good yield (72%) and enantiomeric excess (72:28 e.r.). In addition, the model product 2 was modified by acetylation (31), allowing the determination of the absolute stereoconfiguration as (R). We tested whether this reactivity could be used to prepare a TSH inhibitor described by MSD.³² The product 32 was formed in good yield (35%) and low enantiomeric excess (55:45 e.r.). Nevertheless, this can serve as a starting point for an enzyme engineering campaign to generate this key moiety in three steps, compared to the precedent of six steps.

During the interrogation of the substrate scope, we recognized that this enzyme accepts internal alkenes. Recognizing that allylic amines are easier to prepare than their homo-allylic congener, we prepared both olefin isomers of the allylic amine and subjected them to the reaction conditions. The (E)-isomer (33) yielded the product in 75% yield with diminished enantioselectivity (91:9 e.r.). The (Z)-isomer (34) was less reactive (48%) but afforded slightly higher enantiomeric excess (93:7 e.r.). As the two olefin isomers do not afford opposite stereoisomers, these results suggest that tertiary radical reorientation is faster than C–C bond formation.

Despite obtaining the final mutant from a yield hit in our high-throughput analysis in plates, the expression of soluble enzyme was nearly 15 times lower (5.7 mg of AcHYAR/500 mL of expression volume) than AcHYAM (85 mg of AcHYAM/500 mL expression volume). We hypothesized that interactions of the final T187V mutation with proximal residues could have a detrimental effect on protein folding. To explore this hypothesis, we used DynaMut2 to better understand interactions between side chains.³³ This software predicted interactions between T187V and T378 on the adjacent loop. We hypothesized that point mutagenesis of T378 into the corresponding valine, alanine, tryptophan, or phenylalanine residue would help diminish any detrimental interactions between the polar T378 and the newly apolar isostere T187V. Mutating the residue to T378V improved the yield of soluble protein fourfold (22.6 mg/500 mL of expression volume) and this variant (AcHYAR-T378V) yielded product in 91% yield while retaining high enantioselectivity (94:6 e.r.). With a well-expressing variant in hand, we tested the total turnover number (TON) at 0.1 mol% of enzyme loading in lyophilized lysate. The product was still formed in 27% yield (TON=270) while retaining excellent stereoselectivity (94:6 e.r.).

We were next interested in determining the mechanism of this transformation. In our previous hydroamination work,²⁴ we found that the amine buffer could reduce the flavin cofactor under photoirradiation to afford the flavin hydroquinone (FAD_hq), consistent with earlier observations by Massey.³⁴ The resting state FAD_hq is then photoexcited, enabling single electron transfer to the bound substrate, reducing the styrenyl motif to the corresponding radical anion, followed by rapid protonation to generate a benzylic radical. In our previous study, we calculated that FAD_sq⁻ was insufficiently oxidizing (E⁰ = −250 mV vs. SCE) to transfer an electron from the tertiary benzylic radical (E^ox = 160 mV vs. SCE). Supported by DFT calculations and MD simulations, we proposed that an enzyme templated through-space interaction between the benzylic radical and aniline motif attenuates the oxidation potential of the radical, enabling concomitant oxidation and C–N bond formation. Based on this mechanistic understanding, we questioned whether a similar mechanism is responsible for C–C bond formation in the hydroarylation reaction. We envisioned two unique mechanisms as possible. The first was a stepwise mechanism with C–C bond formation occurring first; subsequent oxidation of the radical would furnish a Wheland intermediate and FAD_hq. Alternatively, a through-space interaction of the benzylic radical with the HOMO of the aniline, held in proximity by the enzyme’s active site, would allow for a quasi-concerted oxidation & bond-forming mechanism.

To probe these hypotheses, we employed DFT calculations to investigate the energetic landscape surrounding the C–C bond-forming event. Two favorable structures were found with the substrate preorganized for cyclization, one with a C–C distance between the benzylic radical and the ortho carbon (r_C–C) of 3.24 Å (Figure 4, A), indicating a weak through-space interaction between the radical and the arene and the other with the C–C bond formed at a distance of 1.57 Å with the radical character delocalized over the aromatic ring (Figure 4, C). Structure A had a calculated oxidation potential of E^ox = −117 mV vs. SCE, suggesting that the through-space interaction with the arene attenuates the potential of the tertiary benzylic radical, enabling a modestly endergonic oxidation by FMN_sq⁻ (E^ox = 0 mV for linear substrate conformation). The structure where the C–C bond is already formed (C) is approximately 6 kcal/mol higher in energy than the semi-closed structure A and is nearly a volt more reducing (E^ox = −1160 mV) than structure A.³⁵ Importantly, the energetic barrier for C–C bond formation via transition state B at 2.10 Å was found to be prohibitively high at 24 kcal/mol suggesting that the reaction does not occur via the stepwise cyclization/oxidation mechanism. In contrast, cyclization of the tertiary carbocation, formed after oxidation of the benzylic radical, has a 3 kcal/mol barrier. Based on these calculations, we propose two possible mechanisms: i) a step-wise oxidation/cyclization mechanism, or ii) a concerted mechanism where C–C bond formation is concomitant with oxidation by FAD_sq⁻ (Figure 4).³⁶ In either case, deprotonation of the resulting Wheland intermediate affords the product and returns the enzyme to its resting state.

Figure 4:

DFT calculations were performed at ωB97XV/6–311+G(d,p) level of theory in MeCN and redox values were calculated according to the Born-Haber relation. The mechanistic proposal outlines the relevant intermediates, based on experimental evidence and calculations.

In conclusion, we have developed a stereoselective hydroarylation reaction that uses the unique environment provided by an enzyme active site. By employing directed evolution, we identified a mutant that efficiently transforms a range of homoallylic anilines into the corresponding tetrahydroquinolines. Calculations corroborate a required through-space attenuation of the oxidation potential, enabling C–C bond formation to occur via a stepwise or concerted oxidation/cyclization mechanism. This mode of reactivity presents a unique pathway to form quaternary carbon stereocenters, avoiding classical carbocations by leveraging a reductive photoredox manifold. This is highlighted by the hydroarylation with electron-deficient heterocycles, which is difficult to carry out using transition metal and Brønsted acid catalysts. We envision this mode of activation as an increasingly general method for forging bonds under photoredox conditions that overcomes traditional challenges in forming congested stereocenters.

Supplementary Material

Experimental procedures, supplemental data tables, NMR, and HPLC data

Data and materials availability:

The data that support the findings in this study are available from the corresponding author upon reasonable request.