Broadening the scope of biocatalytic C–C bond formation

Abstract

The impeccable control over chemo-, site-, and stereoselectivity possible in enzymatic reactions has led to a surge in the development of new biocatalytic methods. Despite carbon-carbon (C–C) bonds providing the central framework for organic molecules, development of biocatalytic methods for their formation has been largely confined to the use of a select few lyases over the last several decades, limiting the types of C–C bond-forming transformations possible through biocatalytic methods. This Review provides an update on the suite of enzymes available for highly selective biocatalytic C–C bond formation. Examples will be discussed in reference to the (1) native activity of enzymes, (2) alteration of activity through protein or substrate engineering for broader applicability, and (3) utility of the biocatalyst for abiotic synthesis.

Graphical Abstract

Discovery and application of enzymes that carry out new reactions are essential for the broader implementation of biocatalysts in organic synthesis. This Review highlights the most recent developments in biocatalytic methods for carbon-carbon bond formation.

Introduction

In nature, enzymes catalyse a myriad of highly selective transformations in the biosynthesis of primary and secondary metabolites in nature.^1–3 Major advances in enzyme discovery and engineering over the last several decades have brought with them a surge in the development and implementation of biocatalytic reactions in organic syntheses executed in both academic and industrial laboratories.^4–6 Enzymes are attractive catalysts in organic synthesis due to (1) the control they exhibit over the chemo-, site-, and stereoselectivity in the reactions they catalyse, (2) the new retrosynthetic disconnections they can offer over small molecule-mediated transformations, and (3) their sustainable footprint, especially in industrial-scale reactions.^6–9

The major barrier in the use of biocatalysts in organic synthesis is the identification and development of a biocatalyst for a given reaction.¹⁰ This can be challenging because, while the three-dimensional control of a protein active site lends itself toward high levels of selectivity in biocatalytic reactions, this same tight regulation can limit the versatility of a biocatalyst through a restricted substrate scope.¹¹ To overcome this limitation, enzymes can be engineered for the desired chemistry through rational design or random mutagenesis.¹² The power of iterative rounds of random mutagenesis and screening was recently recognized with the 2018 Nobel Prize in Chemistry, awarded for the directed evolution of enzymes. In this approach, evolutionary pressure inherent to the development of enzymes in nature is expedited and applied artificially in a laboratory setting to evolve an enzyme for a target transformation or process (Fig. 1a).^13–15 Additionally, advances in bioinformatic tools and machine learning are progressively merging the use of random and rational protein design for increasingly rapid protein engineering.^16–20

Figure 1 |. Engineering enzymes for organic synthesis.

a | Typical iterative workflow of the directed evolution of enzymes for the development of new biocatalysts. b | Five enzymes were evolved by Merck and Codexis for the manufacture of islatravir (8), including galactose oxidase (GOase), pantothenate kinase (PanK), deoxyribose 5-phosphate aldolase (DERA), phosphopentomutase (PPM), and purine nucleoside phosphorylase (PNP).²¹

The biocatalytic route developed by Merck and Codexis for the manufacture of an anti-HIV drug, islatravir (8), demonstrates the tunability of enzymes for a desired transformation achievable with protein engineering.²¹ To overcome the inefficient toggling of protecting groups required to successfully set stereocenters present in non-natural nucleoside 8,²² five enzymes were selected based on the nucleoside salvage pathway present in bacteria in a new retrosynthetic approach (Fig. 1b).²¹ Each of the selected enzymes was engineered for a distinct purpose; for example, the oxidase (GOase) was engineered for improved activity and stereoselectivity whereas the aldolase (DERA) was engineered specifically for acetaldehyde (5) tolerance. Using these five engineered enzymes along with four auxiliary enzymes, islatravir (8) could be synthesized in a three-step biocatalytic sequence with a yield triple that of the original chemical synthesis (Fig. 1b).^21,22

In addition to biocatalytic cascades, such as the islatravir (8) synthesis,^21,23 biocatalysts can be used to catalyse key transformations in synthetic routes that draw on both biocatalytic methods and traditional small molecule reagents.²⁴ The power of using enzymes to address synthetic challenges has been demonstrated over the last several decades with the increased implementation of biocatalytic processes in the synthesis of pharmaceutical compounds.^5,6 However, a major limitation to the broader use of biocatalysts in both academic and industrial labs is the relatively narrow set of biocatalysts that can be reliably designed into a synthesis.

Carbon-carbon (C–C) bonds provide the central framework for all organic molecules and thus there are countless enzymes involved in their formation in primary and secondary metabolite pathways. Yet, the commonly employed biocatalysts for C–C bond formation are primarily restricted to a small subset of lyases, thereby limiting the scope of biocatalytic C–C bond formation to aldol reactions, acyloin condensations, and cyanohydrin formation in organic synthesis.²⁵ Aldolases are by far the most commonly employed lyases for biocatalytic C–C bond formation and have been extensively developed into a versatile class of biocatalysts for site- and stereoselective aldol reactions.^26–28 Thiamine diphosphate (ThDP)-dependent enzymes perform acyloin condensation reactions and have been used as valuable biocatalysts in the synthesis of enantioenriched building blocks.^29–31 Lastly, biocatalytic cyanohydrin formation is achievable through several different mechanisms: (1) the addition of a cyanide ion to a ketone or aldehyde catalysed by a hydroxynitrile lyase^32–34 or (2) the opening of an epoxide catalysed halohydrin dehalogenases.³⁵ The breadth of research on these biocatalysts for C–C bond formation is captured in previous reviews on this topic.^28,36,37

Discovery and application of enzymes that carry out new reactions are essential for the broader implementation of biocatalysts in organic synthesis, including new catalysts for C–C bond formation. Over the last several years, there have been numerous advances in biocatalytic C–C bond formation, using both wild-type enzymes and enzymes engineered for non-natural functions. This Review provides an update on the exciting new advances in biocatalytic C–C bond formation reactions including alkylation, acylation, oxidative C–C coupling, cyclisation, and carbene transfer reactions.

Alkylation and acylation reactions

Alkylation reactions are ubiquitous in nature and are one of the key C–C bond-forming events in natural product biosynthesis. These reactions can occur through various mechanisms, however one of the most common is an S_N2-type reaction utilizing an activated electrophilic cofactor and a nucleophilic substrate (Fig. 2)

Figure 2 |. Biocatalytic alkylation reaction using activated electrophilic cofactors or substrates as alkyl donors.

a | Methylation reactions catalysed by SAM-dependent enzymes can be diversified with the use of synthetic SAM analogues for broader C–C alkylation reactions.⁵⁰ b | The diphosphate prenyl donors and acceptors in prenyltransferase-catalysed reactions can be modified for catalyst- and substrate-controlled site-selective alkylation reactions.^54,55 c | Tryptophan synthase, a PLP-dependent enzyme, has been engineered to catalyse alkylation reactions on substituted indoles, nitro alkanes, and oxindoles.^63–68 d | CylK utilizes an alkyl halide in the Friedel-Crafts alkylation of resorcinol derivatives.^69,70

S-adenosyl-L-methionine (SAM)-dependent enzymes.

SAM (9) is the major methyl donor in biological systems and the second most common enzymatic cofactor after ATP.^38,39 The most common reaction catalysed by enzymes harbouring a SAM cofactor is the transfer of the electrophilic methyl group bound to the sulfonium centre of SAM through an S_N2 reaction (Fig. 2a). However, SAM-dependent enzymes can also perform radical reactions for the transfer of methyl or adenosyl groups,^40,41 including those for C–C bond formation.⁴² Methyltransferases utilizing SAM cofactors are incredibly important in biological systems for applications including the biosynthesis of small molecules, protein repair, gene silencing, and chromatin remodelling.⁴³

Unsurprisingly, given the crucial biological ramifications of these methylation events, the ability of these methyltransferases to selectively methylate their substrates is critical. This selective alkylation of such a diverse range of biomolecules has motivated the use of SAM-dependent methyltransferases as tools for biorthogonal labelling in biotechnology.^44–48 In these systems, synthetic analogues of SAM are used to label nucleic acids or proteins by replacing the sulfonium centre with an aziridine ring or larger alkyl or allyl group.⁴⁹ Upon nucleophilic attack of the target molecule on the SAM analogues, the resulting product will either be trapped by the aziridine analogue or alkylated with the alternative sulfonium substituent.⁴⁹

Although methyltransferases have been used extensively for the modification and labelling of nucleic acids and proteins, methyltransferases using SAM analogues have only sparingly been explored for the biocatalytic C–C alkylation of small molecules in abiotic synthesis. Currently, the only example of biocatalytic C–C alkylation using SAM analogues is for a site-selective Friedel-Crafts alkylation at the unactivated C–H bond of a coumarin substrate to form 10 in Streptomyces secondary metabolite pathways (Fig. 2a).⁵⁰ In these reactions, SAM analogues synthesized with bulkier substituents at the sulfonium centre were used as cofactors for the methyltransferase-catalysed Friedel-Crafts alkylation of several different coumarin derivatives to form coumarin 11 modified with allyl (a-b), propargyl (c-d), and benzyl (e) groups with excellent site-selectivity.⁵⁰

Prenyltransferases.

Using a similar strategy to SAM-dependent methyltransferases, prenyltransferases take advantage of the excellent leaving group capability of the diphosphate group on allyl diphosphate substrates (12) for the transfer of an allylic group to a second more electron-rich substrate.^51,52 The catalyst- and substrate-controlled site-selectivity exhibited by these enzymes have made them attractive biocatalysts for late-stage functionalization of natural products, especially due to the diverse biological effects the addition of a hydrophobic prenyl group can provide to a molecule.⁵³

The investigation of the substrate scopes of several tryptophan prenyltransferases have shown that these enzymes are often promiscuous in their ability to catalyse the site-selective carbon-prenylation of a broad scope of aromatic tryptophan derivatives.⁵² Like SAM-dependent methyltransferases, both the alkyl donating and accepting substrates can be modified to diversify these reactions (Fig. 2b). A recent report demonstrated that a single wild-type prenyltransferase (FgaPT2) could donate alkyl groups from 34 different diphosphate derivatives, each forming a distinct C–C bond to give an alkylated product (13a-c).⁵⁴ Interestingly, these FgaPT2-catalysed reactions exhibited substrate-dependent site-selectivity, with changes to the electronics and bulkiness of the alkyl group being donated conferring a switch in selectivity from the native C4 site of alkylation to a non-natural C5 alkylation. The divergent site-selectivity demonstrated by these reactions adds to the rapidly growing toolbox of site-selective alkylations, such as the C3 alkylations performed by the bacterial prenyltransferase PriB.⁵⁵ This recently discovered enzyme is highly promiscuous, possessing the ability to alkylate a broad range of indole (14) and naphthalene (15) derivatives.⁵⁵

Overall, the relative promiscuity of these enzymes enables rapid substrate engineering through perturbation of the alkyl group being transferred and the substrate being alkylated, making these enzymes an attractive class of potentially robust biocatalysts.

Pyridoxal 5-phosphate (PLP)-dependent enzymes.

PLP is a versatile cofactor present in enzymes that catalyse many transformations including transamination, epimerization, and decarboxylation. Despite this diversity, the shared driving force in all PLP-dependent mechanisms is the “electron sink” character of the PLP cofactor as it uses its π-system to stabilize the negative charge that develops at its Cα position.⁵⁶ Several PLP-dependent enzymes have been identified to catalyse the formation of C–C bond forming reactions, including (1) aldol reactions catalysed by threonine aldolase,⁵⁷ (2) the formation of alpha-amino ketones by a polyketide-like synthase,^58,59 and (3) the formation of tryptophan and tryptophan derivatives by tryptophan synthase.^60,61

In the catalytic mechanism of tryptophan synthase, a serine amino acid condenses onto the PLP cofactor and, upon dehydration, forms the highly electrophilic aminoacrylate intermediate species (16). Activation of this electrophilic cofactor allows for the nucleophilic attack by an indole to forge a new C–C bond, ultimately forming tryptophan (17). Tryptophan synthase is composed of an α subunit (TrpA) that acts as a regulatory subunit to the catalytic β subunit (TrpB), which binds PLP and catalyses the C–C bond formation event.⁶² Through directed evolution of the TrpB subunit, it was identified that the allosteric regulation provided by the TrpA subunit could be circumvented through a few key mutations in TrpB, increasing the biocatalytic utility of this enzyme.⁶³ Since this first discovery, TrpB has been engineered extensively for broader C–C bond forming reactions on substituted indoles (18), nitro alkanes (19), and oxindoles (20) to form quaternary bonds (Fig. 2c).^63–68

Halogenated electrophiles.

Unlike the highly activated electrophilic cofactors described so far, an alkylating enzyme (CylK) was recently discovered that solely relies on a halogenated alkyl chain (21) as the electrophilic alkyl group in an intermolecular Friedel-Crafts alkylation.⁶⁹ The native reaction catalysed by this alkylating enzyme is a dimerization reaction between two halogenated resorcinol derivatives to form 22. In addition to this native reaction, CylK can catalyse the alkylation of a panel of resorcinol derivates (such as 23 and 24) using truncated alkyl halides (25; Fig. 2d).⁷⁰

Intramolecular Friedel-Crafts alkylations.

Intermolecular Friedel-Crafts alkylations occur routinely in primary and secondary metabolite pathways and can be catalysed by some of the enzymes described so far; however, C–C bonds can also be formed in intramolecular Friedel-Crafts alkylations to forge a new ring, such as in the biosynthesis of many isoquinoline alkaloid and lignan natural products (Fig. 3a,c).^71,72

Figure 3 |. Biosynthetic and biocatalytic ring-closing reactions.

a | Several different enzymes catalyse intramolecular alkylation or oxidative coupling reactions to form a new ring in the biosynthesis of isoquinoline alkaloids.⁷¹ b | Noroclaurine synthases (NCSs) have been utilized as biocatalytic Pictet-Spenglerases for the formation of a panel of products.^74,75 c | A non-haem dioxygenase catalyses an intramolecular alkylation reaction to form a new ring in the biosynthesis of podophyllotoxin lignan products.⁸⁵ This biocatalytic reaction has been applied to the chemoenzymatic synthesis of several natural product analogues.^88,89

One critical enzyme in the biosynthesis of isoquinoline alkaloid natural products is norcoclaurine synthase (NCS; Fig. 3a). NCS is a Pictet-Spenglerase that catalyses the condensation of an aldehyde 27 onto an amine 26 to form the iminium intermediate 28. This electrophilic intermediate (28) is then poised for an intramolecular Friedel-Crafts alkylation by nucleophilic attack to forge a C–C bond with the phenolic ring A to form (S)-norcoclaurine (29).⁷³ Through the exploration NCSs for their broader biocatalytic utility for Pictet-Spengler reactions, the substrate scopes of these enzymes have been found to be relatively broad with the ability to accept a wide range of aldehyde derivates to form various tetrahydroisoquinoline products (33).⁷⁴ As such, these enzymes have been applied to the chemoenzymatic syntheses of a range of alkaloid natural products and natural product derivatives.⁷⁴ Additionally, a recent report demonstrated the ability of these enzymes to perform this ring closure with unactivated ketone coupling partners, thereby overcoming the limitation of coupling only to aldehydes.⁷⁵ Using ketone substrates, quaternary centres can be formed in chiral disubstituted- and spiro-tetrahydroisoquinolines (see 34 and 35, respectively; Fig. 3b).⁷⁵

Modification of the NCS product, (S)-norcoclaurine (29), by several methyltransferases and a cytochrome P450 yields (S)-reticuline (30), a branchpoint in the biosynthesis of many isoquinoline alkaloids.^71,76 Isoquinoline alkaloids encompass many pharmacologically valuable natural products such as morphine and berberine. As such, extensive research has focused on the metabolic engineering of microorganisms for the scalable production of these molecules.^77–81 The in vivo production of valuable molecules can be beneficial, especially when the chemical syntheses are not amenable to large-scale synthesis or the enzymes involved in the pathway are challenging to work with in vitro.⁸²

As a branchpoint in the biosynthesis of isoquinoline alkaloids, several different C–C bond-forming reactions can be performed on (S)-reticuline (30) for further diversification. In particular, a fourth ring can be formed either through another intramolecular Friedel-Crafts alkylation to form (S)-scoulerine (31) or through an oxidative coupling reaction (described in more detail in the next section of this Review) to form (S)-corytuberine (32; Fig. 3a). The enzyme responsible for the Friedel-Crafts alkylation is the flavin-dependent berberine bridge enzyme (BBE). In all the examples described thus far, the first step before an alkylation event is the activation of an electrophilic carbon source; however, in the case of the BBE-catalysed alkylation, the unactivated methyl group is directly attacked by the phenol in 30 (ring D).⁸³ This is made possible by the presence of a flavin cofactor in the active site of BBE which accepts a hydride from the methyl group being attacked in a single concerted step.⁸³ This enzyme has not been explored for use as a biocatalyst as extensively as NCS; however, it has been identified that the substituents on ring D in 30 can be modified with successful C–C bond formation. Additionally, the site-selectivity of the bond formation can be biased in cases where a C–F bond is present at a potential alkylation site.⁸⁴

Beyond alkaloids, another intramolecular Friedel-Crafts alkylation was recently identified in the biosynthesis of podophyllotoxin lignans (Fig. 3c).⁸⁵ In particular, a non-haem dioxygenase was identified to catalyse the formation of the final ring in this biosynthetic pathway to form deoxypodophyllotoxin (37).^85,86 Enzymes containing a non-haem iron are most commonly involved in hydroxylation events; however, this class of enzymes can also perform a myriad of other oxidative transformations.⁸⁷ Evaluation of the substrate scope of the C–C coupling non-haem dioxygenase, 2-ODD-PH, demonstrated its ability to perform oxidative C–C alkylations on a variety of derivatives of 36, particularly with modifications to the substituents around rings A and C (Fig. 3c).⁸⁸ This chemistry was used to perform late-stage C–C coupling reactions for the chemoenzymatic synthesis of products such as 38 and 39.⁸⁹

Friedel-Crafts acylations.

Typically, acyl transfers in nature are carrier protein-dependent processes that proceed through S_NAc-type reactions.⁹⁰ However, similarly to the Friedel-Crafts alkylations described in this section so far, one unique bacterial acyltransferase has been identified to catalyse a Friedel-Crafts acylation on resorcinol substrates (Fig. 4).⁹¹ This acyltransferase natively transfers an acetyl group from 43 to the resorcinol substrate (40), either by directly forming the C–C bond in the acetylated product (41), or through an ester intermediate (42).⁹² In the latter mechanism, the acyltransferase can catalyse the biocatalytic equivalent of a Fries-rearrangement with an intramolecular acetyl transfer to form 41.

Figure 4 |. Site-selective acetylation of resorcinol derivatives catalysed by an acyltransferase.

In addition to the native acyl donor (43), various acyl donors (44-49) were successfully employed in a biocatalytic Friedel-Crafts acylation or Fries rearrangement-like reaction to acetylate resorcinol derivates (40).^92–94

Like the previously described intermolecular Friedel-Crafts reactions, both the acyl donor and acceptor could be modified to increase the versatility of these biocatalytic reactions. Specifically, a broad range of commercially available acetyl donors were productively utilized, including activated esters (44-45), thioesters (46-47), and amides (48-49) to the selectively acetylate various resorcinol derivates (40).^92–94

Oxidative C–C coupling reactions

Oxidative coupling reactions perhaps offer the most efficient approach to C–C bond formation as two building blocks are directly stitched together.⁹⁵ However, a major limitation in current synthetic methods for oxidative coupling lies in the control over the chemo-, site-, and stereochemical outcome of the reaction, particularly in the case of intermolecular coupling reactions.^96,97 Nature has evolved enzymes that can control the oxidative dimerization of phenolic substrates with precise control of the site- and stereoselectivity of bond formation. These enzymes include cytochromes P450, laccases, and peroxidases. These three classes of enzymes share the ability to initiate radical mechanisms on phenolic substrates through the abstraction of a hydrogen atom from the phenol. However, the mechanism of hydrogen abstraction and extent of control exerted over the bond formation event in intermolecular and intramolecular reactions vary among these enzymes.

Intramolecular C–C coupling.

Cytochromes P450 catalyse a wide range of oxidative transformations leveraging their iron-haem cofactor.⁹⁸ Often, these oxidative transformations involve a hydrogen atom abstraction from the substrate, followed by oxygenation through a rebound hydroxylation.⁹⁹ However, the mechanism of these oxidative transformations can diverge following initial hydrogen atom abstraction, including C–C bond formation.⁹⁸ The C–C bond formed through an oxidative coupling event in a number of natural products have been attributed to P450-catalysed reactions.¹⁰⁰

Many of these P450-catalysed C–C coupling reactions occur intramolecularly, such as in the biosynthesis of (S)-corytuberine (32).¹⁰¹ In this reaction, the cytochrome P450 CYP80G2 initiates the oxidative mechanism on the substrate 30 with a hydrogen atom abstraction from one of the phenols. Instead of the canonical rebound hydroxylation typically performed by these enzymes, a C–C bond is formed either through a radical-radical coupling or by radical addition to a second phenolic group to form 32 (Fig. 3a). Another key intramolecular C–C coupling catalysed by a P450, is in the biosynthesis of the glycopeptide antibiotic vancomycin.¹⁰² Upon identification of the P450 OxyC responsible for catalysing this reaction, it was applied for the in vitro chemoenzymatic synthesis of vancomycin derivatives.¹⁰³

Intermolecular C–C coupling.

In addition to intramolecular C–C coupling, oxidative coupling reactions can occur in nature between two phenolic substrates in an intermolecular fashion. These reactions are key steps in the biosynthesis of many dimeric biaryl secondary metabolites and can be catalysed by P450s, laccases, or peroxidases (Fig. 5).

Figure 5 |. Intermolecular oxidative coupling reactions in nature.

a | Cytochrome P450-catalysed site-selective oxidative coupling of coumarin substrates to form bicoumarin isomers.¹⁰⁵ b | Laccase-catalysed oxidation of polyphenol substrates followed by dirigent protein-mediated radical coupling to yield enantioenriched gossypol.¹¹⁵ c | Laccase-catalysed atroposelective oxidative coupling of naphthopyrone substrates to form axially chiral biaryl products.¹¹⁶

P450s have been identified as the catalysts for intermolecular C–C coupling reactions for site-selective biaryl bond formation in several secondary metabolite pathways. For example, a group of Aspergillus P450s have been identified that catalyse the dimerization of a coumarin substrate to form up to six distinct axially chiral bicoumarins with catalyst-controlled site-selectivity.¹⁰⁴ Two of these enzymes, KtnC and DesC, were characterized to selectively catalyse the site- and atroposelective C–C coupling to form axially chiral (P)-orlandin (50) and (M)-desertorum (51), respectively (Fig. 5a).¹⁰⁵ This divergent site-selectivity among P450 homologues appears in several other bacterial and fungal secondary metabolite pathways, suggesting high levels of control exerted over the C–C bond formation event in P450 active sites.^106–108

While P450s have only recently demonstrated an ability to catalyse site- and atroposelective intermolecular C–C coupling reactions, laccases and peroxidases have long been used in biotechnological applications for their oxidative coupling capabilities.^109,110 Until recently, however, these enzymes were not known to impart site- or enantioselectivity in their oxidative coupling reactions limiting their utility for selective biocatalytic reactions. At a mechanistic level, these oxidative enzymes are responsible for the generation of radical intermediates but were not implicated in templating the substrates to exert control over the bond-forming event.¹¹¹ In contrast, select radical substrates formed by laccases or peroxidases can rapidly be captured by a second class of protein catalysts, dirigent proteins.^112,113 Dirigent proteins contain two distinct substrate binding sites and are uniquely poised to rapidly bind radical species in an aqueous solution and orient two of the substrates in a productive position with control over the site- and enantioselectivity of the C–C bond formation.^113,114 Dirigent proteins have been identified in plant secondary metabolite pathways, including in the formation of lignans (such as 37) and the biaryl molecule gossypol (52; Fig. 5b).^113,115

Recently, however, the first report of laccases that demonstrated control over the bond-forming event independently of an auxiliary dirigent protein was reported in the biosynthesis of axially chiral fungal ustilaginoidin natural products (Fig. 5c).¹¹⁶ Interestingly, while three of the characterized laccases selectively catalysed the formation of the (P)-atropisomer (53), one of the laccases, UstL, exhibited a catalyst-loading dependence on the atroposelectivity of the product, where the authors noted increasing concentrations of the catalyst gave an increased preference for the (M)-atropisomer (54).¹¹⁶ Although the mechanism behind the selectivity switch observed with UstL is unclear at this point, this unique catalyst-controlled atroposelectivity is unprecedented among laccases and holds promise for the development of stereodivergent biocatalysts.

This growing subset of P450s and laccases that perform oxidative coupling reactions to afford biaryl products with complementary site- and atroposelectivity has been vastly underexplored as biocatalysts despite the promise they offer for these challenging transformations.

Cyclisation reactions

Nature has mastered the construction of diverse scaffolds from common precursors by evolving catalysts capable of controlling the outcome of cyclisation reactions. These reactions vary in complexity with some catalysing the formation of a single C–C bond, whereas others can catalyse the stereoselective formation of a series of new C–C bonds in cascade cyclisation reactions.

Terpene cyclases.

Terpenes constitute the largest natural product family with over 80,000 members.¹¹⁷ Despite the magnitude and structural diversity of this family, all these natural products stem from a similar five-carbon precursor unit. Most of the structural complexity of the products are installed in cascade cyclisation reactions wherein multiple fused rings and stereocenters are installed by a single enzyme (such as in Fig. 6a). The enzymes that catalyse these reactions are referred to as terpene cyclases (or terpene synthases). These enzymatic reactions are considered the most complex in nature, with a typical cyclisation reaction breaking or making new bonds at the majority of the carbon atoms comprising the backbone of the terpene precursor substrate.¹¹⁷ The complexity-generating nature of terpene cyclase-catalysed reactions have influenced the biomimetic syntheses of terpenoid natural products, such as the use of Lewis acid-mediated opening of vinyl epoxides for the cascade polycyclisations of terpenoids, however the stereocontrol of the cyclisation is still lacking in chemical methods.^118,119

Figure 6 |. Cyclisation reactions catalysed by natural and engineered enzymes.

a | Terpene cyclases catalyse stereoselective cascade cyclisation reactions for the generation of polycyclic products.¹²¹ b | Diels-Alderases have been identified to catalyse intramolecular Diels-Alder reactions in secondary metabolite pathways and are hypothesized to be involved in intermolecular cycloaddition reactions.^131,135 c | Engineered flavin-dependent ene-reductases can catalyse non-natural photobiocatalytic radical cyclisation reactions.^141,142

Triterpene cyclases, such as the squalene-hopene cyclase (SHC), have been studied extensively for their structure and mechanism.¹²⁰ In the native SHC-catalysed cascade cyclisation of squalene (55) to form hopene (56), five new C–C bonds are forged to form five fused rings and ten new stereocenters (Fig. 6a). It has been found that SHCs can catalyse a much wider variety of cyclisation reactions;¹²¹ examples of these include (1) stereoselective cascade cyclisations with smaller linear terpenes to form molecules with two or three fused rings,^122,123 (2) Friedel-Crafts alkylations,¹²⁴ (3) Prins-type reactions,^125–127 and (4) isomerisation reactions.¹²¹ Additionally, these enzymes have demonstrated their amenability to evolution campaigns, often with a single amino acid change in the binding pocket of the enzyme resulting in a dramatic increase in activity with a non-native reaction.¹²¹ However, this promiscuity can also lead to the formation of multiple product isomers from a single linear substrate due to the ability of the active site to accommodate multiple conformations of the carbocation skeleton during cyclisation.¹²⁸ Despite this potential limitation of some terpene cyclases, engineered SHC variants have successfully been employed in chemoenzymatic syntheses, such as in the synthesis of (+)-ambrein, a high-commodity natural product that is difficult to isolate from natural sources.¹²⁹

Diels-Alderases.

In the list of classic, complexity-generating C–C bond forming reactions, the Diels-Alder reaction comes close to the top. This transformation has been central to the synthesis of six-membered rings in complex natural products and simple structures alike. This bread-and-butter reaction is embedded in the retrosynthetic logic chemists use to develop strategies toward target molecules, which has driven the development of variants of the Diels-Alder reaction that allow this cycloaddition to be applied intra- or intermolecularly and on substrates with a broad range of electronic properties (see 57 for intramolecular cycloaddition and 59-60 for intermolecular cycloaddition in Fig. 6b). Despite the common occurrence of the cyclohexene motif in natural product cores, the first enzymatic Diels-Alder reactions were carried out with proteins engineered to this non-native reactivity by leveraging the rate enhancement possible through templating of a given diene and dienophile and the acid-base type activation possible in an active site.¹³⁰

While a number of artificial Diels-Alderases have been developed, natural Diels-Alderases have proven to be a much more elusive target and have garnered a great deal of attention over the last several decades.¹³¹ Elements of this controversy arise from the dual action of many “Diels-Alderases” as many enzymes given this label generate one of the Diels-Alder partners in addition to potentially catalysing the cyclization event. This controversy has also been fuelled by debate over what mechanistically constitutes a Diels-Alder cycloaddition, with strict definitions calling for concerted formation of two new C–C bonds and more liberal labelling of Diels-Alderases catalysing these bond formations in a step-wise fashion.

Recently, crystal structures of several bifunctional enzymes were reported that catalyse intramolecular Diels-Alder reactions: (1) SAM-dependent O-methyltransferase LepI in leporin C (62) biosynthesis^132,133 and (2) short-chain dehydrogenases MalC and PhqE in (+)-premalbrancheamide (63) biosynthesis.¹³⁴ Although these enzymes are homologous to functionally diverse enzyme classes, it appears that enzymes have evolved away from their canonical functions to act as a scaffold for stereoselective Diels-Alder reactions. This possesses a challenge in assigning cyclase function based on sequence. Although, several intermolecular [4+2] cycloadditions have been proposed to be catalysed by a Diels-Alderase, these proposals have not been experimentally confirmed.¹³⁵ One example of these is riboflavin synthase, which catalyses the putative intermolecular cycloaddition of two molecules to form a key intermediate 64 in the riboflavin biosynthetic pathway that has been isolated and characterized.¹³⁶ Ultimately, both characterization of the chemistry mediated by putative cyclases and mechanistic investigations are required to accurately label a Diels-Alderase.

Ene-reductases.

Ene-reductases are a class of enzymes that natively catalyse the asymmetric reduction of activated alkenes. These flavin-dependent enzymes deliver a hydride to one face of the alkene substrate. Subsequent protonation of the reduced C–C bond by a conserved tyrosine residue reduces the alkene in a stereo-controlled fashion.¹³⁷ Ene-reductases are robust enzymes, typically with broad substrate scopes, high yields, and excellent selectivity, garnering them a great deal of attention as industrial biocatalysts.^138,139

Beyond their widely studied and utilized canonical function, several ene-reductases were recently employed in the reduction of other functional groups. Ene-reductase variants lacking the conserved tyrosine residue responsible for the protonation event can transfer a hydride to α,β-unsaturated aldehydes and ketones triggering a reductive cyclisation event to form cyclopropanes.¹⁴⁰ Although these reactions only exhibit moderate stereoselectivity, these experiments represent an important proof-of-concept that the reactivity of these transformations could be controlled through protein and substrate engineering beyond the native reactivity observed for ene-reductases.

As well as the non-natural cyclopropanation reactions, several recent reports demonstrated the ability for ene-reductases to catalyse the radical cyclisation of five- to eight-membered rings in photobiocatalytic reactions (Fig. 6c).^141,142 In contrast to the typical mechanism of ene-reductases, where hydride transfer from the flavin cofactor to the substrate initiates the reaction, the delivery of a hydrogen atom quenches a radical species such as 66 to form a cyclised product (69; Fig. 6c).¹⁴³ In the first report of this photobiocatalytic reaction, a panel of chloroacetamide derivatives (65) were exposed to cyan LEDs, causing the formation of a radical and subsequent radical cyclisation and hydride transfer to form products such as 70 and 71.¹⁴¹ A second report demonstrated the effectiveness of this reaction on a panel of chloroacetamide substrates to form diverse oxindole products (72).¹⁴²

In addition to engineering ene-reductases for a completely new type of reaction when combined with photocatalysis, engineered haemoproteins for carbene transfer reactions serve as an excellent case study in engineering enzymes for completely new-to-nature types of transformations.

Carbene transfer reactions

The diversity of transformations achievable through carbene transfer including cyclopropanations, heteroatom-hydrogen (X–H) bond insertions, and rearrangements make carbene transformations valuable in synthetic chemistry.^144–149 Although carbene transfer is not known to occur in nature, oxene transfer is commonly performed by cytochromes P450.⁹⁹ These enzymes harbour a haem cofactor (73) in their active sites, which upon reduction binds molecular oxygen to form a highly reactive iron-oxene (74) intermediate species.⁹⁹ As was previously described, this reactive iron-oxene species (74) typically performs a hydrogen atom abstraction, after which it can perform a myriad of oxidative transformations such as a rebound hydroxylation or insertion of oxygen into a double bond to form an epoxide (Fig. 7a).^98,150

Figure 7 |. Engineering non-natural carbene transferases for biocatalytic C–C bond formation.

a | Natural iron-oxene formation in the haem cofactor of haemoproteins. b | Engineering artificial carbene-transferases through iron-carbene formation.^{151,168–170} c | Scope of cyclopropanation reactions catalysed by carbene transferases^{158–160,162–164} and representative use in chemoenzymatic synthesis.¹⁶⁶ d | Carbene transfer into alkyne forms highly strained cyclopropene and bicyclobutane products.¹⁶⁵ e | Carbene insertion into C–H bonds forms alkylated products.^{168,169,171–174} f | Carbene transferases have catalysed C–C bond formation through rearrangement reactions and Wittig reactions.^175–177

In 2013, Arnold and co-workers published the first of a series of reports describing the utilization of the bacterial P450 BM3 for carbene transfer.¹⁵¹ Arnold discovered that replacement of molecular oxygen with a diazoacetate reagent (75) allowed for the formation of an iron-carbene species (77) in the P450 haem cofactor. In the presence of an olefin (78), carbene transfer could occur to form a cyclopropane (79) with the stereo-control provided by an enzyme active site, marking the first instance of a non-natural “carbene transferase” (Fig. 7b).

Engineering carbene transferases.

Following Arnold’s initial report, several research groups have led the way in engineering P450 BM3 and other haemoproteins into carbene transferases for cyclopropanation of olefins, alkylation through insertion into C–H bonds, and C–C bond formation through rearrangement reactions.¹⁵² Engineering efforts have focused on mutagenesis and alterations to the haem cofactor and coordinating metal ion to modulate oxidation potential and overcome obstacles including (1) limited substrate scope, (2) access to a single isomer of the desired product, and (3) competition with the native iron-oxene formation.¹⁵² The pursuit of this new-to-nature biocatalyst serves as an excellent example for how enzymes can be engineered into refined biocatalysts for a given reaction.

In 2013, Arnold’s group identified that mutagenesis of the cysteine residue, which serves as the axial ligand to the P450 haem cofactor, to a serine increased the reduction potential of the resting state enzyme thereby abolishing monooxygenase activity and giving a preference to the formation of a carbene over an oxene.¹⁵³ This point mutation was then applied to this highly conserved residue in other P450s, forming a suite of enzymes dubbed “P411s” due to the change in their absorbance from 450 nm to 411 nm.¹⁵⁴ Additionally, nucleophilic residues in the active site causing mechanism-based inhibition were identified and removed to further promote carbene transfer to a desired substrate.¹⁵⁵ The Fasan group ventured beyond P450s to other haemoproteins and successfully engineered the first myoglobin capable of carbene transfer.¹⁵⁶ For the myoglobin-derived carbene transferases, modification of the haem cofactor through installation of alternative substitution patterns increased the efficiency of iron-carbene formation and enabled reactions to be performed aerobically without inhibition of carbene transfer by oxygen.¹⁵⁷

Olefin cyclopropanation.

Over the last five years, haemoproteins have been engineered for modified selectivity and substrate scope of olefin cyclopropanations performed by carbene transferases through directed evolution and rational mutagenesis of the active site.¹⁵² These engineering efforts have provided access to stereo-divergent cyclopropane products with control over both the diastereo- and enantioselectivity in carbene transfer reactions.

The substrate scope of these reactions has moved beyond phenyl-substituted olefins (forming 82) to form cyclopropanes substituted with heteroatoms (83) or aliphatic chains (84) to access more diverse cyclopropanes (Fig. 7c).^158–161 Furthermore, the electron withdrawing group on the diazo reagent could be modified from an ester moiety to a nitrile or trifluoromethyl substituent to form the corresponding cyclopropanes (82).^162,163 Additionally, these cyclopropanations can occur intramolecularly to form products such as 85.¹⁶⁴ Recently, this cyclopropanation chemistry was extended beyond alkenes to alkynes (88), in which a strained cyclopropene (89) was generated.¹⁶⁵ Cyclopropene products (89) are primed to undergo a second carbene transfer to form a bicyclobutane product (90) or be derivatized with chemical methods (Fig. 7d).¹⁶⁵

These reactions are generally performed in whole E. coli cells, requiring no purification of the active catalysts. Additionally, these reactions can typically be performed on gram-scale and have been applied to the syntheses of several cyclopropane-containing natural products and pharmacologically-relevant molecules, such as the TRPV1 inhibitor (86).^166,167

Carbene insertion into C–H bonds.

Beyond cyclopropanation, these engineered haemoproteins can perform insertion into C–H bonds to form various types of products (Fig. 7e). In 2016, Hartwig and co-workers found that replacing the iron in the P450 or myoglobin haem with iridium promoted carbene insertion into C–H bonds to form products such as 91.^168,169 This demonstrated the ability to alter the reactivity of an enzyme by altering the catalytic metal ion without the need for extensive evolution. Shortly after these initial findings, the Fasan group demonstrated that several first-row transitions metals such as cobalt and manganese could inflict similar changes in reactivity to enable C–H insertions.¹⁷⁰

More recently, several groups have demonstrated the ability to perform sp³ C–H alkylations on unactivated substrates using evolved P411 or myoglobin variants containing the native iron-haem cofactor.¹⁷¹ These C–H insertion reactions can be performed on a wide range of substrates (92). Further evolution broadened the substrate scope even more to alkylate unprotected indoles (93)^172,173 and N-aryl pyrrolidine substrates (94).¹⁷⁴ For the latter substrates, trifluoroethyl groups were installed through insertion into α-amino C–H bonds.¹⁷⁴ These reactions can either be performed with purified enzyme or in whole E. coli cells expressing the active catalyst as described for the cyclopropanation reactions. Additionally, the scalability and relevance of haemoprotein-catalysed C–H insertions were demonstrated in several chemoenzymatic syntheses, such as in the synthesis of the anti-inflammatory drug indomethacin¹⁷² and enantio-complementary alkaloids.¹⁷¹

Other reactions catalysed by carbene transferases.

Although cyclopropanations and C–H insertions have been the most well-established C–C bond-forming reactions catalysed by engineered carbene transferases, extensive substrate and protein engineering have resulted in the discovery of other reactivity (Fig. 7f). For example, exposure of the diazoester reagent (95) and an allylic sulfide (96) to an engineered myoglobin led to the formation of a sulfur ylide intermediate that then underwent a [2,3]-sigmatropic rearrangement to form 97.¹⁷⁵ Alternatively, exposure of the diazoester reagent (95) and an engineered myoglobin to triphenylphosphine promoted the formation of the proposed intermediate 98, which, when in the presence of an aldehyde (99) could undergo a Wittig reaction to form olefin 100.^176,177

Conclusion and outlook

Enzymes offer an attractive alternative to many small molecule-mediated transformations due to the high levels of chemo-, site-, and stereoselectivity they impart in their reactions. However, a major obstacle to the routine incorporation of biocatalysts in organic synthesis currently lies in the relative lack of diversity and availability of biocatalysts for a given transformation. Despite the surge in the development of new biocatalytic methods over the last several decades, additional biocatalysts need to be developed to increase the library of biocatalysts, particularly for C–C bond formation. Until recently, biocatalytic C–C bond formation has primarily been limited to the use of a small subset of enzymes, limiting the versatility of biocatalytic C–C bond formation. This Review highlights the major advances over the last several years in enzyme discovery and the development of biocatalysts for more versatile C–C bond-forming reactions including alkylation, acylation, oxidative coupling, cyclisation, and carbene transfer reactions. Technological advances have accelerated the discovery of novel biocatalysts, and as more synthetic chemist embrace biocatalytic methods, this area is ripe for development of transformations that will build molecular scaffolds. We anticipate the C–C bond forming transformations outlined in this Review will be enabling for synthetic chemists brave enough to incorporate biocatalysts into their syntheses and direct chemists to additional resources that bridge the educational gap between organic chemistry and biocatalysis.^7,178–180

Glossary terms

Lyase

A class of enzymes characterized by their ability to break or form a new chemical bond using mechanisms other than hydrolysis or oxidation.

Atroposelective

A preference for the formation of a specific stereoisomer (or atropisomer) containing an axis of chirality caused by hindered rotation around a single bond that results in only one of two possible stable conformations.