Enzymatic halogenation and dehalogenation reactions: pervasive and mechanistically diverse

Abstract

Naturally produced halogenated compounds are ubiquitous across all domains of life where they perform a multitude of biological functions and adopt a diversity of chemical structures. Accordingly, a diverse collection of enzyme catalysts to install and remove halogens from organic scaffolds has evolved in nature. Accounting for the different chemical properties of the four halogen atoms (fluorine, chlorine, bromine, and iodine) and the diversity and chemical reactivity of their organic substrates, enzymes performing biosynthetic and degradative halogenation chemistry utilize numerous mechanistic strategies involving oxidation, reduction, and substitution. Biosynthetic halogenation reactions range from simple aromatic substitutions to stereoselective C-H functionalizations on remote carbon centers and can initiate the formation of simple to complex ring structures. Dehalogenating enzymes, on the other hand, are best known for removing halogen atoms from man-made organohalogens, yet also function naturally, albeit rarely, in metabolic pathways. This review details the scope and mechanism of nature’s halogenation and dehalogenation enzymatic strategies, highlights gaps in our understanding, and posits where new advances in the field might arise in the near future.

TOC Image

1. Introduction

Halogens are biologically available in the environment as halides (F⁻, Cl⁻, Br⁻, and I⁻). A remarkable diversity of halogenated organic natural products are synthesized in nature ranging from single-carbon halogenated methanes, simple halogenated aromatic phenols, more complex terpenoids, polyketides, oligopeptides, and alkaloids, all the way to some of the most complex and synthetically challenging to access molecules such as palau’amine.^1,2 The majority of organohalogen natural products are brominated or chlorinated with over 1,000 examples each. Fewer examples of iodinated compounds, and even rarer examples of fluorinated compounds, currently exist.² While chlorination is represented in both marine and terrestrial environments, bromination is almost exclusively a feature of marine natural products.^2,3 Remarkably, despite the fact that brominated natural products outnumber their chlorinated counterparts, chloride occurs in seawater at a concentration nearly three orders of magnitude greater than that of bromide at 500 mM to 0.9 mM. Biological fluorination is exceedingly rare despite the great natural abundance of fluorine, which is likely due to the extreme electronegativity of fluorine and the high enthalpic cost associated with fluoride desolvation.^4,5 On the other hand, the scarcity of organoiodine natural products (~180 reported),⁶ that still outnumber their organofluorine counterparts by over tenfold, is explained by the extremely low natural availability of iodide.^4,7,8

For natural products that are halogenated, halogenation can be of critical importance to their bioactivities. This includes both, activities of halogenated natural products in their native physiological context, as well as bioactivities that make halogenated natural products attractive pharmaceuticals. For example, in a natural physiological context, activity of the thyroid hormones triiodothyronine (1, Figure 1) and prohormone, thyroxine (2) is critically dependent on the number of iodine atoms appended to the bicyclic core.⁹ As a therapeutic, the antibiotic vancomycin (3) exhibits a marked drop in bioactivity following the removal of one or both of the chlorine atoms decorating its glycopeptide core.¹⁰ Halogenation plays a direct role in the mechanism of action of the chlorinated anticancer candidate salinosporamide A (4), which forms a covalent adduct with the active site threonine residue of the 20S proteasome facilitated by a chlorine as a leaving group.^11,12 In addition, halogen identity also affects the potency of halogenated molecules. For example, replacing the two chlorines of balhimycin (5), a differently glycosylated variant of 3, to bromines profoundly alters its antimicrobial profile.¹³ Another well-known example of a clinically used halogenated natural product is the antibiotic chloramphenicol (6).¹⁴ Furthermore, fluorination, while exceedingly rare in biology, is a well-proven strategy for increasing the efficacy of pharmaceuticals.^7,15 Halogenation strategies involved in the biosynthesis of the abovementioned natural products, among several others, will be discussed as a part of this review.

Figure 1.

Examples of halogenated natural products relevant to human health and disease.

In addition to naturally produced halogenated molecules, the environment also bears a heavy load of halogenated anthropogenic molecules that are often recalcitrant to degradation and are toxic to humans and wildlife. Indeed, all molecules on the list of persistent organic pollutants regulated by the Stockholm convention, are polyhalogenated. As a result, degradation and bioremediation of organohalogen pollutants has brought renewed interest to studies that have suggested a native role for microbial dehalogenation as part of a biogeochemical halogen cycle in marine and terrestrial environments.^16,17 A large number of biocatalysts capable of removing halogens from anthropogenic compounds through processes coupled to respiration or carbon metabolism have been described. However, in contrast to the abundance of reported halogenated natural products, the biochemical transformations leading to their natural formation and degradation remain largely unexplored. While the link between biological halogenation and dehalogenation may be tenuous, the authors of this review hope that the inclusion of a brief overview of dehalogenation will stimulate the investigation of the cognate process of dehalogenation from the fresh perspective of natural product biosynthesis.

1.1 Historical perspective on natural product halogenation

The first natural product halogenating biocatalyst reported nearly fifty years ago was a hydrogen peroxide (H₂O₂) and heme-dependent chloroperoxidase (CPO) secreted from the terrestrial fungus Caldariomyces fumago, and implicated in the biosynthesis of halogenated cyclopentanediol fungal halometabolite caldariomycin (7, Figure 2A).^18–20 CPO has since been thoroughly studied and even marketed as an electrophilic halogenating reagent with broad substrate specificity.²¹ The structure of the enzyme has been determined (Figure 2A).²² The operative mechanism, as has been reviewed extensively elsewhere,^23–26 is similar to heme-dependent hydroxylases and oxygenases. Briefly, a cysteine thiolate ligated heme Fe(III)-porphyrin is activated by hydrogen peroxide to generate the oxo-Fe(IV) species. In a departure from P450 hydroxylases, nucleophilic addition of a halide would generate the catalytic Fe(III)-hypohalite species. Halogenation can conceivably proceed via two routes: the CPO can bind the substrate proximal to the Fe(III)-hypohalite leading to the formal delivery of the electrophilic halenium ion to an electron rich substrate with, or without, an intermediary transfer of the halenium to an active site residue side chain, a scheme reminiscent of the halogenation strategy employed by flavin-dependent halogenases (vide infra). A second route, that has been favored in literature involves the release of hypohalite by the enzyme to non-selectively halogenates electron-rich substrate centers. This proposed route is also suggested for vanadium-dependent haloperoxidases (V-HPOs) as discussed in Section 3. Control exercised by the enzyme over its catalytic cycle seems to be rather relaxed as the CPO catalyzes a host of other reactions in addition to halogenation, including oxidation, sulfoxidation, epoxidation, and hydroxylation. Whether the physiological role of CPO indeed involves the production of 7, or whether halogenation is a side activity demonstrated by this fairly catalytically promiscuous enzyme are still open questions. As reviewed by Drennan and coworkers, the halide and substrate binding sites for heme-dependent haloperoxidases have not been identified.²⁴

Figure 2. Heme-dependent haloperoxidases in biological halogenation reactions.

(A) Structure (PDB: 2CPO) and mechanism of CPO that participates in the biosynthesis of 7. For the sake of simplicity, throughout the review, crystal structures of only the monomeric units of multimeric enzymes are shown. (B) Halogenation of 8 leads to loss in absorption at 277 nm. The MCD-assay has been widely used to monitor formation of freely diffusible hypohalite by halogenases. (C) Diiodination of two tyrosyl side chains of the thyroglobulin protein, followed by oxidative coupling, β-elimination, rearomatization, and proteolytic cleavage leads to the production of 2. Note that halogenation and bi-radical coupling is postulated to be affected within the same enzyme active site.

The hypothetical release of hypohalite by haloperoxidases can be efficiently monitored spectrophotometrically by utilizing monochlorodimedone (8, MCD, Figure 2B) that exhibits a loss in absorption at 290 nm following electrophilic halogenation.¹⁸ The MCD-assay relies on the ability of a candidate enzyme to oxidize a halide substrate (X⁻ = Cl⁻, Br⁻, and I⁻) to form a diffusible hypohalite species, and hence inherently selects for promiscuous oxidative halogenation enzyme catalysts.²⁷ Utilizing the MCD-assay, numerous haloperoxidases have been discovered, including V-HPOs found in seaweeds some twenty years after the initial report of the first CPO.²⁸

The notion that a freely diffusible hypohalite is released by heme-dependent haloperoxidases as the effective halogenating agent without subsequent control over substrate halogenation is perhaps challenged by the human thyroid epithelial heme-dependent iodoperoxidase that selectively diiodinates two tyrosine side chains (Tyr5 and Tyr130) ortho to the phenoxyl harbored within the 1000-residue thyroglobulin protein. Subsequent oxidative bi-radical coupling of the diiodinated tyrosyl side chains by the same peroxidase, followed by scavenging of the two amino acids from thyroglobulin by dedicated proteases releases 2 that is subsequently deiodinated to its biologically active form 1. The enzymatic deiodination of 2 to furnish 1 is covered in a later section in this review (Section 6).

The paucity of alternative biological halogenation enzymes and strategies discovered over the decades that followed the discovery of haloperoxidases led to the initial assumption that haloperoxidases were the primary halogenating catalysts in the biosynthesis of halogenated natural products.²⁹ In other words, absence of evidence became evidence for absence. Indeed, at the time of the discovery of the first CPO, this argument was helped by the fact that biological halogenation was largely considered to be an accident of biology as only thirty halogenated natural products had been described.³⁰ However, with the subsequent discovery of thousands of bioactive natural products bearing halogens attached to both electron rich and aliphatic scaffolds in regiospecific configurations, it became clear that halogenated natural products had well defined roles in biology and ecology, and that they were biosynthesized by dedicated enzymatic processes.^1,2 Moreover, the discovery of fluorinated natural products, despite the inability of any known haloperoxidase to accept fluoride as a substrate, further challenged the notion that these enzymes were the sole strategy for biological halogenation.⁷ Hence, the broad structural diversity of organohalogen natural products signaled a wealth of substrate-specific biocatalysts utilizing alternative halogenation strategies beyond promiscuous haloperoxidases.²⁷ One of the main obstructions to the discovery of new halogenation biocatalysts has been the inability to link enzymes with their natural substrates. However, progress in DNA sequencing and computational assembly of genomes, particularly bacterial genomes, over the last two decades have led to a renaissance in the discovery of natural product biosynthetic pathways and tailoring enzymes.³¹ This is also reflected in the current inventory of halogenating enzymes that is heavily dominated by bacterial natural product halogenases.

The marine environment has been a prolific source of halogenated natural products.^32,33 A cornucopia of halogenated marine natural products has been discovered, aided in part by the distinctive isotopic signature associated with chlorine and bromine atoms.³⁴ Several instances in which marine natural products and the mechanisms for their production have led to advancement in halogenation biochemistry will be highlighted throughout this review.

1.2 Strategies for natural product halogenation

Halogenases exploit three oxidation states of halogen atoms, and can thus be divided into three broad groups. The first group that will be discussed in this review are halogenases that oxidize the halide (X⁻) to halenium (X⁺),³⁵ formally envisaged as hypohalite (XO⁻) in aqueous media. The halenium ion in turn catalyzes electrophilic substitution reactions on electron-rich substrates, the majority of which are aromatic. The second class of enzymes catalyzes 1e⁻ oxidations of the halide to the X^• radical and can thus catalyze halogenation on unactivated aliphatic carbon centers, a transformation that is difficult to realize in synthetic organic chemistry. Both the above classes of oxidative enzymes unsurprisingly cannot utilize fluoride due to its electronegativity. Fluoride incorporation, among other halides, falls under the domain of the third class of halogenases that utilize halides as nucleophiles to catalyze S_N2 substitutions on electron deficient carbon centers. Hence, it is evident that the wide inventory of halogenating catalysts has evolved to serve a variety of substrates. All three classes of halogenases are discussed with examples presented for their participation in natural product biosynthesis. The review also emphasizes the progression of the structure-mechanism studies for halogenases and the remaining mechanistic questions for these fascinating catalysts. Structural and biochemical characterization of natural product halogenases, combined with biomolecular, chemical, and process engineering promises to transcend these biological catalysts to laboratory and industrial small molecule synthetic schemes. Recent engineering and development efforts with halogenases are also presented. The review covers literature as current to 2016 with a focus on natural product biosynthesis.

2. Flavin-dependent halogenases

Perhaps the most extensively characterized class of halogenating enzymes is the flavin adenine dinucleotide (FAD) dependent halogenases (FDHs). FDHs are ubiquitous across all domains of life. FDHs catalyze the 2e⁻ oxidation of the halide to the hypohalite, and thus, for the most part, the activity of FDHs is restricted to aromatic substitution reactions. The discussion regarding FDHs, in this review, is divided to reflect the diversity of their substrates and their participation in natural product biosynthetic schemes. Despite the diversity of substrates and the wide repertoire of natural product biosynthetic reactions they catalyze, FDHs follow a conserved mechanism involving halide oxidation followed by regiospecific delivery of the halenium to aromatic substrates.

2.1 Structure and mechanism

Mechanistically and structurally, FDHs resemble FAD-dependent oxygenases (Figure 3). Multiple routes of the redox biochemistry associated with the flavin-cofactor in natural product biosynthetic schemes have been reviewed recently.^40,41 Over the last decade, a consensus mechanistic proposal for FDHs has emerged based primarily on the biochemical characterization and crystal structures of L-tryptophan FDHs. Briefly, a 2e⁻ transfer from NAD(P)H to an oxidized flavin cofactor (Fl_ox), catalyzed by a flavin-reductase, generates the reduced flavin cofactor (Fl_red, Figure 4). The flavin-reductase may, or may not, be encoded with the FDH in the natural product biosynthetic gene cluster. To date, FDHs have not been shown to make specific contacts with the flavin-reductase and the activity of numerous FDHs has been reconstituted using the E. coli flavin-reductases SsuE⁴² and Fre⁴³ as reaction partners. Fl_red diffuses to the FDH active site where it reacts with molecular oxygen (O₂) to generate the Fl-C4a-OOH transient species. Until this stage, the reaction scheme for FDHs and FAD-dependent oxygenases is indistinguishable. FAD-dependent oxygenases position the aromatic substrate such that the distal oxygen of the Fl-C4a-OOH, which can also be thought of as the electrophilic ‘OH⁺’ equivalent, is directly transferred to hydroxylate electron-rich aromatic substrates, thereby leaving the Fl-C4a-OH species to resolve via the loss of water to regenerate Fl_ox. FDHs use the cosubstrate halide anions (X⁻, X = Cl, Br, I) to attack the distal Fl-C4a-OOH oxygen atom, leading to the scission of the labile oxygen-oxygen single bond to generate the electrophilic hypohalite and the Fl-C4a-OH. The hypohalite is thus the effective halenium delivering species. Each of the three flavin-cofactor states, Fl-C4a-OOH, Fl-C4a-OH, and Fl_ox, have been spectroscopically detected in FDHs.⁴⁴ The kinetics for the formation of these species has been shown to be unaffected by the addition of the substrate, suggesting that substrate engagement either occurs after the flavin redox chemistry, or in a pocket distant to the cofactor. These findings are in concert with the substrate binding site as identified first in the co-crystal structure of the tryptophan-7-chlorinase PrnA,³⁶ and corroborated since with the co-crystal structures of tryptophan-7-chlorinase RebH,^45,46 and tryptophan-5-chlorinase PyrH.⁴⁷ In each of these crystal structures, the substrate binds in a pocket distant to the site of generation of the hypohalite. The hypohalite is not freely diffusible from the enzyme active site, but is rather delivered regiospecifically to the aromatic substrate via the formation of a haloamine species at the side chain primary amine of a highly conserved lysine residue, the mutation of which renders the FDH catalytically inactive (Figure 4).⁴⁵ Whether the hypohalite is delivered via a covalent haloamine intermediate (as shown in Figure 4), or a tightly coordinated hypohalous acid,⁴⁸ is presently not clear. Furthermore, the dynamics via which the hypohalite reaches the lysine side chain, while not halogenating the two sterically proximal and highly conserved tryptophan side chains is also not fully understood. While a halide binding site has been identified at the re face of the cofactor isoalloxazine ring, whether this a productive site for halide binding has been debated in the literature.^24,40

Figure 3. Structural similarity of FDHs with flavin-dependent oxygenases.

(A) Overlay of crystal structures of FDH PrnA (PDB: 2AQJ) (in blue) and kynurenine-3-monooxygenase (PDB: 5FN0) (in magenta). The structures were overlaid so as to align the cofactor FAD isoalloxazine rings. FAD is shown in stick-ball representation. Chloride ions bound to PrnA is shown as an orange sphere. L-Trp (bound to 2AQJ) and kynurenine-3-monooxygenasae inhibitor (GSK180, bound to 5FN0) are shown as spheres with carbon atoms colored yellow. (B) Zoomed in view of the active sites, with the PrnA catalytic lysine (vide infra) shown in stick-ball representation with carbon atoms colored yellow. (C) Crystal structure of FDH Bmp2 (PDB: 5BVA) (in green) overlaid with 4-hydroxybenzoate hydroxylase (PDB: 1YKJ) (in cyan). 4-hydroxybenzoate (PHBA, bound to 1YKJ) is shown as spheres with carbon atoms colored yellow. (D) Zoomed in view of the active sites, with the Bmp2 catalytic lysine (vide infra) shown in stick-ball representation with carbon atoms colored yellow.

Figure 4. Reaction mechanism for FDHs.

(A) The reaction scheme for FDHs and flavin-dependent oxygenases is identical until the formation of Fl-C4a-OOH. FDHs resolve this intermediate by the displacement of the distal oxygen atom by the halide anion, from where the hypohalite is captured by the catalytic lysine side chain by the formation of a haloamine intermediate, finally affecting an electrophilic aromatic substitution on an electron-rich substrate. (B) The relative positioning of the flavin cofactor isoalloxazine ring and the FDH catalytic lysine side chain as identified in the recently described crystal structure of the FDH Bmp2 (PDB: 5BVA). In the Bmp2 crystal structure, an ethylene glycol (EG) molecule was found in the vicinity of the cofactor proximal to the two tryptophan amino acids that are conserved in all FDH sequences. The dashed line represents the distance of 2.7 Å between the isoalloxazine N5 and one of the EG oxygen atoms. The EG binding site likely represents the region where Fl_red engages molecular oxygen, finally leading to the formation of the hypohalite that is then transferred to the lysine side chain.

The above discussion is restricted to two-component FDHs, which, just like two-component flavin-dependent oxygenases, require the flavin-reductase as a reaction partner to transfer two electrons from NAD(P)H to Fl_ox and generate Fl_red. However, flavin-dependent oxygenases also occur as ‘stand-alone’ single-component systems that have a distinct NAD(P)H binding domain and can catalyze the reduction of Fl_ox to Fl_red, and the subsequent substrate oxygenation all within the same polypeptide. To date, only one single-component FDH has been reported with the discovery of Bmp5 from epiphytic marine bacteria of the genera Pseudoalteromonas and Marinomonas.^49,50 Bmp5 catalyzes the bromination of 4-hydroxybenzoate to 3-bromo-4-hydroxybenzoate, followed by decarboxylative-bromination to yield the marine natural product 2,4-dibromophenol (9, Figure 5A). This two-step reaction scheme for Bmp5 combines the activities of two previously reported single-component flavin-dependent oxygenases: 4-hydroxybenzoate hydroxylase⁵¹ that yields 3,4-dihydroxybenzoate as the product (Figure 5B), and 4-hydroxybenzoate decarboxylative hydroxylase⁵² that yields hydroquinone as the product (Figure 5C). Discrete single-component flavin-dependent hydroxylases that hydroxylate distinct positons of the phenoxyl ring derived from 4-hydroxybenzoate are also represented in microbial production of ubiquinone.⁵³ Just as for the single-component flavin-dependent oxygenases, Bmp5 does not require the presence of a flavin reductase for in vitro activity.

Figure 5. Resemblance of bromination activity of single-component halogenase Bmp5 to single-component monooxygenases.

(A) The two-step conversion of 4-hydroxybenzoate to 9 catalyzed by Bmp5. (B) Hydroxylation and (C) decarboxylative-hydroxylation of 4-hydroxybenzoate catalyzed by single-component monooxygenases. Aryl-rings bearing the halogen adducts are highlighted in blue in this section.

2.2 Small molecule halogenases

To date, the inventory of FDHs acting on small molecule substrates is populated, for the most part, by bacterial FDHs chlorinating the amino acid L-tryptophan. The tryptophan-7-chlorinase PrnA, encoded within the biosynthetic gene cluster^54,55 for the production of Pseudomonas derived antifungal natural product pyrrolnitrin (10), was the first FDH for which the in vitro biochemical activity was reconstituted using cell free extracts (Figure 6).⁵⁶ The activity of PrnA was subsequently demonstrated using purified enzyme, and the dependence of FDH activity on a flavin-reductase was identified.⁵⁷ PrnA has been crystallized in various catalytic states,³⁶ and together with the crystallographic and kinetic studies with the indolocarbazole antitumor natural product rebeccamycin (11) tryptophan-7-chlorinase RebH,^44,45,58 forms the basis for much of our understanding about FDH mechanistic enzymology. The celebrated discovery, in vitro enzymatic activity reconstitution, and the structural characterization of L-tryptophan halogenases has been reviewed extensively.^{23,24,26,41,59–61} Indolocarbazoles can be converted to indolotryptolines,⁶² and tryptophan halogenases such as AbeH, BorH, and ClaH have been detected in the biosynthetic gene clusters for indolotryptoline natural products BE-54017 (12),⁶³ borregomycin (13),⁶⁴ and cladoniamide (14),⁶⁵ respectively (Figure 6). Apart from indolocarbazoles and indolotryptolines, the tryptophan-5-chlorinase PyrH that participates in the biosynthesis of pyrroindomycin B (15) has been biochemically and structurally characterized, together with the biochemical⁶⁶ and crystallographic⁶⁷ characterization of the tryptohan-6-chlorinase SttH that is involved in an as yet unidentified biochemical pathway in Streptomyces toxytricini. Another tryptophan-6-chlorinase, ThdH, that participates in the biosynthesis of the Streptomyces derived plant growth regulating hormone theinodolin (16) has been characterized,⁶⁸ while the contribution of an unnamed tryptophan-6-chlorinase AORI_5336 has been implicated in the biosynthesis of plant growth-regulating compound LYXLF2 (17) from the actinobacterium Amycolatopsis orientalis that also produces 3.⁶⁹

Figure 6. Regiospecific tryptophan halogenases and their participation in natural product biosynthesis.

The halogenated indole rings are shaded blue. Note that the chlorination on the pyrrole ring of 10 is catalyzed not by PrnA but by PrnC, a second FDH encoded with the pyrrolnitrin biosynthetic gene cluster.

With a repertoire of enzymes that catalyze the regioselective chlorination of tryptophan at three different positions, efforts started to understand the structural bases for this regioselectivity and to engineer tryptophan FDHs as tailored catalysts using mutagenesis and activity screening assays. The reader is directed to several excellent recent reviews dedicated to this topic.^26,70–72 Thus, these efforts are not covered in detail here. Combined with efforts directed towards improving the thermostability and organic solvent tolerance of FDHs,^73,74 enzyme immobilization to affect gram-scale product transformations,⁷⁵ and the development of high throughput assays to screen through mutant libraries⁷⁶ hold promise to push FDHs towards industrially viable catalysts.

Perhaps more relevant from a natural product biosynthesis perspective, a second engineering approach asked the question whether tryptophan FDHs could be used in vivo to generate novel natural products, or intermediate structures to be used in synthetic derivatization via cross-coupling schemes. Regiospecifically halogenating aryl rings can be challenging via purely synthetic means, especially for natural products where other electron rich aryl rings might also be present. Tryptophan FDHs have been successfully used to address these challenges. The Pseudomonas FDH PrnA was conjugated into the genome of the natural product pacidamycin producer Streptomyces coeruleorubidus that led to the production of a halogenated derivative of pacidamycin that was chlorinated only on the indole, and not on the phenyl ring.⁷⁷ Chloropacidamycin was then used as the substrate in cross-coupling reactions with boronic acids to generate a variety of derivatives. In another approach, tryptophan halogenases were expressed in the medicinal plant Catharanthus roseus to drive the production of chlorinated and brominated derivatives of highly complex plant alkaloids⁷⁸ that were then amenable to Suzuki-Miyaura cross-coupling reactions with boronic acid substrates.⁷⁹ A recent report describing the use of regioselective tryptophan halogenases followed by Boc-protection of the main chain amine open the possibility of incorporating halogenated tryptophan intermediates in peptide synthesis and subsequent downstream cross-coupling reactions.⁸⁰

All FDHs discussed above derive from bacterial sources. This observation is not reflective of the lack of halogenative potential among other domains of life, but rather the bias in DNA sequencing efforts. For example, filter feeding benthic marine invertebrates such as sponges and marine algae are prolific producers of halogenated natural products. However, to date, only a solitary halogenating enzyme has been identified from a marine sponge within a genetic purview of halogenated natural product biosynthesis.⁸¹ This bias is slowly starting to shift. FDHs have been found in halogenated fungal natural product gene clusters (Figure 7), such as the Hsp90 inhibitor radicicol (18) chlorinase RadH from Chaetomium chiversi,⁸² also called as Rdc2 from a different producer Pochonia chlamydosporia.⁸³ In a series of in vitro assays, the fungal FDH Rdc2 could be paired with the E. coli flavin-reductase Fre, and in a substrate promiscuous fashion, chlorinate a series of lactones on the electron rich resorcylic ring.^84,85 Substrate promiscuity was also reported for the halogenation of the mushroom toxin melleolide F (19) at the resorcylic acid ring by each of the five different FDHs identified in the mushroom cDNA.⁸⁶

Figure 7.

Structures of eukaryotic halogenated natural products biosynthesized by the action of FDHs.

Identification of the physiological substrates and the precise timing for halogenation in natural product biosynthetic schemes can be challenging, and the abovementioned substrate tolerance hints towards the possibility that the physiological substrates for these FDHs are yet to be identified. Noteworthy here is the observation that tryptophan halogenases also demonstrate broad substrate tolerance, albeit with reduced catalytic efficiencies, towards a wide swathe of electron rich substrates.^87–89 Advances in the identification of the physiological substrate for a fungal FDH was provided by the recent study examining the biosynthesis of chaetoviridin A (20) from Chaetomiun globosum via the caz biosynthetic gene cluster.⁹⁰ The postulated small molecule substrate for the FDH CazI accumulates in the culture extracts when the cazI gene is deleted, and the chlorination of this accumulated small molecule was experimentally demonstrated by the FDH CazI.⁹¹ However, halogenation occurring en route modular polyketide elongation on a substrate tethered to a carrier protein (CP) cannot be entirely discounted based on these findings, as also applicable for the halogenation of the resorcylic ring of the Streotomycete natural product venemycin.⁹² The applicability of using the small molecule substrate to experimentally demonstrate the dichlorinating activity of amoeba Dictyostelium discoideum FDH ChlA⁹³ can be called into question as well, as halogenation on the two electron rich positions can be easily envisaged while the polyketide chain is still bound to the upstream carrier protein to generate the Dictyostelium differentiation inducing factor-1 (DIF-1, 21).⁹⁴

2.3 Halogenases for substrates acylated to carrier proteins

CPs are ubiquitous small proteins that undergo a post-translation condensation of coenzyme A (CoA) derived phosphopantetheine to a highly conserved serine side chain hydroxyl. By virtue of the phosphopantetheine terminal thiol, CPs act as molecular shuttles for carboxylic acids that are transiently acylated to it by the formation of a thioester bond (Figure 8A). CP-mediated modifications are ubiquitous in natural product biosynthesis, particularly as it relates to modular extension of peptide and polyketide chains by non-ribosomal peptide synthetases (NRPSs) and polyketide synthases (PKSs), respectively.

Figure 8. Halogenation of acyl-S-CP substrates.

(A) Post-translational acylation of apo-CP with CoA-derived phosphopantetheine generates holo-ACP that then undergoes thioesterification with L-proline and oxidation of the prolyl heterocycle to generate pyrrolyl-S-CP. The action of six different pyrrolyl-S-CP FDHs with corresponding natural product structures is shown. Note that the fourth bromine atom in 24 is also installed by Bmp2, while the dibromophenol moiety of 29 is generated by aforementioned single-component FDH Bmp5. Note that 30 is not derived from the action of Bmp2 and Bmp5. (B) Representative L-tyrosine derived halogenated natural products. Chlorination of β-hydroxytyrosine by SgcC3 is shown. The timing of halogenation for other natural products shown here has not been experimentally determined.

The first report of a FDH acting on a CP-tethered substrate described the dichlorination of the pyrrole ring en route the production of the pseudomonad antifungal natural product pyoluteorin (22, Figure 8A).⁹⁷ PltA did not chlorinate pyrrole-2-carboxylic acid, thus demonstrating the absolute requirement of the substrate to be acylated to a CP. Subsequently, the activity of the FDH SgcC3 was demonstrated to catalyze the chlorination of the CP-acylated β-tyrosine intermediate in the biosynthetic scheme for the Streptomyces globisporus enediyne antitumor antibiotic C-1027 (23, Figure 8A).⁹⁶ Just like PltA, enzyme SgcC3 only accepted substrates acylated to the CP, though the substrate specificity was relaxed to include both (R)- and (S)-enantiomers of β-tyrosine indicating flexibility in the substrate recruitment site.

Since these reports, the detection of FDHs that work on CP-tethered substrates has blossomed to include the tetrabromopyrrole (24), marinopyrrole A (25), chlorizidine (26), pyrrolomycin (27), and hormaomycin (28) pyrrolyl-S-CP halogenases Bmp2,^38,40,49 Mpy16,⁹⁸ Clz5,⁹⁹ Pyr29,¹⁰⁰ and HrmQ,¹⁰¹ respectively, though the activities for only Bmp2 and Mpy16 have been reconstituted in vitro. Derived from 24, Bmp2 also participates in the biosynthesis of pentabromopseudilin (29).⁴⁹ In the biosynthesis of each of the abovementioned natural products, the CP carries L-proline through the oxidation of the prolyl heterocycle to pyrrole, followed by halogenation of the pyrrole residue by a FDH. This short but highly conserved biosynthetic route, first elucidated for the biosynthesis of 22,^95,97,102 has served as a beacon for genome mining and discovery of the biosynthetic gene clusters for bacterial natural products bearing a halopyrrole moieties, one that is likely to extend further to illuminate the biosynthetic routes for the exceptionally abundant and pharmacologically interesting marine pyrrole aminoimidazole alkaloids.¹⁰³ Furthermore, the abovementioned FDHs can be used to modify other pyrrole containing natural product biosynthetic pathways to generate halogenated derivatives, as was reported for the engineering of the clorobiocin pathway by the introduction of HrmQ.¹⁰⁴ Note that the pyrrole-3-chlorinase PrnC participating in the biosynthesis of 10 is not postulated to follow the CP-mediated route,⁵⁵ while the physiological substrate for the pentachloropsuedilin (30) FDH HalB is yet to be deciphered.¹⁰⁵ Additional tyrosyl-S-CP FDHs have also been discovered, such as the marine proteobacterial brominase AltN participating in the biosynthesis of bromoalterochromides, such as dibromoalterochromide A (31).^106,107 The myxobacterial chlorinase CndH, that likely chlorinates the tyrosyl ring in chondrochlorens has been identified with the crystal structure of the enzyme determined as well.^108,109 However, the physiological substrate and the timing of halogenation remains to be elucidated. The cyanobacterial FDH AerJ that participates in the biosynthesis halogenated congeners of aeruginosins, such as aeruginosin 101 (32), has been postulated to act on a tyrosyl-S-CP substrate^110,111 and will be discussed in a later section together with FDHs that encode halogenation of other cyanobacterial products such as cryptophycin 1 (33), cyanopeptolin 954 (34), and anabaenopeptilide 90-B (35, Figure 8B).

While numerous crystal structures of acyl-S-CP utilizing FDHs have been reported (PltA,¹¹² Mpy16,³⁸ Bmp2,³⁸ CndH¹⁰⁸), key mechanistic questions remain unresolved. Primary among these is the catalytic base that abstracts the proton to rearomatize the product during the electrophilic aromatic substitution by the halenium. Based on the structure of PrnA,³⁶ a conserved glutamate residue side chain carboxylate was proposed to serve this role (PrnA-Glu346). While alternate roles for PrnA-Glu346 have been proposed in coordinating hypohalous acid in conjunction with PrnA-Lys79,⁴⁸ no glutamate or aspartate residues are located in analogous positions in the PltA, Myp16, Bmp2, or CndH crystal structures (Figure 9). In these pyrrolyl-S-CP halogenases, a phenylalanine residue replaces the PrnA glutamate. In PrnA, a Glu→Gln mutation of the proposed catalytic base leads to reduction, rather than abolishment of enzymatic activity, while a Glu→Asp mutation led to complete loss of enzymatic activity without changing the enzyme structure or the substrate binding site.^36,48 Whether acyl-S-CP FDHs, that in fact possess retarded turnover numbers, pay this kinetic penalty remains to be experimentally elucidated. Considering that the active site is deeply buried within the enzyme core, an earlier proposal that the catalytic base can be donated by the CP¹⁰⁸ seems rather unlikely without invoking large scale conformational changes in the halogenase structure.

Figure 9. Proposed CP binding site in acyl-S-CP halogenases.

Active site of L-tryptophan chlorinase PrnA and pyrrolyl-S-CP halogenases Mpy16 and Bmp2 demonstrating conserved positioning of the FAD cofactor and the postulated haloamine bearing Lys side chain. However, the catalytic base in PrnA- Glu346, is replaced with Phe323 and Phe300 in Mpy16 and Bmp2, respectively. The product, 7-chlorotryptophan is also shown in the PrnA structure to identify the substrate recruitment site.

It should be noted that no crystal structures of acyl-S-CP FDHs have been reported in complex with their CP-tethered substrates. While the requisite positioning of the CP relative to the FDH has not been experimentally determined, the recently reported crystal structures of PltA,¹¹² Mpy16,³⁸ and Bmp2³⁸ allow us to advance a postulate. In each of the three crystal structures, a positively charged region on the surface of the enzyme (in blue, Figure 10) can be identified in a concave cavity above the FAD binding site. Electrostatic recruitment of acidic CPs to positively charged surfaces on enzymes is well documented.^113,114 Furthermore, in the crystal structure of Mpy16, starting from this concave cavity, a chain of water molecules can be identified that likely represent the path traversed by the phosphopantetheine arm of the acyl-S-CP. This water filled cavity terminates at the reaction site that is defined by the catalytic haloamine bearing lysine side chain (Mpy16-Lys72) and is positioned analogously to the tryptophan binding site in the reported PrnA co-crystal structure (Figure 3).³⁶

Figure 10. Proposed CP binding site in acyl-S-CP halogenases.

Semi-transparent electrostatic surface representation of pyrrolyl-S-CP halogenases PltA (PDB: 5DBJ), Bmp2 (PDB: 5BVA) and Mpy16 (PDB: 5BUK) showing conserved positioning of a positively charged (in blue) concave cavity in the vicinity of the FAD cofactor (in stick-ball representation, carbon atoms colored yellow). In Mpy16, a chain of water molecules (in red) starting at the positively charged cavity traverses the FAD binding site terminating at the halogenation active site defined by the catalytic lysine side chain.

The acyl-S-CP FDHs, such as AltN, PltA, Mpy16, Clz5, and Bmp2, among others, catalyze multiple halogenations upon their CP-loaded aromatic substrates. Recently, the remarkable tetrahalogenating activity of the marine bacterial FDH Bmp2 was reported.³⁸ Bmp2 brominates all four carbon atoms of the CP-tethered pyrrole ring setting up a thioesterase mediated hydrolytic release of a 2,3,4,5-tetrabromo-pyrrole-2-carboxylic acid that undergoes a non-catalytic decarboxylation to yield the coral settlement chemical cue 24 (Figure 8A).¹¹⁵ Multiple halogenations, thought of as relaxation in the regiospecificity for halogenation by acyl-S-CP FDHs, is in stark contrast to small molecule halogenases such as PrnA,³⁶ PyrH,¹¹⁶ RebH,⁵⁸ and SttH⁶⁷ that display remarkable regiocontrol in halogenating tryptophan.⁴⁷ While it is not clear why acyl-S-CP halogenases have lost the regiocontrol exercised by tryptophan FDHs, engineering the Bmp2 enzyme via structure guided site directed mutagenesis has revealed that halogenation regiocontrol in both classes of FDHs is encoded within the halogenase substrate binding active site.^38,47,67,117 In light of the above discussion, an unresolved mechanistic question presents itself. Are multiple halogenation reactions catalyzed by acyl-S-CP FDHs processive, in that, does the acyl-S-CP substrate remain bound to the enzyme during successive halogenation events, or distributive, so that the acyl-S-CP substrate is released by the enzyme and has to find the enzyme active site again for the subsequent halogenation? Such a question has also been asked for other natural product biosynthetic enzymes such as the lanthipeptide dehydratases and cyclases that catalyze multiple reactions on spatially distinct sites on a substrate peptide.^118–121 Regeneration of the acyl-S-CP FDH between subsequent halogenations would minimally require the following events to occur: (i) loss of FAD, (ii) diffusion of FADH₂ back into the FDH active site, (iii) reengagement of O₂, and (iv) binding of the halide anion. The abovementioned hypothesis for the site for CP binding proximal to the FAD binding site precludes that FAD and FADH₂ can be exchanged while the CP is still bound, though a different site for exchange of FAD/FADH₂ cannot be ruled out. Moreover, reduction of O₂ at FAD-C4a, followed by the oxidation of the halide and generation of the postulated lysine side chain haloamine can be impeded by the phosphopantetheine arm of the acyl-S-CP substrate that would sterically hinder these processes to occur. Hence, it is likely that acyl-S-CP FDHs reside as ‘cocked guns’ with the reduction of O₂, oxidation of the halide, and generation of the haloamine occurring prior to substrate engagement. In an event of productive substrate binding, likely a thermodynamically controlled process, substrate halogenation would occur. Disengagement of the CP would then allow for FAD to be exchanged for FADH₂, and the enzyme to be regenerated for the next catalytic cycle. This hypothetical order of events is in line with prior stopped flow kinetic characterization of the RebH-catalyzed chlorination of tryptophan where the presence or absence of substrate did not influence the dynamics of the FAD redox chemistry, and that the substrate chlorination occurred subsequent to the flavin redox reactions.⁴⁴ Clearly, a co-crystal structure of an acyl-S-CP FDH in complex with its physiological substrate will go a long way in verifying or refuting the above hypotheses. Advances in the biosynthesis of acyl-coenzyme A thioesters,^122,123 mass spectrometry based analytical techniques¹²⁴ that circumvent alkaline offloading of substrates and products from acyl-S-CPs prior to the characterization, and synthesis of molecular probes to interrogate flavin-dependent enzymes¹²⁵ will undoubtedly aid in answering these mechanistic conundrums that have persisted for more than a decade.

2.4 Halogenases acting en route modular elongation

Each of the acyl-S-CP FDHs discussed in the previous section, with the exception of Bmp2 and SgcC3, halogenate the initiating building block for the biosynthesis of natural products via downstream NRPS or PKS modular assembly line pathways. For example, the 5-chloropyrrolyl-S-CP product of HrmQ is incorporated into the downstream NRPS derived scaffold of 28,¹⁰¹ the 4,5-dichloropyrrolyl-S-CP products of PltA, Pyr29, Mpy16, and Clz5 are incorporated in respective downstream PKS pathways (Figure 8A), and the 3-bromo- or the 3,5-dibromo-p-coumaroyl-S-CP product of AltN is postulated to be elongated via iterative fatty acid biosynthetic enzymes, finally getting incorporated in a NRPS pathway (Figure 8B).¹⁰⁷ While the tetrabromination of the pyrrole ring by Bmp2 catalyzes the decarboxylative-offloading of 24,³⁸ chlorination of CP-loaded β-tyrosine by the SgcC3 also does not involve downstream modular NRPS or PKS elongations, rather, aryl hydroxylation,¹²⁶ followed by direct transesterification¹²⁷ to the enediyne core to furnish 23 (Figure 8B).^128,129 However, when halogenated building blocks occur not as the first building block incorporated in the NRPS or PKS chain, several different routes via which halogenation occurs can be envisaged.

The simplest mechanism would be the accumulation of a pool of halogenated precursor amino acids or polyketide building blocks, and their eventual incorporation into the assembly line by dedicated adenylation and ketosynthase domains that preferentially use these halogenated precursors over their corresponding deshalo congeners. For NRPS derived peptides, such as the recently reported actinomycete cryptic natural product taromycin A (36) that was accessed via transformation associated recombination-based direct capture of the biosynthetic gene cluster and refactoring in a heterologous host,¹³⁰ this strategy is likely operative. A single FDH, Tar14, encoded with the taromycin biosynthetic gene cluster, likely generates 6-chlorotryptophan that is then incorporated directly into the growing peptide chain by the NRPS Tar8 module-1 adenylation domain (Figure 11A). Sequence comparison revealed the closest structural homolog of Tar14 to be the recently reported tryptophan-6-chlorinase SttH,⁶⁷ followed by the tryptophan halogenases PrnA, RebH, and PyrH. 6-Chlorotryptophan generated by Tar14 is envisaged to be also converted into 4-chlorokynurenine, that is in turn incorporated by the NRPS Tar10 module-13 adenylation domain. Rationalizing based on the meta-positioning of the chlorine relative to the ortho- and para- directing aniline in 4-chlorokynurenine amino acid,⁸⁸ it is unlikely that Tar14 halogenates both tryptophan and kynurenine. Furthermore, as deschlorotaromycin congeners are not produced at detectable levels together with 36, the Tar8 module-1 and Tar10 module-13 adenylation domains are likely specific for the recognition and activation of 6-chlorotryptophan and 4-chlorokynurenine, respectively, in contrast to the corresponding adenylation domains encoded within the biosynthetic gene cluster of closely related lipopeptide antibiotic daptomycin (37) that conceivably are specific for the incorporation of tryptophan and kynurenine (Figure 11A). This chemical logic for the incorporation of halogenated building blocks in NRPS peptides is likely conserved in the biosynthesis of kutznerides,^131,132 such as kutzneride E (38), actinobacterial depsipeptides bearing a 6,7-dichlorotryptophan derived amino acid building block that is putatively recognized and activated by the NRPS KtzH module-6 adenylation domain (Figure 11B).¹³³ The two FDHs encoded with the kutzneride biosynthetic gene cluster,¹³³ KtzQ and KtzR, bear greatest sequence homology to tryptophan halogenases SttH, PrnA, RebH, and PyrH. Whether the 1,2-epoxidation of the indole ring, followed by cyclization via the main chain nitrogen atom to generate the tricyclic pyrroloindoline motif occurs post or prior to incorporation in the NRPS peptide remains to be discerned. Chlorinated pyrroloindolines, postulated to be synthesized by as yet unidentified FDHs, also occur in the NRPS-derived chloptosin¹³⁴ (39) and other structurally related peptides isolated from Streptomyces alboflavus 313 (Figure 11C).¹³⁵ Similarly, while the abovementioned hypotheses for the action of the FDHs Tar14, KtzQ, and KtzR have not been experimentally verified, adenylation domains that recognize and activate halogenated precursors in NRPS assembly lines are attractive ‘plug-and-play’ elements that can be used to regiospecifically engineer natural product scaffolds to bear halogens, a modification that is likely challenging to be affected using chemical derivatization techniques.

Figure 11. Incorporation of halogenated tryptophan in NRPS-derived peptide natural products.

(A) 6-chlorotryptophan, generated by FDH Tar14 is converted to 4-chlorokyunerine. Both amino acids are incorporated in 36. (B) FDHs KtzQ and KtzR generate 6,7-dichlorotryptophan that is incorporated as a pyrroloindoline in the kutznerides family of natural products, such as 38. (C) The natural product 39 also bears the chlorinated pyrroloindoline motif that is generated via an as yet unidentified enzymatic route. (D) The FDH CmdE encoded in the biosynthetic gene cluster for the production of 45 bears homology to acyl-S-CP utilizing FDHs rather than Tar14, KtzQ, and KtzR (vide infra). It is thus likely that the mechanism for incorporation of the 2-chlorotryptophan moiety in 45, and in closely related jasplakindolide (46), might differ from that for 36 and 38.

Perhaps the most celebrated example of halogenation in a NRPS-derived natural product is the dichlorination for the glycopeptide antibiotic-of-last-resort, 3 (Figure 1). Though the emergence of vancomycin resistant pathogenic bacteria dilutes this proud distinction,¹³⁶ it is well established that the aryl-chlorination of the two β-hydroxytyrosine building blocks by the FDH VhaA is critical to the antibiotic efficacy of 3.^10,137–139 For a structurally related natural product, deletion of the FDH participating in the biosynthesis of 5, BhaA,¹⁴⁰ from the biosynthetic gene cluster¹⁴¹ led to the production of the deschloro derivative of 5.¹⁴² Further gene deletion and precursor feeding experiments¹⁴³ established that the module-2 and module-6 adenylation domains incorporate β-hydroxytyrosine and not the 3-chloro-β-hydroxytyrosine amino acid precursor in 5. Thus, contrary to the biosynthetic proposal for 36 and 38 presented above, it is unlikely that chlorination of the β-hydroxytyrosines in 3 and 5 occurs prior to the NRPS assembly of the respective heptapeptides. At this stage, the physiological substrate for the VhaA and BhaA FDHs, and the timing of the halogenation relative to the peptide assembly remained elusive. In the temporal vicinity of these findings, biosynthetic gene clusters encoding production of structurally analogous bacterial NRPS derived chlorinated peptides such as chloroeremomycin (40),¹⁴⁴ complestatin (41),¹⁴⁵ A47934 (42),¹⁴⁶ A40926 (43),¹⁴⁷ and teicoplanin (44)¹⁴⁸ were reported (Figure 12). In each of these gene clusters, only a single FDH was located, as more recently also observed in environmental DNA (eDNA) derived vancomycin-like eDNA gene cluster (VEG) and teicoplanin-like eDNA gene clusters (TEG).^149,150 With the assumption that halogenation in these NRPS peptides is intertwined with the NRPS peptide assembly, the rules of regiospecificity for FDHs as discussed in the preceding sections need to be further relaxed, such that the single FDH encoded within the biosynthetic gene cluster for 41, ComH, needs to perform a remarkable six halogenations on three phenyl rings.

Figure 12. Halogenated NRPS derived glycopeptide antibiotics.

A CP loaded hexa-peptide was used as a substrate for the in vitro reconstitution of the activity for the FDH VhaA. The VhaA reaction product would require further NRPS elongation by the 3,5-dihydroxyphenylglycine amino acid, oxidative coupling of the aryl rings, glycosylation, and offloading from the NRPS assembly line for maturation into 3. The order or amino acid addition during the NRPS assembly is denoted by numerals for 3 and halogenated aryl rings are shaded blue.

The longstanding biosynthetic riddle regarding the timing of halogenation in glycopeptide antibiotic biosynthesis was recently resolved with the in vitro characterization of the FDH VhaA.¹⁵¹ This study demonstrated, for the first time, that halogenation in biosynthesis of 3 occurs en route the NRPS assembly of the linear heptapeptide. A biomimetic CP-loaded hexapeptide substrate was dichlorinated by VhaA. Whether the physiological substrate for VhaA is the CP-loaded hexapeptide (as used in these investigations) or the heptapeptide (formed after the addition of the terminal 3,5-dihydroxyphenylglycine amino acid by NRPS module-7) remains to be determined. Nevertheless, this study demonstrated the challenge nascent in generating the substrates that are required to characterize these FDHs. Interestingly, VhaA halogenates the aryl rings at positions 2 and 6 of the substrate, while analogous phenoxyl activated aryl rings at positions 4 and 5 remain unmodified. This regiospecific decision making by VhaA can be attributed, in part, to the β-hydroxyl moieties at positions 2 and 6 of the hexapeptide in the absence of which chlorination did not occur. Halogenation also did not transpire when a CP-loaded dipeptide substrate comprising of amino acids at positions 1 and 2 (Figure 12) were used as substrate for VhaA, demonstrating that halogenation occurs on a relatively mature NRPS-derived peptide. No crystal structures of the NRPS glycopeptide FDHs are presently available. The underlying reason why the biosynthetic logic for 3 involving the incorporation of chlorinated β-hydroxytyrosines differs from that for the incorporation of chlorinated tryptophan, and its derivatives, in 36 and 38 is presently not clear.

NRPS derived cyanobacterial natural products, such as aeruginosins and cyanopeptolins, constitute both halogenated and nonhalogenated congeners (Figure 8B).¹⁵² Cyanobacteria of the genera Microcystis and Planktothrix have been shown to harbor the FDHs AerJ and McnD within aeruginosin and cyanopeptolin biosynthetic gene clusters, respectively, the presence of which corresponds to the production of halogenated congeners such as 32 and 34.^110,111 While the activity of the two halogenases has not been reconstituted in vitro and physiological substrates have not been experimentally established, chemical logic suggests that the aeruginosin FDH AerJ mono- or di-chlorinates the first building block of the NRPS peptide, the L-tyrosine-derived 4-hydroxyphenyl lactate while it is acylated to the AerA CP domain. Hence, in effect, AerJ could resemble the β-tyrosine chlorinase SgcC3 in its activity (Figure 8B), though halogenation at a different stage of NRPS elongation cannot be presently ruled out. Chlorination by McnD is however localized to a L-tyrosine-derived late stage building block. Hence, as before, whether chlorination of the L-tyrosine aryl ring occurs during the NRPS assembly of 34, or whether a chlorinated tyrosine amino acid is incorporated by the associated adenylation domain cannot be rationalized based on chemical logic alone. This biosynthetic conundrum also presents itself en route to the production of chlorinated anabaenopeptilides, such as 35, by cyanobacteria of the genus Anabaena¹⁵³ where the timing for halogenation by the FDH ApdC relative to the NRPS assembly has not been discerned, as well as during the production of chlorinated cryptophycins¹⁵⁴ such as 33 by cyanobacteria of the genus Nostoc that likely involves the chlorination of the methoxytyrosine aryl ring by the FDH CrpH (Figure 8B).¹⁵⁵ Sequence analyses of the cyanobacterial FDHs AerJ, McnD, ApdC, and CrpH reveal closest structural homologs to FDHs acting on acyl-S-CP substrates, pointing towards substrates for these enzymes to be loaded onto CPs as well. A close sequence homolog to CrpH is the myxobacterial tryptophan-2-halogenase CmdE involved in the biosynthesis of chlorinated cytotoxic NRPS derived chondramides, such as chondramide B (45, Figure 11D).¹⁵⁶ Whether the biosynthetic scheme for the incorporation of chlorinated tryptophan residue in 45 differs from that of 36 and 38 remains to be experimentally verified. 45 shares close structural homology to the marine sponge Jaspis derived brominated jasplakinolides,^157,158 such as 46 (Figure 11D), that also bear brominated tryptophan building blocks in a NRPS-derived scaffold. However, halogenation at the 2-position of tryptophan in 45 and 46 posits towards the presence of a tryptophan halogenase, such as CmdE, that differs in its regiospecificity for halogenation from all other tryptophan halogenases discussed previously. Interestingly, a shift in halogenation regiospecificity, including halogenation at the 2-position on the indole ring, has been reported for tryptophan-7-chlorinases with changes in the substrate chemical structure.⁸⁹

Other notable halogenated natural products in which the timing of halogenation by FDHs has not yet been elucidated include the NRPS-derived antibiotics ramoplanin (47) and enduracidin (48) that target cell wall biosynthesis,^159,160 and the unique 3-amino-5-hydroxybenzoic acid (AHBA) starter unit initiated PKS-derived chlorinated ansamacrolactams such as napthomycin A (49) and ansamitocin AP-1 (50) that inhibit RNA polymerase (Figure 13).¹⁶¹ Notably, the ansamitocin and napthomycin FDHs-Asm12¹⁶² and Nat1, respectively, could functionally replace each other in their respective deletion strains,¹³⁸ possibly implying broad substrate tolerance. Deletion of the asm12 gene demonstrated the accumulation of mature deschloro-ansamitocin by the Actinosynnema pretiosum, implying that either halogenation could likely be late stage biosynthetic step, or that other Asm enzymes are promiscuous enough to accommodate both halogenated and deshalo intermediates in the ansamitocin biosynthetic scheme.¹⁶³ A combinatorial replacement of FDHs implicated in the biosynthesis of 48 and 47, End30¹³⁹ and Ram20, respectively, led to the production of novel derivatives of these antibiotics.^164–166 However, unlike 3, chlorination is not essential for the bioactivity of 47.¹⁶⁷ In a recent study using degenerate primers to screen for FDHs in mangrove actinomycetes, gene clusters for the production of 48 and ansamacrolactams were detected and novel producers of these antibiotics were identified.¹⁶⁸ Using a similar approach but for Arctic marine actinomycetes, a homolog of the FDH Asm12 was identified, among several other FDHs that likely participate in the biosynthesis of as yet unidentified halogenated natural products.¹⁶⁹

Figure 13. Chemical structure of chlorinated natural products 47–50.

The halogenated rings are shaded in blue.

Another intriguing knowledge gap exists in the timing of halogenation by the FDH CmlS en route the biosynthesis of 6 (Figure 1), a commonly employed antibiotic. Though the activity of CmlS has not been reconstituted, intriguingly, CmlS is the only FDH known to date that demonstrates a covalently tethered FAD cofactor to the enzyme active site,¹⁷⁰ calling into question the hypothesis that reduced and oxidized flavin cofactors are diffusible from the FDH active site. Even though the biosynthetic gene cluster for the production of 6 has been known for more than a decade,¹⁷¹ the molecular details, and indeed the genetic players involved in the biosynthesis, are still being worked out.¹⁷²

Though outside the purview of halogenated natural products biosynthesized via modular NRPS or PKS pathways, it is interesting to speculate on the timing of halogenation relative to the maturation of the lanthipeptide antibiotic microbisporicin (51)^173,174 bearing the 5-chlorotryptophan residue that is likely synthesized by the FDH MibH encoded within the biosynthetic gene cluster (Figure 14).¹⁷⁵ While the activity of MibH has not been reconstituted in vitro and the identity of the physiological substrate remains elusive, fidelity of peptide synthesis by the ribosome precludes that L-tryptophan will be the physiological substrate for MibH. Hence, like VhaA that accepts a peptidic late stage biosynthetic intermediate as the substrate, MibH would likely accept either the microbisporisin prepeptide as a substrate for halogenation, or a biosynthetic intermediate en route ring closures and removal of the leader peptide. Either way, MibH promises to expand the substrate scope for FDHs to now include ribosomally synthesized post-translationally modified peptides (RiPPs).¹⁷⁶ Already, feeding of potassium bromide to the producer strain of 51 has revealed the production of brominated variants of 51 with improved antimicrobial efficacy.¹⁷⁷

Figure 14. Chemical structure of the lanthipeptide antibiotic 51.

The halogenated indole ring is shaded in blue.

2.5 Predictive functional assignments

Can we use the inventory of FDHs presented above to glean information for as yet experimentally uncharacterized FDHs in natural product biosynthetic gene clusters? A phylogenetic analysis reveals that these FDHs have a propensity to cluster according to their substrate specificity. For instance, as shown in Figure 15, all FDHs halogenating pyrrolyl-S-CP substrates (red nodes), regardless of the source bacterial phyla, cluster together. Similarly, tryptophan (blue nodes) and CP-tethered phenoxyl FDHs (in purple) are organized in their respective clades. Notable is the absence of the myxobacterial chondramide tryptophan halogenase CmdE that instead clades with the fungal FDHs Rdc2, RadH, and CazI, together with the cyanobacterial FDH CrpH. As has been mentioned previously, whether or not CmdE catalyzes the halogenation of free tryptophan remains to be verified. If not, it begets the question whether Rdc2, RadH, and CazI catalyze halogenation of small molecule or CP-tethered substrates. It should be noted that the small molecules used for the activity reconstitution of Rdc2 and CazI bear resorcylic rings that are highly susceptible to electrophilic aromatic substitution by hypohalites. In themselves, the next nearest neighbors of Rdc2, RadH, CazI, CrpH, and CmdE are FDHs that catalyze halogenation for phenolic CP-tethered substrates. This analysis certainly has its limitations. Why CrpH clusters away from other cyanobacterial phenolic FDHs such as AerJ, McnD, and ApdC is presently not clear.

Figure 15. Neighbor-joining tree showing the relatedness of FDHs.

Phylogenetic analysis was performed using Mega. The scale bar indicates 0.2 changes per amino acid. Full length primary sequences of select FDHs that have been biochemically characterized, or identified within genetic context of natural product biosynthetic gene clusters are included as discussed in the text. The FDH Amm3 is implicated in the biosynthesis of the ammosamide alkaloids.

Another curious case is the clustering together of the FDH PrnC, implicated in the chlorination of the pyrrole ring of 10, with the FDH HalB (as yet unexplained contribution to the biosynthesis of 30) and the chlorinases Mpy10 and Mpy11 participating in the biosynthesis of 25 (green nodes). While the physiological substrate for HalB is yet to be identified, the contribution of Mpy10 and Mpy11 in the atroposelective cross-coupling of two monodeoxypyoluteorin monomers to generate 25 has been demonstrated in vivo.⁹⁸ In a proposed mechanistic scheme, a cryptic halogenation at either the 3-position of the pyrrole ring, or on the pyrrole nitrogen should set up a cross-coupling reaction with the elimination of the halogen. In such a scenario, both Mpy10 and Mpy11 being present in the neighborhood of PrnC in a phylogenetic analysis makes biochemical sense. However, why does this scheme need the presence of two FDHs? Deletion of either mpy10 or mpy11 leads to loss in production of 25.⁹⁸ Is one of the Mpy10/Mpy11 FDHs masquerading as a cross-coupling catalyst that also introduces atroposelectivity in the final natural product outcome? A comprehensive in vitro biochemical characterization of Mpy10 and Mpy11 enzymes is required to answer these open questions.

3. Vanadium-dependent haloperoxidases

With the majority of biogenic organohalogens being marine in origin, it was originally thought that heme-dependent haloperoxidases were responsible for catalyzing the bulk of the halogenation events. Using the MCD-assay that was employed in the initial characterization of CPO (Figure 2), the first V-HPO was isolated from the marine brown alga Ascophyllum nodosum.¹⁸⁰ Upon its characterization, stoichiometric amounts of vanadium were identified instead of the proposed heme as the enzyme cofactor. To date, characterized V-HPOs have been primarily of marine origin. Vanadium-dependent bromoperoxidases (V-BPOs) are predominantly isolated from marine algae, while vanadium-dependent chloroperoxidases (V-CPOs) are found in fungi and marine-derived bacteria.

3.1 Structure and mechanism

V-HPOs utilize a chemical oxidant, such as hydrogen peroxide, to catalyze the oxidation of halides and, as for FDHs, are named after the most electronegative halide they are able to oxidize. Thus, V-CPOs can oxidize chloride, bromide and iodide, whereas V-BPOs only oxidize bromide and iodide. Because hydrogen peroxide lacks the thermodynamic potential to oxidize fluoride, vanadium-dependent fluoroperoxidases have not been found in nature.¹⁸¹ Vanadate is a required prosthetic group and is coordinated to the protein in a trigonal bipyramidal fashion through the side chain imidazole nitrogen atom of a conserved histidine residue (Figure 16). The overall negative charge of the vanadate decreases through an extensive hydrogen bonding network between vanadate’s three equatorial oxygen atoms and conserved amino acid residues in the active site.^182–184 This vanadate coordination geometry is also observed for the transition-state mimics for di-metal ion-dependent phosphate and phosphonate esterases.^185–188 As with the heme-dependent haloperoxidases discussed previously in the biosynthesis of 2 and 7, V-HPOs utilize hydrogen peroxide to catalyze the two-electron oxidation of halides to the corresponding reactive hypohalites. However, unlike the heme-dependent haloperoxidases, V-HPOs do not succumb to oxidative inactivation during turnover, as vanadate maintains the V(v) oxidation state throughout the catalytic cycle and is not redox active. After formation of the hypohalite intermediate, if the appropriate nucleophile is present, the reactive X⁺ species will react with the substrate to form a halogenated product with the overall stoichiometry being one halogenated product for every one equivalent of hydrogen peroxide consumed (Figure 16A). If the correct substrate is not present, the oxidized species can react with another equivalent of hydrogen peroxide to regenerate the halide and produce dioxygen in the singlet state.¹⁸⁹ Whether hydrogen peroxide is the physiological halide oxidant has not been established, however, no other oxidants have yet been identified for V-HPOs and the V-HPO biochemistry has relied on using hydrogen peroxide in in vitro biochemical reactions.¹⁹⁰ Unlike FDHs, crystal structures of V-HPOs have failed to reveal either the halide binding site, or the site at which substrates are bound by these enzymes. Moreover, the order of halide oxidation relative to substrate binding has not been revealed by kinetic studies. These shortcomings have precluded, for the most part, a rational engineering of V-CPOs and V-BPOs for tailored applications.

Figure 16. Overall mechanism and vanadate coordination in V-HPOs.

(A) Proposed catalytic scheme for generating the halogenating agent X⁺ in V-HPOs (adapted from Winter and Moore, 2009). In the first step of catalysis, hydrogen peroxide is coordinated to the vanadium center in a side-on manner. The peroxide bond is broken through a nucleophilic attack on the partially positive oxygen by a halide resulting in the release of the hypohalite. If the appropriate substrate (R) is present, a halogenated compound will be formed with the loss of water and regeneration of vanadate. If the appropriate substrate is not present, the hypohalite can react with hydrogen peroxide to regenerate the halide and form dioxygen in the singlet state. (B) Active site of native V-CPO from Curvularia inaequalis (PDB accession number 1IDQ). Active site residues are shown as yellow ball and sticks and vanadate is shown coordinated to the conserved His₄₉₆ residue.

With respect to natural product biosynthesis, investigations involving V-HPOs have traversed two broad areas: first being the transformation of electron rich substrates to their halogenated products by marine eukaryotic V-BPOs. As illustrated below, these investigations lacked the genetic context of a dedicated natural product biosynthetic gene cluster and thus utilized electron rich substrates as biogenic precursors. The identities of the precursors were selected based on chemical logic dictating the biomimetic scheme via which the targeted natural product could be biosynthesized. The second, and more recent line of investigation, involves marine bacterial V-CPOs within confirmed genetic context of natural product biosynthetic schemes and gene clusters. The discussion below is thus divided to follow these two broad directions.

3.2 Halogenation by marine Eukarya

As has been mentioned previously, given the halide concentration in seawater, the majority of naturally occurring organohalogens, including nearly all brominated compounds, are of marine origin.¹⁹³ Many of these halometabolites are produced by marine algae; with a majority of the compounds being isolated from species belonging to red macroalgae.^{34,189,194–198} It was hypothesized that V-BPOs could be responsible for the observed chemistry, and surveys of red, green and brown algae for bromoperoxidase activity demonstrated that red algae belonging to the genus Corallina had the highest activity.^199,200 Initial characterization studies, however, with the V-BPO from C. pilulifer and its capability of halogenating anisole and prochiral aromatic compounds failed to show any regio- or stereospecific bromination selectivity.²⁰¹ Because of its lack of selectivity, it was speculated that V-BPOs generated a hypobromite intermediate that was freely diffusible and would carry out bromination reactions outside the enzyme’s active site. However, several subsequent characterization studies discussed below have shown that V-BPOs can indeed halogenate a range of organic compounds in a regio- and stereospecific manner.

Halogenated sesquiterpenes are one of the largest groups of natural products isolated from marine algae, and as a consequence, it was initially hypothesized that most halogenated cyclic sesquiterpenes were biosynthesized via a halenium-induced cyclization of an acyclic terpene precursor.^202,203 V-BPOs isolated from marine red algae (e.g. C. officinalis, Laurencia pacifica, and Plocamium cartilagineum) were shown to catalyze the cyclization and asymmetric bromination of the acyclic sesquiterpene (E)-(+)-nerolidol (52) to yield the α-, β-, γ-snyderols (53–55) and (+)-3β-bromo-8-epicaparrapi oxides (56–57) marine natural products (Figure 17).¹⁹⁴ After the selective bromination of the C10-C11 olefin, the bromonium-intermediate is attacked by the internal olefin of 52 to yield the tertiary carbocation. This intermediate can then undergo one of three different elimination reactions leading to 53–55. Additionally, quenching of the reaction with the tertiary alcohol of 52 produces 56–57. Single diastereomers of 54 and 55 were produced in the enzyme reaction, whereas, in the synthetic reaction with 2,4,4,6-tetrabromocyclohexa-2,5-dienone, two diastereomers of each were formed. This study established the role of V-BPOs in the biosynthesis of brominated cyclic sesquiterpenes from marine red algae.

Figure 17.

V-BPO-catalyzed cyclization of 52 to form the brominated sesquiterpene marine natural products. Co-VBPO: C. officinalis V-BPO.

C₁₅ acetogenins are nonterpenoid cyclic ether metabolites containing varying oxane ring systems and a conjugated enyne or bromoallene terminus.²⁰⁴ Eight-membered cyclic ethers are the most abundant C₁₅ acetogenins, and Laurencia spp. are common producers of these halogenated metabolites. The eight-membered cyclic ethers can be divided into two subclasses: the lauthisan-type, which contains the metabolite laurencin (58, Figure 18), and the laurenan-type, which contains the metabolites laureatin (59) and laurallene (60).²⁰⁴ It was proposed that linear laurediols were a precursor to both subtypes, and that the assembly of the cyclic ether metabolites could proceed through a bromoetherification of the linear precursor. Incubation of (3E, 6R, 7R)-laurediol (61) with a crude preparation of a V-BPO from the red alga Laurencia nipponica produced the cyclic bromoether deacetyllaurencin (62, Figure 18A).²⁰⁵ When (3Z, 6S, 7S)-laurediol (63) was used as the substrate, the C₁₅ acetogenin prelaureatin (64), which belongs to the laurenan class of cyclic ethers, was observed (Figure 18B).²⁰⁵ While yields of the brominated cyclic metabolites were low, results from the experiment established that cyclization of the acyclic laurediol precursors to the eight-membered bromoethers was initiated by a bromonium ion. The partially purified V-BPO from L. nipponica was also shown to convert (Z)-prelaureatin (65) to 59 and isolaureatin (66). (Figure 18C).²⁰⁶ However, unlike the lactoperoxidase that could also catalyze the bromonium-ion induced cyclization of (E)-prelaureatin to form the bromoallene-containing natural product laurallene, no turnover with the (E)-isomer was observed with the V-BPO thereby further supporting the notion that V-BPOs possess inherent substrate selectivity. Recently, results of the deacetyllaurencin study were validated using the more stable trimethylsilyl-capped (3E, 6R, 7R)-laurediol precursor and purified V-BPO from L. nipponica.²⁰⁷

Figure 18. V-BPO-catalyzed chemoenzymatic synthesis of brominated acetogenins.

(A) Bromolactonization of 61 to 62 using the L. nipponica V-BPO (Ln-VBPO). (B) Bromolactonization of 63 using partially purified Ln-VBPO. (C) Conversion of 65 to 59 and 66 using a partially purified Ln-VBPO.

Quorum sensing is a process used by bacteria to monitor cell density and regulate phenotypic responses such as biofilm formation.²⁰⁸ Many bacteria secrete small signaling molecules, such as 3-oxo-acyl homoserine lactones to signal cell density and to coordinate gene regulation within a population. In the ocean, all surfaces are susceptible to biofouling and in order to survive, marine algae must have access to sunlight and nutrients. It was proposed that for some macroalgal species, V-HPOs are present on the plant surface and produce hypohalous acids that function as anti-fouling agents.²⁰⁹ Studies of gene expression in the brown alga Laminaria digitata demonstrated that V-BPOs and vanadium-dependent iodoperoxidases (V-IPOs) are up-regulated upon defense elicitation.²¹⁰ Interestingly, the 3-oxo-acyl homoserine lactone signaling molecules are readily susceptible to electrophilic halogenations at the C2 position.²¹¹ When incubated with the red alga Delisea pulchra, dibromination at the C2 position of 3-oxo-acyl homoserine lactone 67 was observed (Figure 19A). The resulting α,α-3-oxo-hexanoylhomoserine lactone 68 disrupts bacterial quorum sensing, suggesting that V-BPOs are localized at the alga’s surface to defend against microbial colonization.²¹² D. pulchra also produces a series of bromofuranone natural products that inhibit quorum sensing in select bacteria by interfering with receptor-mediated phenotypic responses.²¹² As the precursors for the naturally occurring bromofuranones are unknown, 4-pentynoic acid (69) was used as a substrate for biomimetic characterization experiments. The V-BPO isolated from D. pulchra was shown to catalyze the bromolactonization of 69 to form 5E-bromomethylidenetetrahydro-2-furanone (70, Figure 19B). Compared to the naturally occurring bromofuranones (Figure 19C),^213,214 70 lacks a degree of unsaturation within the furanone ring. However, the enzymatically synthesized 70 was still able to inhibit quorum sensing in Agrobacterium tumefaciens. Altogether, results from these studies demonstrated that to mitigate microbial fouling on its surface, the red alga D. pulchra has evolved a two-pronged approach for disrupting bacterial quorum sensing, that is, brominative inactivation of 3-oxo-homoserine lactones and synthesis of brominated furanone inhibitors of quorum sensing. These results posit towards the ecological roles of halogenases in small molecule mediation of marine chemical ecology.

Figure 19. V-BPO-catalyzed chemoenzymatic synthesis of brominated furanones.

(A) Bromination of acylhomoserine lactone 67 to 68 was demonstrated using whole pieces of D. pulchra. (B) Bromolactonization of 69 to 70. (C) Bromofuranones isolated from D. pulchra.

3.3 Biosynthetic schemes in marine bacteria

The first V-CPOs were identified from the terrestrial fungus Curvularia inaequalis in 1987^215,216 and marine fungus Embellisia didymospora in 1998.²¹⁷ However, no halogenated metabolites were reported from either source. It was therefore suggested that fungal V-CPOs generate hypochlorous acid for the chlorination and degradation of lignin.²¹⁸ While there are numerous chlorinated natural products, especially from the marine environment, it wasn’t until 2007 that the first V-CPO affiliated with a natural product was identified.²¹⁹

Meroterpenoids are natural products of mixed polyketide-terpenoid origin and while they are commonly isolated from fungi and higher plants, only a small group of metabolites have been identified from bacteria.^220–222 The biosynthetic gene cluster responsible for the synthesis of meroterpenoids belonging to the napyradiomycin family of chlorinated dihydroquinones, including the tri-chlorinated napyradiomycin A80915C (71), was identified in Streptomyces aculeolatus NRRL 18422 and the marine-derived Streptomyces sp. CNQ-525.²¹⁹ Through heterologous expression, the nap biosynthetic gene cluster from S. sp. CNQ-525 was shown to be wholly responsible for the synthesis of the chlorinated meroterpenoid natural products. Three putative V-HPO genes were annotated within the nap biosynthetic gene cluster, which was consistent with the structural inspection of the chlorinated napyradiomycins in which their terpenoid fragments might undergo chloronium ion-induced cyclization biochemistry in a manner reminiscent of snyderol biosynthesis as depicted in Figure 17.

Structural and functional comparisons of V-HPOs from eukaryotic organisms have shown that seven amino acids are required for halogenating activity.^{217,223–225} All three nap V-HPO enzymes contain His-496 (which covalently binds to the vanadate cofactor) and five of the six residues that participate in the hydrogen bonding of vanadate: Lys-353, Arg-360, Ser-402, Gly-403, and Arg-490 (numbering from C. inaequalis). Conserved residue His-404, which is proposed to form a hydrogen bond to the apical oxygen of the cofactor, is replaced with Ser, as seen in NapH1 and NapH4, or Phe as seen in NapH3. Mutagenesis studies have shown that the V-CPO mutant H404A from C. inaequalis loses chlorinating activity²²⁵ and a similar natural exchange is observed in NapH3. Hence, NapH3 is most likely not involved in halogenation and may have a different biosynthetic function in the assembly of the napyradiomycins. NapH1 and NapH4, on the other hand, contain a hydrophilic Ser residue at this position that can still hydrogen bond with the apical oxygen atom and were therefore predicted to catalyze the chloronium-induced cyclization of the two terpene units. It is worth pointing out that based on the MCD-assay, NapH1 was predicted to be a V-BPO.²²⁶ Additionally, a critical review of the MCD-assay was written after attempts to identify the enzymes involved in the synthesis of the ambigols and tjipanazoles in the cyanobacterium Fischerella ambigua yielded false positive results. Caution should therefore be exercised when using the MCD-assay as the sole source for identifying a putative V-HPO and/or assigning its physiological halide specificity.²²⁷ However, the utility of the MCD assay in predicting halogenating potential for a haloperoxidase enzyme at an early stage in the characterization process remains unquestioned. Ultimately, the proposed role of NapH1 in the biosynthesis of the napyradiomycins was established through in vitro chemoenzymatic conversion of the meroterpenoid SF2415B1 (72) to SF2415B3 (73, Figure 20).²²⁶ Single diastereomers of 73 were observed confirming that NapH1 mediates the stereoselective chlorination-cyclization of the hemiterpene moiety in which the chloronium ion is quenched by a nucleophilic attack from the hydroxyl group on the dihydroquinone core. It is still unclear if NapH4 or NapH1 is responsible for the second oxidative cyclization event to convert 73 to 71. As the V-CPOs were identified within a biosynthetic operon, the napyradiomycin system provided a unique opportunity to characterize V-HPOs in their biosynthetic context.

Figure 20.

Chloronium-mediated cyclization of the napyradiomycin intermediate 71 to 72 catalyzed by the bacterial V-CPO NapH1. Tetrahydroxynaphthalene is the PKS-derived building block and is assembled by a type III PKS encoded within the biosynthetic gene cluster.

Recently, a new group of chlorinated meroterpenoids, merochlorins A−D (74–77), were identified from the marine-derived bacterium Streptomyces sp. CNH-189.^228,229 While 74 contains a highly unusual bicycle[2.2.1]-heptanone ring system, 75 and 76 possess distinctive carbon skeletons that are unprecedented in natural products (Figure 21). A close inspection of the structures suggested that the merochlorins were possibly assembled by a V-CPO-dependent chloroetherification reaction similar to the napyradiomycins. Through genome sequencing and heterologous expression, the merochlorin biosynthetic cluster was identified and found to contain two V-CPO homologs (Mcl24 and Mcl40). Through differential expression of the biosynthetic pathway, V-CPO Mcl40 was implicated in the chloronium-induced macrocyclization of the isosesquilavandulyl terpenoid moiety to convert 77 to 76, however, characterization of this enzyme in vitro has yet to be achieved (Figure 21).²²⁸

Figure 21. Chloronium-mediated cyclization of the merochlorin meroterpenoids.

Mcl40 and Mcl24 are characterized V-CPOs from the merochlorin biosynthetic cluster, whereas Mcl23 is a characterized prenyltransferase that attaches the isosesquilavandulyl terpenoid moiety onto the tetrahydroxynapthalene-derived polyketide building block.

Subsequent in vitro studies revealed the role of the V-CPO Mcl24. Surprisingly, only four enzymes (Mcl17, 22–24) are required for the complete biosynthesis of 74 and 75 from the biosynthetic precursors malonyl-CoA, dimethylallyl pyrophosphate (DMAPP), and geranyl pyrophosphate (GPP).²³⁰ The prenyl diphosphate synthase Mcl22 was shown to catalyze the unusual “head-to-torso” coupling of DMAPP and GPP to form the newly established sesquiterpene isosesquilavandulyl pyrophosphate. The prenyltransferase Mcl23 then attaches the isosesquilavandulyl moiety to the aromatic tetrahydroxynaphthalene scaffold, previously formed through the activity of the type III polyketide synthase Mcl17, to yield the triene intermediate pre-merochlorin (78).²³⁰ The identification of 78 suggested an oxidative cyclization event must occur in order to form the polycyclic cores of 74 and 75. This was confirmed by in vitro characterization of Mcl24, wherein incubation with enzymatically prepared 78 afforded both 74 and 75. Ultimately, the entire pathway was reconstituted in a one-pot in vitro reaction.²³⁰

In an accompanying study, the novel Mcl24 transformation was further interrogated.²³¹ Mcl24 was inactive when assayed with other simple terpene alcohols and phenolic and naphtholic substrate analogs, establishing enzyme specificity and confirming its role in a biosynthetic pathway, in contrast to the non-specific halogenation observed in many V-HPOs characterized to date. As with NapH1, Mcl24 was unable to oxidize 8 in the presence of potassium chloride, and oxidation only occurred with addition of potassium bromide underscoring the importance of exhibiting caution when conducting assays with this model substrate. The mechanism for intramolecular cyclization was hypothesized to take place via a chloronium-induced oxidative dearomatization. To further probe the enzyme mechanism, a chemical chlorination reaction was designed to induce oxidative dearomatization/cyclization of 78. The most efficient reaction conditions included N-chlorosuccinimide and diisopropylamine.²³¹ Formation of a lysine chloramine intermediate has been observed in the FDH RebH, as previously described.⁴⁵ Therefore, it is possible that V-HPO enzymes have evolved to direct regiospecific chlorination, perhaps via a conserved lysine residue in the vicinity of the vanadate cofactor. In addition, the product distribution using the biomimetic preparation of merochlorins was compared to that of the Mcl24 enzyme reaction with 78 as the starting substrate. Structural characterization of the products identified three major compounds of the synthetic reaction as minor components in the enzymatic reaction (isochloro-merochlorin B and deschloro-merochlorin A and B), while the major components merochlorin A and B in the enzymatic reaction were only minor components in the biomimetic synthesis – essentially a reversal of selectivity.²³¹ Re-isolation of the merochlorin series from the natural producing strain confirmed the presence of these minor compounds, albeit in small amounts, demonstrating the utility of biomimetic synthesis.

Based upon both the enzymatic and synthetic studies, a hypothetical mechanism was posited that begins with selective C2 chlorination of 78. This is followed by an additional O-chlorination forming an aromatic hypochlorite intermediate, which then rearranges to a benzylic carbocation intermediate upon chloride loss. The formation of a benzylic carbocation intermediate had previously been postulated in separate studies on the total syntheses of both merochlorin A and B.^232,233 Finally, differential cation-induced terpene cyclizations result in either merochlorin A or B. This study established the novel activity of the V-CPO Mcl24, in that, it not only mediates the site-selective chlorination of the tetrahydroxynaphthalene core, but sets off the cascade of oxidative dearomatization and isosesquilavandulyl cyclization events to yield the complex scaffolds of the merochlorins.

To date, bacterial NapH1 and Mcl24 remain the only V-HPOs that have been characterized within the genetic and biochemical context of natural product biosynthesis. This reflection is tempered, in no small part, by the advances in sequencing and assembly of bacterial draft genomes, and clustering of natural product biosynthetic genes in bacterial genomes that allow for heterologous expression and biochemical rationale for substrates to mature at a faster rate than for eukaryotic halogenases discussed previously. Recently, a V-BPO was identified in the sequenced genome of marine cyanobacterium Synechococcus sp. CC9311.^234,235 While the physiological substrate for the V-BPO has not been identified, the genetic proximity of the V-BPO to phosphoenolpyruvate mutase enzyme lends speculation to its involvement in the biosynthesis of marine halogenated phosphonate natural products.^236,237

3.4 Structure-sequence relationships

There are currently five V-HPO crystal structures available and include the V-CPO from the fungus C. inaequalis,^182,238,192 V-BPO from the brown alga Ascophyllum nodosum,¹⁸⁴ V-BPO from the red alga C. officinalis,¹⁸³ V-IPO from the marine bacterium Zobellia galactanivorans,²³⁹ and the V-CPO NapH1 from S. sp. CNQ-525 (Table 1). While the proteins share very little sequence identity and differ dramatically in their reported oligomeric states, their structures are almost superimposable when overlaid around the vanadate center. This core structure consists of mainly α-helices with two four-helix bundles as the tertiary structural motif and an almost identical arrangement of amino acids surrounding the vanadate active site (Figures 16B, 22).^{183,192,238,239} Given the similarities in active site residues between the three sub-classes of V-HPOs, several site-directed mutagenesis studies have targeted subtle differences between the three classes in hopes of bringing to light the residue(s) responsible for halide selectivity and binding.^{183,201,225,238–243} As mentioned previously, no halide or substrate co-crystal structures are available for this class of halogenases. One such difference that has been targeted is Phe397 (numbering from C. inaequalis). Phe is present at this position in the fungal V-CPO and bacterial V-IPO structures, whereas V-BPOs and the V-CPO NapH1 contain His. It was proposed that while this residue does not directly interact with vanadate, the His side chain could affect the overall oxidation of the peroxo-vanadate complex by acting as a proton donor and acceptor during catalysis.^201,244 Interestingly, while the bacterial V-IPO contains Phe at this position and is unable to oxidize chloride in the thymol blue assay,²⁴⁵ the bacterial V-CPO NapH1 contains His and was unable to oxidize chloride in the MCD-assay. However, when provided its natural substrate, 72, NapH1 catalyzed the stereoselective chlorination-cyclization of the hemiterpene moiety. With the crystal structure of the V-CPO NapH1 in hand, further co-crystallization experiments with napyradiomycin intermediates will provide further information on biosynthetic function and identify residues involved in substrate selectivity.

Table 1.

PDB accession codes for crystal structures of halogenases

Enzyme	Natural product	Organism	PDB id (mutant enzyme)	Physiological substrate	Structurally described cofactors and ligands
FAD-dependent halogenases
PrnA	10	Pseudomonas fluorescens	2APG 2AQJ 2AR8 2ARD 4Z44 (F454K) 4Z43 (E450K) 2JKC (E346D)	L-tryptophan	FAD, Cl− FAD, Cl−, Trp FAD, Cl−, 7-Cl-Trp FAD FAD, Cl− FAD, Cl− FAD, Cl−, Trp
RebH	11	Lechevalieria aerocolonigenes	2OAM 2OAL 2E4G 2OA1 2O9Z 4LU6	L-tryptophan	Apo FAD, Cl− Trp FAD, Cl−, Trp Apo Apo
PyrH	15	Streptomyces rugosporus	2WEU 2WET 2WES (E46Q)	L-tryptophan	Trp FAD, Cl−, Trp FAD, Cl−
SttH	Unidentified	Streptomyces toxytricini	5HY5	L-tryptophan	FAD, Cl−
Bmp2	24	Pseudoalteromonas PS-5	5BVA 2BUL (Y302S F306V A345W)	pyrrolyl-S-CP	FAD FAD
Mpy16	25	Streptomyces sp. CNQ-418	5BUK	pyrrolyl-S-CP	FAD
PltA	22	Pseudomonas protegens	5DBJ	pyrrolyl-S-CP	FAD, Cl−
CndH	chondrochlorens	Chondromyces crocatus	3E1T	Unidentified	FAD, Cl−
CmlS	6	Streptomyces venezuelae	3I3L	Unidentified	FAD
Vanadium-dependent haloperoxidases
Fungus (V-CPO)	Unidentified	Curvularia inaequalis	1IDQ 1IDU 1VNE (D292A) 1VNF (R360A) 1VNG (H404A) 1VNH (H496A) 1VNI 1VNS 1VNC	Unidentified	VO4 VO4 VO4 VO4 VO4 VO4 VO4 SO4 VO4
Red algae (V-BPO)	54–57 and brominated furanones	Corallina officinalis	1QHB	Unidentified	PO4
Brown alga (V-BPO)	Unidentified	Ascophyllum nodosum	1QI9 5AA6 (A68V V87I V103M A203V A204V L327M)	Unidentified	PO4, I− VO4
zobellia_1262 (V-IPO)	Unidentified	Zobellia galactanivorans	4USZ 4CIT	Unidentified	VO4 VO4
NapH1	73	Streptomyces sp. CNQ-525	3W36 3W35	72	VO4 Apo
Red algae (V-BPO)	54–57 and brominated furanones	Corallina pilulifera	1UP8	Unidentified	PO4
Non-heme iron-dependent halogenases
SyrB2	82	Pseudomonas syringae	2FCU 2FCV 2FCT	L-threonine-S-CP	α-KG α-KG, Fe(II), Br− α-KG, Fe(II), Cl−
CytC3	83	Streptomyces sp.	3GJA 3GJB	aminobutyryl-S-CP	Apo α-KG, Fe(II)
CurHal	88	Moorea producens	3NNF 3NNL 3NNM 3NNJ	(S)-3-hydroxy-3-methylglutaryl-CP	α-KG, Fe(II), Cl− α-KG, Fe(II), Cl− Apo Apo
WelO5	92	Hapalosiphon welwitschii	5IQT 5IQS 5IQU 5IQV	12-epi-fischerindole U	α-KG, Fe (II), Cl−, 12-epi-fischerindole U α-KG, Fe(II), Cl− α-KG, Fe (II), 12-epi-fischerindole U α-KG, Fe (II), Cl−, 12-epi-fischerindole U, nitric oxide
SAM-dependent halogenases
FlA	94, 95	Streptomyces cattleya	1RQR 1RQP 2C2W	SAM, F−	5′-FDA, L-Met SAM 5-ClDA
SalL	4	Salinispora tropica	2Q6I 2Q6K 2Q6O (Y70T) 2Q6L (Y70T G131S)	SAM, Cl−	5′-ClDA, L-Met Adenosine SAM, Cl− 5′-ClDA, L-Met
MJ1651	Unidentified	Methanococcus jannaschii	2F4N	SAM, H2O†
PH0463	Unidentified	Pyrococcus horikoshii	1WU8	SAM, H2O†	Adenosine
TTHA0338	Unidentified	Thermus thermophilus	2CW5	SAM, H2O†
AtHTMT	97–99	Arabidopsis thaliana	3LCC	SAM, X−	SAH, Cl−

^†SAM has been shown to be a substrate in vitro. It remains to be demonstrated if SAM is indeed the physiological substrate

Figure 22. Comparison of V-HPO structures.

(A) V-CPO from C. inaequalis (PDB: 1IDQ); (B) V-CPO from S. sp. CNQ-525 (PDB: 3W36); (C) V-BPO from A. nodosum (PDB: 1QI9); (D) V-IPO from Z. galactanivorans (PDB: 4USZ); and (E) V-BPO from Corallina officinalis (PDB: 1QHB). (Top panel): Ribbon structures with active site residues highlighted in yellow. (Bottom panel): Electrostatic surface representation showing a cross section of the cavity leading to the active site with bound vanadate or phosphate. Active site side chains, vanadate and phosphate are shown in ball and stick representation.

The high conservation among the V-HPOs around the active site suggests that they arose from a common ancestor and may share an evolutionary relationship with the acid phosphatases.¹⁹¹ Phylogenetic analysis of V-BPOs, V-CPOs, V-IPOs, and bacterial non-specific acid phosphatases shows that the enzymes form clades based on proposed enzymatic function (Figure 23). The three V-CPOs from the napyradiomycin biosynthetic cluster (NapH1, NapH3 and NapH4) and two V-CPOs from the merochlorin cluster (Mcl24 and Mcl40) do not clade with the fungal or other postulated bacterial-derived V-CPOs that are proposed to synthesize hypohalous acid. Instead, these enzymes form a distinct clade that is consistent with their proposed dedicated roles in natural product biosynthesis. While NapH1 and Mcl40, which have both been shown to facilitate chlorolactonization reactions, are closely related, the multitasking V-CPO Mcl24 is surprisingly more related to the non-halogenating V-CPO NapH3. This suggests that NapH3 may too catalyze unprecedented biochemical reactions and characterization studies will help elucidate this new chemistry. With genome sequencing becoming a common laboratory technique, new prokaryotic V-HPO homologs possessing new biological functions are emerging. Several promising homologs have been identified and clade with the bacterial V-CPOs involved in natural product biosynthesis. A closer investigation of these putative clusters may reveal new biological functions that go beyond the known chemistry of currently characterized V-HPOs.

Figure 23. Neighbor-joining tree showing the relatedness of V-BPOs, V-CPOs, V-IPOs and acid phosphatases identified from fungi, algae and bacteria.

Phylogenetic analysis was performed using Mega. The scale bar indicates 0.2 changes per amino acid. Sequences in which crystal structures are available are in bold. Sequence identification codes include Ci_VCPO from Curvularia inaequalis (GenBank accession number CAA59686); Rb_VCPO from Rhodopirellula baltica SH1 (CAD72609); Ed_VCPO from Embellisia didymospora (CAA72622); NapH1 VCPO from Streptomyces sp. CNQ-525 (ABS50486); NapH3 VCPO from Streptomyces sp. CNQ-525 (ABS50491); NapH4 VCPO from Streptomyces sp. CNQ-525 (ABS50492); Mcl40 VCPO from S. sp. CNH-189 (469658154); Mcl24 V-CPO from S. sp. CNH-189 (469658138); Cs_VCPO from Cellulophaga sp. MED134 (ZP_01050453); Ss1_VCPO1 from Streptomyces sp. CNQ-766 (WP_018841016); Ss2 VCPO from Streptomyces sp. CNQ-766 (WP_018836294); Ss3 VCPO from Streptomyces sp. CNQ-329(WP_027774460); Ss4_VCPO from Streptomyces sp. CNT-371 (WP_027745869); Ss5_VCPO from Streptomyces sp. CNQ-865 (WP_027769314); Ss6_VCPO from Streptomyces sp. CNT-371 (WP_027744001); Ss7_VCPO from Streptomyces sp. CNQ-509 (WP_027745869); Ss8_VCPO from Streptomyces sp. SBT349 (WP_049575625); Ss9_VCPO from Streptomyces sp. CNQ-509 (WP_047020372); Ss10_VCPO from Streptomyces sp. CNS-335 (WP_018842950); Ss11_VCPO from Streptomyces sp. MMG1121 (WP_053666624); Sac_VCPO from Saccharomonospora saliphila (WP_019819575); Am_VCPO from Actinopolyspora mortivallis (WP_019852972); An_VBPO from Ascophyllum nodosum (P81701); Co_VBPO from Corallina officinalis (AAM46061); Cp1_VBPO from Corallina pilulifera (BAA31261); Cp2_VBPO from Corallina pilulifera (BAA31262); Sys_VBPO from Synechococcus sp. CC9311 (YP_731869); Fd_VBPO from Fucus distichus (AAC35279); Ld1_VBPO from Laminaria digitata (CAD37191); Ld2_VBPO from Laminaria digitata (CAD37192); Ld1_VIPO from Laminaria digitata (CAF04025); Ld3_VIPO from Laminaria digitata (CAQ51446); Am1_VIPO from Algoriphagus marincola HL-49 (KPQ20079); Am2_VIPO from Algoriphagus marincola HL-49 (KPQ13775); Zg1_VIPO from Zobellia galactanivorans (4USZ); Zg2_VIPO from Zobellia galactanivorans (CAZ96246); Pi_ACP acid phosphatase from Prevotella intermedia, phoC (AB017537); Kp_ACP acid phosphatase from Klebsiella pneumoniae, phoC (AJ250377); St_ACP acid phosphatase from Salmonella typhimurium, phoN (X63599); Ps_ACP acid phosphatase from Providencia stuartii¸ phoN (X64820); and Eb_ACP acid phosphatase from Escherichia blattae (AB020481).

It was originally thought that V-HPOs were non-specific halogenases. However, with recent in vitro and in vivo characterization experiments, the enzymes have received increasing attention due to their ability to not only halogenate a range of organic compounds, but because they have the capability to do so in a regio- and stereospecific manner. Marine bacteria, and in particular actinomycetes, produce a wide range of halogenated secondary metabolites and represent an untapped resource for the discovery of new halogenating enzymes. Whole genome sequence in combination with phylogenetic analysis will provide a powerful platform for identifying promising candidates.

4. Non-heme iron-dependent halogenases

As shown for the heme-dependent HPOs, FDHs, and V-HPO halogenating enzymes, convergent strategies have evolved for addition of an oxidized electrophilic halogen atom to electron rich centers. However, this chemical logic cannot account for natural products that are halogenated on unactivated alkyl carbon atoms. Therefore, a separate halogenation strategy must be employed. To accomplish a reaction of this type, radical chemistry can be invoked negating the need for prior functionalization of an aliphatic substrate, and indeed, the α-ketoglutarate-dependent non-heme iron (NHFe) halogenases have evolved to utilize this strategy during halogenation. The reaction is dependent upon O₂, α-ketoglutarate, and Fe(II) for halogenation of the substrate, ultimately forming succinate, CO₂, and the halogenated product. The reducing power of α-ketoglutarate is harnessed through oxidative decarboxylation of this co-substrate to drive the overall oxidation reaction. Just like FDHs that borrow their underlying mechanistic logic from oxygenases and hydroxylases, NHFe halogenases are closely related to NHFe hydroxylases that catalyze analogous hydroxylations on unactivated carbon atoms. The evolution of our understanding for NHFe halogenases is however opposite of that of FDHs, in that, NHFe halogenases acting on CP-tethered substrates were discovered and characterized earlier than NHFe halogenases acting on free small substrates. But, like for FDHs, a crystallographic description of substrate engagement in the halogenase active site had to await the recent discovery of a free-standing small molecule NHFe halogenase, as will be described below.

The structural and mechanistic description of the NHFe halogenase reaction is provided in a later section (Section 4.d) in light of the evolution of the discovery and understanding of these remarkable enzymatic catalysts. Furthermore, a mechanistic switch between NHFe halogenases and NHFe hydroxylases has recently been realized with the characterization of a small molecule NHFe halogenase, a conversion that is as yet unrealized for FDHs, and will be described together with the structure-mechanism description of NHFe halogenases in Section 4.d.

Natural products synthesized by NHFe halogenases can be separated into two distinct groups based on the presence of a visible chlorine atom, or a cyclopropyl ring present in the final product, the latter of which has been termed cryptic chlorination due to the requirement of a chlorination event prior to cyclopropyl ring formation. The loss of a chloride ion facilitates ring closure, and chlorine is no longer visible in the final product.

4.1 Halogenation leading to visible chlorination

The first notion that radical chemistry may play a role in halogenation was introduced during the investigation of the biosynthesis of barbamide (79), a marine natural product containing a trichloromethyl group (Figure 24).²⁴⁶ Studies using various ¹³C- and ²H-labeled leucine substrates fed to the native producer, the marine cyanobacterium Moorea producens (formerly Lyngbya majuscula), established that no prior activation of the methyl group was necessary for the trichlorination reaction to occur.²⁴⁷ This observation suggested a different mechanistic route to chlorination than what had been reported previously. Subsequently, in vitro reconstitution of halogenation activity was accomplished by the sequential action of the NHFe halogenases BarB2 and BarB1 on the substrate L-leucine attached to the CP BarA.²⁴⁸ Additional BarB2 and BarB1 homologs (DysB2 and DysB1) were identified in cyanobacterial symbionts known to produce the hexachlorinated natural products dysidenin (80) and dysideathiazole (81),⁸¹ however, recapitulation of activity has yet to be accomplished. The first example of in vitro reconstitution of enzymatic NHFe halogenase activity was demonstrated in the context of the phytotoxin syringomycin E (82) biosynthesis in the organism Pseudomonas syringae pv. syringae B301D. This nine-residue lipopeptide harbors a 4-chloro-L-threonine residue formed through radical mediated halogenation catalyzed by the NHFe annotated SyrB2. Recombinant expression, purification, and reconstitution of SyrB2 with α-ketoglutarate and Fe(II) afforded ~0.85–0.95 mol Fe per monomer. Activity was recapitulated by incubation of SyrB2 with the tethered amino acid substrate L-threonine-S-SyrB1 to form 4-chloro-L-threonine-S-SyrB1and the reaction was found to be dependent upon O₂, α-ketoglutarate, Fe, and chloride.²⁴⁹

Figure 24. Natural products bearing halogen atoms on non-activated aliphatic carbon centers.

Representative natural products shown here are biosynthesized via the action of NHFe halogenases as described in the text.

In the biosynthesis of dichloroaminobutyrate, also known as the actinomycete natural product armentomycin (83), the NHFe halogenase CytC3 is responsible for the formation of both 4-chloro- and 4,4-dichloroaminobutyryl-S-CytC2 from aminobutyryl-S-CytC2, whereby the thioesterase CytC4 releases 83 from the CytC2 CP.²⁵⁰ The chlorine substituent in the natural product hectochlorin (84) was recently confirmed to be the result of activity by the enzyme HctB that catalyzes the conversion of a CP-tethered fatty acyl moiety, hexanoyl-S-HctB, to 5,5-dichlorohexanoyl-S-HctB.²⁵¹ Note that the halogenase and the CP domains are encoded within the same peptide in HctB. A similar transformation has been hypothesized in the biosynthesis for the structurally related lyngbyabellins, such as lyngbyabellin A (85) shown in Figure 24.²⁵² Finally, as has been discussed previously, kutznerides are a series of cyclic depsipeptides that harbor multiple chlorine atoms on non-proteinogenic residues (Figure 11, representative molecule 38). One residue of particular interest harbors a 5-chloropiperizate moiety. Initial identification and genetic characterization of the biosynthetic gene cluster suggested a single NHFe halogenase, in addition to two putative FDHs.¹³³ Subsequent analysis of the entire genome sequence revealed an additional NHFe halogenase, KthP, just upstream of the initially proposed cluster that catalyzes the conversion of (3S)-piperazyl-S-KtzC to (3S,5S)-5-chloropiperazyl-S-KtzC.²⁵³ Each of the abovementioned NHFe halogenases work on CP-tethered substrates, and as discussed for pyrrolyl-S-CP dependent FDHs, catalyze one or up to two halogenation events on their respective substrate carbon centers. How an NHFe halogenase tailors this activity is not known, but the discrimination is likely encoded within the halogenase active site itself as demonstrated for FDHs.³⁸

4.2 Cryptic halogenation leading to cyclopropane ring formation

While addition of a halogen substituent is the outcome of a NHFe halogenase reaction, the same class of enzymes are involved in biosynthetic pathways where halogenation may not be obvious. In addition to the evident multiple chlorine atoms adorning the kutznerides (Figure 11), the biosynthetic pathway contains an additional NHFe halogenase that is responsible for the biosynthesis of the 2-(1-methylcyclopropyl)glycine residue. Characterization in vitro showed that the NHFe halogenase KtzD catalyzes formation of 4-chloro-L-isoleucine-S-KtzC from L-isoleucine-S-KtzC.²⁵⁴ However, subsequent cyclopropyl ring formation catalyzed by the flavoprotein KtzA, affords (1S,2R)-allocoronamic acid tethered to KtzC. This intramolecular elimination erases the previous evidence of chlorination justifying the use of “cryptic” to describe the initial chlorination event. An analogous pathway of coronamic acid (86) biosynthesis is present in the biosynthesis of the phytotoxin coronatine (87) (Figure 25A). The NHFe halogenase CmaD converts L-allo-isoleucine-S-CmaD to 4-chloro-L-allo-isoleucine-S-CmaD.²⁵⁵ In this instance however, subsequent cyclopropyl ring formation is catalyzed by the zinc-dependent enzyme CmaC.²⁵⁶ Activation via cryptic halogenation to assist the formation of the cyclopropane rings in natural products mentioned above is in line with the underlying chemical logic for N-C coupling and ring closure in marinopyrrole (25) and merochlorin (74–76) biosynthesis, respectively. However, a completely different logic of methylation of an alkene by a radical S-adenosyl-L-methionine (SAM) dependent methyltransferase and cyclopropane ring closure is followed in the biosynthesis of mycolic acids²⁵⁷ and the recently described natural product yatakemycin,²⁵⁸ while a radical modification of the valine side chain to afford a norcoronamic acid building block in NRPS-derived natural products has also been reported.²⁵⁹ The reader is directed to an excellent review on this topic.²⁶⁰

Figure 25. Halogenation assisted biosynthesis of cyclopropane rings and vinyl chlorides in natural products.

(A) Structures of 86 and 87. (B) The biosynthetic origin of both the vinyl chloride (Jam, 88 and 89) and cyclopropane group (Cur, 87) arises from the initial action of an NHFe halogenase (Hal). A conserved dehydratase domain (ECH₁, 94% sequence identity between Cur/Jam) is responsible for an identical dehydration reaction, however, the pathways then deviate through differing activities of the respective decarboxylase domains (ECH₂). In jamaicamide biosynthesis, this results in the vinyl chloride moiety present in both 89 and 90. In the biosynthesis of 88, the product of the ECH₂ domain is further modified by a novel NADPH-dependent enoyl reductase (ER) forming the cyclopropyl moiety.

The biosyntheses of the jamaicamides and curacin in the marine cyanobacterium M. producens provides a unique example of evolutionary interplay as the pathways bear striking homology, yet result in both forms of chemistry discussed above (Figure 25B). The mixed polyketide-nonribosomal peptide natural products differ in the presence of either a cyclopropane or a vinyl chloride moiety in curacin A (88) or jamaicamides A (89) or B (90), respectively. Surprisingly, each NHFe halogenase domain is located in a multidomain protein and the halogenase domains share 92% sequence identity. A thorough biosynthetic investigation revealed the activity of the predicted NHFe halogenase (Cur Hal) involved in the biosynthesis of 88 to be chlorination at the aliphatic carbon of a (S)-3-hydroxy-3-methylglutaryl-CP ((S)-HMG-CP) substrate.²⁶¹ While the activity of the jamaicamide halogenase was not investigated due to the high sequence identity to the Cur Hal, the structural diversity between the two natural products was determined to arise due to the difference in activity of the downstream decarboxylase and enoyl reductase domains, the latter of which was found to catalyze a novel NADPH-dependent intramolecular nucleophilic substitution to afford the cyclopropane ring in 88.

4.3 Halogenation of small molecules

Recently, the welwitindolinone alkaloid biosynthetic gene cluster from the cyanobacterium Hapalosiphon welwitschii was identified and characterized.^262,263 Welwitindolinone alkaloids are biosynthetically related to hapalindole, ambiguine, and fischerindole alkaloids, and bear a chlorine at C-13 (Figure 26).²⁶⁴ Bioinformatic analysis of the cluster identified a putative NHFe halogenase (WelO5) that was subsequently characterized in vitro to catalyze the monochlorination of the aliphatic C-13 of 12-epi-fischerindole U and 12-epi-hapalindole C to form 12-epi-fischerindole G (91) and 12-epi-hapalindole E (92), respectively.²⁶⁵ This is the first reported example of an NHFe halogenase catalyzing the chlorination of a small molecule, specifically a substrate not tethered to a CP, expanding the scope of NHFe halogenase catalysis. An additional biosynthetic pathway from the cyanobacterium Fischerella ambigua has been characterized and shown to be responsible for the production of the structurally related ambiguines, such as ambiguine K (93) shown in Figure 26.²⁶⁶ An NHFe halogenase is also present in this pathway (AmbO5) that shares ~79% sequence identity with WelO5 and has recently been shown to exhibit chlorination activity on a variety of deschloroambiguines.²⁶⁷

Figure 26.

Alkaloids with chlorine atoms installed via the action of NHFe halogenases.

4.4 Structure and mechanism

To date, SyrB2, CytC3, Cur Hal, and WelO5 have been structurally characterized by X-ray crystallography.^268–271 All exhibit an antiparallel β-sandwich, or “jelly roll”, topology characteristic of the cupin fold superfamily and similarly observed in NHFe hydroxylases (Figure 27).^272,273 In addition, all structures have been obtained with Fe, chloride, and α-ketoglutarate bound in the active site. In the case of NHFe hydroxylases, a conserved 2 His, 1 carboxylate (either Asp or Glu side chains) motif occupies three positions of the octahedral coordination geometry of Fe, forming a ‘facial triad’.²⁷⁴ The remaining coordination positions are occupied by bidentate coordination with α-ketoglutarate and a water molecule that will ultimately become displaced by molecular oxygen during the catalytic cycle. This coordination sphere is retained in NHFe halogenases with one exception; replacement of the coordinating carboxylate residue by the halide.

Figure 27. Structural conservation between NHFe halogenases and hydroxylases.

Crystal structures of NHFe halogenases (A) SyrB2 (PBD: 2FCT), (B) CytC3 (PBD: 3GJB), (C) Cur Hal (PBD: 3NNF), (D) WelO5 (PBD: 5IQT), and (E) the NHFe hydroxylase TauD (PDB: 1GY9) are shown in cartoon representation with Fe(II) as orange spheres and α-ketoglutarate in stick-ball representation with carbon atoms colored blue. Note the conservation of the anti-parallel β-sandwich motif (in grey) that binds the Fe(II) ion and α-ketoglutarate. For the sake of simplicity, active site comparisons with bound substrates and halide ions are shown separately in Figure 28B–D.

No crystal structures of NHFe halogenases that utilize CP-tethered substrates have been obtained in complex with their substrate, perhaps due to many of the inherent experimental difficulties with such an effort that have also been encountered with FDHs. The only structural information available regarding the interaction between enzyme and a CP-tethered substrate is an NMR structure of a CP domain from the curacin biosynthetic pathway and mapping of the CP residues that modulate Cur Hal activity based on mutational analyses.²⁷⁶ However, the discovery of WelO5 has facilitated the attainment of a co-crystal structure with the substrate 12-epi-fischerindole U,²⁷¹ and furthered the elucidation of the mechanism of halogenation/hydroxylation control, as discussed below.

NHFe halogenases exhibit significant conformational changes during the course of catalysis. Studies on the curacin halogenase (Cur Hal) in the biosynthesis of 88 revealed a conformational switch, wherein a 27-residue lid covers the active site (closed form) triggered by binding of α-ketoglutarate.²⁷⁰ This closed form is proposed to allow for (S)-HMG-CP substrate binding and subsequent catalysis. In the absence of α-ketoglutarate, the enzyme is in the open form and the lid region is completely disordered. In the case of WelO5, the enzyme adopts a unique open form prior to substrate binding and the active site is exposed to solvent.²⁷¹ Upon substrate binding, an external helix shifts to occlude the active site from solvent, forming the mainly hydrophobic active site necessary to bind the similarly hydrophobic substrate.

NHFe halogenases so far characterized have only been implicated in chlorination of natural products. However, SyrB2 exhibits bromination ability when assayed with significantly higher concentrations of NaBr, yet the enzyme shows a 180-fold preference for chloride during incorporation.²⁷⁷ This observation supports previous isolation studies of bromosyringomycin E (brominated derivative of 82) production in vivo when the producing bacterium is grown in a medium with an excess of NaBr.²⁷⁸ HctB also exhibits bromination activity when assayed with its native substrate and high concentrations of KBr (1 M).²⁵¹ Bromination activity is also present in WelO5, as pre-saturation dialysis with NaBr results in formation of (13R)-bromo-12-epi-fischerindole U, yet the enzyme prefers chlorine over bromine at an approximate ratio of 500:1.²⁷⁹ In no instance has an NHFe halogenase been reported to catalyze iodine or fluorine incorporation.

In addition to flexibility in regards to Cl/Br incorporation, NHFe halogenases often catalyze multiple chlorination events within biosynthetic pathways. In the biosynthesis of 79, BarB2 was shown to catalyze both the mono- and dichlorination of L-leucine-S-BarA, while BarB1 catalyzes the final chlorination to obtain (2S,4S)-5,5,5-trichloroleucine-S-BarA.²⁴⁸ Interestingly, both enzymes possess the capacity to di- and trichlorinate, albeit in trace amounts, confirming the necessity of both halogenases in biosynthesis. Furthermore, the previously described CytC3 and HctB can catalyze both mono- and dichlorination, while SyrB2 has shown the ability to catalyze a dichlorination when assayed in equimolar amounts with native substrate.²⁷⁷

NHFe halogenases possess a significant degree of promiscuity, both in the substrate scope and the reaction catalyzed. Remarkably, a single point mutation switches the activity of WelO5 from halogenation to hydroxylation.²⁷¹ The mutation (G166D) provided the necessary carboxylate ligand, re-establishing the ‘facial triad’ observed in NHFe hydroxylases, and switched the activity of WelO5 to solely hydroxylation; a conversion that is as yet unrealized for FDHs. Both HctB and SyrB2 exhibit hydroxylation of their native substrates.^251,280 However, the propensity for hydroxylation is negligible compared to halogenation during multiple turnovers, and attempts to convert the activity to solely hydroxylation through mutations have been unsuccessful for these enzymes.

SyrB2 can also chlorinate and hydroxylate different substrate analogs with the chemoselectivity modulated based upon the specific substrate.²⁸⁰ Remarkably, SyrB2 harbors nitration and azidation activity towards the non-native substrate L-2-aminobutyrate-S-SyrB1 when assayed in the presence of the corresponding N₃⁻ and NO₂⁻ anions.²⁸¹ A single point mutation or a residue in the active site known to stabilize halide binding (A118G) significantly decreased the affinity for Cl⁻, while enhancing nitration and azidation activity. Studies on the substrate scope of NHFe halogenases acting on freestanding substrates, WelO5 and AmbO5, revealed a relaxed substrate specificity of AmbO5 towards a variety of hapalindole-type molecules.²⁶⁷ In addition, the substrate scope could be modulated by generating a variety of WelO5/AmbO5 chimeras, all while retaining regio- and stereoselectivity.

The mechanism of NHFe halogenases (Figure 28A) closely mirrors that of the NHFe hydroxylases.^282,283 Briefly, NHFe hydroxylases catalyze hydroxylation reactions by first binding α-ketoglutarate then substrate, which displaces the coordinated water. Next, O₂ binds in the vacated water position, followed by subsequent oxidative decarboxylation of α-ketoglutarate forming the reactive Fe(IV)-oxo (ferryl) species. This intermediate abstracts a hydrogen atom from the substrate to form Fe(III)-OH and the substrate radical, which subsequently combines with the hydroxyl radical from the newly formed Fe(III)-OH species (termed radical rebound) resulting in the hydroxylated product and Fe(II) with an open coordination position. NHFe halogenase X-ray crystal structures clearly lack the canonical carboxylate ligand found in NHFe hydroxylases, with the coordination position instead occupied by a halide, and would therefore utilize a halo-ferryl intermediate for oxidation (Figures 28B–D). Furthermore, after H-atom abstraction by the haloferryl intermediate, recombination with the hydroxyl radical is suppressed in favor of transfer of the halogen to the substrate radical. Indeed, analogous haloferryl species have been directly observed in the reaction mechanisms of CytC3 and SyrB2 using a combination of X-ray absorption, UV/visible absorption, and Mössbauer spectroscopy.^284–286 Surprisingly, the haloferryl intermediate exhibits a half-life orders of magnitude longer than any other NHFe hydroxylase previously characterized.²⁸⁶

Figure 28. Reaction mechanism and active site comparisons of NHFe-dependent enzymes.

(A) Scheme of the general reaction mechanism for both NHFe halogenases and hydroxylases. Comparison of the active sites of (B) TauD, (C) Cur Hal, and (D) WelO5, with substrate-bound, where available. The conserved 2 His, 1 carboxylate coordination of Fe (orange sphere) observed in the NHFe hydroxylase TauD is analogous to a 2His, 1 halide (green sphere) arrangement, wherein the halide replaces the carboxylate in the same coordination position. The substrate is bound in an orientation that is perpendicular to the plane in which the halide resides. The side chains of the active site amino acids are shown in stick-ball representation. Taurine (panel B), 12-epi-fischerindole U (panel D), and α-ketoglutarate (panel B-D) are shown in stick-ball representation with the carbon atoms colored yellow.

Undoubtedly, one of most intriguing questions surrounding the NHFe halogenase mechanism is the ability of the enzyme to selectively control for halogenation over rebound hydroxylation. This “radical rebound” is facilitated by the proximity of the resultant Fe(III)-OH species to the newly formed substrate radical. However, generation of the haloferryl intermediate in NHFe halogenases results in five-coordinate trigonal bipyramidal geometry with the halide in an equatorial coordination position, perpendicular to the axial oxo group and the presumed point of H-atom abstraction. Therefore, the question as to how the enzyme discriminates between the two reactivities has been an active area of investigation. A significant body of research on NHFe halogenases acting on CP-tethered substrates supports the positioning of the substrate, coupled with the sluggish rate of H-atom abstraction,^285,286 as the major factors dictating halogenation over hydroxylation. Kinetic and deuterium-labeling studies on the promiscuous activity of SyrB2 revealed this enzyme capable of utilizing SyrB1-tethered L-threonine, L-2-aminobutyrate, and L-norvaline as substrates to catalyze varying degrees of both chlorination and hydroxylation.²⁸⁰ Comparison of the kinetic parameters and identification of the precise location of chlorination/hydroxylation on the various substrates proved these enzymes position the substrate in an orientation such that the target carbon is closer to the halide than the reactive Fe(IV)-oxo species.²⁸⁰ This observation aligns with the sluggish H-atom abstraction rates and the extended lifetime of the haloferryl intermediate, as the enzyme positions the substrate to favor radical halogen transfer, while sacrificing the efficiency of H-atom abstraction. The concept that substrate positioning is the main determinant in halogenation is also supported by additional computational and electron paramagnetic resonance spectroscopy studies.^287–290 Furthermore, geometries and molecular orbitals accessed within the catalytic trajectory may play an additional role in selectivity.²⁹¹

The recent X-ray crystal structure of WelO5 with substrate bound (Figure 28D) provides invaluable insight into the catalytic mechanism, one that entails rearrangement of the reactive haloferryl species during turnover. The structure reveals that the target carbon of halogenation is actually positioned closer to the coordination position initially occupied by the water molecule and replaced by oxygen during turnover, and further from the halide,²⁷¹ prior to formation of the reactive haloferryl intermediate. Therefore, to negate radical rebound, the reactive oxo species is hypothesized to rotate into the same plane as the halide at some point during formation, in the space vacated by a coordinating carbon of α-ketoglutarate lost as CO₂ during conversion to succinate. This rearrangement results in a configuration in which the target carbon is now closer to the halide and further from the reactive oxo intermediate prior to H-atom abstraction. An additional outer-sphere hydrogen bond between the oxo group and a serine residue (S189) is presumed to stabilize the rotated conformation, as mutation of this residue (S189A) results in a mixture of halogenation and hydroxylation activity. Active site rearrangement has previously been observed in the case of the NHFe enzyme clavaminate synthase,²⁹² and was also predicted to occur in NHFe halogenase catalysis through computational analyses.²⁹¹ While this rearrangement takes places after oxygen binding in WelO5, it remains to be determined if it takes place prior to or after oxygen binding in NHFe halogenases acting on CP-tethered substrates. It must be noted that this rearrangement is not at odds with previous reports, but rather reinforces the conclusion that the target carbon is closer to the halide than the reactive oxo species during H-atom abstraction, and substrate positioning is critical for discrimination between the two competing activities.

NHFe halogenases have been the focus of much attention due to their unique mechanism of halogenation on unreactive substrates. A combination of kinetic and spectroscopic studies, in addition to structural validation in the form of an X-ray crystal structure with substrate bound, provides a convincing mechanistic paradigm. Further understanding of active site dynamics and outer-sphere hydrogen bonding participants will be necessary to fully ascertain how these enzymes control substrate preference, multiple halogenations, and ultimately catalyze halogenation over hydroxylation. In addition, enzymatic halogenation has long been implicated in the development of “green catalysts” for organic chemistry.^71,293 With attainment of a substrate-bound structure, and the knowledge that NHFe halogenases harbor other reactivities such as nitration and azidation, this enzyme class has the potential for development into potent biocatalysts through further engineering.

5. S-Adenosyl-L-methionine-dependent halogenases

Each class of halogenases discussed previously activate the halide via oxidation, either a two-electron oxidation for FDHs or haloperoxidases (heme- and vanadate-dependent), or a single-electron oxidation (NHFe halogenases). Nucleophilic halogenases, on the other hand, utilize halogens without oxidation, as halides. Two classes of nucleophilic halogenases are known, the 5′-halo-5′-deoxyadenosine synthases and the halide methyltransferases. Both use SAM as co-substrate, catalyzing nucleophilic substitution reactions. The structure-mechanism of these two enzyme classes are described below.

5.1 5′-Halo-5′-deoxyadenosine synthases

The first crystal structure of a nucleophilic halogenase to be reported was that of the fluorinating enzyme 5′-fluoro-5′-deoxyadenosine synthase (FlA) in 2004, from the soil bacterium Streptomyces cattleya.²⁹⁴ FlA participates in the biosynthesis of fluoroacetate (94) and fluorothreonine (95) and is the only fluorinase characterized to date (Figure 29). In addition to S. cattleya, the enzyme has been identified in other bacteria, including the marine actinomycete Streptomyces xinghaiensis NRRL B-24674,²⁹⁵ the soil isolate Streptomyces sp. MA37, the hospital pathogen Nocardia brasiliensis HUJEG-1, and the soil strain Actinoplanes sp. N902-109.²⁹⁶ FlA catalyzes a S_N2-type displacement reaction in which fluoride attacks the electrophilic C-5′ carbon of SAM to generate 5′-fluoro-5′-deoxyadenosine (5′-FlDA) with concomitant displacement of L-methionine.²⁹⁷

Figure 29. Reaction catalyzed by nucleophilic, SAM-dependent halogenases of the 5′-halo-5′-deoxyadenosine synthase class.

Fluorinase FlA catalyzes the first committed step in the biosynthesis of 94 and 95 in the soil bacterium Streptomyces cattleya. Analogously, chlorinase SalL catalyzes the first committed step towards chloroethylmalonyl-CoA, a precursor to 96.

While the fluorinase FlA also accepts chloride to a much lesser extent as later shown in coupled enzyme assays,²⁹⁸ the first physiological 5′-chloro-5′-deoxyadenosine (5′-ClDA) synthase was reported in 2008 in association with the biosynthesis of the potent proteasome inhibitor salinosporamide A (96).^299,300 The chlorinase SalL has 35% identity to fluorinase FlA, yet does not accommodate fluoride as a substrate but rather chloride, bromide and iodide (in this order of preference).²⁹⁹ The SalL product 5′-ClDA is converted in a series of enzymatic reactions to the novel halogenated PKS substrate chloroethylmalonyl-CoA that provides the key reactive substrate to the PKS-NRPS derived 96 that is presently in clinical trials to treat multiple myeloma and other cancers (Figure 29).^301,302

Comparison of the crystal structures of the two enzymes^294,299 and of active site mutants^299,303 (Figure 30), along with stereochemical,²⁹⁷ theoretical,³⁰⁴ and kinetic studies,^299,303 helped shed light on the nucleophilic halogenating mechanism and on halide specificity. The reader is also directed to two previous reviews on this topic.^24,305 The first indication of a S_N2 nucleophilic substitution mechanism arose from the observation that the reaction proceeds with inversion of configuration at the C-5′ carbon.^297,303 The crystal structures of both fluorinase (Figure 30A–B) and chlorinase (Figure 30C–D) also support a S_N2 mechanism. For fluorinase, comparison of the substrate and product complexes aligns the C-F bond of the product 164° anti to the C5′-S bond of the substrate.²⁹⁴ Analogously, for a virtually inactive chlorinase active site mutant (Y70T) in which substrates SAM and chloride were trapped in the active site, a nearly 180° angle was observed between chloride and the C5′-S bond of SAM (Figure 30C–D).²⁹⁹

Figure 30. Crystal structures of 5′-halo-5′-deoxyadenosine synthases.

(A) Overview structure of fluorinase FlA (PDB: 1RQR) displaying monomers colored in pink, green and cyan (top view). Note that the active site lies in the interface between monomers, with three active sites per homotrimer (ligands 5′-FDA and L-met shown as spheres and colored by element). The fluorinase has been shown to be a hexamer in solution (dimer of trimers), according to gel filtration, whereas the chlorinase is a trimer in solution according to ultracentrifugation studies. (B) Fluorinase active site with bound SAM (PDB: 1RQP); chains are colored the same way as in panel A, with the ligand colored by element and hydrogen bonds with key enzyme residues shown as yellow dashes. (C) Chlorinase SalL Y70T mutant (PDB: 2Q6O) active site with bound substrates SAM (colored by element) and chloride (green sphere). (D) Chlorinase SalL Y70T mutant active site showing a view through the chloride ion. Note that the chloride makes a hydrogen bond with the amide backbone of Gly131. The water molecule shown would be displaced by the Tyr70 residue in the wild-type enzyme. (E) Overview of one monomer each of fluorinase FlA (PDB: 1RQR, pink) and chlorinase SalL G131S Y70T double mutant (PDB: 2Q6L, wheat) highlighting a 23-residue loop in FlA (magenta, position 93 to 115) which is absent in SalL (yellow, position 87 to 90) and sits just above the active site. (F) Close up view of the active sites of panel E. Overlay of the active sites of fluorinase FlA (PDB: 1RQR, pink) and chlorinase SalL G131S Y70T double mutant (PDB: 2Q6L, wheat) with bound products 5′-FDA and 5′-ClDA, respectively, highlighting the displacement of the chlorinase loop carrying Ser131 away from the product when compared to fluorinase.

In addition, a common theme that emerges is catalysis by desolvation.⁵ Fluoride in particular is a potent nucleophile in its desolvated state. However, it is tightly hydrated in aqueous solutions and, thus, effectively inert. In their desolvated state, the order of nucleophilicity for halides is F⁻ > Cl⁻ > Br⁻ > I⁻.²⁴ Crystallographic^294,299,303 and QM/MM theory studies³⁰⁴ support complete desolvation of fluoride prior to catalysis. Fluoride makes H-bond contacts with Ser158 (to both the backbone -NH and the side chain OH, as observed in the crystal structure²⁹⁴) and with Thr80 (as it progresses towards the transition state and as predicted in theory studies³⁰⁴), which should help offset the energy of desolvation and assist with stabilization of the desolvated fluoride for catalysis, respectively. It is proposed that fluoride diffuses into the enzyme and binds first, and that subsequent SAM binding further drives dehydration of fluoride.³⁰³

The major change in the active site of chlorinase²⁹⁹ compared to fluorinase²⁹⁴ is replacement of Ser158 with a glycine (Gly131 in SalL), eliminating one H-bond with the halide and increasing the size of the halide binding pocket, which matches the halide discrimination of SalL for larger halides (Figure 30C–D). Other notable differences around the active site include Tyr70 (Thr75 in fluorinase) and Trp129 (Phe156 in fluorinase). Thus, the halide binding pocket of SalL is large enough to accommodate Cl⁻, Br⁻, and I⁻ and the relative enzyme efficiency (k_cat/K_m) matches the nucleophilicity of the desolvated halides (Cl⁻ > Br⁻ > I⁻). While no halogenating activity could be detected for a G131S mutant of SalL, a Y70T mutation resulted in a 100-fold reduction in activity while affording the observation of SAM and chloride trapped in the active site. Importantly, chloride is observed interacting with water molecules in the Y70T mutant (Figure 30C). Thus, the Y70T SalL structure also supports the hypothesis that excluding water from the active site is important to activate halides for nucleophilic substitution.

The inability of SalL to catalyze fluorination may be partially explained by the serine to glycine substitution mentioned above (Ser158 in FlA and Gly131 in SalL) since the H-bond donor side chain is eliminated. Yet, an Y70T G131S mutant of SalL – i.e. a more “fluorinase-like” mutant – is still unable to catalyze fluorination. Another notable difference between the two enzymes is the presence of an additional 23-residue loop in fluorinase that sits just above the active site (Figure 30E) and appears to contribute to the size of the halide binding pocket. In fact, a structural overlay of fluorinase and SalL Y70T G131S shows a 1.4-Å displacement of the loop carrying Ser131 (SalL numbering) away from the product that results in disruption of the H-bonding network in the SalL double mutant (Figure 30F).²⁹⁹ Thus, the halide binding site for nucleophilic halogenases appears to have evolved to fully dehydrate halide ions and to position halides for S_N2 nucleophilic substitution. Halide specificity seems to depend on the size of the halide binding pocket – which is determined by active site residues and by the presence or absence of an extra loop in the N-terminus – and on H-bond donors.

FlA and SalL belong to a large family of >200 homologous proteins referred to as DUF62 in various bacteria and archaea. Protein sequence alignments show that Gly131 (SalL numbering) is conserved in most cases, while Tyr70 is replaced by nonpolar, aliphatic residues including Val, Gly and Ile.^306–308 On the basis of structure–function studies of FlA and SalL, these DUF62 proteins would most likely not act as nucleophilic halogenases due to the predicted larger size of the halide binding pocket and likely presence of water in the active site. In fact, two DUF62 proteins have been biochemically analyzed, i.e. archaeal PH0463 from Pyrococcus horikoshii and marine bacterial SARE1364 from Salinispora arenicola CNS-205.^306–308 Both enzymes showed SAM hydroxide adenosyltransferase activity (K_m SAM = 1.6 μM and k_cat = 0.5 min⁻¹ for SARE1364³⁰⁶ and K_m SAM = 39 μM and k_cat = 0.14 s⁻¹ for PH0463³⁰⁷) but no significant halogenase activity in vitro. Only iodide was accepted by SARE1364 but with K_m for iodide (20 mM, which is 40,000-fold higher than the concentration of iodide in seawater) and k_cat (0.5 h⁻¹) that do not appear biologically relevant.³⁰⁶ Moreover, DUF62 proteins contain a strictly conserved Asp-Arg-His triad that is absent in fluorinase and chlorinase and is proposed to be involved in water activation for SAM hydrolysis (Figure 31).^306–308 It remains to be shown if this SAM hydrolase activity measured in vitro is physiologically relevant. It seems counterintuitive to hydrolyze SAM as it is a high-energy, essential metabolite, biosynthesized in an ATP-dependent reaction and widely used as the methyl donor in biological methylation reactions.^309,310 Yet, biochemical studies oppose halogenase activity and the active site is tailored for acid-base catalysis using water as a nucleophile. It is also conceivable that the physiological substrate is not SAM itself but a related substrate that has yet to be identified. In any case, similarly to other halogenases that appear to have evolved from hydroxylases,^25,311 5′-halo-5′-deoxyadenosine synthases may have originated from a common enzyme that utilizes hydroxide from water instead of halides as nucleophiles.

Figure 31. SAM hydroxide adenosyltransferases.

(A) Overview structure of DUF-62 PH0463 from archaeon Pyrococcus horikoshii (PDB: 1WU8) showing the same fold as 5′-halo-5′-deoxyadenosine synthases. The three monomers of the homotrimer are colored magenta, green and cyan. (B) Previously proposed mechanism for water activation involving a conserved Asp-Arg-His triad not found in 5′-halo-5′-deoxyadenosine synthases.

Halogenation often improves biological activity. For example, 96 is 500-fold more active against cancer cell lines than its deschloro analog.^302,312 Fluorine, in particular, is extremely desirable in medicinal chemistry with about 20% of available drugs being fluorinated.³¹³ However, fluorinated natural products are extremely rare with only a handful known.³¹⁴ Thus, nucleophilic halogenases, and fluorinase in particular, are attractive targets for biotechnological applications. Three studies have been reported that provide proof of concept for engineering fluorometabolite biosynthesis. First, the incorporation of fluoroacetyl-CoA as a starter unit in aromatic polyketide biosynthesis in vitro was demonstrated.³¹⁵ Second, the fluorinase from soil S. cattleya was engineered in the marine bacterium S. tropica to achieve the production of fluorosalinosporamide directly from fluoride ion by fermentation (via fluoroethylmalonyl-CoA as an extender unit).^316,317 Thirdly, the site-selective incorporation of fluoromalonate into a simplified, mini-PKS model system was realized.³¹⁸ These studies set the stage for broader applicability of nucleophilic halogenases towards engineering halometabolites of biomedical utility.^71,319,320 Furthermore, both FlA and SalL have been used for the biosynthesis of SAM and its derivatives,^321,322 complementing other recent approaches^323,324 to generate derivative of this ubiquitous biological cofactor to enable synthetic biological efforts.

5.2 Halide methyltransferases

Biosynthesis of halomethanes, MeCl, MeBr, and MeI (97–99), has received attention due to their role in ozone depletion,³²⁵ and, more recently, due to their application as end products (e.g. solvents) and as intermediates in the synthesis of more complex organic compounds (e.g. fuels).³²⁶ SAM-dependent halide methyltransferases present in plants, algae, bacteria, and fungi have been proposed to be responsible for the presence of halomethanes in the atmosphere. Indeed, halomethane emission has been correlated with halide methyltransferase activity in e.g. agricultural plants.³²⁷

A comprehensive study analyzing 89 recombinant, putative halide methyltransferases from bacteria, fungi, plants, and archaea showed that 56% of the enzymes were active with chloride, 85% with bromide, and 69% with iodide (6% of the proteins tested showed no activity).³²⁶ A halide methyltransferase from the halophytic plant Batis maritima exhibited the highest activity on each halide and several enzymes displayed unique specificities for given halides. Moreover, the highest activity was observed for the production of 99, which was about 10-fold higher than that for 97 and 98.

Halide methyltransferases catalyze a nucleophilic substitution reaction between SAM and halide ions to form methyl halides and S-adenosylhomocysteine (SAH) (Figure 32A). A carbocation route based on the halide preference (I⁻ > Br⁻ > Cl⁻) of an enzyme from the marine alga Endocladia muricata has been proposed.³²⁸ In addition to halide ions chloride, bromide, and iodide, thiocyanate and bisulfide have been shown to act as nucleophiles in some cases, e.g. for enzymes from radish Raphanus sativus and cabbage Brassica oleracea.^327,329

Figure 32. SAM-dependent halide methyltransferase.

(A) Reaction catalyzed by SAM-dependent halide methyltransferases. (B) Overview structure of SAM-dependent halide methyltransferase from A. thaliana (PDB: 3LCC) with SAH ligand shown as spheres. (C) Halide methyltransferase active site. Waters coordinating to Tyr172 are shown as red spheres; SAH is shown with carbon atoms colored yellow.

The crystal structure of a halide methyltransferase from the plant Arabidopsis thaliana³³¹ co-crystallized with SAH has been reported (Figure 32B–C).³³⁰ A methyl group modeled at the sulfonium center to represent SAM occupies an open cavity consistent with the promiscuity of the enzyme towards larger halides and thiocyanate, with catalytic efficiency in the order NCS⁻ > I⁻ > Br⁻ > Cl⁻ (F⁻ not accepted). The model proposed by the authors is that ordered H-bonding through a bridging water molecule to the side chain of Tyr172 helps orient the nucleophile, particularly smaller nucleophiles. Consistent with this model, a Y172F mutant shows reduced efficiency only with chloride.

In conclusion, much remains to be understood regarding halide methyltransferases, their mechanism, their physiological role and the implications of methyl halide emissions for atmospheric chemistry. Nonetheless, halide methyltransferases are promising biocatalysts for the production of halomethanes, which are of interest as final products and as intermediates in the synthesis of complex organic compounds.^325,326

6. Enzymatic dehalogenation

In contrast to the halogenating enzymes described in this review with dedicated physiological roles in natural product biosynthesis schemes, the field of biological dehalogenation has largely focused on the identification of enzymes on the basis of their ability to act on anthropogenic compounds toward the end goal of bioremediation. This anthropocentric approach has succeeded in identifying numerous dehalogenases utilizing diverse strategies—reductive, oxidative, and hydrolytic, as will be discussed in this section—to dehalogenate a wide range of molecules. However, whether dehalogenases evolved as a result of recent industrial input of polyhalogenated persistent organic pollutants in the environment, or are coincident to polyhalogenated natural products that bare resemblance to man-made molecules, remains uncertain.³³² What is clear is that contextual and mechanistic disconnects exist between what is known about biological halogenation versus dehalogenation, which is exacerbated by the fact that halogenation in the context of natural product synthesis appears to be a biosynthetic end-point in that halogenated carbon centers rarely undergo further modification. This is in sharp contrast to the ubiquity of halogenated starting units and intermediates in organic syntheses in which halogenation is routinely utilized as a carbon center activating strategy. Moreover, in contrast to halogenation biocatalysts described from the context of secondary metabolite biosynthesis, the vast majority of dehalogenation biocatalysts identified to date are implicated in primary metabolic processes such as respiration. This divide is likely just as much a reflection of the different approaches taken to study these enzymes as it is due to the inherent challenge in confirming the physiological nature of dehalogenation. While dehalogenation has been studied almost exclusively from the perspective of activity-guided screens, the renaissance in biological halogenation has been sparked by the ability to study halogenation in the unequivocal contexts of biosynthetic pathways. To bridge the gap between halogenation and dehalogenation in natural products, this section will begin with a description of the few dehalogenases known from physiological processes, namely, two classes of deiodinases from the thyroid hormone system, and a recently reported debrominase from the context of the biosynthesis of a microbial natural product. This section also provides an overview on progress in reductive dehalogenation from the context of halorespiration of anthropogenic compounds due to the recent spate of structure-function studies in this area, while the remaining dehalogenation strategies will be given sparer treatment in recognition of comprehensive review articles that cover these areas as highlighted below.

6.1 Natural product reductive dehalogenases

Reductive dehalogenases (RDHs) catalyze net 2e⁻ reductive elimination reactions that result either in the cleavage of a carbon-halogen bond and formation of a carbon-hydrogen bond and a halide (α-elimination), or the breakage of two adjacent halogen substituents resulting in the release of two halides and the formation of an unsaturated product (β-elimination).¹⁶ Perhaps the best-studied class of RDHs are cobalamin (B₁₂)-dependent enzymes for their potential applications in bioremediation. However, physiological and biotechnological studies have also revealed several additional classes of non-B₁₂-dependent reductive dehalogenases from aerobic organisms. Indeed, despite a perpetuating notion in the literature on the rarity of biological reductive dehalogenation in aerobic organisms, the aerobic realm has actually afforded a greater diversity of enzymes catalyzing reductive dehalogenations.^333–335 This section will emphasize the current mechanistic understanding of dehalogenation enzymes utilizing reductive strategies beginning with a description of the few known examples of natural product dehalogenases.

6.1.1 Reductive deiodinases from the thyroid hormone system

The earliest reported examples of RDHs for which physiological substrates are known are selenoprotein iodothyronine (thyroid hormone) deiodinases and flavoprotein iodotyrosine deiodinases that, respectively, modulate thyroid hormone activity and salvage of the scarce micronutrient iodide from iodotyrosine waste products of thyroglobulin proteolysis.³³⁶ These enzymes are also the only known examples of dehalogenases from mammals. Due to the importance of thyroid hormone as “master control of metabolism”, both classes of deiodinases play an essential role in human health and dysfunctions in their activities lead to diseases such as hypo- and hyper-thyroidism and cancers. Remarkably, although both of these enzymes act on similar substrates, they arise from distinct structural families and utilize entirely different electron donors. Recent crystal structures of the catalytic domains of these enzymes has provided preliminary insight into their reaction mechanisms and substrate binding modes.^337,338 This section summarizes structure-function studies on these two distinct classes of physiological deiodinases from the thyroid system.

Thyroid hormone is mainly released from the thyroid gland as the prohormone thyroxine (2, Figure 1), and is activated and deactivated by a series of deiodinases that catalyze monodeiodination of the aromatic rings of the thyroid hormone. The first suggestion of deiodinases from thyroid physiology came about in 1955 with the discovery of 1, the outer-ring triiodinated variant of 2, and active form of the thyroid hormone. Almost twenty years later, proof of extrathyroidal production of 1 was demonstrated, and several years on in 1975, the first thyronine 5′-deiodinase was described shortly followed by two additional classes of iodothyronine deiodinases.^339–341 At present, iodothyronine deiodinases are classified into one of three types (Type I–III, abbreviated Dio1–3) on the basis of their regiospecificities with respect to monodeiodination of the inner and outer rings of thyroid hormone (Figure 33A).³⁴² Dio1 or Dio2 both catalyze outer-ring deiodination of 2 to active form 1. This can further undergo deiodination of the inner ring by Dio1 or Dio3 to afford 3,3′-T2 (100), which is thought to also be an active form of thyroid hormone. Initial deiodination of the inner ring of 2 by Dio1 or Dio3 leads to the inactive metabolite reverse T3 (rT3) that can be converted to 100 by outer-ring deiodination by either Dio1 or Dio2.³⁴³ The catalysis of deiodination is selenocysteine(Sec)-dependent and mutation of the active site Sec to cysteine (Cys) leads to a 100-fold reduction in enzyme activity.³⁴⁴ The favored mechanistic route for dehalogenation of 2 proceeds through the direct attack of the selenocysteine selenolate of the iodine leaving group to form a selenyl iodide species analogous to the Cob(II)alamin halide intermediary formed in the recently proposed mechanism for B₁₂-dependent dehalogenases (Figure 33B). An alternative proposal that proceeds through formation of an organoselenium adduct via a keto-enol intermediate. However, keto-enol tautomerization is only viable for outer-ring deiodination, while individual dehalogenases are able to perform deiodination of both inner and outer rings of 2. Hence, assuming dehalogenation of the inner- and outer- rings of 2 proceed through the same mechanism, a keto-mechanism would not be viable. Due to the high degree of amino acid similarity of Dio1–3, it is likely that all three types of iodothyronine dehalogenases act by the same catalytic mechanism.³⁴³

Figure 33. Selenoprotein iodothyronine deiodinase structure-mechanism.

(A) Transformations catalyzed by Dio1–3. (B) Peroxiredoxin-like mechanism proposed for dehalogenation of 2 with residues corresponding to mouse Dio3 in panel ‘C’. Direct attack of the leaving iodine by the selenolate of Sec170 leads to formation of a selenyl iodide intermediate. The selenyl-iodine bond is broken by the addition of the thiolate from the Cys239 side chain to form an endogenous selenyl-sulfide bond in line with the mechanism of 2-Cys peroxiredoxins. The selenyl-sulfide bond is subsequently reduced by attack of the thiolate side chain of a third cysteine residue (Cys168) leading to formation of an endogenous disulfide bridge that is then reduced by a small protein thioredoxin to complete the catalytic cycle. (C) Crystal structure of the mouse Dio3cat (PDB: 4TR3) in green, and 2-Cys peroxiredoxin PtGPx5 (PDB: 2P5Q) in yellow showing the side chains of the catalytic selenocysteine and cysteine residues. Note that the selenocysteine (Sec170) in Dio3cat is mutated to a cysteine in the crystal structure.

The integral membrane nature of Dio1–3 and Sec-dependency has hindered their recombinant heterologous production and consequently, structure-function characterization of iodothyronine deiodinases. Indeed, nearly forty years separate the first in vitro characterization of a thyroid hormone deiodinase and the first reported crystal structure of the catalytic domain of mouse Dio3 (Dio3cat) (Figure 33C).³³⁷ While the mechanism of thyroid hormone deiodinases remains uncertain, the peroxiredoxin-like fold of Dio3cat has provided further insights into the reaction mechanisms of selenocysteine-dependent deiodinases. Superposition of a structure of Dio3cat on the structure of prototypical 2-Cys peroxiredoxin PtGPx5 reveals that Sec170 (by proxy of a substituted Cys) of Dio3cat overlays with the anticipated peroxidatic Cys residue (the electron donating residue in peroxiredoxins) (Figure 33C).³⁴⁶ Moreover, a second Cys residue involved in resolving the protein-substrate complex in peroxiredoxins is also conserved in Dio3cat (Cys239). The Dio3cat structure further reveals that the side chain thiol of Cys170 points into an elongated cleft that is proposed to be the substrate-binding site, consistent with its role as the electron donor.³³⁷ A substrate-binding model constructed by superposition of the binding residues of a thyroxine receptor co-crystal structure on the proposed binding residue of Dio3cat further shows the substrate orientated with the departing 5-iodine of 2 in proximity to the Cys170 side chain thiol as anticipated for an E1-type haloelimination leading to formation of the proposed Sec170 selenyl iodide species. In a mechanism, akin to recycling of the active site cysteine in peroxiredoxins, an endogenous cysteine thiolate would resolve the selenyl iodide bond to form an endogenous selenyl-sulfide bond (Figure 33B). A third Cys residue conserved in Dio3, but not in peroxiredoxins, that is shown to enhance the deiodinase activity, is then thought to resolve the endogenous selenyl sulfide bond, returning Sec170 to its reduced state and resulting in formation of an endogenous Cys-Cys disulfide bond. The newly formed disulfide bond would then be reduced by an exogenous small protein thioredoxin to complete the catalytic cycle (Figure 33B).³³⁷ Although the mechanism for selenoprotein deiodinases remains unresolved, recent structural advances have shed light on a peroxiredoxin-like mechanism for dehalogenation in line with the favored mechanism for reductive dehalogenation.

Continued structure-function studies of iodothyronine deiodinases promise to aid in the development of small molecule modulators of thyroid function as therapeutics in the treatment of thyroid diseases. Moreover, the fact that the polyhalogenated diphenyl core of 2 closely resembles the structure of polybrominated diphenyl ether (PBDE) anthropogenic and natural persistent organic compounds, such as from marine bacteria⁴⁹ and sponges,^347–349 suggests that deiodinases might also be engineered for bioremediation function, a role predominantly entertained for bacterial B₁₂-dependent RDHs.^350,351 Structural understanding of thyroid hormone deiodinases may also provide insight into the mechanisms of toxicity of endocrine-disrupting PBDE molecules. For further details on advances in selenoprotein deiodinase structure and mechanism the reader is referred to a recent review of the topic.³⁴³

The synthesis of thyroid hormone from the proteolysis of the precursor protein thyroglobulin leads to release mono- and di-iodinated L-tyrosine (101 and 102, respectively). Molecules 101 and 102 are thought to be waste products that are present in six- to seven-fold excess over the thyroid hormone itself.³³⁶ To salvage scarce iodide from organic iodine, mammals have evolved a remarkable enzyme, iodotyrosine deiodinase (IYD), to catalyze the complete reductive deiodination of 102.³⁵² Dysfunction of IYD results in hypothyroidism, presently the leading congenital endocrine disorder affecting 1 in 3000 newborns.³⁵³ Iodotyrosine deiodinating activity in the thyroid gland was first reported in 1950s, and further investigations of IYD demonstrated its NADPH-dependent/O₂-independent catalytic requirement (Figure 34A).^352,354,355 Over half a century later, the first IYD gene sequence- human DEHAL1 was identified, and specific mutation-associated disease states were revealed shortly thereafter.^{353,356–359} Intriguingly, despite their critical role in thyroid physiology, IYDs have also recently been identified in Eukarya that lack thyroid systems as well as in bacteria.³⁶⁰

Figure 34. Structure-function of IYD.

(A) Net reaction catalyzed by IYD. (B) mIYD (PDB: 3GFD) co-crystal structure with the substrate 101 (left), and close-up of active site (right) showing FMN cofactor (yellow) and 101 (blue). The substrate 101 bridges active site residues of mIYD with the isoalloxazine ring of the FMN cofactor. (C) Single electron transfer mechanism proposed for deiodination of iodotyrosines beginning with keto-enol tautomerization followed by a single electron transfer from FMN hydroquinone (hq) to release iodide and form phenoxyl radical. The phenoxyl radical is then reduced by a second single electron transfer from the FMN semiquinone (sq) to afford the deiodinated tyrosine.

Like iodothyronine deiodinases, IYDs are transmembrane proteins. Unlike their thyroid hormone-acting counterparts, IYDs utilize a flavin mononucleotide (FMN) cofactor rather than a genetically encoded selenocysteine electron donor.³⁵⁶ In contrast to the majority of flavoenzymes characterized to date that catalyze O₂-dependent oxidative chemical reactions (including halogenation as discussed earlier in this review), IYDs catalyze an O₂-independent reductive reaction.^40,41 Moreover, apart from a cursory report of a flavoprotein dehalogenases, IYD is the only confirmed example of a flavoprotein that directly catalyzes reductive dehalogenation.^352,361 Mammalian IYDs possesses a highly conserved three-domain architecture consisting of a membrane anchor region (~20 amino acids) and intermediate domain with no homology to any known protein domains (~50 amino acids), and the C-terminal Nitro-FMN-like domain (~200 amino acids).³⁶² Notably, IYD lacks a nicotinamide binding domain and relies on an external reductase to deliver NADPH. On the basis of sequence homology, IYD has been assigned to the Nitro-FMN superfamily of enzymes, which includes nitroreductases that catalyze the reduction of nitroaromatic compounds, flavin reductase P (FRP) which provides reduced FMN to luciferase in bacterial bioluminescence, and FMN-dependent NADH oxidase, all of which catalyze O₂-dependent two-electron redox chemistry. The C-terminal domain of IYD shows highest homology to bacterial Thermus thermophilius FMN-dependent NADH oxidase.³⁶³

Following the progress in molecular understanding of IYD, the unbound as well as co-crystal structures with 101 and 102 of the catalytic domain of mouse IYD (mIYD, Figure 34B) were recently reported.³³⁸ mIYD shows closest structural homology to BluB flavin destructase which catalyzes the O₂-dependent degradation of its FMN cofactor via serial single-electron transfer mechanism to form the lower axial ligand of vitamin B₁₂.³⁶⁴ Upon binding of the substrate, an active site lid sequesters the substrate from solvent in mIYD. The amino acid residues forming the active site lid align with the corresponding active site lid-forming residues in BluB and are distinct from the residues forming the active site lid in other members of the nitro-FMN superfamily. Hence, based on their FMN binding mode and alignment of amino acids forming their active site lids, BluB and mIYD form their own subfamily within the nitro-FMN superfamily. In contrast to other members of the nitro-FMN superfamily, the side chain residues of the active sites of mIYD and BluB do not coordinate with the pyrimidine moiety of FMN, hinting at the alternative redox mechanism employed by these enzymes. Moreover, the co-crystal structure of mIYD with 101 and 102 revealed polar interactions between the FMN pyrimidine and amino side chain residues with the zwitterionic side of the substrate, suggesting that the substrate plays a direct role in modulating the reactivity of the FMN cofactor (Figure 34B). It was recently demonstrated that titration of mIYD with the substrate mimic iodotyrosine leads to a switch from the two-electron oxidized and reduced forms of the IYD-bound FMN cofactor to the neutral one-electron reduced semiquinone. Hence, preliminary experimental evidence favors a serial single-electron transfer mechanism.³⁶⁵ While no direct interaction with the substrate halogen substituents were observed, the substrate C-I bond is optimally positioned for electron transfer just above C4a in the FMN isoalloxazine ring. Consistent with lack of halogen binding, the identity of the halogen (with the exception of fluorine) of the halotyrosine substrate does not significantly affect substrate binding.³⁶⁵ However, halogenation does appear to play a significant role in substrate recognition, potentially due to reduction of the pK_a of the phenol hydroxyl, which is suggested by the mIYD structure to coordinate the ribityl 2′-hydroxyl group of the FMN cofactor.³³⁸

While the mechanism of IYD remains speculative, its similarities to BluB and mIYD hint at a serial single-electron transfer strategy in contrast to the two-electron chemistry catalyzed by other members of the nitro-FMN superfamily. With respect to the nature of the substrate radical-generating species, earlier studies showed high affinity of IYD for substrates intended to mimic the keto form of the iodotyrosine.³⁶⁶ Taken together, these observations lead to a mechanistic proposal proceeding through a stabilized keto form of the iodotyrosine phenoxyl which undergoes halide elimination concomitant to a first single-electron transfer to generate the phenoxy radical, which is then reduced by a second single electron transfer by the neutral semiquinone to the deiodinated phenol (Figure 34C).³⁶⁵ In line with recent progress in the molecular understanding of IYDs, an improved understanding of its enzyme mechanism is surely on its way.

The structure of mIYD has had important implications for human health, providing a basis for mechanistic understanding of disease states. For example, mutations associated with hyperthyroidism in human DEHAL1 mapped to key residues involved in coordination of the FMN cofactor in mIYD.^338,352 In addition to its importance to human health, the widespread distribution among diverse phyla and oxygen-insensitivity make IYDs an attractive and underappreciated contender to bioremediation applications. Indeed, it has been shown that IYDs are also capable of dechlorination and debromination of halotyrosines.³⁶⁷ Moreover, further investigation of halotyrosine deiodinases in the context of the synthesis and degradation of abundant halotyrosine bacterial natural products could provide an under-appreciated link to global halogen cycling, and in turn expand the substrate scope of these unusual flavoproteins. While anaerobic dehalogenation continues to be a focal point for microbial bioremediation due to the status quo of B₁₂-dependent RDHs, IYDs open the door to aerobic strategies. Illustratively, RDH-mediated dehalogenation activity has been investigated in the anaerobic microbiome fraction of the marine sponge Aplysina aerophoba which harbors a diversity of bromotyrosine natural products.³⁶⁸ However, the aerobic communities where IYD-like enzymology might be present were neglected by this study.

6.1.2 Reductive dehalogenase from microbial natural product biosynthesis

Tremendous attention has been devoted to identifying microbial enzymes capable of dehalogenating anthropogenic pollutants toward the goal of bioremediation. Remarkably, the study of enzymatic dehalogenation as it relates to the formation and degradation of naturally produced halogenated molecules, as discussed in detail in the previous sections, remains largely uncharted. The bioaccumulation and recalcitrance of many anthropogenic pollutants to degradation is due to the high degree of halogenation that renders their carbon scaffolds inaccessible to oxidative catabolism.^369–371 An analogous problem manifests itself in the biosynthesis of the highly brominated marine microbial natural product 29 in which dehalogenation of the intermediate 24 is required to unmask the C-2 position of the pyrrole to furnish 2,3,4-tribromopyrrole (103), which is the physiological substrate for the CYP450-Bmp7 catalyzed oxidative coupling to 9 to afford 29. (Figure 35A).^38,49,372 This requisite dehalogenative transformation was recently shown to be enzymatically catalyzed by the unique thioredoxin-like reductive dehalogenase, Bmp8, encoded in the biosynthetic gene cluster responsible for the production of 29.³⁷² Of note, Bmp8 is the first, and to date only example of a dehalogenase enzyme that performs a dedicated role in the biosynthesis of a natural product. Moreover, characterization of the catalytic role of Bmp8 enabled a complete in vitro reconstitution of all enzymatic reactions in the biosynthesis of 29, a half century after its initial report as a remarkable marine microbial natural product bearing 70% of its molecular weight as halogen atoms.^373,374

Figure 35. Dehalogenation in natural products biosynthesis.

(A) The oxidative coupling reaction in the final step in the biosynthesis of 29 necessitates the dehalogenation of the highly brominated intermediate 24 to 103 catalyzed by the dehalogenase Bmp8. (B) Scheme depicting the role of the active site CXXC Cys residues in the Bmp8 reaction mechanism. The attacking Cys residue is denoted ‘CysA’, while the resolving Cys residue denoted ‘CysR’. (C) Proposed role for a Bmp8-like dehalogenase in the biosynthesis of 104.

Bmp8 was found to utilize a cofactor-independent reaction mechanism akin to the homologous thioredoxin AhpD. The primary amino acid sequence of Bmp8 possesses a signature thioredoxin motif consisting of a cysteine residue separated by two amino acid residues, followed by another cysteine residue (i.e. Cys-X-X-Cys, CXXC, Figure 35). As in the reaction mechanisms for thioredoxins, the side chain thiolates of the C- and N- terminal cysteine residues for the CXXC motif were found to act as the “attacking” electron donor that forms a thioether bond with the substrate and “resolving” nucleophile that liberates the substrate leading to the formation of an endogenous disulfide bond, respectively (Figure 35B). Indirect evidence suggests that the electron acceptor in the reaction catalyzed by Bmp8 is the C-2 bromine of 24. Attack at the C-2 bromine by the cysteinyl thiolate would result in the formation of a transient cysteinyl bromide (Figure 35B). The cysteinyl bromide is either directly thiolyzed by the side chain thiolate of the resolving active site cysteine to form an intramolecular disulfide bond (route ‘i’ in Figure 35B). In an alternate route, the cysteinyl bromide can be first hydrolyzed to form a sulfenic acid species, followed by disulfide bond formation and release of water (route ‘ii’ in Figure 35B) by the resolving cysteine. The proposed mechanism for Bmp8 is analogous to that presently favored for the peroxiredoxin-like thyroid hormone deiodinase which invokes the formation of selenyl iodide intermediate resulting from direct attack of the an active selenocysteinyl selenolate (in place of a cysteinyl thiolate) on the departing iodide, as previously discussed in Section 6.1.1.³⁴³ Notably, Bmp8 bears no sequence homology to thyroid hormone deiodinases, suggesting convergent evolution of their respective molecular mechanisms.

In line with cryptic halogenation events catalyzed by NHFe halogenases that participate in the construction of cyclopropane rings in natural product scaffolds (Section 4.2), Bmp8 too encrypts halogenation at the C-2 position of 24.³⁸ However, in contrast to NHFE halogenase-mediated cyclopropane ring formation, dehalogenation that occurs en route the biosynthesis of 103 and 29 is explicitly catalyzed by a dedicated dehalogenase enzyme. In light of the focus of this review on natural product halogenation biochemistry, it is worth highlighting the intimate link between halogenation and dehalogenation presented by the biosynthesis of 29. As detailed in Section 2.3, with the exception of Bmp2, all pyrrolyl-S-CP FDHs described to date halogenate pyrroles in a regioselective manner. In contrast, Bmp2 halogenates all four carbon atoms of the acyl-pyrrole in a regiopromiscuous manner, setting up a non-catalytic decarboxylation reaction to liberate 24. Therefore, the conversion of 24 to 103 by Bmp8 effectively confers back-end regiocontrol on the halogenation of the pyrrole by Bmp2, providing an alternative biosynthetic strategy for cryptic halogenation in natural product biosynthesis.

As the first example of a dehalogenase in natural products biosynthesis, Bmp8 expands the repertoire natural product biosynthetic logic to include not only enzymes that add halogens, but also enzymes that specifically eliminate them to facilitate downstream transformations. One example where the biochemical logic illuminated by the characterization of Bmp8 might be applicable is toward the biosynthesis of the N-C coupled bipyrrole natural product ‘Q1’ (104), a heptachlorinated molecule ubiquitously detected in the marine food web.^375,376 The fully halogenated structure of 104 consists of a tetrachlorinated pyrrole with N-1 coupling to the C-2 of a trichlorinated pyrrole (Figure 35C). In light of the discovery of Bmp8, one might envisage a pathway for the biosynthesis of 104 in which tetrachloropyrrole (105) is biosynthesized in a similar manner to its brominated counterpart 24 and enzymatically dehalogenated by a Bmp8-like dehalogenase. Subsequent oxidative coupling to another molecule of 105 can lead to the formation of 104 (Figure 35C). Lastly, an important distinction is to be made between Bmp8 and other microbial dehalogenases characterized to date, in that, the definition of Bmp8 as a dehalogenase is derived from an explicit biosynthetic context. In contrast, other microbial “dehalogenases” discussed later in this section are so-named because of their ability to dehalogenate compounds, which may or may not be their physiological substrates. In this respect, dedicated natural product biosynthetic dehalogenases tailored by evolution to act efficiently on manifold organic scaffolds could lead the way to more efficient dehalogenation biocatalysts than those discovered outside of a natural context.

6.2 Reductive dehalogenases acting on anthropogenic substrates

Over the course of last two decades, numerous bacterial RDHs have been characterized via functional proteomics or transcriptional assays, revealing a repertoire of anthropogenic substrates, which include halogenated phenols, dioxins, biphenyls, and aliphatic hydrocarbons.¹⁶ Meanwhile, many hundreds more homologous sequences await functional characterization.³⁷⁹ The following section describes the discovery and structure-function details for two classes of RDHs that act on anthropogenic substrates.

6.2.1 Reductive dehalogenases from halorespiring prokaryotes

Anaerobic halorespiring bacteria utilize a broad spectrum of organohalogens as terminal electron acceptors in energy-conserving and catabolic processes using metalloenzymes. With the exception of a single reported example of heme-dependent RDH, the majority of RDHs characterized from phylogenetically diverse genera of halorespiring bacteria (and archaea) are B_12/iron-sulfur (Fe-S) cluster-containing RDHs and are classified as their own subclass of oxidoreductases.^377–379 Cytoplasmic B₁₂/Fe-S-containing RDHs are also found in the context of organohalogen catabolic pathways in microaerophilic bacteria.³⁸⁰ These RDHs contain two Fe-S clusters (usually two 4Fe-4S clusters, but sometimes one 4Fe-4S and one 3Fe-4S). Respiratory dehalogenases possess N-terminal signal sequences used for transport of the pre-folded protein to the periplasm, while catabolic dehalogenases possess an NAD(P)H binding domain that is absent in respiratory dehalogenases.³⁸⁰ In addition to significant sequence variation, B₁₂-dependent RDHs are highly dissimilar at the amino acid level from the other major subfamilies of B₁₂-dependent methyltransferases and isomerases that catalyze radical Co(III)-C chemistry.³⁷⁸

The sub-families of B₁₂-dependent RDHs utilize a spectrum of organohalogens, but individual RDHs possess strict substrate specificities and regiocontrol with respect to haloelimination.³⁷⁸ Historically the detailed mechanistic characterization of B₁₂-dependent RDHs has been hindered by slow growth rates of the native halorespiring microbes that harbor them, as well as challenges in their heterologous expression, and the instability of the proteins under aerobic conditions.¹⁶ Functional heterologous expression of a B₁₂-dependent trichloroethene (106) dehalogenase PceA from Desulfitobacterium hafniense strain Y51 was only recently reported in the non-dehalogenating B₁₂-producing bacterium Shimwellia blattae over a decade after its initial characterization.^381,382

Two recent breakthrough structure-function studies of oxygen-tolerant B₁₂-dependent dehalogenases—a canonical respiratory PceA functionally characterized nearly two decades prior, and a newly characterized catabolic 2,6-dibromophenol dehalogenase (NpRdhA)—are strongly suggestive of dehalogenation mechanism invoking unprecedented Co-halogen chemistry (Figure 36A–C).^383–385 Previously, three mechanistic routes were suggested for B₁₂-dependent RDHs beginning with the coordinated Co in its initial Co(I) state (Figure 36D). The first mechanistic proposal (‘Route I’ in Figure 34D) invokes formation of Co(III)-C bond with the organic backbone of the substrate. This mechanism is inspired by Co(III)-C chemistries catalyzed by B₁₂-dependent methyltransferases and isomerases, and is unsupported experimentally in RDHs.³⁷⁸ A second route (‘Route II’ in Figure 36D) supported by limited experimental evidence is initiated by a single electron transfer from Co(I) to the organohalogen substrate leading to the formation of a transient substrate radical.³⁸⁶ Halide elimination is followed by single electron transfer leading to the reduced substrate, or by addition of the organic substrate to cobalt to again form an organocobalt intermediate (‘Route III’ in Figure 36D). A co-crystal structure of PceA with substrate 103 shows the dichlorinated end of 106 in proximity to the coordinated Co, with the leaving chlorine positioned axially to the Co (Figure 36C). Further, iodide was shown to form a complex with Co(II) at the position aligning with the putative chlorine leaving group, suggestive of a weak halide-binding site.³⁸³ While a co-crystal structure of NpRdhA with the substrate 3,5-dibromo-4-hydroxybenzoic acid (107) has not been reported yet, computational docking of the substrate in the enzyme active site places the departing bromine atom in proximity to the Co ion. Further, EPR analysis of NpRdhA with 107 strongly suggested the formation of a Co(II)-bromine adduct.³⁸⁴ Active site sterics in PceA and NpRdhA exclude the formation of an organocobalt intermediates with respective to substrates, but still accommodate the previously proposed radical mechanism (‘Route II’ in Figure 36D). However, structure-function evidence strongly favors a new paradigm of B₁₂-mediated Co-halogen chemistry. The unusual ‘base-off’ configuration of the norpseudo-B₁₂ and B₁₂ cofactors of PceA and NpRdhA, respectively, further alludes to unconventional B₁₂-mediated chemistry. Hence, the structure-function studies of PceA and NpRdhA disfavors Co-C chemistry (‘Routes I and III’ in Figure 36D), while suggesting an unanticipated alternative mechanism (‘Route IV’ in Figure 36D) proceeding through a Co(III)-halide intermediate.

Figure 36. Structure-mechanism B₁₂-dependent RDHs.

(A) Dehalogenation reactions catalyzed by PceA (top) and NpRdhA (bottom). (B) Overall crystal structures of PceA (PDB: 4UR0) in magenta and NpRdhA (PDB: 4RAS) in green. (C) Active sites of PceA and NpRdhA showing the B₁₂ cofactor in yellow and the Fe-S clusters. In the PceA structure, two alternate conformations of 106 are modeled above the Co, while the similar site in NpRdhA is occupied by a chloride ion shown as a grey sphere. (D) Scheme showing hypothesized mechanistic routes for dehalogenation by B₁₂-dependent RDHs illustrated with NpRdhA substrate 107.

B₁₂-dependent RDHs exhibit strict substrate specificities and regioselectivities with respect to haloelimination. For example, PceA selectively dehalogenates the dihalogenated end of 106, while NpRdhA exhibits substrate specificity for ortho-substituted halophenols, with a strong preference for 2,6-dihalogenated phenols.^383,384 In PceA the substrate first encounters a “letterbox” shaped substrate filter that leads into a narrow hydrophobic channel that opens up into an amphiphilic substrate binding pocket situated at β-face of the B₁₂ cofactor confirmed by co-crystallization of PceA with 106 (Figure 36C). Investigation of the brominated analog of the product cis-dichloroethene- cis-dibromoethene demonstrated a vacancy at the position of the chlorine nearest the coordinated Co ion in the co-crystal structure with 106, consistent with the regioselectivity of dehalogenation.³⁸³

Over three decades after the discovery bacterial reductive dehalogenation and two decades after the first report of the first RDH sequence, structural and mechanistic understanding of B₁₂-dependent RDHs remains at its infancy.^387–389 Further structural studies sampling from the wide sequence space of B₁₂-dependent RDHs promise to shed light on the basis for substrate selectivity, and aid in the engineering of dehalogenases with expanded substrate scopes. In addition to anticipated differences for respiratory versus catabolic dehalogenases, the structures of PceA and NpRdhA exhibited striking similarities. Intriguingly, the B₁₂ binding domains of both RDHs are structurally homologous to that of the human B₁₂-trafficking chaperone CblC, suggesting, for the first time, a distant common ancestor for B₁₂-dependent RDHs.^383,384 As B₁₂-dependent RDHs vary widely in the primary amino acid sequence homology, the extent of their structural diversity remains to be explored through further structure-function studies from within the subfamily. While initial reports have indicated the potential for reductive dehalogenation in pristine unpolluted environments, nothing is known of the physiological substrates of RDHs.³⁸⁹ Study of the native transformation of RDHs may further shed insight into their evolutionary origins. Finally, the reader is referred to recent reviews on the B₁₂-dependent RDHs for a more extensive treatment of the topic.^16,379

6.2.2 Glutathione S-transferase-like reductive dehalogenases

An additional well-studied class of reductive dehalogenase is a glutathione S-transferase (GST) superfamily protein capable of dehalogenating the toxic intermediate tetrachlorohydroquinone (TCHQ, 108). The product 108 is generated by the oxidative dehalogenation of the anthropogenic compound pentachlorophenol (109) followed by hydrogenation in the first two steps of the pathway catabolizing 109 in the aerobic bacterium Sphingomonas chlorophenolica (Figure 37A).^390–392 TCHQ-dehalogenase, PcpC, catalyzes the dechlorination of 108 to trichlorohydroquinone (TriCHQ), and then to dichlorohydroquinone (DCHQ), which subsequently undergoes enzymatic oxidative ring opening, followed by two rounds of abiotic dehalogenation to feeding into the catabolic β-ketoadipate pathway (Figure 37A).³⁹³ The primary sequence of the active site of PcpC is identical to that of glutathione-dependent malelylacetoacetate (MAA) isomerase involved in the catabolism of aromatic amino acids,³⁹⁴ suggestive of the recent recruitment MAA isomerase into a bona fide pentachlorophenol degradation pathway.

Figure 37. Bacterial degradation of 109.

(A) Bacterial pathway for degradation of 109. (B) Mechanism for dehalogenation of 109 by GST-like PcpC. Keto-enol tautomerization is followed by conjugation to glutathione (GS⁻). Attack by an active site cysteine thiolate forms a protein-glutathione thioester and the dehalogenated product. The catalytic cycle is completed by the resolution of the protein-glutathione thioester by an exogenous glutathione thiolate.

A series of elegant studies have elucidated the thiol-mediated redox mechanism utilized by TCHQ-dehalogenase.^394–398 Mirroring the initial steps proposed for dehalogenation of iodotyrosine by IYD, 108 is deprotonated to afford a keto intermediate that then loses a chloride as a leaving group to afford a diketo species, which is subsequently conjugated to a glutathione molecule by nucleophilic addition (Figure 37B). The substrate-glutathione thioester is degraded by attack of an active site cysteine thiolate. The active site cysteine-glutathione thioester is subsequently resolved by attack by a glutathione thiolate to return the enzyme to its catalytically active state. A second glutathione-dependent GST-like dehalogenase, 2,5-dichloroquinone dehalogenase LinD, has been characterized from Sphigobium spp., but to a lesser extent than PcpC.³⁹⁹ GST-like dehalogenases add to a growing number of reductive dehalogenases from aerobic organisms that challenge the status quo of reductive dehalogenases from anaerobic environments. The recruitment of primary metabolic enzymes to perform “non-essential” functions is a theme shared between the pathways of secondary metabolism and catabolic pathways involved in the degradation of xenobiotics, and hints at a vast diversity of strategies recruited by Nature to cope with the manifold organohalogens of both anthropogenic and natural origin.³⁹³

6.3 Oxidative dehalogenases

Oxidative dehalogenation involves the replacement of a halogen with a hydroxyl group derived from molecular oxygen or hydrogen peroxide. A diverse array of enzymes catalyze oxidative haloaromatic dehalogenations, including flavin-dependent oxygenases, cytochrome P450 oxidases, and dehaloperoxidases.^400–402 The best studied oxidative dehalogenase is the pentachlorophenol-4-monooxgygenase, PcpB, an O₂/NADPH-dependent flavoprotein monooxygenase that catalyzes the conversion of 109 to tetrachloroquinone in the bacterial catabolism of 109 (Figure 37A).⁴⁰³ In contrast to the other dehalogenating enzymes described in this review, many enzymes that catalyze oxidative dehalogenases are promiscuous against a broad range of substrates.^404,405 Hence, in the absence of a physiological context, it is unclear whether the dehalogenation activity of many oxidative dehalogenases is an incidental laboratory observation or evolved specifically for the dehalogenation of man-made organopollutants. For example, the oxidative dehalogenation of 109 by PcpB is highly inefficient and the enzyme shows substrate promiscuity for the downstream product 108, suggesting its function in the degradation pathway of 109 could be happenstance.⁴⁰⁶ To maintain the focus of the review to address dehalogenation mechanisms in the context of substrate-specific enzymes, the reader is referred to a series of recent reviews on the topic of oxidative enzymatic dehalogenation.^400–402

6.4 Hydrolytic dehalogenases

In contrast to previously described dehalogenation strategies, hydrolytic dehalogenases (HDHs) catalyze the redox neutral substitution of a halogen by a hydroxyl derived either from water or a vicinal alcohol.³³⁴ At present, HDHs that dehalogenate a variety of anthropogenic pollutants comprise the most diverse sub-class of dehalogenase enzymes. Compared to their RDH counterparts, the characterization of HDHs has proceeded at a rapid pace due to their facile heterologous expression. Indeed, the first reported HDH crystal structure, that of an α/β-hydrolase-superfamily haloalkane dehalogenase, preceded that of B₁₂-dependent RDHs by nearly a quarter of a century.⁴⁰⁷ In subsequent years, structures of haloacid, 4-chorobenzoyl-CoA, haloalcohol, and 3-chloroacrylic acid dehalogenases have joined the gamut of structurally described HDHs.³³⁴

The mechanisms of HDHs can be divided on the basis of substrate specificity of the enzymes. Haloalkane, haloacid, and 4-chlorobenzoyl-CoA HDHs are united by the formation of a covalent bond between the substrate and an active site aspartate residue that results in elimination of a halide in the first half reaction (Figure 38A,B).³³⁴ The aspartyl-substrate oxoester is then hydrolyzed in the second half reaction by an activated water molecule that is deprotonated by an active site base. Recognition of the substrate halogen by an active site halide binding site is critical to substrate binding and stabilization of the transition state in the first half-reaction.⁴⁰⁸ Interestingly, the only salient difference between the abovementioned HDHs and a canonical hydrolase is the exchange of the hydrolytic serine/cysteine residue (as in a Ser-His-Asp/Glu catalytic triad) with an aspartate. A remarkable bacterial HDH hydrolyzes the carbon-fluorine bond in the highly toxic natural product fluoroacetate.^409,410 Despite sequence similarity to other members of the haloacid dehalogenases, fluoroacetate dehalogenase (FAcD) shows highest affinity for the fluorinated substrate in comparison to other differentially halogenated acetates.¹⁷ While the mechanism of FAcD proceeds via the anticipated aspartyl oxoester, its unique ability among haloacid dehalogenases to catalyze hydrolysis of a C-F bond has been attributed to an active site halogen binding site tailored to the small ionic radius of fluorine.⁴¹⁰

Figure 38. Structure-mechanism for HDHs with a catalytic triad.

(A) Proposed reaction mechanism for haloalkane/haloacid HDHs. (B) Overall structure (left) and active site (right) for a recently reported structure of a haloalkane HDH from a marine Rhodobacteraceae sp (PDB: 4C6H). The catalytic Asp residue is shown with the carbon atoms colored cyan. The His and Glu residue comprising of the catalytic triad are shown with carbon atoms colored green. The in situ generated product from the substrate 1-bromohexane used in the crystallization trials, 1-hexanol, is shown bound to the active site in two alternate conformations in yellow. A chloride ion bound in a hydrophobic pocket in the vicinity of the product is shown as a grey sphere and likely denotes the binding site for the substrate halogen atom. (C) Proposed reaction mechanism for 4-chlorobenzoyl-CoA HDH.

In contrast to the HDHs that catalyze a two-step intermolecular substitution reaction, haloalcohol dehalogenases catalyze single-step dehalogenative intramolecular substitution reactions for halohydrin substrates. In this mechanism the alcohol vicinal to departing halogen substituent is deprotonated by a Ser-Tyr-Lys/Arg catalytic triad followed by intramolecular substitution of the vicinal halogen leading to formation of an epoxide (Figure 39A–C).^334,412 Halogenation of alkyl centers renders them inherently activated to inter- or intramolecular substitution reactions, a strategy routinely exploited in synthetic organic chemistry as well as natural product biosynthetic routes such as that for merochlorin biosynthesis. The dehalogenative substitution reactions catalyzed by HDHs are in stark contrast with the intricate radical chemistry employed by NHFe halogenases to halogenate unactivated alkyl centers. Interestingly, haloalcohol dehalogenases belong to the short chain dehydrogenase/reductase (SDR) superfamily of dinucleotide-binding redox enzymes. However, haloalcohol dehalogenases lack the dinucleotide binding motif, and structural analysis reveals that this domain has been substituted by bulky amino acid residue side chains that compose a halogen binding site.^334,412 In addition to dehalogenation, the epoxide cyanolysis and azidolysis side activity of the HDHs have been utilized to generate β-hydroxynitriles and azides in conjunction with site-directed enzyme engineering efforts.^413–416

Figure 39. Structure-mechanism for SDR-like HDHs.

(A) Proposed reaction mechanism for halohydrin dehalogenases and recruitment of a cyanide anion to resolve the dehalogenated epoxide product to generate β-hydroxynitriles. (B) Structural similarity of between halohydrin HDHs (in green, PDB: 1PWZ) and lanthipeptide oxidoreductase SDR ElxO (in brown, PDB: 4QEC). NAD+ bound to ElxO is shown in yellow sticks. (C) Active site of the halohydrin HDH HheC with an epoxide product bound to the active site (in yellow), and the side chains of the catalytic Tyr (in cyan), and the Ser and Arg (in green) residues. A chloride ion bound in the vicinity of the product is shown as a grey sphere.

Yet another dehalogenation mechanism is employed by the 3-chloroacrylic acid dehalogenase, CaaD, that adds catalytic base-activated water to the conjugated substrate double bond leading to the formation of a tetrahedral transition state that then leads to the departure of the halide.⁴¹⁹ In essence, CaaD combines the two half reactions catalyzed by the haloalkane/haloacid HDHs into a single concerted reaction. While the role of the CaaD active site residues and the operative reaction mechanism has been a matter of extensive computational and experimental studies,^420–424 CaaD is the first member of the tautomerase superfamily with hydratase activity.⁴²⁵ Moreover, CaaD dehalogenases acting upon both cis- and trans-3-chloroacryclic acid substrates have been characterized. A more recent addition to the group of hydrolytic dehalogenases is the 2-haloacrylate hydratase that unexpectedly requires reduced flavin (FADH₂) for catalytic proficiency. While the role of the flavin in the reaction mechanism remains unknown, it has been suggested to either act as an acid-base catalyst as in the flavin-dependent isopentenyl diphosphate isomerase, or a transient electron donor as in the mechanism of flavin-dependent chorismate synthase.⁴²⁶ Further investigations of this intriguing enzyme promise to shed light on its surprising catalytic requirement.

A remarkable diversity of enzymes catalyzes variations on the theme of hydrolytic dehalogenation from the haloalkane dehalogenases that differ only modestly from their hydrolase counterparts, to haloacid dehalogenases that define a superfamily of their own. Hence, a careful study of these varied enzymes promises to provide insights into the evolution of enzymatic dehalogenation, in particular addressing the question of modern versus ancient origins of evolved dehalogenase function.⁴⁰⁴ For detailed account of hydrolytic dehalogenation, the reader is referred to several dedicated reviews on the topic.^17,334

7. Future perspectives

Natural product biosynthetic schemes continue to provide examples of novel chemical transformations catalyzed by enzymes.^427,428 Halogenation biochemistry is no different. Historically, as has been the underlying motivation of this review article, natural product chemistry has led to the discovery and characterization of novel halogenation biocatalysts, which, in turn, have inspired the development of novel halogenating chemical catalysts as well.⁴²⁹ This process has not been exhausted. A recent report describing the structure elucidation and biosynthetic gene cluster discovery of cyanobacterial bartolosides has uncovered a potentially novel diiron-dependent halogenase,^430,431 with homologs present in other cyanobacteria, including in a cyanobacterial gene cluster encoding the production of structurally analogous halogenated cyclophane natural products.⁴³² Furthermore, a recent report challenges our understanding of the catalytic scope of FDHs which were previously thought to only halogenate aromatic rings or electron rich activated aliphatic carbon centers. Discovery and characterization of a fungal bifunctional methyltransferase-halogenase enzyme established the gem-dichlorination on a non-activated carbon atom,⁴³³ a transformation previously within the realm of NHFe halogenases proceeding with the halide radical route. Another recently described FDH, Amm3,¹⁷⁹ is a close homolog of the RiPP FDH MibH, perhaps postulating towards the involvement of peptidic substrates in the elaboration of marine halogenated pyrroloquinoline alkaloids,⁴³⁴ an assertion that is supported by the presence of numerous biosynthetic enzyme encoding genes with resemblance to RiPP dehydratases in the recently described amm biosynthetic gene cluster.¹⁷⁹

Further novel halogenases await discovery. A recent report describing the regio-, stereo-, and enantioselective dihalogenation of allylic alcohols during the course of the synthesis of halogenated terpenes such as the anticancer candidate molecule halomon posits an as yet undiscovered enzymatic halogenation strategy for the biosyntheses of these ubiquitous marine macroalgal natural products.⁴³⁵ Enantioselective dichlorination has also allowed access to chlorosulfolipid marine natural products,⁴³⁶ among other recent efforts,^437,438 bringing into focus the exquisite control exercised by these as yet undiscovered marine halogenases to furnish structures that are otherwise unrealized by the enzymatic halogenation routes characterized thus far. Without a doubt, polyhalogenated natural products continue to inspire synthetic chemistry efforts.⁴³⁹ Directed evolution of halogenases too is expanding the substrate scope of halogenating enzymes, pushing them as viable complementary catalysts in complex synthetic schemes.^440–442

The inventory of experimentally characterized halogenases is dwarfed by both known halogenated natural products, as well as putative halogenases that are detected in sequence databases. While sequence databases are expanding at an exponential pace, the sequence space explored by halogenases is restricted, for the most part, to culturable bacteria. Two sets of contemporary technological advances promise to affect a paradigm shift in where the halogenases of the future will come from. First are techniques such as metagenomics and single-cell genomics that allow the interrogation of complex microbial communities without a priori constraints of culturing or enrichment.^443–445 Coupling these to advances in synthesis of large DNA sequences, direct capture of natural product gene clusters from a complex pool of heterogeneous DNA, and iteratively teaching computational algorithms to find novel classes of natural product gene clusters will revolutionize our access to sequence space that has not yet been explored. Particularly with respect to natural product halogenation biochemistry, for instance, filter-feeding benthic marine invertebrates such as sponges are one of the most prolific producers where halogenated natural products can constitute up to 10% of the sponge dry weight.^{34,347,446,447} Marine sponges are the ‘petri-dishes of the oceans’ harboring an extraordinary diversity of endosymbiotic microbiome that has been postulated, and in many cases experimentally verified, to be the source of the natural product diversity.^448–451 It is thus noteworthy that while the inventory of FDHs from sponge-microbiome consortia has attracted recent interest,^452,453 to the best of our knowledge, no sponge-derived FDH, V-CPO, V-BPO, or SAM-dependent halogenase has been confirmed within genomic or biochemical context for the production of a halogenated natural product. A singular instance of homologs of NHFe halogenases detected in sponge microbiomes have been paired with natural product biosynthetic chemistry.⁸¹

The second advance is the interrogation of halogenation biochemistry, at the genetic and biochemical levels, for the production of halogenated natural products by eukaryotic sources. Notable examples from halogenases encoded within fungal natural product biosynthetic gene clusters have been mentioned above. Again, looking towards to the marine environment, an area that has been genetically unexplored, is the production of halogenated natural products by macroalgae and seaweeds. In particular, marine red and brown algae have been celebrated producers of bioactive halogenated natural products, some with curious accumulation of iodide from the seawater with postulated roles in self-defense against biofouling.⁴⁵⁴ In addition to pharmaceutical candidates, several thousand tons of ozone-damaging polyhalogenated methanes, such as bromoform and dibromomethane, are produced by marine eukaryotes every year that saturate the marine layer and contribute significantly to ozone-depletion.^455–457 Nearly forty years ago, the pioneering work of Hager and coworkers hinted towards the contribution of hydrogen peroxide-dependent marine macroalgal haloperoxidases in the production of bromoform using biomimetic substrates.^458,459 Despite the environmental and ecological significance, scant progress^460–462 has been made to identify the genetic and biochemical bases for the production of polyhalogenated methanes in the oceans, and the physiological substrates for the abovementioned macroalgal haloperoxidases remain unidentified. In fact, while numerous other marine algal haloperoxidases have also been identified and their biochemical activity characterized with respect to biomimetic substrates and model bioconversions,^189,194,212 to date, the dedicated genetic and biochemical contribution of even a single macroalgal halogenase in a natural product biosynthetic scheme has not been established. An untapped reservoir of eukaryotic halogenases presents itself that should become accessible as the scientific community realigns its focus to include increasingly more eukaryotic sources to interrogate for secondary metabolic schemes. A recent report describing the contribution of bromide as an essential micronutrient for the posttranslational sulfilimine formation bears testimony to the presence of halogenation biochemistry in all kingdoms of life.⁴⁶³

In light of halogenation biochemistry discussed thus far, it is interesting to note that the molecular basis for halide specificity for halogenases has not been elucidated. Numerous examples of marine brominated natural products have been presented, including 24, 29, 31, and 103, for which the biosynthetic gene clusters are known and experimentally validated.^50,107 However, the molecular sieve that allows marine brominases to selectively incorporate bromine in natural product structures against a greater than 500-fold excess of chloride in seawater remains elusive. Characterizing this selectivity promises to open avenues for generating differentially halogenated natural products, further pushing forward the already impressive contribution of halogen atoms in conferring favorable bioactivities to natural product scaffolds.

Contrasting the cornucopia of halogenases discovered from natural products biosynthesis described in this review, to date, only a solitary dehalogenating enzyme- Bmp8, has been characterized from the context of a microbial natural product biosynthetic pathway. This is in addition to the two physiological deiodinases described from thyroid hormone maturation and iodide salvage. A common feature of these physiological dehalogenase enzymes is that they partially undo the handiwork of halogenation biocatalysts. Furthermore, apart from fluoroacetate hydrolytic dehalogenase, the enzymatic degradation of the >5,000 known organohalogen natural products remains almost entirely unexplored as a source of dehalogenating enzymes. What is perhaps most striking of all is that, for the most part, the cofactors utilized by known halogenation enzymes is a mutually exclusive set from the cofactors involved in enzymatic dehalogenation. Even in the rare case of the flavoenzyme reductive iodotyrosine dehalogenase, the FMN co-factor serve a reductive role in contrast to the oxidative role of the flavin co-factor found in FDHs. This co-factor gap might be inherent to chemical biology, but could just as well hint at the infancy of the field of biological dehalogenation. Despite the fact that enzymatic dehalogenation once superseded biological halogenation with respect to the diversity of enzyme strategies, the tables have quickly turned in the post-genomic era.⁴⁶⁴ In part, this knowledge gap maybe due to fact that the search for dehalogenases has, and continues to be biased toward a narrow spectrum of anthropogenic substrates. While clearly no dearth of dehalogenases has been discovered to date, how many more might be found if natural product biosynthesis and biodegradation are taken into account? Hinting at an answer to this question, studies have implicated a compelling biological contribution to global halogen cycling, which includes both the incorporation and removal of halogens from organic scaffolds.^465,466 Moreover, recent metagenomics surveys of pristine unpolluted terrestrial environments for sequences related to known classes of halogenating and dehalogenating enzymes have revealed an astounding diversity of biocatalysts with unknown substrates.^467,468

Biographies

Vinayak Agarwal received his Bachelors and Masters degrees in Biochemical Engineering from the Indian Institute of Technology, Delhi, before moving to the University of Illinois Urbana-Champaign for graduate studies with Dr. Satish K. Nair at the Center for Biophysics and Computational Biology. Graduating in 2012, he joined the laboratory of Dr. Bradley S. Moore at UC San Diego to interrogate marine halogenation biosynthetic pathways. He has received postdoctoral funding award from the Helen Hay Whitney Foundation (2014–2016) and the NIH Pathway to Independence Award (2016).

Zachary D. Miles obtained his Bachelor of Science degree from the University of Wisconsin-Madison in 2009. Soon after, he started his graduate work at the University of Arizona in the department of Chemistry and Biochemistry under the guidance of Prof. Vahe Bandarian, where he received both a National Science Foundation IGERT Program in Genomics Fellowship and a National Institutes of Health Biological Chemistry Training Program Fellowship. After graduating with his Ph.D. in 2014, he joined the laboratory of Prof. Bradley Moore and is currently studying the biosynthesis of marine meroterpenoids.

Jaclyn Winter received her B.Sc. in Chemistry and Molecular Genetics from the State University of New York at Fredonia in 2004 and completed her Ph.D. at UC San Diego under the mentorship of Dr. Bradley S. Moore in 2010. After graduate school, she completed postdoctoral training with Dr. Christian Hertweck at the Hans Knöll Institute and Leibniz Institute for Natural Products Research and Infection Biology in Jena, Germany (2010–2011) and with Dr. Yi Tang at UC Los Angeles (2011–2014). During her graduate and postdoctoral studies, she was a recipient of a NIH Marine Biotechnology Predoctoral Ruth L. Kirschstein National Research Service Award and the L’Oréal/AAAS USA Postdoctoral Fellowship for Women in Science. She is currently an Assistant Professor in Medicinal Chemistry in the College of Pharmacy at the University of Utah, Salt Lake City, UT.

Alessandra S. Eustáquio graduated from the University of São Paulo, Brazil, with a degree in Pharmacy and Biochemistry, before obtaining a PhD from the University of Tübingen, Germany, mentored by Prof. Lutz Heide. After being a Life Sciences Research Foundation postdoctoral fellow at Prof. Bradley S. Moore’s laboratory at the University of California San Diego, Alessandra joined Pfizer’s Natural Products group as a Principal Scientist, where she worked with natural product drug development for four years. Alessandra has been an Assistant Professor in the Department of Medicinal Chemistry and Pharmacognosy of the University of Illinois at Chicago since August 2015.

Abrahim El Gamal graduated from Stanford University in 2011 with a B.S. in Chemistry and M.S. in Chemical Engineering. Shortly thereafter he moved to the University of California, San Diego where he completed his Ph.D. work in the laboratory of Dr. Bradley S. Moore in the summer of 2016. During his graduate career, Abrahim was a recipient of a U.S. National Institutes of Health Marine Biotechnology Training Grant pre-doctoral fellowship. He is currently a postdoc with Dr. Chaitan Khosla at Stanford University.

Bradley Moore is Professor of Marine Chemical Biology at the Scripps Institution of Oceanography and Professor and Chair of Pharmaceutical Chemistry the Skaggs School of Pharmacy and Pharmaceutical Sciences at UC San Diego. He holds degrees in chemistry from the University of Hawaii (B.S. 1988) and Washington (Ph.D. 1994), was a postdoc at the University of Zurich (1994–1995), and held prior faculty appointments at the University of Washington (1996–1999) and Arizona (1999–2005). His research interests involve exploring and exploiting marine microbial genomes to discover new biosynthetic enzymes, secondary metabolic pathways, and natural products for their biomedical and biotechnological utility.