Stereospecific Enzymatic Conversion of Boronic Acids to Amines

Abstract

Boronic acids and esters are highly regarded for their safety, unique reactivity, and versatility in synthesizing a wide range of small molecules, bioconjugates, and materials. They are not exploited in biocatalytic synthesis, however, because enzymes that can make, break, or modify carbon-boron bonds are rare. We wish to combine the advantages of boronic acids and esters for molecular assembly with biocatalysis, which offers the potential for unsurpassed selectivity and efficiency. Here we introduce an engineered protoglobin nitrene transferase that catalyzes the new-to-nature amination of boronic acids using hydroxylamine. Initially targeting aryl boronic acids, we show that the engineered enzyme can produce a wide array of anilines with high yields and total turnover numbers (up to 99% yield and >4000 TTN), with water and boric acid as the only byproducts. We also demonstrate that the enzyme is effective with bench-stable boronic esters, which hydrolyze in situ to their corresponding boronic acids. Exploring the enzyme’s capacity for enantioselective catalysis, we found that a racemic alkyl boronic ester affords an enantioenriched alkyl amine, a transformation not achieved with chemocatalysts. The formation of an exclusively un-rearranged product during the amination of a boronic ester radical clock and the reaction’s stereospecificity support a two-electron process akin to a 1,2-metallate shift mechanism. The developed transformation enables new biocatalytic routes for synthesizing chiral amines.

Graphical Abstract

Introduction

Boronic acids and esters are prevalent functional groups in synthesis, prized for their unique reactivity, stability, and nontoxicity. These qualities sustain their prominence in chemical research, leading to a wealth of strategies for introducing boron at nearly any given carbon within a small molecule.^1–4 Likewise, there are numerous methods that can elaborate the carbon–boron (C–B) bond in organoborons, such as the venerable Suzuki-Miyaura and Chan-Evans-Lam coupling reactions. These reactions facilitate the conversion of C–B bonds into C–C, C–N, and C–O bonds and are routinely used in the pharmaceutical industry.^5–7 Beyond small molecules, boronic acids have been introduced into proteins through post-translational modification or genetic code expansion, allowing new biological functions and a handle for site-selective transformations.^8,9 Boronic acids have also been featured in sidechains of polymers, where their Lewis acidity and biocompatibility have been leveraged to create biomarkers, drug carriers, and chemical sensors.¹⁰

Biocatalysis is rapidly emerging as an enabling technology in chemical synthesis, driven largely by the desire to capitalize on the efficiency and selectivity of enzymes to streamline syntheses and reduce waste.¹¹ We sought to bring boron’s synthetic utility to biocatalysis by making enzymes that can process abiotic boronic acids and esters just as synthetic chemists have done. Certain enzymes exhibit promiscuous activity toward boronic acids and esters: some Baeyer-Villiger monooxygenases (BVMOs), for example, catalyze the conversion of aryl and alkyl boronic acids into their corresponding phenols and alcohols, respectively (Scheme 1A).^12–14 This transformation proceeds through the formation of a boronate complex with the catalytic flavin-C(4a)-hydroperoxide intermediate. Another notable example is the metabolism by cytochrome P450 3A4 of bortezomib, a proteasome inhibitor that bears a chiral α-amino boronic acid (Scheme 1B).^15,16 Bortezomib is deborylated by a presumed heme-bound oxidant to produce an epimeric mixture of alcohols. There are no reports, however, of enzymes that catalyze the amination of boronic acids and esters.

Scheme 1.

Examples of enzymes with promiscuous activity for processing boronic acids to alcohols and a proposed strategy to use nitrene transferases to form primary amines.

Already a decade ago, we and Fasan showed that engineered heme enzymes can generate Fe-nitrenes, the isoelectronic and isolobal analog of a cytochrome P450’s iron-oxo species or Compound I.^18,19 More recently, we have engineered new-to-nature ‘nitrene transferases’ that can engage a wide array of hydrocarbons in selective C–H functionalization and addition to alkenes using hydroxylamine derivatives as a nitrogen source.^20,21 We reasoned that nitrene transferases could possibly aminate boronic acids, thereby introducing a new substrate pool to this emergent class of biocatalysts.

We focused first on the enzymatic primary amination of aryl and alkyl boronic acids to produce anilines and aliphatic amines, respectively (Scheme 1C). Similar chemical transformations can be realized through metal-catalyzed cross-coupling and direct amination employing hydroxylamine derivatives.^22–24 For metal-catalyzed reactions, high catalyst loadings are required to overcome catalyst inhibition.^24–28 The aniline and amine products from these reactions can undergo further cross-coupling with other organoborons or halides present in the reactants. Meanwhile, direct amination with hydroxylamine derivatives provides a catalyst-free alternative, but it generates significant waste due to the need for electron-withdrawing groups on oxygen for activation and regioselective N–O bond cleavage.^{22,23,29–35} These reactions typically require strong bases and elevated temperatures to proceed. Recently, photochemistry and main-group catalysis have emerged as promising approaches to overcome some of the challenges associated with aryl boronic acids and esters.^36,37

We set out to engineer a heme enzyme capable of using simple hydroxylamine for the amination of aryl and alkyl boronic acids (Scheme 1C).³⁸ Such an enzyme would need no activating groups on the aminating reagent or stoichiometric bases and would generate water and boric acid as the only byproducts. It is known, however, that aryl boronic acids preferentially react with hydroxylamine to give phenols — an effective enzyme would have to kinetically override this inherent reactivity.³⁹ In addition, the protein scaffold of a good enzyme would protect the heme cofactor from inhibition by the amine product, thus enabling high turnover numbers and reactivity.

Results and Discussion

At the outset of this work, the mechanism underlying the proposed biocatalytic transformation was unclear due to the absence of obvious chemocatalytic analogs. The closest chemocatalytic equivalent we identified involves C–N bond formation from azides and aryl boronic acids catalyzed by a synthetic iron-heme system.⁴⁰ We hypothesized that the enzymatic amination of aryl and alkyl boronic acids would share a similar mechanism, and thus an enzyme that can convert aryl boronic acids into anilines might also be able to aminate alkyl boronic acids. We first evaluated a curated collection of diverse protoglobin enzymes expressed in Escherichia coli (E. coli) whole cells for the amination of 2-naphthylboronic acid (1) with hydroxylamine (NH₂OH) hydrochloride (Figure 1A). Compound 2 was chosen as the target product due to its strong chromogenic features, which allowed us to screen for reactivity using UV/Vis absorbance.⁴¹ Protoglobins are an emergent protein family for new-to-nature transformations with several promising features.^42–44 Protoglobins from extremophiles are highly thermostable, which is useful for process development as well as for directed evolution to enhance activity, which sometimes decreases protein stability.⁴⁵ They also show robust expression in E. coli.³⁸ Screening of this collection using high-performance liquid chromatography-mass spectrometry (HPLC-MS) revealed several protoglobins capable of promoting the reaction. Noteworthy was a variant of a protoglobin from the strictly aerobic hyperthermophile Aeropyrum pernix (ApPgb) that contained two mutations known to confer carbene transferase activity by opening the active site (W59A and Y60G) and a mutation that affects stereoselectivity in carbene transfer reactions (I149N).⁴⁴ Comprising 195 amino acids, ApPgb features a heme group coordinated by a histidine residue, nestled in a 3-on-3 α-helical sandwich arrangement (Figure 1B).⁴⁶ This ApPgb variant, henceforth referred to as ApPgb-HYA-5292 (G0) (Hydroxylamine-dependent Aminase), catalyzed the reaction in <1% yield and 9 ± 1 TTN. With this enzyme as a starting point, we turned to improving this novel function using directed evolution (DE).

Figure 1.

(A) Directed evolution of ApPgb-HYA-5292 (G0) to improve reaction yield and total turnover number (TTN) for the amination of 1 with hydroxylamine hydrochloride. Reaction conditions with resting whole cells (OD₆₀₀: 30): 1 (2.5 mM), NH₂OH–HCl (7.5 mM), M9-N buffer (pH 7.5), 5% v/v EtOH, D-glucose, aerobic, room temperature (RT), 16–24 h. Reaction conditions with clarified E. coli lysate: 1 (2.5 mM), NH₂OH–HCl (7.5 mM), Na₂S₂O₄ (2.5 mM), KPi buffer (pH 8.0, 50 mM), 5% v/v EtOH, aerobic, RT, 16–24 h. Reaction conditions with purified protein: ApPgb-HYA-5294 (G2) (0.02 mol%), 1 (2.5 mM), NH₂OH–HCl (7.5 mM), Na₂S₂O₄ (2.5 mM), KPi buffer (pH 8.0, 50 mM), 5% v/v EtOH, aerobic, RT, 16–24 h. Yields were determined by HPLC using a calibration curve with internal standard. (B) The mutated amino acids in ApPgb-HYA-5294 (G2) are shown as blue spheres on a homology model of wild-type ApPgb, which is based on the crystal structure of Methanosarcina acetivorans protoglobin (PDB: 3ZJN).¹⁷ The total turnover number is defined as the ratio of product concentration to enzyme concentration.

Two screening assays were used to search for improved variants: a chromogenic assay that measured the absorbance (A) of 2 with single-point measurements at 350 nm (A₂/A₁ = 3.4 at 350 nm) and a mass-based assay using HPLC-MS. DE was performed by generating libraries of ApPgb-HYA-5292 (G0) mutants using error-prone polymerase chain reaction (epPCR) mutagenesis and site-saturation mutagenesis (SSM) and screening in 96-well plates. From these libraries, top-performing variants based on yield were recombined by shuffling their sequences using staggered extension process (StEP) recombination.⁴⁷ Two rounds of DE under aerobic conditions produced ApPgb-HYA-5294 (G2), containing four new mutations (M39V, S123P, K125R, and P148L), which promoted the reaction in 26% yield and 150 ± 10 TTN (Figure 1A). At this stage we evaluated the reaction using clarified E. coli lysate and found that addition of sodium dithionite (Na₂S₂O₄) as a reductant greatly improved reaction efficiency, procuring 2 in quantitative assay yield (99%) and 580 ± 1 TTN. We also examined the reaction using purified protein under identical conditions and discovered that with just 0.5 μM of ApPgb-HYA-5294 (G2), complete consumption of 1 could be achieved, resulting in a 79% yield of 2 and 3900 ± 340 TTN. The remaining mass balance was found to be sp² amination of 2 at the 1-position, affording naphthalene-1,2-diamine (3a); no starting material was observed. Interestingly, while phenol 3b could be observed in reactions performed using whole cells, this product is largely absent when clarified lysate or purified protein were used. This is consistent with the susceptibility of boronic acids to oxidation by reactive oxygen species that arise from cellular respiration.⁴⁸ Finally, to showcase the scalability of the engineered enzyme, we explored the amination of 1 using clarified E. coli lysate at a 1-mmol scale and obtained 2 with a 94% isolated yield and 1030 TTN.

We next evaluated the evolved enzyme’s reactivity with other aryl boronic acids (Table 1). Due to the sensitivity of anilines to oxygen, we performed these reactions anaerobically, using clarified E. coli lysate (0.5 mol% of ApPgb-HYA-5294). Aryl boronic acids with substitution in the meta and para positions were suitable substrates for this enzymatic reaction. These include nitro– (4), trifluoromethyl– (5 and 15), cyano– (6), methyl ester– (7), bromo– (8), fluoro– (9 and 10), chloro– (11 and 17), methyl– (12), t-butyl– (13), methyl sulfide– (16), methoxy– (18 and 21), and phenyl– (20) substituted aryl boronic acids. Sterically encumbered naphthalen-1-amine (19) was also accessed. Notably, methyl ester 7, which would hydrolyze under basic conditions, and polyhalogenated 11, primed for further metal-catalyzed cross-coupling reactions, were tolerated in this enzymatic reaction. o-Methyl-substituted 22 failed to produce the desired amine in useful yields.

Table 1.

Substrate studies of ApPgb-HYA-5294 (G2)-catalyzed amination of boronic acids using hydroxylamine^a

^aReaction conditions with clarified E. coli lysate: ApPgb-HYA-5294 (G2) (0.5 mol%), 1 (2.5 mM), NH₂OH–HCl (7.5 mM), Na₂S₂O₄ (2.5 mM), KPi buffer (pH 8.0, 50 mM), 5% v/v EtOH, Coy Chamber (O₂, < 30 ppm), RT, 16–24 h. Yields were determined by HPLC using a calibration curve with internal standard and are the average of at least two experiments.

The use of boronic esters in lieu of boronic acids has greatly expanded the selection of organic fragments that can be used in cross-coupling methodologies.^5,49 These include boronic acid pinacol esters and potassium trifluoroborate salts, among others, which are known to be more stable than their corresponding boronic acids. We found that an aryl boronic acid pinacol ester 24 and potassium trifluoroborate salt 25 were suitable reactants for amination, giving comparable yields to the corresponding boronic acid 23. Boronic esters are known to hydrolyze into their respective boronic acids under aqueous conditions. Consequently, we hypothesize that prior to enzymatic amination, boronic esters undergo hydrolysis to their corresponding boronic acid, a notion supported by their detection through HPLC-MS.^50,51

After demonstrating the ability of aryl boronic acids and esters to undergo enzymatic amination, our focus shifted to explore whether alkyl organoborons could also be used in this reaction. Indeed, we found that primary (26) and secondary (27–29) alkyl boronic acid pinacol esters undergo enzymatic amination (Scheme 2). In the case of a racemic boronic acid pinacol ester, the enzyme demonstrated an impressive level of control, giving (S)-configured amine 28 in 87:13 enantiomeric ratio (er). To our knowledge, the synthesis of 28 is a unique instance of converting a racemic alkyl organoboron directly into an enantioenriched amine, bypassing the need to synthesize an enantioenriched alkyl boronic acid pinacol ester, which is typically necessary for producing chiral amines from organoborons.²² This enzyme offers a foundation for creating selective amination processes with alkyl organoborons, with directed evolution providing a pathway to improve both efficiency and selectivity.

Scheme 2. Enzymatic amination of a primary and secondary boronic acid pinacol ester^a.

^a Reaction conditions with purified protein: ApPgb-HYA-5294 (G2) (1 mol%), boronic acid pinacol ester (2.5 mM), NH₂OH–HCl (12.5 mM), Na₂S₂O₄ (12.5 mM), KPi buffer (pH 8.0, 50 mM), 5% v/v DMSO, anaerobic, RT, 16 h. ^b Changes to reaction conditions: ApPgb-HYA-5294 (G2) (2.5 mol%), NH₂OH–HCl (7.5 mM), Na₂S₂O₄ (2.5 mM), 5% v/v EtOH. Yields were determined by HPLC using a calibration curve with internal standard.

At this stage we considered the possible mechanisms for amination of boronic acids by ApPgb-HYA-5294 (G2). We started by examining the role of the protein scaffold and heme b. Control studies demonstrated that no combinations of heme b, reductant, or denatured ApPgb-HYA-5294 (G2), led to product formation. Purified ApPgb-HYA-5294 (G2), existing in its Fe (III) state as observed using UV-Vis spectroscopy (Figure 2B), is also inactive. Collectively, these studies indicate an Fe (II) heme in the engineered protoglobin scaffold is necessary for amination of boronic acids with hydroxylamine. These control studies also underscore the critical role of the protein scaffold in substrate recognition and facilitating the bond-forming step.

Figure 2.

(A) Proposed one- and two-electron mechanisms for the amination of boronic acids. (B) UV-Vis spectra of ApPgb-HYA-5294 (G2) in its ferric and ferrous (with addition of Na₂S₂O₄) state. The ferric state features a Soret peak at 411 nm and a Q-band at 541 nm, whereas the ferrous state features a red-shifted Soret peak at 432 nm and a Q-band at 562 nm. (C) Reaction conditions with purified protein: ApPgb-HYA-5294 (G2) (62.5 μM), boronic acid pinacol ester (2.5 mM), NH₂OH–HCl (7.5 mM), Na₂S₂O₄ (2.5 mM), KPi buffer (pH 8.0, 50 mM), 5% v/v EtOH, anaerobic, RT, 16–24 h. Compound 31 was detected using HPLC-MS.⁴¹ Yields were determined by HPLC-MS using a calibration curve with internal standard.

We then directed attention toward understanding the generation of the product. Although metal-catalyzed amidations of organoborons have been demonstrated previously using hydroxylamine derivatives, these are understood to operate under at least one inner sphere cross-coupling step (oxidative addition, transmetallation, or reductive elimination).^52–55 In the case of heme enzymes, the sole vacant axial coordination site is dedicated to nitrene formation. Accordingly, the bond-forming event must involve outer sphere interaction of the nitrene with boronic acid. Within this framework, we can propose two mechanisms. The first involves nucleophilic attack of the nitrogen lone pair in the Fe-nitrene to the empty p orbital in the boronic acid, generating a boronate complex. This intermediate can undergo the 1,2-metallate shift mechanism typically observed in the reaction of boronic acids with peroxides and hydroxylamine derivatives (Figure 2A, blue box), forging the C–N bond.⁵⁶ Secondly, we imagined that the Fe-nitrene, which we have observed behaving as an open-shell intermediate in C–H functionalization and addition to alkenes,^20,38 could engage boronic acids through a bimolecular homolytic substitution (S_H2) mechanism, generating a carbon radical that can rebound on nitrogen to form the C–N bond (Figure 2A, red box).^57–59 To distinguish between these possibilities, we investigated the amination of diagnostic substrate 30 (Figure 2C), which was expected to undergo ring-opening if the latter mechanism is operative.⁶⁰ The cyclopropyl ring in 30 remained unscathed, however, supporting a 1,2-metallate shift mechanism. Nonetheless, the possibility of a fast rebound, as informed by the rearrangement rate constant of the primary radical of 30 (>1.3 × 10⁸ s⁻¹),⁶¹ cannot be discounted.

A hallmark of 1,2-metallate shift mechanisms is that they are stereoretentive, maintaining the chirality of the carbon migrating group during the boron-to-nitrogen rearrangement process.^33,62 If the enzymatic reaction indeed occurs via this mechanism, ApPgb-HYA-5294 (G2) provides an ideal case for testing the stereospecificity of the amination process, given that it produced amine product 28 as a scalemic mixture (87:13 er). If this mechanism holds true, the scalemic mixture of amine 28 should originate independently from different enantiomers of racemic 32. To test this hypothesis, we subjected enantiopure R- and S-configured 32 to the enzymatic reaction in parallel and observed the nearly exclusive formation of R- and S-configured 28, respectively (Figure 2C). S-configured 28 was also obtained in a higher yield, consistent with the result obtained with racemic 32.

The experimental results and literature precedents support the mechanism depicted in Scheme 3. Single-electron reduction of ApPgb-HYA-5294 (G2) by Na₂S₂O₄ generates the active Fe (II) heme enzyme. Hydroxylamine, generated in situ from deprotonation of hydroxylamine hydrochloride (pKa = 5.97)⁶³ in KPi buffer medium (pH = 8.0), diffuses into the hydrophobic cavity of ApPgb where the heme rests. We speculate that the protein scaffold oversees the regioselective coordination of hydroxylamine to heme and N–O bond cleavage for the formation of Fe-nitrene and water. The boronic acid substrate interacts with the nitrene in the heme cavity, forming product through boronate complex formation and 1,2-metallate shift. Eventual hydrolysis of the product delivers the desired amine and boric acid.

Scheme 3.

Proposed mechanism for enzymatic amination of boronic acids with hydroxylamine.

In this work, we have engineered enzymes that can cleave the C–B bonds in boronic acids and esters and convert them into amines — one of the most important functional groups in chemistry. The diverse methods available for producing boronic acids and esters, combined with these engineered enzymes, have the potential to simplify and streamline the synthesis of amines. We have elucidated important features of the mechanism of this new-to-nature reaction, laying the ground-work for designing additional pathways to synthesize high-value amine products. Notably, the engineered enzyme was shown to generate enantioenriched amines from a racemic mixture of an alkyl boronic ester, a capability not reported with any known chemocatalysts. Lastly, the revelation that the enzyme likely processes boronic acids via a two-electron mechanism akin to a 1,2-metallate shift, establishes a new paradigm in nitrene transferases, offering guidance for the discovery of new transformations. This proposed mechanism, initiated by nucleophilic addition by Fe-nitrene, represents an underappreciated modality in nitrene transferases and may have broader implications in other enzymatic reactions.

Conclusion

To summarize, we have introduced a new-to-nature protoglobin nitrene transferase that processes aryl and alkyl boronic acids and esters into amines through a proposed stereospecific 1,2-metallate shift mechanism. The reaction operates under ambient conditions, low enzyme loading, and uses simple hydroxylamine as a nitrogen source, giving water and boric acid as the only byproducts. This enzyme avoids producing phenols from aryl boronic acids, the preferred outcome in reactions with hydroxylamine, and offers a chemoselective approach without activating groups. We anticipate that this enzyme will find use in kinetic resolutions, desymmetrization reactions, and other processes aimed at synthesizing chiral amines from boronic acids and esters.