Biocatalytic Synthesis of α-Amino Esters via Nitrene C−H Insertion

Abstract

α-Amino esters are precursors to noncanonical amino acids used in developing small-molecule therapeutics, biologics, and tools in chemical biology. α-C−H amination of abundant and inexpensive carboxylic acid esters through nitrene transfer presents a direct approach to α-amino esters. Methods for nitrene-mediated amination of the protic α-C−H bonds in carboxylic acid esters, however, are underdeveloped. This gap arises because hydrogen atom abstraction (HAA) of protic C−H bonds by electrophilic metal-nitrenoids is slow: metal-nitrenoids preferentially react with polarity-matched, hydridic C−H bonds, even when weaker protic C−H bonds are present. This study describes the discovery and evolution of highly stable protoglobin nitrene transferases that catalyze the enantioselective intermolecular amination of the α-C−H bonds in carboxylic acid esters. We developed a high-throughput assay to evaluate the activity and enantioselectivity of mutant enzymes together with their sequences using the Every Variant Sequencing (evSeq) method. The assay enabled the identification of enantiodivergent enzymes that function at ambient conditions in Escherichia coli whole cells and whose activities can be enhanced by directed evolution for the amination of a range of substrates.

Graphical Abstract

Chiral α-amino esters are a valuable source of noncanonical amino acids (ncAAs) used to develop therapeutics, introduce new functions to biomolecules, and facilitate catalysis.^1–4 Accordingly, many chemo- and biocatalytic approaches have been devised that exploit diverse precursors for their synthesis.^5–14 α-C−H amination of abundant carboxylic acid esters offers a practical approach for synthesizing α-amino esters.¹⁵ Enantioselective α-amination of carbonyls has been achieved by leveraging the innate acidity of the α-C−H bonds: deprotonation facilitates the formation of nucleophilic intermediates that can be joined to electrophilic amines.^16–22 Similarly, carbonyl-containing compounds can be rendered electrophilic and coupled with nucleophilic amines.^23,24 None of these approaches afford primary α-amino esters, however, and instead lead to products that require additional manipulations to remove activating or protecting groups from the amine.

Leveraging recent developments in hydroxylamine-derived reagents,²⁵ we demonstrated that Fe-nitrenoids generated with engineered cytochrome P450s can be used for chiral amine synthesis, converting benzylic, allylic, propargylic, (with bond dissociation energy (BDE) <85 kcal/mol) and even unactivated C−H bonds (>95 kcal/mol) into primary amines.^26–29 These transformations proceed through a stepwise radical pathway: an Fe-nitrenoid performs hydrogen atom abstraction (HAA) on the hydrocarbon, forming a carbon-centered radical that rebounds on nitrogen, leading to primary amines. We wondered whether nitrene transferases could enable access to α-amino esters through α-C−H amination of carboxylic acid esters. While amination of hydridic C−H bonds is well established in nitrene transfer using chemo- and biocatalysts, the functionalization of protic C−H bonds, which have a lower BDE, is paradoxically underdeveloped (Scheme 1A).^30–32 This disparity arises because, like metal-oxos, metal-nitrenoids are subject to polar effects during HAA.^33–37 In the case of α-C−H bonds to electron-withdrawing groups (EWGs), electrophilic metal-nitrenoids are polarity mismatched, leading to slower rates of HAA (Scheme 1B). Consistent with this observation, most reported approaches for C−H amination α to EWGs by metal-nitrenoids rely on intramolecular strategies. Du Bois and Zhang, for example, reported Rh- and Co-catalyzed intramolecular amination of protic C−H bonds, respectively.^38–40 More recently, Meggers demonstrated intermolecular, directed enantioselective α-C−H amination of carboxylic acids that proceeds through a cyclic, intramolecular transition state.^41,42 Zhang’s Co-catalyzed amination of carboxylic acid esters with aryl azides is, to our knowledge, the only nondirected, intermolecular example for enantioselective amination of protic α-C−H bonds.⁴³ A chiral porphyrin ligand which facilitated attractive noncovalent interactions with the reaction partners was recognized as a critical design feature.

Scheme 1. Background and proposal.

Enzymes can exert precise control over substrate binding and positioning through multiple noncovalent interactions, determining selectivity outcomes and kinetics in biocatalytic reactions. We postulated that if we could identify a nitrene transferase capable of amination of protic C−H bonds, directed evolution (DE) could improve this novel function. Here, we describe the discovery and evolution of heme enzymes capable of catalyzing intermolecular α-C−H primary amination of carboxylic acid esters (Scheme 1C).

We began by assembling two heme protein collections shown useful for nitrene and carbene transfer reactions. The first comprised 96 engineered P450 nitrene transferases, including enzymes tailored for unactivated C−H amination.⁴⁴ The second featured 60 smaller heme proteins, globins and cytochromes c from extremophiles, previously evolved for carbene transfer.^45–48 Both collections were expressed in Escherichia coli and subjected to high-throughput screening with alkyl and benzylic carboxylic acid esters and various nitrene precursors in a whole-cell format. This screening process identified only one protein with low activity for the α-C−H amination of ethyl 2-(4-fluorophenyl)acetate (1) using O-pivaloylhydroxylamine triflic acid (2) (Figure 1A), underscoring the difficulty of functionalizing protic C−H bonds through nitrene transfer.

Figure 1.

(A) Model reaction. (B) Homology model snapshots of l-ApPgb-αEsA-G0. (C) Directed evolution lineage of l-ApPgb-αEsA. Sequence and activity for SSM libraries of sites (D) F73 and (E) F93 in l-ApPgb-αEsA-G8, with amino acids ordered by side chain properties (charged, polar, and nonpolar). Low/no activity variants had unmeasurable selectivity. (F) Directed evolution lineage of d-ApPgb-αEsA. (G) Mutations in ApPgb and directed evolution map. Abbreviations: epPCR = error-prone polymerase chain reaction; SSM = site-saturation mutagenesis; StEP = staggered extension process.

The identified biocatalyst is a thermostable protoglobin from Aeropyrum pernix (ApPgb) (Figure 1B).⁴⁹ This dimeric gas-binding protein, with 195 amino acids per monomer, contains three mutations (W59A, Y60G, and F145G) that replace aromatic residues with smaller, nonpolar ones, granting substrates access to the heme distal cavity.^50–52 The heme group of wild-type ApPgb is enclosed within a 3-on-3 α-helical sandwich arrangement and is coordinated by a histidine axial ligand (H120). Two nonpolar tunnels, one formed by α-helices G/B and the second by α-helices B/E, provide access to the distal cavity.⁵² Previous work by our group on engineering ApPgb carbene transferases suggested that substrates enter the distal cavity through the B/E helix tunnel.^46,47 The ApPgb variant afforded (S)-3a in 3% yield and 46% enantiomeric excess (ee). Renamed l-ApPgb-αEsA-G0 (α-Ester Aminase), this enzyme served as the initial variant for DE to enhance yield and selectivity for α-C−H amination of carboxylic acid esters.

The initial rounds of DE focused on identifying mutations that improved the yield of (S)-3a. After eight rounds of DE under aerobic conditions, we identified l-ApPgb-αEsA-G8 capable of catalyzing the reaction in 41% yield and 84% ee (Figure 1C and G). Mutations obtained during DE include C45A, which removes a surface cysteine residue that can form a disulfide bond between α-helices B/E with C102 in the native protein⁴⁹ and likely alters the entry tunnel into the active site. Additionally, mutations H136N and I137T are adjacently located in the G/F flexible loop near the proximal cavity, which we have observed to host enabling mutations in other engineering campaigns.⁵³ Five other beneficial mutations were identified: K36E, T97V, F156L, K159E, and F175L.

Although yield increased, enantioselectivity plateaued at 84% ee by the fourth generation. Thus, we shifted the evolution strategy to prioritize selectivity improvements using a high-throughput assay. The assay involved treating the reaction products with Marfey’s reagent, which converts the enantiomers of 3 into diastereomers that can be analyzed using achiral HPLC-MS.⁵⁴ This quantitative assay allowed simultaneous assessment of both activity and selectivity. We also sequenced every variant in the libraries using Every Variant Sequencing (evSeq) to gain biochemical insights into the effects of mutations.⁵⁵ We used site-saturation mutagenesis (SSM) to interrogate active site residues F73, known to affect enantioselectivity in cyclopropanation reactions,⁴⁵ and F93, which stabilizes a carbene intermediate,⁴⁹ and found dramatic changes in activity and enantioselectivity (Figure 1D and E). Two main observations can be gleaned from these data: 1) enzyme activity was maintained only with nonpolar amino acids,⁵⁶ and 2) residues 73 and 93 are involved in substrate recognition, as substantial changes in enantioselectivity were observed. Although site 93 did not offer variants with improved enantioselectivity towards (S)-3a, moving from larger phenylalanine (F) to smaller residues like serine (S), glycine (G), and alanine (A) inverted the enzyme’s enantioselectivity, likely altering how 1 approaches the nitrene. This is the first time we have observed enantiodivergence in enzymatic intermolecular nitrene transfer reactions. Conversely, site 73 offered two mutations that increased enantioselectivity (F73M and F73W). We advanced with l-ApPgb-αEsA-G9 (F73W), which catalyzed the reaction with a 90% ee. Further DE rounds focused on residues near W73 and F93, culminating in l-ApPgb-αEsA-G11, which includes G60S and R90G mutations in the B and E helixes, respectively. This variant delivered (S)-3a in 50% yield and 96% ee. Overall, l-ApPgb-αEsA-G11 was evolved over 11 DE rounds, amassing 12 missense and 8 synonymous mutations (Figure 1G).

Given the importance of d-ncAAs in drug discovery, we evaluated variant l-ApPgb-αEsA-G8 F93A (renamed d-ApPgb-αEsA-G0), which produced (R)-3b in 19% yield and 70% ee (Figure 1F) under standard, non-screening conditions. To improve this enzyme, we performed two rounds of DE, providing d-ApPgb-αEsA-G2 (with mutations W62G and L86G), which increased (R)-3b yield to 62% and 72% ee. Mutation W62G, appearing in helix B above the heme distal cavity, gave a substantial improvement in yield. This mutation is part of a sequence where neighboring residues 60 and 61 are also glycine, potentially transforming the B helix into a flexible loop.⁵¹ This likely contributes to the observed activity boost by enlarging the cavity around the enzyme’s active site. Starting from l-ApPgb-αEsA-G0, d-ApPgb-αEsA-G2 accumulated 12 missense and 8 synonymous mutations (Figure 1G). While further evolution using our developed strategy can improve the enantioselectivity of the reaction, it is best to focus on a desired target at this stage. We sought instead to understand the range of possible targets accessible to the engineered enzymes.

l-ApPgb-αEsA-G11 and d-ApPgb-αEsA-G2 showed a broad substrate scope across different carboxylic acid esters (Scheme 2A and 2B). α-Amino esters containing methyl, ethyl, and isopropyl esters (4-6) were accessible. For isopropyl ester 6, both enzymes favored the formation of l-α-amino ester. Ethyl phenylacetates with substitutions on the ortho, meta, and para positions were suitable substrates for this process. These include fluoro− (3), chloro− (7 and 18), bromo− (8 and 12), trifluoromethyl− (9), di-halogenated− (10 and 11), methyl− (13 and 17), methoxy− (14), 1,3-benzodioxolo− (15), and trifluoromethoxy− (16) substitution. Naphthalene and heterocycles such as thiophene and pyridine (19-21) were also accepted. α,α-Disubstituted α-amino ester 22, which can be used as a proteolysis-resistant ncAA in peptides,⁵⁷ was also obtained. Excitingly, both enzymes displayed trace activity in the amination of an alkyl carboxylic acid ester (23), establishing starting points for future DE efforts to generate enzymes for the synthesis of alkyl α-amino esters. We had found no enzymes active towards alkyl carboxylic acid esters when screening for a starting enzyme for this project and thus these evolved enzymes have entered a new, previously unknown reaction space. l-ApPgb-αEsA-G11 gave a 2:1 mixture of 23 and tertiary amine 24, whereas d-ApPgb-αEsA-G2 provided only the former. The observed promiscuous activity for β-amination indicates that it may be possible to evolve site-selectivity toward other constitutional ncAAs. Unreacted carboxylic acid ester constitutes the remaining mass balance of these enzymatic reactions.

Scheme 2. Substrate scope^a.

^aYields were determined by HPLC using a calibration curve with internal standard.

We performed the model reaction on 1-mmol scale with l-ApPgb-αEsA-G11 and isolated benzoyl-protected 3 in a 55% yield (2 steps) and 96% ee (Scheme 2C). Lastly, we demonstrated that the evolved enzymes could access an advanced intermediate (25a) for synthesizing clopidogrel, a World Health Organization’s essential medicine, and its enantiomer (25b) (Scheme 2D).^58,59 Overall, l-ApPgb-αEsA-G11 exhibited excellent enantioselectivity for halogen-substituted aryl carboxylic acid esters and good enantioselectivity for electron-rich substitutions. d-ApPgb-αEsA-G2 had modest enantioselectivities but higher yields, making it an excellent starting point for further evolution to create more selective and efficient enzymes for targeted d-α-amino esters.

DE campaigns create libraries of enzyme variants that exhibit distinct substrate specificities and selectivities.⁵⁶ To identify enzymes with superior selectivity for electron-rich substituted aryl carboxylic acid esters, we curated a collection of 70 enzymes variants identified during DE that present distinct active site residues and examined them for producing 14 using the developed assay (Figure 2A). Several enzymes showed better activity and selectivity than l-ApPgb-αEsA-G11 (B8). Notably, l-ApPgb-αEsA-G10–3 (B7) furnished 14 in a 45% yield and 86% ee (Figure 2B). We also identified l-ApPgb-αEsA-G9 for synthesizing 13a and 15a with improved yield and selectivity. This assembled library serves as a foundation for finding enzymes for the amination of desired targets, with DE offering a solution for further optimization.

Figure 2:

(A) Enzyme variant survey for amination of electron-rich aryl carboxylic acid esters; each matrix cell represents a unique variant. (B) Identified improved variants.

Herein, we developed protoglobin nitrene transferases that catalyze the enantioselective intermolecular α-C−H primary amination of carboxylic acid esters, enabling the synthesis of unprotected chiral α-amino esters. A high-throughput assay, critical in evolving a highly enantioselective enzyme for l-α-amino ester synthesis and discovering one favoring d-α-amino esters, assessed the activity and enantioselectivity of mutant enzymes. The engineered enzymes offer a valuable substrate scope, including access to an intermediate used for synthesizing an active pharmaceutical ingredient. Notably, they operate aerobically and at room temperature, unlike previous nitrene transferases that require anaerobic environments or sub-ambient temperatures to perform optimally. This work expands the substrates known to undergo enzymatic amination and showcases the extent to which an enzyme’s active site can be re-shaped through DE. We anticipate that these enzymes will facilitate the functionalization of other protic C−H bonds for synthesizing important chiral amines.