From Ground-State to Excited-State Activation Modes: Flavin-Dependent “Ene”-Reductases Catalyzed Non-Natural Radical Reactions

Graphical abstract

1. Introduction

Enzymes are powerful tools for chemical synthesis because they provide unparalleled levels of efficiency and selectivity.^5,6 Over the past 30 years, the biocatalytic toolbox has rapidly expanded thanks to the revolutionary advances in enzyme discovery, design, and engineering, transforming the field from a curiosity to an essential tool for industrial chemical synthesis.^7–10 As therapeutics move from relatively flat molecules to ones with a higher degree of three-dimensionality, there is a need for catalyst architectures capable of precisely controlling the stereochemical outcome of reactions. While enzymes are ideally suited for stereoselective reactions, biocatalysts currently catalyze fewer types of reactions when compared to small molecule catalysts. Consequently, there is interest in identifying enzymes to catalyze new types of reactions. While this often involves mining nature for unique reactivity, an alternative approach is to extract new functions from known enzymes.^11–13

In 2015, our group was interested in developing strategies for generating radical intermediates within protein active sites. As high-energy intermediates, radicals can facilitate a litany of valuable transformations.¹⁴ While the reactivity of these intermediates is well studied,¹⁵ controlling the selectivity of radical reactions remains challenging for small-molecule catalysts.^16,17 By generating radicals within protein active sites, we can use directed evolution to create enzymes that provide unparalleled levels of chemo, regio, and enantioselectivity. However, at the time, there were no general strategies for forming non-native free radicals within protein active sites. Nature has various enzymes, such as radical S-adenosylmethionine (SAM) enzymes and cytochrome P450 monooxygenases (P450s), to precisely control free radicals in the biosynthesis of natural products.^18–22 We sought to develop radical initiation strategies that mirror the mechanisms commonly used by synthetic organic chemists.²³ As electron transfer reactions are central to radical formation in electrosynthesis and photoredox catalysis,^24–27 we questioned whether these mechanisms could be used with proteins.

2. Enzyme Selection and Radical Initiation Mechanisms

When beginning our study, we sought proteins with cofactors capable of facilitating electron transfer events. We were immediately attracted to flavin-dependent enzymes because of the mechanistic versatility of this cofactor, and targeted enzyme scaffolds that were reported to be substrate promiscuous, thermally stable, and tolerant to organic solvents, as these features would promote evolvability and catalytic promiscuity.^28,29 Flavin-dependent ‘ene’-reductases (EREDs) satisfied these requirements. While this enzyme family plays various roles in nature, it is most widely used for the asymmetric reduction of electronically activated alkenes.^30,31 Reduction occurs via hydride transfer from flavin mononucleotide hydroquinone (FMN_hq) to the electrophilic β-position of the alkene, followed by protonation of the resulting enolate by a conserved tyrosine to afford the hydrogenated product and flavin mononucleotide quinone (FMN_ox).

Beyond the hydride transfer mechanism available to EREDs, flavoenzymes, such as ferredoxin reductases and P450 reductases, mediate single-electron transfer (SET) reactions.³² We recognized that if EREDs could use FMN_hq as a single electron reductant, they could catalyze reductive radical transformations. Indeed, Miura et al. reported the single-electron reduction of menadione to the corresponding radical anion using old yellow enzyme 1 (OYE1).³³ However, as FMN_hq has only a modest reduction potential (E_1/2 = −215 mV vs. saturated calomel electrode (SCE)), this mechanism would only be effective for highly oxidizing molecules.^1,34 For less activated substrates, more reducing conditions are required. Flavin-dependent DNA photolyase absorbs blue light to access the more reducing singlet excited state of FMN_hq.³⁵ This mechanism mirrors photoredox catalysis, where a photocatalyst is excited with visible light to access an excited state with enhanced redox properties.^26,36 If this photoexcitation mechanism were available to EREDs, it would expand the types of substrates which could serve as radical precursors

Ground-State Electron Transfer

We began by developing an ERED to catalyze an asymmetric hydrodehalogenation reaction. The simplicity of the reaction allowed us to study the fundamental features of the radical initiation and termination events. Moreover, at the time, few small molecule reagents or catalysts could stereoselectively deliver a hydrogen atom to prochiral radicals. As a model reaction, we targeted the hydrodebromination of acyclic α-bromoesters 1 (E_p/2 = −750 mV). While the initial electron transfer from ground-state FMN_hq is endergonic by approximately 500 mV, this step is coupled to irreversible mesolytic cleavage of the C–Br bond, driving the reaction forward (Figure 1).¹ We recognized that binding to the protein active site could activate the substrate for reduction; however, at the time, we did not have a clear mechanistic hypothesis on how this would occur. After testing a small panel of EREDs and Baeyer-Villager monooxygenases (BVMOs), we found that many enzymes could catalyze the dehalogenation, albeit with variable yields and enantioselectivities. The most selective catalyst was a homolog from Gluconobacter oxydans (GluER), which afforded the (S)-enantiomer of product 2 in 57% yield and 88:12 er. We conducted site-saturation mutagenesis on the conserved tyrosine (Y177), which is understood to be the proton source in alkene reduction. As the O−H bond of tyrosine is weak (BDE ~ 88 kcal/mol), it could also serve as a hydrogen atom donor. Mutation of this residue to phenylalanine (GluER-Y177F) produced an improved catalyst, which furnished the dehalogenated product 2 in 89% yield and 97:3 er.

Figure 1.

ERED-catalyzed asymmetric hydrodehalogenation.¹

Isotope incorporation experiments were conducted to determine the mechanism of radical termination. We considered two possibilities: i) reduction of the α-acyl radical and stereoselective protonation, or ii) asymmetric hydrogen atom transfer from the neutral flavin semiquinone (FMN_sq) (BDE_N–H = 59 kcal/mol).^37,38 Isotopically labeling of the hydridic N5–H position of flavin can distinguish these two mechanistic outcomes. In situ labeling was achieved using a cofactor turnover system composed of NADP⁺ and an engineered glucose dehydrogenase (GDH-105) with d₁-glucose as the deuterium source. When used under the reaction conditions, we observe dehalogenation in 71% yield with 81% deuterium incorporation at the α-position,³⁹ supporting FMN_sq functions as a hydrogen atom source to terminate the α-acyl radical (Figure 1).

Regarding radical initiation, substrates with more positive reduction potentials, such as ketones, were reactive. In comparison, ones with lower reduction potentials, such as amides or non-phenyl acetic ester derivatives, were not consumed, suggesting that this electron transfer mechanism is only effective for substrates within a specific redox window.

Having demonstrated the viability of EREDs as chiral catalysts for radical reactivity and possessing a basic understanding of radical initiation and termination mechanisms, we shifted our attention to asymmetric radical cyclization. Chiral Lewis acids are commonly used for this type of transformation; however, they are typically used in stoichiometric quantities because their complexation to substrates does not significantly lower the activation barrier for radical formation. Moreover, Lewis acid catalysts usually require bidentate substrates to achieve high levels of enantioselectivity.^16,17 We envisioned that the multiple non-covalent interactions involved in substrate binding and the high affinity of protein for the flavin cofactor would enable the use of simple substrates and help localize radical formation to the protein active site.

As a model reaction, we explored the cyclization of α-bromoketone 5, a substrate lacks extraneous functional groups for cyclization (Figure 2).³⁴ Several EREDs produced cyclic ketone 6, with the most promising being the wild-type nicotinamide-dependent cyclohexanone reductase (NCR) from Zymomonas mobiles, which provided 6 in 20% yield (79:21 er) with 19% yield of acyclic hydrodehalogenated product. Amino acid residues that line the active site of NCR were targeted in the iterative saturation mutagenesis (ISM) campaign to improve efficiency and selectivity.⁴⁰ After four rounds of protein engineering, a quadruple mutant (Y343W/F269W/W342A/I231S, named NCR-C9) delivered the product in 92% yield and 95:5 er. This variant was ideally suited for 5-exo-trig cyclizations; however, members of the evolutionary series could catalyze 6-endo-trig and 7-exo-trig cyclizations. The 5-exo-trig cyclization preference of NCR-C9 was mainly attributed to the mutations that facilitate the preorganization of substrates favoring the transition states for such cyclization. A striking feature of this work is the enhanced conversion of starting material to product despite the lack of significant changes in the oxidation potential of flavin cofactor, suggesting that the enzyme activates the substrate for reduction.

Figure 2.

Engineering EREDs for radical cyclization.³⁴

Photoexcitation for Challenging Electron Transfers

A limitation of the previous method was its limited substrate scope, only being effective for α-bromoketones. To overcome this issue, we proposed using the excited state of FMN_hq as a single-electron reductant (E_1/2* = −2.26 V vs. SCE).⁴¹ We targeted the radical cyclization of α-chloroamide 10 (E_p/2 = −1.65 V vs SCE) as a model reaction to afford β-stereogenic lactam 11 (Figure 3).² In the absence of light, none of the EREDs tested could react with the starting material. When irradiated with violet light, which best aligned with the absorption of FMN_hq, multiple enzymes produced the cyclized product with GluER-T36A affording the product with the highest yield. Optimization of various reaction parameters revealed cyan LEDs (λ_emission = 500 nm) to be better suited for this reaction, providing product 11 in 92% yield and 94:6 er.

Figure 3.

Photoenzymatic lactam cyclization.²

The mechanistic rationale for improving cyan light performance was not immediately apparent, so we conducted UV-vis spectroscopy studies to identify the species responsible for absorbing light in this region. Mixing reduced GluER-T36A with substrate 10 led to a broad absorption feature (λ_max = 495 nm), which disappeared upon adding a competitive binder sodium benzoate (NaOBz). This feature is suggestive of a charge-transfer (CT) complex between 10 and FMN_hq, accounting for the wavelength preference of the reaction. No CT complex was observed when the analog lacking the alkene was used for the same UV-vis spectroscopy study, indicating its importance to the complex. Importantly, this complex only forms in the presence of the enzyme, indicating the protein is essential for templating this complex and thus gating radical formation to the active site. This observation, coupled with additional mechanistic studies, suggests that radical formation occurs via photoexcitation of the enzyme-templated CT complex to generate an α-acyl radical 12, which undergoes a 5-exo-trig cyclization to afford radical 13, which is terminated via hydrogen atom transfer from FMN_sq to furnish product (Figure 3).²

EREDs are remarkably versatile catalysts for this type of cyclization. While traditional methods produce a mixture of the cyclized and hydrodehalogenated product, GluER-T36A catalyzes 5-exo-trig cyclization in high yield and good to excellent levels of stereoselectivity and negligible formation of the undesired hydrodehalogenation. When using trisubstituted alkenes, the products were formed with high enantio- and diastereoselectivity, highlighting that enzymes can precisely control the stereoselectivity of the C–C bond-formation and HAT steps. 5-endo-trig, 6-, 7-, and 8-exo-trig cyclizations are also possible using other homologs (Figure 3). Importantly, these non-traditional cyclizations are challenging to achieve using traditional synthetic methods, suggesting that the enzyme preorganizes the substrate for cyclization, which is determinantal for templating a CT complex for radical initiation and facilitating a productive transition state.

The model cyclization uses light inefficiently, with wild-type GluER exhibiting low quantum efficiency (Φ = 0.024). By comparison, natural photoenzymes such as DNA photolyase display a quantum yield of Φ = 0.85.⁴² To improve the photon efficiency, we subjected GluER-T36A to directed evolution (Figure 4).⁴³ After developing a high-throughput platform for photoenzyme engineering, we conducted a single round of error-prone mutagenesis and identified a triple variant (T36A/K317M/Y343F, namely GluER-G6) which afforded 3.5-fold improved yield and 4-fold increase in quantum yield (Φ = 0.105) accounting for the yield improvement (Figure 4).

Figure 4.

Engineering EREDs for improved photon efficiency.⁴³

Transient absorption spectroscopy supports a change in reaction mechanism over the engineering campaign. While wild-type GluER displays a radical lifetime of 700 ps, it is decreased to <20 ps with GluER-G6. We hypothesize that with GluER-G6, radical initiation, cyclization, and termination occur in a concerted asynchronous transition state, whereas GluER has a more step-wise mechanism. We propose that the enzyme templates a hyperconjugative interaction between the π-system of the alkene and σ*_C–Cl (Figure 4).⁴³ This interaction is essential for CT complex formation, accounting for the lack of an observable complex when the substrate lacks an alkene. Moreover, it preorganized the substrate for cyclization, accounting for the lack of hydrodehalogenated product observed in these reactions.

Having established a biocatalytic radical cyclization, we next focused on different substrates and mechanisms of radical termination. Given the importance of the alkene in CT complex formation, we questioned whether other π-systems could serve as SOMOphilic coupling partners. We explored the use of oximes because of the ability to form enantioenriched α-tertiary amines, an increasingly important motif in pharmaceutically important molecules. When using O-benzyl oximes 17, we found that GluER-T36A provided product 18 in high yield but low enantioselectivity, whereas wild-type NCR afforded 18 with good enantioselectivity but low yield (Figure 5).⁴⁴ We subjected both enzymes to iterative rounds of site saturation mutagenesis, ultimately identifying a quadruple GluER mutant (T36A/K317M/Y343F/F269V) and a double NCR mutant (D294W/Y343W), which afford the product in high levels of yield and enantioselectivity for this and other cyclization modes.

Figure 5.

Engineering EREDs for synthesis of α-tertiary hydroxylamines.⁴⁴

Toward changing the mechanism of radical termination, we questioned whether allyl silane 20 could be used to achieve an anullative allylation. We found that GluER-G6 could catalyze the desired reaction in 92% yield and 99:1 er, with <5% yield of the reductive cyclization product.⁴⁵ While our initial hypothesis was that radical termination occurred via radical/polar crossover to form a β-silyl cation followed by elimination, transient absorption spectroscopy revealed two new species that decay with a lifetime of 38 ns and 150 ns before generating flavin quinone. We interpret these results to indicate that radical termination occurs via β-scission of trimethylsilyl radical (or silicon-ate) and the resulting silyl radical abstracting a hydrogen atom from FMN_sq (Figure 6).

Figure 6.

Photoenzymatic allylic alkylation via β-scission.⁴⁵

We also questioned whether the protein scaffold could template CT complexes with unactivated alkyl iodides, substrates that typically do not form EDA complexes in solution. When alkyl iodide 25 was subjected to a panel of EREDs, we found that GluER-Y177F could catalyze the intramolecular Giese reaction in good yield and enantioselectivity by avoiding undesired native reduction of starting material, while the (R)-enantiomer could be obtained using OYE2-Y197F.⁴⁶ Spectroscopic studies supported the formation of a weak CT complex between 25 and FMN_hq. GluER-Y177F tolerates various ester substituents of the acrylates to provide products with good yields and enantioselectivities. Notably, N-hydroxyphthalimide (NHPI) esters 30 can also be used for alkyl radical formation, expanding the types of radical precursors available in photoenzymatic systems (Figure 7).

Figure 7.

Cyclization via unactivated alkyl radicals.⁴⁶

Intermolecular Radical Reactions

A particular challenge in radical biocatalysis is intermolecular reactions. As radical intermediates are short-lived, both coupling partners must be bound within the protein active site before radical formation to avoid hydrodehalogenation of the alkyl halide. Consequently, we were interested in understanding how the protein would control the electron transfer events when looking for intermolecular reactivity. Our initial studies focused on coupling N,N-dimethylchloroacetamide 31 with α-methylstyrene 32 to afford γ-stereogenic product 33 (Figure 8).³ We found that GluER-T36A affords the (R)-enantiomeric 33 in 99% yield and 99:1 er, while the ERED from Nostoc punctiforme (NostocER) provided the (S)-enantiomer of 33 in 96% yield and 10:90 er. Mechanistic studies revealed that tyrosine 219 competed with FMN_sq as a hydrogen atom source for radical termination. Mutating Y219 to phenylalanine (Y219F) resulted in product formation in 99% yield and 95:5 er. In all cases, less than 5% yield of the hydrodehalogenated product was formed, indicating a control mechanism for radical termination. Spectroscopic studies revealed that a weakly absorbing CT complex was formed when α-chloroacetamide 31 was added to the reduced enzyme. However, upon adding alkene 32, the absorbance of the feature increased, indicating the generation of a higher-order enzyme-templated ternary CT complex between 31, 32, and FMN_hq. As both the alkyl halide and alkene are essential components of this highly absorbing complex, radical formation can only occur when both are present within the active site, thus disfavoring hydrodehalogenation of the alkyl halide. As in intramolecular chemistry, we hypothesize that a bimolecular complex involving a hyperconjugative interaction between π-bond of 32 and σ*_C–Cl of 31 serves as the electron acceptor for these CT complexes (Figure 8). These enzymes accept various tertiary and secondary α-chloroacetamides. Styrenyl alkenes and vinyl pyridines are the most reactive and selective coupling partners, however, aliphatic alkenes, such as allylic alcohols, allylic amines, and vinyl ethers, are also reactive (Figure 8).³

Figure 8.

Photoenzymatic intermolecular radical hydroalkylation.³

While exploring the scope of alkyl halides, we found α-chloroacetophenone 38a to be reactive, however, it behaved differently than α-chloroamides 31. While α-chloroamides were completely unreactive in the dark, α-chloroacetophenone afforded product in 3% yield with 97:3 er using GluER-T36A.³⁴ A test of other homologs revealed NCR to be more reactive, affording product 39 in 49% yield and 99:1 er in the dark. As flavin redox properties are similar for these enzymes, we hypothesize that the protein scaffold activates the substrate for reduction by modifying its binding orientation within the active site. A more electronically activated α-bromoacetophenone 38b was used, which afforded a product with 99:1 selectivity for the (R)-enantiomer and quantitative yield with <15% formation of the hydrodehalogenated product (Figure 9). Initial rate analysis was conducted to determine how the presence of the alkene influences the rate of alkyl halide consumption. We found that the consumption of 38b was 2.5-fold faster in the presence of alkene, suggesting that the alkene makes α-bromoacetophenone easier to reduce. Like the photochemistry described above, we propose that a complex where the π-bond of 32 hyperconjugatively engages with the σ*_C–Br of 38b is a better electron acceptor than the alkyl halide alone (Figure 9). In enzymes that provide a substantial interaction between the alkene and alkyl halide, the reduction potential is increased to the point where ground-state electron transfer from FMN_hq is feasible. Photoexcitation is required for enzymes that provide a weaker magnitude of interaction or for substrates with lower reduction potentials.³ This hypothesis accounts for the difference in reactivity between our group and concurrent work by Zhao using similar substrates but different enzymes.⁴⁷

Figure 9.

Ground-state intermolecular radical hydroalkylation.³⁴

Beyond α-halocarbonyl compounds, EREDs can also use chloromethylpyridines as radical precursors, enabling the synthesis of heterocycle motifs found in many pharmaceuticals and agrochemicals. We tested a series of GluER homologs and discovered that the ERED from G. morbifer (MonstER) could couple 4-(chloromethyl)pyridine 43 with α-methylstyrene in 71% yield with 94:6 er using violet light (Figure 10).⁴⁸ In these reactions, the remaining mass balance is the hydrodehalogenated starting material. Presumably, the π-system of the chloromethylpyridine enables CT-complex formation without the alkene. To understand the differences between the two possible complexes, we tested various wavelengths of light. We observe a 3:1 ratio of the coupled product and hydrodehalogenation using violet light. In contrast, using blue LEDs, hydrodehalogenation is formed in 27% yield with only 17% of the coupled product. This result suggests that the two types of complexes absorb in different regions of the UV-vis spectrum, providing a new mechanism for controlling reaction outcomes (Figure 10).

Figure 10.

Alkene hydroalkylation by pyridylmethyl radicals.⁴⁸

Beyond alkenes, EREDs can also catalyze regioselective arene alkylation in a redox-neutral reaction.⁴⁹ Radical termination occurs via substrate oxidation rather than hydrogen atom transfer in these reactions. We were interested in this family of reactivity because of the opportunity to alkylate positions that are challenging to functionalize using traditional methods. Indole was an attractive target because it has six possible alkylation sites, with most methods favoring C2 alkylation. When looking at the alkylation of indole with N,N-dimethylchloroacetamide 31, we found that an ERED from Aspergillus nidulans (AspER) exclusively alkylated the C2-position in 48% yield, while GluER-T36A provided mixed regioisomers [C4/(C2 + C3) = 1:2] (Figure 11).⁵⁰ As methods for C4-alkylation of indoles are elusive,⁵¹ we sought to generate a catalyst to provide this product selectively. After five rounds of SSM, a highly evolved mutant (GluER-T36A/Y343F/T25L/T231V/G270M/Q232W/T268D, named PagER) was identified to afford the C4-regioisomer in high-yield and regioselectivity (90% yield, C4/others > 9:1). Mechanistic studies showed that radical initiation in AspER occurs from ground-state FMN_sq, accounting for the C2-selectivity. Note that the indoyl radical was terminated by an oxidation/deprotonation sequence. We previously found that ground-state anionic FMN_sq can initiate intramolecular arylation reactions.⁴⁹ Conversely, PagER relies on a CT-state mechanism in which presumably the α-chloroamide is positioned over the C4-position of indole. Interestingly, PagER displays dark reactivity while other members of the evolutionary series do not, suggesting that protein engineering is strengthening the proposed hyperconjugative interaction between alkyl halide and indole, forming a more oxidizing electron acceptor. This and other engineered enzymes are effective for alkylating indoles, anilines, and quinonlines (Figure 11).⁵⁰

Figure 11.

Engineering EREDs for regioselective alkylation of arenes.⁵⁰

Next, we explored the possibility of using EREDs to prepare α-tertiary amines. We imagined that the redox-neutral coupling of nitronates with alkyl halides would be attractive for preparing tertiary nitroalkanes, precursors to α-tertiary amines.^52,53 Numerous EREDs could catalyze the C-alkylation of 2-nitropropylbenzene 51 with α-chloroamide 31, however most formed product as a racemate, highlighting the challenge of setting stereocenters in C–C bond-forming events (Figure 12A).⁵⁴ After extensive screening, we found that the ERED from Geobacillus kaustophilus (GkOYE) gave 52 in 33% yield and 78:22 er. Three rounds of ISM yielded a triple GkOYE mutant (D73C/A104H/Y264W, namely GkOYE-G7), delivering the product in 96% yield and 96:4 er. A high-order CT complex between 31, nitrone 55 (as a mimic of nitronate), and FMN_hq within GkOYE-G7 was observed, accounting for the selective radical initiation. GkOYE-G7 well accepted various α-benzyl nitroalkanes and simple linear or cyclic aliphatic nitroalkanes. Heterocycles, including pyridine and pyrazine, were also tolerated in this reaction. Beyond amide, α-halo ketone, ester, and sulfone could also serve as radical precursors (Figure 12A).⁵⁴

Figure 12.

Photoenzymatic C-alkylation of nitroalkanes and cross-electrophile coupling.^4,54

When considering the thermochemistry of nitroradical anion oxidation by FMN_sq, we found a report by Kornblum demonstrating that these species undergo mesolytic cleavage of the C–N bond to afford a nitrite and a carbon-centered radical.⁵⁵ While this mechanistic step is elusive in small molecule catalysis, we hypothesized that the appropriate enzyme might favor this step over electron transfer. This would allow the reductive coupling of alkyl halides and nitroalkanes. Cross-electrophile coupling (XEC) offers an attractive method for constructing Csp³–Csp³ bonds,⁵⁶ however, metal-catalyzed variations of this reaction often form homo-coupled products and struggle to control the stereoselectivity.^57–59 We sought to develop a biocatalytic XEC to coupling nitroalkanes with alkyl halides.

As a model, we explored the coupling of α-chloroamide 31 with 1-nitroethylbenzene 57 (Figure 12B).⁴ Pleasingly, the ERED from Caulobacter segnis (CsER) afford the XEC product (S)-58 in 92% yield and 95:5 er. In contrast, GluER-T36A furnished the opposite enantiomer with no homocoupling products observed in either case. We proposed a mechanism in which reduction of the 31 forms an alkyl radical 34 that can add to the in situ-generated nitronate 59 to give a nitro radical anion 60. Enzyme-mediated homolytic cleavage of the C–N bond forms nitrite and radical 61, which can be terminated via HAT to afford product 58. In support of this mechanism, when nitromethylbenzene 62 was used as a coupling partner, the denitrated XEC product 64 and the nitro-containing product 63 were formed. Resubjection of 63 to the reaction conditions did not yield 64, indicating that C–N bond cleavage rather than reduction is responsible for the denitration. This method is effective for an array of α-aryl nitronates and various α-halocarbonyl compounds (Figure 12B).

One of the most appealing features of biocatalysts is their excellent specificity. While CsER is efficient for catalyzing the reductive XEC between 31 and 68 to provide β-stereogenic amides as the major products (a/b > 50:1), the engineered GkOYE-G7 is superior for catalyzing C-alkylation of nitroalkanes affording tertiary nitroalkanes as the main products (b/a > 25:1, Figure 12C).⁵⁴ We hypothesize that specific polar interactions established by the nitro radical anion within the active site of GkOYE-G7 or CsER contribute to their chemoselectivity. This unparalleled chemoselectivity of biocatalysts is challenging to achieve using small molecule catalysts.

3. Conclusion and Perspective

Over the past eight years, our group and others have unlocked considerable non-natural radical reactivities with flavin-dependent EREDs.^60–69 Through these studies, we found that the protein active site can template interactions between multiple substrates and the flavin cofactor, resulting in new photochemistry and electron transfer mechanisms. This emergent activity helps to gate radical formation to the protein active site, enabling unparalleled levels of activity and selectivity. Beyond enabling electron transfer, the active site also helps to promote elusive reaction mechanisms that are challenging to achieve in solution. Thanks to the power of directed evolution, these features can be optimized and tuned for desired mechanisms and selectivity patterns.

The principles and activation modes discovered during our studies are only the beginning of what is possible with radical biocatalysts. While the chemistry described relied on reductive radical formation, numerous opportunities exist to identify new chemistry using oxidative mechanisms with flavoproteins.^70–72 Our group and others have shown exogenous photocatalysts can aid in unlocking new reactivity with nicotinamide,⁷³ flavin,^74–78 pyridoxal,⁷⁹ and thiamin diphosphate-dependent enzymes.⁸⁰ Beyond electron transfer, other groups have demonstrated the power of atom transfer mechanisms for radical initiation.^81–83 We hope this account will serve as a resource for others interested in repurposing existing enzymes for radical chemistry.

CONSPECTUS:

Enzymes are desired catalysts for chemical synthesis because they can be engineered to provide unparalleled levels of efficiency and selectivity. Yet, despite the astonishing array of reactions catalyzed by natural enzymes, many reactivity patterns found in small molecule catalysts have no counterpart in the living world. With a detailed understanding of the mechanisms utilized by small molecule catalysts, we can identify existing enzymes with the potential to catalyze reactions currently unknown in nature. Over the past eight years, our group has demonstrated that flavin-dependent “ene”-reductases (EREDs) can catalyze various radical-mediated reactions with unparalleled levels of selectivity, solving long-standing challenges in asymmetric synthesis.

This account will present our development of EREDs as general catalysts for asymmetric radical reactions. While we have developed multiple mechanisms for generating radicals within protein active sites, this account will focus on examples where flavin mononucleotide hydroquinone (FMN_hq) serves as an electron transfer radical initiator. While our initial mechanistic hypotheses were rooted in electron-transfer-based radical initiation mechanisms commonly used by synthetic organic chemists, we ultimately uncovered emergent mechanisms of radical initiation that appear unique to the protein active site. We will begin by covering intramolecular reactions and discussing how the protein activates the substrate for reduction by altering the redox-potential of alkyl halides and templating charge transfer complex between the substrate and flavin-cofactor. Protein engineering has been used to modify the fundamental photophysics of these reactions, highlighting the opportunity to tune these systems further using directed evolution. This section will highlight the range of coupling partners and radical termination mechanisms available to intramolecular reactions.

The next section will focus on intermolecular reactions and the role of enzyme-templated ternary charge transfer complexes between the cofactor, alkyl halide, and coupling partner in gating electron transfer to ensure that it only occurs when both substrates are bound within the protein active site. We will highlight the synthetic applications available to this activation mode, including olefin hydroalkylation, carbohydroxylation, arene functionalization, and nitronate alkylation. This section will also discuss how the protein can favor mechanistic steps that are elusive in solution for the asymmetric reductive coupling of alkyl halides and nitroalkanes. We are aware of several recent EREDs-catalyzed photoenzymatic transformations from other groups. We will discuss results from these papers in the context of understanding the nuances of radical initiation with various substrates.

These biocatalytic asymmetric radical reactions often complement the state-of-the-art small-molecule-catalyzed reactions, making EREDs a valuable addition to a chemis’s synthetic toolbox. Moreover, the underlying principles studied with these systems are potentially operative with other cofactor-dependent proteins, opening the door to different types of enzyme-catalyzed radical reactions. We anticipate this Account will serve as a guide and inspire broad interest in repurposing existing enzymes to access new transformations.

Biographies

Biographies

Haigen Fu received his Ph.D. in 2019 from the University of Groningen under the supervision of Prof. Gerrit J. Poelarends. He was a Postdoctoral Fellow with Prof. Todd K. Hyster at Princeton University and Cornell University from 2020 to 2022. In 2023, he started as an Assistant Professor at the Chinese Academy of Medical Sciences & Peking Union Medical College. His group focuses on developing new biocatalytic methods for accelerating drug discovery and preparation.

Todd K. Hyster is a Full Professor of Chemistry at Princeton University. He received his B.S. in Chemistry from the University of Minnesota. He did his Ph.D. studies with Tomislav Rovis at Colorado State University. As part of his Ph.D., he was a Marie Curie Fellow with Thomas Ward at the University of Basel. He was an NIH Postdoctoral Fellow with Prof. Frances Arnold at Caltech. He started his independent career at Princeton University in 2015. His group has developed new methods in photoenzymatic catalysis.