Activation modes in biocatalytic radical cyclization reactions

Abstract Radical cyclizations are essential reactions in the biosynthesis of secondary metabolites and the chemical synthesis of societally valuable molecules. In this review, we highlight the general mechanisms utilized in biocatalytic radical cyclizations. We specifically highlight cytochrome P450 monooxygenases (P450s) involved in the biosynthesis of mycocyclosin and vancomycin, nonheme iron- and α-ketoglutarate-dependent dioxygenases (Fe/αKGDs) used in the biosynthesis of kainic acid, scopolamine, and isopenicillin N, and radical S-adenosylmethionine (SAM) enzymes that facilitate the biosynthesis of oxetanocin A, menaquinone, and F420. Beyond natural mechanisms, we also examine repurposed flavin-dependent “ene”-reductases (ERED) for non-natural radical cyclization. Overall, these general mechanisms underscore the opportunity for enzymes to augment and enhance the synthesis of complex molecules using radical mechanisms.


General Mechanism
Cytochrome P450 monooxygenases (P450s) are a large superfamily of iron and heme-dependent enzymes that catalyze oxidative transformations of a variety of endogenous and exogenous substrates, including xenobiotic metabolism and biosynthesis of steroids, lipids, vitamins, and natural products (Cochrane & Vederas, 2014;Denisov et al., 2005;Rudolf et al., 2017;Urlacher & Girhard, 2019;Whitehouse et al., 2012). P450s are best-known to catalyze hydroxylation reactions: the insertion of a single oxygen atom into a C-H bond of a substrate. In addition, they facilitate a diverse array of other reactions, including epoxidation, N, S-oxidation, N-, O-, S-dealkylation, C-C bond cleavage, and Baeyer-Villiger-type oxidation. Most recently, they have been employed to catalyze non-natural C-C and C-N bond-forming reactions via carbene and nitrene transfer reactions, making them one of the most favorable enzyme families for chemical synthesis (Bernhardt & Urlacher, 2014;Chen & Arnold, 2020;Fasan, 2012;Wei et al., 2018).
A simplified catalytic cycle of a P450s-catalyzed hydroxylation is shown in Scheme 1. The multistep catalytic cycle starts with binding of a substrate (R-H) to the enzyme active site that induces a spin shift of the ferric iron, allowing Fe III -to-Fe II reduction by a first electron derived from NAD(P)H via the redox Scheme 1. Catalytic cycle of cytochromes P450. R-H stands for the substrate. R-OH is the resulting hydroxylated product (some P450s can use the peroxide shunt pathway to directly produce Compound 0 from the substrate-bound high-spin Fe III state using H 2 O 2 , bypassing the first three catalytic steps).
partners (Mclean et al., 2015). Binding of molecular oxygen to heme-Fe II followed by a second electron transfer and protonation forms the hydroperoxy-ferric complex (Fe III -OOH, Compound 0). Protonation of the terminal oxygen and subsequently loss of a water leads to the formation of a high valent oxo-ferryl π -cation radical intermediate (Fe IV ═O, Compound I) (Rittle & Green, 2010). Compound I can abstract a hydrogen atom from the substrate (R-H) to form a carbon-centered radical species (R • ), which rapidly rebounds with the equivalent of hydroxyl radical (Fe IV −OH, Compound II) to generate the hydroxylated product (R-OH) (Ortiz de Montellano, 2010). Release of the product and re-coordination by water regenerates the ferric resting state of the catalyst. In addition to C-H bonds, Compound I can abstract a hydrogen atom from the O-H bond of phenols or the N-H bond of aniline to yield oxygen-centered or nitrogen-centered radicals (Denisov et al., 2005;Whitehouse et al., 2012). Notably, for certain P450 biocatalysts, the radical rebound step is slower than intramolecular radical rearrangement or diradical recombination (e.g., cyclization, ring expansion and fusion) (red part in Scheme 1) (Guengerich & Yoshimoto, 2018;Tang et al., 2017), allowing the formation of complexity-added products without incorporation of any oxygen atom (Tang et al., 2017;Walsh & Moore, 2019;Walsh & Tang, 2018).

P450-Catalyzed Biosynthesis of Mycocyclosin (3)
The biosynthesis of mycocyclosin (3) is a representative example of a P450-catalyzed radical cyclization (Belin et al., 2009). Two sequential genes are responsible for the biosynthesis of mycocyclosin. The first enzyme is a cyclodipeptide synthase (Rv2275), which catalyzes the dipeptide bond formation of cyclo-Tyr-Tyr (2) using two molecules of tyrosyl-tRNA Tyr (1) as substrates (Vetting et al., 2010). Subsequent coupling of the phenol rings of cyclo-Tyr-Tyr (2) is catalyzed by cytochrome P450 CYP121 (Rv2276) to provide mycocyclosin (3, Scheme 2A) (Belin et al., 2009). A diradical combination mechanism is proposed for the intramolecular cyclization step (Scheme 2B) (Belin et al., 2009). Mechanistically, this occurs via formation of Compound I (Fe IV ═O, Scheme 1), which abstracts a phenolic hydrogen atom from 2 to form the tyrosyl radical intermediate 4 and Compound II (Fe IV -OH). Subsequently, a second phenolic hydrogen atom of 4 is abstracted by the resulting Compound II to yield the O-diradical 5, which isomerizes to provide C-diradical 6. The following intramolecular diradical combination of C-diradical 6 forges the new C-C bond 7 and gives the final product mycocyclosin (3) after rearomatization (Belin et al., 2009;Dornevil et al., 2017;Dumas et al., 2014). Notably, P450 CYP121 is also found to be essential for M. tuberculosis growth by in vitro gene knockout studies (Mclean et al., 2008), which makes it an intriguing therapeutic target for antituberculosis (Kishk et al., 2019).
Diradical combination mechanisms were proposed for the cyclization reactions by structural and mechanistic studies (Scheme 3) (Pylypenko et al., 2003;Tang et al., 2017;Walsh & Tang, 2018;Woithe et al., 2007;Zerbe et al., 2002). As shown in Scheme 3, the OxyB-catalyzed C−O−D cyclization starts with the abstraction of a phenolic hydrogen atom from the C-ring of 8 by Compound I, yielding a single radical intermediate 9 and Compound II. Subsequently, Compound II abstracts a phenolic hydrogen atom from the D-ring of 9 to give the O-diradical 10. Delocalization of the phenoxy radical on D-ring to its ortho carbon followed by diradical combination forges the new C-O bond between the C and D ring in 11 , providing 11 after tautomerization. The mechanism of the A−B cyclization process catalyzed by OxyC is similar to the one involved in the biosynthesis of mycocyclosin (Scheme 2B). Recently, an alternative mechanism involving a single radical intermediate was also proposed for the OxyC (Scheme 3) (Forneris & Seyedsayamdost, 2018).

General Mechanism
The mononuclear nonheme iron-and α-ketoglutarate (αKG)dependent dioxygenases (Fe/αKGDs) require iron (II) as metallocofactor and α-KG as co-substrate (Hausinger, 2015;Herr & Hausinger, 2018;Loenarz & Schofield, 2008). Structurally, Fe/αKGD enzymes share a conserved double-stranded β-helix (DSBH) fold that coordinates the Fe center with two histidine residues and one carboxylate from either a glutamic acid or an aspartic acid residue (2-His-1-carboxylate facial triad) (Hegg & Que, 1997). Fe/αKGDs catalyze a variety of oxidative reactions, including hydroxylation, halogenation, cyclization, desaturation, epimerization, C-C bond cleavage, and epoxidation, as such playing an important role in the biosynthesis of secondary metabolites (Gao et al., 2018;Hausinger, 2015;Herr & Hausinger, 2018;Krebs et al., 2007;Loenarz & Schofield, 2008Wu et al., 2016;Zwick & Renata, 2020). The putative mechanism of Fe/αKGDs-catalyzed hydroxylation was shown in Scheme 4 (Hausinger 2015;Krebs et al., 2007;Martinez & Hausinger, 2015). The catalytic cycle starts with the binding of co-substrate α-KG to the Fe II center, during which two of the three metal-bound water molecules are replaced. Upon binding of the primary substrate (R-H) to the enzyme active site, the third metal-bound water is removed, allowing the binding of molecular oxygen to form a Fe III -superoxo intermediate. The distal oxygen atom of the Fe III -superoxo species attacks C2 of α-KG to yield a peroxohemiketal bicyclic intermediate, followed by oxidative decarboxylation to release CO 2 and provide a Fe IV -oxo species (also termed as the ferryl intermediate).
Like Compound I in P450s-catalyzed hydroxylation, this ferryl species (Fe IV ═O) abstracts a hydrogen atom from the primary substrate (R-H) to generate a radical intermediate (R • ) and the Fe III -OH species. The radical (R • ) can rebound with the hydroxyl radical to give the final product (R-OH), with concomitant formation of Fe II . After the release of the hydroxylated product (R-OH) and succinate, the resting state (Fe II ) of Fe/αKGD is regenerated by re-coordinating with three water molecules, thus completing the catalytic cycle (Hausinger 2015;Martinez & Hausinger 2015). Notably, in some cases, instead of the hydroxyl radical rebound step, the substrate centered radical (R • ) can undergo a competing intramolecular radical cyclization to form the cyclic products (red part in Scheme 4) (Tang et al., 2017;Walsh & Moore, 2019;Walsh & Tang, 2018).
for the treatment of Ascaris infections for centuries in Asia (Higa & Kuniyoshi, 2000). Kainic acid (17), a cyclic analog of l-glutamic acid, was identified as a potent ionotropic glutamate receptor (iGluR) agonist and serves as an important pharmacological tool in many neurophysiological studies (Lodge, 2009;Werner et al., 1991;Zheng et al., 2011). Since its discovery in the 1950s, the interesting structural features and important biological activities of kainic acid have attracted the attention of synthetic chemists, leading to the development of numerous synthetic routes (Stathakis et al., 2012). Recently, a concise two-enzyme (KabA and KabC) biosynthetic pathway was reported by Moore and coworkers (Chekan et al., 2019). Specifically, N-prenylation of l-glutamic acid (15) with dimethylallyl pyrophosphate (14) is catalyzed by a N-prenyltransferase KabA to provide the prekainic acid (16), which then undergoes an oxidative cyclization catalyzed by a Fe/αKGD enzyme KabC to form the final product kainic acid (17, Scheme 5A) (Chekan et al., 2019). As shown in Scheme 5B, a mechanism involving radical cyclization was proposed for the KabC-catalyzed C-C bond forming step (Chekan et al., 2019 (17) is generated via either a hydrogen atom transfer pathway (route a, Scheme 5B) or an oxidation/deprotonation pathway (route b, Scheme 5B) (Chekan et al., 2019;Dunham et al., 2018).
species and a valinyl radical (34). The valinyl radical then attacks the coordinated sulfur atom to provide the thiazolidine ring of the final product isopenicillin N (28) and restore the metal to the resting Fe II state (Scheme 7B) (Tamanaha et al., 2016).

General Mechanism
In addition to iron-dependent enzymes, nature has evolved another elegant approach to perform radical cyclization reactions using a highly reactive 5 -deoxyadenosyl radical (5 -dAdo·) (Scheme 8). A small group of adenosylcobalamin (AdoCbl)dependent enzymes (Brown, 2005;Matthews, 2009) and a more recently recognized but much larger class of radical SAM enzymes (Broderick et al., 2014;Brown, 2005;Challand et al., 2011;Frey & Booker, 2001;Marsh & Román Meléndez, 2012;Matthews, 2009) are both able to initiate radical reactions by abstracting a hydrogen atom from a C-H bond of the substrate. The general mechanisms of radical cyclization reactions catalyzed by these two classes of enzymes are shown in Scheme 8. The Co-C bond in AdoCbl has a relatively low bond dissociation energy (BDE) of ∼30 kcal/mol (Yokoyama & Lilla, 2018) and the corresponding 5 -dAdo· is usually generated by direct homolytic cleavage. In contrast, the BDE of the C-S bond in SAM is much higher (∼60 kcal/mol) (Yokoyama & Lilla, 2018). As a result, the correspond-ing 5 -dAdo· could only be formed through a single-electron reduction, commonly facilitated by a reduced iron-sulfur cluster. The 5 -dAdo· formed in the enzyme active site can engage in a hydrogen atom transfer (HAT) with an enzyme-bound substrate (R-H) to yield 5 -deoxyadenosine (Ado) and a radical intermediate (R·) which can undergo a cyclization reaction. Such reactions catalyzed by radical SAM enzymes are independent of molecular oxygen and can be either oxidative or reductive quenched, forging a new C sp3 -C sp3 or C sp3 -C sp2 bond, respectively. The BDE for the 5 -dAdo· C5 -H bond is 94-101 kcal/mol (Luo, 2003), which is higher than sp 3 (t-butyl C-H BDE: 96.5 kcal/mol) or activated sp 2 (benzylic C-H BDE: 90 kcal/mol, C-H bond α to ether: 92 kcal/mol) C-H bonds. This renders these enzymes as powerful catalysts in biosynthetic pathways to install C-C bonds in unconventional positions and provide natural products and cofactors with great structural diversity.

Cobalamin and SAM-Containing Enzymes in the Biosynthesis of Oxetanocin A
Enzymes only dependent on a Cobalamin cofactor catalyze various radical rearrangements (Banerjee, 2003;Wolthers et al., 2008) or elimination reactions (Sandala et al., 2006;Wetmore et al., 2002), but rarely radical cyclizations. Enzymes that rely on both cobalamin and SAM cofactors facilitate radical cyclizations in complex molecule biosynthesis. For instance, in the biosynthesis of oxetanocin A (OXT-A), an antiviral with a unique Scheme 15. Proposed mechanism of the biocatalytic radical cyclization of alkyl iodides by photoexcited flavin proteins.
four-membered ring structure (Nakamura et al., 1986;Shimada et al., 1986), a cobalamin-dependent radical SAM enzyme OxsB is proposed to catalyze the key ring contraction step from 2 -dAMP (36) to dehydro-OXT-A phosphate (37) (Scheme 9) (Bridwell-Rabb et al., 2017). The reaction initiates by single-electron shuttling from Co(I) to the iron-sulfur cluster [4Fe-4S] 2+ , which generates Co(II) and [4Fe-4S] + . Subsequent reduction of SAM by [4Fe-4S] + forms the active 5 -dAdo·, which abstracts the H2 in 2 -dAMP (36). 36 then undergoes C3 -C4 bond cleavage, followed by a radical cyclization reaction in which the resulting C4 radical attacks C2 to form the four-member ring intermediate 42. It is proposed that Co(II) could then act as an electron acceptor and oxidize 42 to provide 37, which regenerates Co(I) and completes the catalytic cycle. OXT-A is subsequently formed after reduction and hydrolysis.

SAM-Containing Enzymes in the Biosynthesis of Menaquinone
In terms of SAM-containing enzyme-catalyzed radical cyclization reactions, a representative example is the "futalosine pathway" biosynthesis of menaquinone (Hiratsuka et al., 2008). Menaquinone is a lipid-soluble small molecule that serves as an electron shuttle in the bacterial electron transport chain (Nowicka & Kruk, 2010) and also an essential vitamin in humans (vitamin K2), playing a critical role in blood coagulation (Cranenburg et al., 2007) and bone formation (Plaza & Lamson, 2005). The proposed biosynthetic pathway of menaquinone from futalosine (44) is shown in Scheme 10 (Cooper et al., 2013). Once formed, futalosine (44) is converted to dehypoxanthine futalosine (DHFL, 45) by the hydrolase MqnB, followed by a radical cyclization that provides cyclic dehypoxanthine futalosine (CDHFL, 46) by the radical SAM enzyme MqnC. The 5 dAdo· formed in MqnC abstracts the H4 in 45 through HAT and provides intermediate 47. The resulting carbon-center radical attacks the phenyl ring at the position para to the carboxylate to form a C-C bond, generating intermediate 48. Further oxidation and deprotonation of 48 afford CDHFL, which is the precursor of metaquinone.

SAM-Containing Enzymes in the Biosynthesis of F420
SAM-containing enzymes also catalyze radical C-N bond formation as in the case of F420 biosynthesis. F420 is a naturally occurring deazaflavin cofactor in which the N5 of the flavin ring is replaced with a methine. It functions as a potent two-electron reductant in cells (Walsh, 1986). The physiological functions of F420 and F420 dependent enzymes include anti-TB prodrug activation (Singh et al., 2008), resistant to oxidative stress (Purwantini & Mukhopadhyay, 2009), and biosynthesis of clinically important natural products (Coats et al., 1989;Nakano et al., 2004). The precursor of F420, F0 is formed by a reaction between 5-amino-6-ribitylamino-2,4-pyrimidinone (ARP) and tyrosine that is cat-Scheme 16. Representative substrate scope of the biocatalytic radical cyclization of alkyl iodides.
alyzed by an F0 synthase (Decamps et al., 2012). In archaea and cyanobacteria, F0 synthase is encoded by two separate genes cofG and cofH. CofH is a radical SAM enzyme facilitating the formation of CofH product 51 from l-tyrosine and ARP (Scheme 11) (Philmus et al., 2015), Enzyme CofG is another radical SAM enzyme that catalyzes a radical cyclization reaction. Based on one of the proposed mechanism, CofG abstracts a hydrogen atom from the 7-position of 51 to form the C7 radical 54. After tautomerization, the carboncenter radical in 55 attacks the imine N6 and form the C9-N6 bond. The resulting intermediate 56 is then oxidatively quenched and eliminates an ammonia to provide F0 (Mehta et al., 2015).

Design of Biocatalytic Platform For Non-Natural Radical Cyclizations
Inspired by enantioselective radical cyclization reactions existing in nature, Hyster and coworkers sought to develop biocatalytic strategies to realize asymmetric cyclization reactions mediated by non-natural radical intermediates (Beigasiewicz et al., 2018Clayman & Hyster, 2020;Black et al., 2019;Emmanuel et al., 2016;Hyster 2020;Nakano et al., 2019Nakano et al., , 2020Sandoval et al., 2017Sandoval et al., , 2019. As many synthetic radical reactions are initiated via radical dehalogenation, they sought to develop mechanisms to carry out this fundamental mechanism. Inspired by the ability of flavin-dependent DNA photolyase to cleave weak bonds using single-electron reductions, Hyster and coworkers questioned whether substrate promiscuous enzymes would display the same reactivity patterns (Scheme 12) (Brettel & Byrdin, 2010). The group targeted EREDs (Heckenbichler et al., 2018;Toogood & Scrutton, 2018) as attractive scaffolds for the desired reactivity because of their ease of handling, substrate promiscuity, and evolvability render them one of the most ubiquitous families of enzymes in chemical synthesis.

ERED-Catalyzed Reductive Radical Cyclization Reactions
As a model for this reactivity, Hyster and coworkers targeted the development of a biocatalytic radical cyclization of αchloroamides to afford β-stereogenic lactams . The lactam motif is prevalent in medicinally valuable molecules (Vitaku et al., 2014), and the proposed synthesis would be distinct from existing biocatalytic approaches for generating N-heterocycles (France et al., 2016). Although this cyclization is well known in the radical literature, it is plagued by the preferential formation of the hydrodehalogenated and oligomerized product, and there are no known catalytic asymmetric variants (Curran & Tamine, 1991;Hiroi & Ishii, 2000 (Ghisla et al., 1974;Massey et al., 1978;Warren et al., 2012). After investigation, they found that the cyclization occurs effectively when Gluconobacter oxydans ene-reductase (GluER) was used as the catalyst, and the reaction was irradiated with cyan light (497 nm) (Scheme 14). A variety of five-, six-, seven-, and eight-membered lactams with different substituent patterns were readily accessed. UV-vis and transient Scheme 17. Proposed mechanism of the biocatalytic radical cyclization of α-halo-β-amides for the preparation of 3,3-disubstituted oxindoles catalyzed by photoexcited flavin proteins.
absorption spectroscopy established that radical formation occurs via excitation of an electron donor-acceptor complex that forms exclusively within the enzyme active site. This enzyme templated complex has a broad absorption band at λ = 500 nm, accounting for the reaction's wavelength preference. This represents a novel biocatalytic electron transfer mechanism that is distinct from the initially envisioned mechanism. In addition to α-chloroamides, this strategy could also be applied to unactivated alkyl iodides (Clayman & Hyster, 2020). In contrast to α-halocarbonyl compounds, which possess comparably low reduction potentials and produce electrophilic radicals, unactivated alkyl iodides are more challenging to reduce and generate nucleophilic radicals. Similar to the case with α-chloroamides, Hyster and coworker found that upon binding to the protein active site, the substrate forms a charge-transfer complex (CT complex) with the fully reduced FMN hq . Photoexcitation of this CT complex facilitates the electron transfer between the alkyl iodide substrate and FMN hq , generating the primary alkyl radical, which involves the following radical cyclization process (Scheme 15). A variety of esters, amides, and ketones with an α-chiral center are efficiently synthesized. The reaction accommodates different substituents at the α-position, including alkyl substituents, an acetamide, an alkoxyl group, a fluorine atom, etc. In addition to 5-exo-trig cyclization to form a five-membered ring, 6-exo-trig cyclization could also be realized to provide a tetrahydropyran ring (Scheme 16).

ERED-Catalyzed Redox-Neutral Radical Cyclization Reaction
Hyster and coworkers developed a redox-neutral radical cyclization process to prepare 3,3-disubstituted oxindoles from the transformations discussed before, which are reductive radical cyclization reaction α-halo-β-amides (Black et al., 2019). 3,3-Disubstituted oxindoles are prevalent in medicinally valuable molecules, and there are no known methods for rendering this radical cyclization asymmetric (Ju et al., 2012;Zhou et al., 2010). The reaction was catalyzed by a EREDs (12-oxophytodienoate reductase, OPR1) and facilitated by cyan light. tolerates a variety of substituents at the α-position of the amide (Scheme 18). Several ester substituents are accepted as well. Substrates with electron-donating and electron-withdrawing groups on the aromatic ring undergo the transformation successfully. When the electron-withdrawing ester group is removed from the substrate, the desired oxindole product is still observed.

Conclusion
Insights as to how nature facilitates radical chemistry on structurally complex molecules in a highly selective manner highlight the prospect for enzymes to address fundamental challenges in the synthetic literature. These opportunities lie in the development of new enzyme enabled syntheses, where the synthesis of complex molecules and their analogs can be streamlined through the inclusion of enzymatic steps. Alternatively, by identifying the general strategies that nature uses to form and harness radical intermediates, new small molecule or enzymatic catalysts can be developed, which take inspiration from their analogs in nature. The fingerprints of enzymatic inspiration can be seen in the development of new photoenzymatic systems for radical reactions. Alternatively, the advent of small molecule hydrogen-bonding catalysts, which bind to radicals in similar strategies to enzymes, highlights how developments in one area of synthesis can spur innovations in another. We are optimistic that further discoveries in the biocatalytic radical cyclization arena will have broad implications in chemical synthesis.

Funding
This work was financially supported by the NIGMS (R01 GM127703).