CHAPTER 1: Biochemical Diversity of 2-Oxoglutarate-Dependent Oxygenases

This chapter summarizes the diverse array of biochemical transformations that are catalysed by Fe(ii)- and 2-oxoglutarate (2OG, also known as α-ketoglutarate)-dependent oxygenases.^1–5  As described more comprehensively in Chapter 2, most of these enzymes coordinate their active site metal at one end of a double-stranded β-helix fold (Figure 1.1A)^6–8  using a 2-His-1-carboxylate motif, bind 2OG within the core, and interact with their primary substrates using less conserved regions of the core and additional loops. Most representatives of this enzyme family catalyse hydroxylation reactions (Figure 1.1B), but desaturation, ring formation, ring expansion, halogenation and other types of chemistry are known, as described for several examples in later sections of this chapter. The mechanisms of these enzymes are detailed in Chapter 3, supported by chemical model investigations summarized in Chapter 4. A highly simplified mechanistic scheme for hydroxylases based on investigations of TauD (taurine hydroxylase), a well-characterized taurine-degrading enzyme (see Section 1.6.4),⁹  is depicted in Figure 1.1C. As illustrated, the binding of 2OG displaces two waters from an octahedral Fe(ii) site (converting state a to state b); a third water is lost as the primary substrate binds (generating state c); the newly vacant coordination site is used to react with O₂ to form an Fe(iii)–superoxo species (state d); oxidative decarboxylation of 2OG yields an Fe(iv)–oxo species (state e, also known as the ferryl intermediate); this powerful oxidant abstracts a hydrogen atom from the substrate to generate a radical and Fe(iii)–hydroxide (state f); and hydroxyl radical rebound (or more complex chemistry)¹⁰  completes substrate hydroxylation and recycles the enzyme to the initial Fe(ii) state. Such enzymes are dioxygenases because both atoms of oxygen are incorporated into substrates, with one atom ending up in succinate and the second in the hydroxylated product (which decomposes spontaneously in some reactions).¹¹  These enzymes are correctly referred to as primary substrate:2OG dioxygenases or primary substrate hydroxylases; it is incorrect to cite these enzymes as primary substrate dioxygenases or 2OG dioxygenases because these names imply both atoms of oxygen become incorporated into a single substrate. The other types of chemistries associated with this family of enzymes are also likely to utilize the Fe(iv)-oxo intermediate.

The total number of 2OG-dependent oxygenases found in biology is immense. For example, humans and other animals are thought to possess about 80 such enzymes.⁷  Even more are predicted to be present in plants such as the model organism Arabidopsis thaliana.¹²  Some 2OG-dependent oxygenases are widely distributed throughout aerobic life forms, whereas highly specialized enzymatic catalysts have evolved in various microorganisms or plants for synthesis or degradation of compounds found in their unique niches. In the following sections, the vast array of reactions catalysed by 2OG-dependent oxygenases are divided into seven categories. Section 1.2 describes the 2OG-dependent enzymes acting on proteins. These reactions may lead to structural consequences, have oxygen-sensing functions, alter histone properties, or possess other roles. Section 1.3 summarizes the metabolism of 2OG-dependent oxygenases acting on polynucleotides. The functions of these enzymes include DNA or RNA repair of alkylation damage, roles in transcriptional regulation, biosynthesis of base J in DNA or of specific tRNAs, and demethylation of 5-methylcytosine. Enzymes involved in lipid-related metabolism are discussed in Section 1.4. Examples include reactions related to carnitine biosynthesis, the degradation of phytanic acids, and the decoration of ornithine lipids or lipid A. Plant-specific representatives and close relatives are highlighted in Section 1.5. Of interest are reactions related to synthesis of flavonoids and anthocyanins, gibberellins, alkaloids, and other metabolites found predominantly in plants. Section 1.6 covers enzymes that act on a variety of small molecules including free amino acids, nucleobases or nucleosides, herbicides, sulfonates/sulfates and phosphonates. In some cases, the products resulting from these transformations are incorporated into antibiotics whereas other reactions are involved in metabolite recycling or alternative biochemical pathways. Additional 2OG-dependent oxygenases utilized for antibiotic biosynthesis are provided in Section 1.7. Examples include several halogenating enzymes and other representatives derived from bacterial and fungal sources. Finally, Section 1.8 covers enzymes that are related in structure or mechanism to the 2OG-dependent oxygenases. These include isopenicillin N synthase and the plant-specific ethylene-forming enzyme, which contain the 2OG oxygenase fold yet fail to utilize 2OG, and two enzymes that share a distinct fold and use the alternative oxo-acid 4-hydroxyphenylpyruvate.

This section describes 2OG-dependent dioxygenases that act directly on protein side chains or at modified sites in proteins and catalyse the reactions summarized in Figure 1.2.

The earliest investigations of the 2OG-dependent dioxygenases were focused on enzymes acting on protein substrates with structural roles. For example, seminal studies from the 1960s demonstrated that prolyl 4R-hydroxylase (Figure 1.2A) utilizes 2OG as a cosubstrate for its function in collagen synthesis.^13,14  Similar findings were observed for two other enzymes needed to functionalize collagen, prolyl 3S-hydroxylase (Figure 1.2B)¹⁵  and procollagen lysyl 5R-hydroxylase (PLOD, Figure 1.2C).¹⁶  By 1982, a remarkably prescient mechanism was proposed for prolyl 4R-hydroxylase and contained many of the features shown in Figure 1.2C, including coordination of Fe(ii) by a facial triad, bidentate coordination of 2OG, and the generation of a reactive metal–oxo intermediate.¹⁷  Collagen, the most abundant protein in mammals, is first synthesized as procollagen, which undergoes extensive modifications including the formation of 4R-hydroxyproline (4Hyp, accounting for 10% of its residues), 3S-hydroxyproline (3Hyp, ∼1% of its residues), and 5R-hydroxylysine (Hyl, 0.5–7% of its residues) prior to assembly into its triple helical structure and export to the extracellular matrix.¹⁸  4Hyp is also found in elastin and other human structural proteins, plant cell wall components, and various other proteins or peptides of algae, selected bacteria, and even a virus.¹⁹  Similarly, 3Hyp and Hyl have been detected in non-collagen proteins. The collagen-specific prolyl 4R-hydroxylases form heterodimers with protein disulfide isomerase, but only the homomeric enzymes from the bacterium Bacillus anthracis and the alga Chlamydomonas reinhardtii have been crystallized.^20–22  These structures reveal the basis for stereospecificity and substrate specificity. The procollagen-specific 2OG-dependent dioxygenases are described in greater detail in Chapter 5.

An O₂-sensing role for 2OG-dependent dioxygenases was uncovered in 2001.^23,24  A key component of this signalling pathway is the hypoxia-inducible factor (HIF-1) which directs the transcription of several genes under low O₂ conditions. The two subunits of this heterodimer (HIF-α and HIF-β in humans) are constitutively synthesized, but human HIF-α is modified in cells under normoxic conditions by three HIF-α-specific prolyl 4R-hydroxylases (PHDs, Figure 1.2A). These enzymes act in a tissue-specific manner to oxidize Pro-402 and Pro-564 of HIF-1α, which results in enhanced affinity of the protein towards the von Hipple–Lindau (VHL) tumour suppressor protein, elongin B and elongin C,^25,26  leading to polyubiquitinylation and proteosomal destruction of the transcription factor. The structure has been described for PHD2²⁷  and for PHD2 in complex with HIF-1α,²⁸  providing great insight into the geometric aspects of catalysis.

Independent of the PHD-specific process, a separate hydroxylation event involving Asn-803 of HIF-1α occurs in the presence of oxygen.²⁹  This modification is carried out by another 2OG oxygenase: factor-inhibiting HIF (FIH), an asparaginyl 3S-hydroxylase (Figure 1.2F).³⁰  The addition of a single atom of oxygen to HIF-1α results in loss of interaction with the p300/CBP transcription coactivators and abrogation of hypoxic signalling.^31,32  Several structures have been reported for FIH, including its interaction with a peptide derived from the coactivator.^33–35 

The O₂-sensing functions of the PHD and FIH 2OG-dependent dioxygenases have been summarized in several reviews,^36–38  and their structures and biological activities are described further in Chapters 2 and 6.

Hydroxylation of ribosomal proteins has been demonstrated for 2OG-dependent enzymes, as first identified for cells from humans and Escherichia coli.³⁹  The human proteins NO66 and MINA53 catalyse 3S-hydroxylation of histidyl residues (Figure 1.2G) in Rpl8 and Prl27a, respectively, while the homologous enzyme YcfD of E. coli catalyses arginyl 3R-hydroxylation of Rpl16 (Figure 1.2H). The precise biological functions of these modifications are unclear, but when the ycfD strain was provided with low nutrient concentrations its growth rate was reduced compared to a wild-type strain. The crystal structure was first determined for the E. coli version of YcfD,⁴⁰  but additional structures quickly became available for truncated versions of NO66 and MINA53 along with the YcfD homologue of the thermophile Rhodothermus marinus and the same YcfD with bound substrate.⁴¹  The structure of the yeast enzyme, known as Tpa1, had also been reported prior to identifying its role in ribosome hydroxylation.⁴²  A 2OG and Fe(ii)-dependent oxygenase domain protein (OGFOD1) is widely found in eukaryotes, where it hydroxylates a particular prolyl residue in small ribosomal protein S23 (RPS23).⁴³  In mammals, loss of the enzyme leads to the formation of stress granules, stoppage of translation and growth inhibition. Whereas the human enzyme catalyses prolyl 3S-hydroxylation (Figure 1.2B), the corresponding enzymes from Schizosaccharomyces pombe, Saccharomyces cerevisiae (Tpa1, see above), and Ostreococcus tauri catalyse dihydroxylation of the cognate residues.⁴⁴  Furthermore, a Drosophila version of 2OGFOD1, known as Sudestada1, was studied; RNAi-mediated reduction in levels in tissue culture cells leads to smaller cell size, fewer cells and decreased translation efficiency.⁴⁵ 

Some 2OG-dependent dioxygenases utilize protein substrates beyond those described above, often without a clear role being identified for the modification. Two Jumonji domain (JMJD)-containing proteins and several other examples are illustrated here.

The nuclear protein JMJD6 hydroxylates the C5 position of a lysyl residue in a protein associated with RNA splicing.⁴⁶  Of special interest, the stereospecificity of the enzyme is opposite to that of PLOD, yielding lysyl 5S-hydroxylysine (Figure 1.2D).⁴⁷  Alternatively, JMJD6 was proposed to catalyse demethylation of methylated argininyl residues,⁴⁸  and recent work extends this theme to support such a role in regulation of transcriptional pause release.⁴⁹  The crystal structure of the catalytic domain of JMJD6 is known.^50,51  Another Jumonji domain protein, JMJD4, was shown to catalyse C4 hydroxylation of a lysyl residue (not depicted in Figure 1.2 because the enantiospecificity is not reported) in the NIKS motif of eukaryotic release factor 1.⁵²  Hydroxylation at this site is needed for efficient translation termination.

The hydroxylation of side chains of aspartic acid and asparagine residues (Figure 1.2E) in epidermal growth factor-like domains of various vitamin K-dependent proteins, coagulation factors and complement proteins is also carried out by 2OG-dependent dioxygenases.^53,54  The resulting 3R-hydroxyaspartyl (Hya) and 3R-hydroxyasparaginyl (Hyn) groups are thought to be important because mice lacking aspartyl (asparaginyl) β-hydroxylase activity exhibit developmental defects and greater susceptibility to intestinal neoplasia.⁵⁵  In addition, overexpression of the gene encoding the enzyme is linked to cellular transformation, at least in the case of biliary epithelial cells.⁵⁶  Of related interest, a bacterial enzyme, CinX from Streptomyces cinnamoneus, catalyses the same type of reaction to modify an aspartyl group in a 15-residue peptide during the synthesis of cinnamycin, an antimicrobial peptide known as a lantibiotic.⁵⁷ 

Aside from its O₂-sensing function related to hydroxylation of HIF-α Asn-803, FIH hydroxylates ankyrin repeat domains (ARDs) of endogenous Notch receptors.⁵⁸  The ARD modifications appear to stabilize the domains and probably affect protein–protein interactions.⁵⁹  In these roles, FIH catalyses aspartyl 3S-hydroxylase (Figure 1.2F)⁶⁰  and histidyl 3S-hydroxylase (Figure 1.2G)⁶¹  activities using Asp and His residues in ARD domains. FIH also hydroxylates members of the apoptosis-stimulating p53-binding protein (ASPP) family, such as modification of ASPP2 at Asn-986.⁶²  An alternative function of type I collagen prolyl 4-hydroxylase involves oxygen addition to Pro-700 of Argonaute 2, an essential component of the RNA-induced silencing complex for RNA interference.⁶³  This modification is also thought to promote stabilization of the protein.

In addition to catalysing the hydroxylation of target proteins as described above, many 2OG-dependent oxygenases catalyse self-hydroxylation reactions. This process is exemplified by TfdA and TauD, described in Sections 1.6.3 and 1.6.4, which generate hydroxytryptophan and dihydroxyphenylalanine from Trp and Tyr, respectively.^64–68  It is unclear whether such modifications have any beneficial/regulatory role, but the abundance of aromatic residues near the active sites of many 2OG-dependent enzymes led to the proposal of a sacrificial function for these side chains, thus sparing the enzymes from more deleterious chemistry.⁶⁹  Other examples of auto-catalysed oxidative modifications are found in procollagen 4R-hydroxylase that adds oxygen atoms to unknown sites of four of its own peptides;⁷⁰  JMJD6 that catalyses lysyl 5S-hydroxylation of one of its own Lys residues in the absence of its RNA splicing factor, perhaps as a regulatory mechanism;⁷¹  FIH that catalyses the formation of hydroxytryptophan when HIF-1α is absent;⁷²  AlkB and ALKBH3 (both described in Section 1.3) that hydroxylate Trp and Leu residues, respectively, at their active sites.^73,74  Two reviews of autocatalysed oxidative modifications in this enzyme family have been published.^75,76 

Another group of 2OG-dependent dioxygenases using protein substrates are those acting on N^ε-methylated lysine residues of histones. Eukaryotic DNA wraps around the histone core (composed of two each of histone H2A, H2B, H3 and H4) leaving the histone amino terminal regions accessible; these regions undergo several types of post-translational modification including the addition of methyl groups on specific lysyl or arginyl residues. The positions and extents of these modification have long been known to play a critical role in gene regulation.⁷⁷  In 2006, several research groups demonstrated that 2OG-dependent enzymes containing JmjC (jumonji domain C) domains catalyse demethylation reactions by hydroxylating the target methyl groups, with the hemiaminal intermediates subsequently decomposing with release of formaldehyde (Figure 1.2I).^78–82  The enzymes act on particular residues of specific histones with explicit levels of methylation. Unlike the flavin-dependent N^ε-methyl lysyl demethylase, the 2OG oxygenases can act on all three N^ε-methylation states of the modified residue. The first crystal structure for one of these enzymes was published that same year,⁸³  and follow-up studies have clarified the basis of the exquisite substrate specificities.^84–88  A molecular threading mechanism for the peptide substrate was proposed in the case of methylated H3K36 histone-specific KDM2A protein,⁸⁹  and the degree of methylation was suggested to influence whether the ferryl group was aligned according to the previously described ‘in line’ or ‘off line’ modes.¹  Additional discussion of this important family of enzymes is provided in Chapter 7 and selected reviews.^90–92 

Several proteins other than histones are methylated in cells, so corresponding demethylases are likely to be identified. As one example, ALKBH4 was shown to mediate the removal of the methyl group from Lys84 in cytoplasmic actin, a reaction that regulates actin dynamics.⁹³  Of additional interest, ALKBH4 was also found to associate with several proteins associated with chromatin or involved in transcription, although the functions of these interactions remain unknown.⁹⁴ 

In 2002, the E. coli enzyme AlkB was shown to directly repair alkylation damage to DNA by 2OG-dependent oxidation of 1-methyladenine (1meA) and 3-methylcytosine (3meC) lesions, resulting in the spontaneous loss of formaldehyde and restoration of the native base (Figure 1.3A and B).^95,96  Soon thereafter, AlkB-specific chemistry was expanded to include oxidative repair of alkylated RNA⁹⁷  and the range of lesions corrected was extended to 1-methylguanine (1meG, Figure 1.3C),⁹⁸  3-methylthymine (3meT, Figure 1.3D),^98,99  N⁶-methyladenine (N⁶meA, Figure 1.3E),¹⁰⁰  N²-methylguanine and N⁴-methylcytosine (not shown),¹⁰¹  as well as bases with slightly larger alkyl groups and exocyclic adducts (not shown). Several structures of AlkB have provided keen insights into the mechanism of substrate binding and substrate specificity.^102–105  Furthermore, the hemiaminal intermediate has been trapped in the protein by binding 3meT and exposing to oxygen.¹⁰⁶ 

AlkB-like proteins are widely distributed in other bacteria,¹⁰⁷  viruses¹⁰⁸  and many types of eukaryotes. Of special interest, mammals have nine homologues of AlkB that are referred to as ALKBH (or ABH) followed by numbers 1–8 along with FTO (sometimes referred to as ALKBH9).^109–111  No direct evidence of polynucleotide demethylation has been identified for several of these proteins, including ALKBH4 mentioned earlier in Section 1.2.6. In contrast, ALKBH2 and ALKBH3 were shown to repair 1meA and 3meC in DNA,¹¹²  with ALKBH3 also active with alkylated RNA.⁹⁷  Consistent with this differential specificity, knockout mouse studies demonstrated that ALKBH2 is the primary demethylase for repairing alkylated DNA.¹¹³  The structures of ALKBH2 and ALKBH3 reveal the basis of their distinct specificities.^{74,103,114–116}  ALKBH1 has a more restricted specificity, acting only on 3meC in DNA and RNA,¹¹⁷  but it also was proposed to be a methylated-histone demethylase¹¹⁸  and is associated with lyase activity at abasic sites.^119,120  ALKBH5 exhibits N⁶meA demethylase activity (Figure 1.3E) for alkylated mRNA.¹²¹  This modification on RNA can affect its processing and has regulatory implications, such as controlling the length of the circadian clock.¹²²  Human and zebrafish ALKBH5 structures reveal the basis for nucleic acid recognition and catalysis.^123–125  The reaction catalysed by ALKBH8 is described in the next section. Meanwhile, the chemistries catalysed by ALKBH6 and ALKBH7 remain unknown, but the structure of ALKBH7 has been determined and shown to lack a nucleotide recognition lid consistent with a potential protein hydroxylation role.¹²⁶ 

FTO is named for the fat mass and obesity-associated gene, whose inactivation protects mice from obesity.¹²⁷  Early studies reported the enzyme acts on 3meT in DNA,¹²⁸  or 3meT and 3-methyluracil (3meU, Figure 1.3F) in DNA and RNA;¹²⁹  however, more recent studies have shown that N⁶meA in RNA is a major substrate.¹³⁰  The RNA-associated N⁶-hydroxymethyladenine resulting from the latter activity is sufficiently stable that it can be further hydroxylated by FTO to form N⁶-formyladenine (the reaction shown in parentheses in Figure 1.3E).¹³¹  The structure of FTO has been determined in the presence of 3meT mononucleotide and with several inhibitors.^132,133 

Further details on polynucleotide repair by AlkB and ALKBH species are found in Chapter 8, while ALKBH5 and FTO are described more fully in Chapter 9.

In addition to the oxidative demethylations of alkylated bases in DNA and RNA discussed above, 2OG-dependent dioxygenases catalyse several additional types of reaction using these substrates (Figure 1.4).

Two types of tRNA modifications have been shown to require 2OG-dependent hydroxylases. Working with an accessory protein, ALKBH8 first utilizes its RNA recognition and methyltransferase motifs to convert 5-carboxymethyluridine (cm⁵U) to 5-methoxycarbonylmethyluridine (mcm⁵U) at position 34 in the anticodon loop of human tRNA.¹³⁴  This modified tRNA then serves as the substrate for the hydroxylase domain of ALKBH8 which forms 5S-methoxycarbonylhydroxymethyluridine (mchm⁵U) (Figure 1.4A).^135,136  The structure of the RNA recognition and hydroxylase domains of ALKBH8 has been reported.¹³⁷  Curiously, protozoan ALKBH8 catalyses both tRNA modification and DNA repair reactions.¹³⁸  Additional discussion of the biosynthesis of this unique base modification can be found in Chapter 10. A second tRNA-related hydroxylase acts on 7-(α-amino-α-carboxypropyl)wyosine known as wybutosine and abbreviated yW (Figure 1.4B), a tricyclic base that abuts the anticodon loop of tRNA^Phe in eukaryotes and archaea.¹³⁹  Among the numerous derivatives formed from this base are hydroxylated species generated by TYW5 (tRNA-yW-synthesizing enzyme 5), a member of the JmjC domain family (see Section 1.2.4), whose structure has been reported.¹⁴⁰ 

A family of t̲en-e̲leven t̲ranslocation (TET) proteins contributes to the epigenetic regulation of a wide range of genes in many eukaryotes by controlling the methylation status of cytosine in CpG islands of DNA. For example, TET1 was shown to convert 5-methylcytosine (5meC) to 5-hydroxymethylcytosine (5hmC) in 2009 (Figure 1.4C).¹⁴¹  Soon thereafter this result was confirmed and extended to TET2 and TET3.¹⁴²  More recently, TET enzymes were shown to catalyse subsequent reactions to create 5-formylcytosine (5fC)¹⁴³  and 5-carboxylcytosine (5caC),^144,145  including in a fungus.¹⁴⁶  In addition, the TET-catalysed oxidation of thymine to 5-hydroxymethyluracil in DNA of embryonic stem cells has been reported.¹⁴⁷  The various oxidized derivatives may have regulatory functions,¹⁴⁸  such as the hypoxic gene induction noted in neuroblastoma.¹⁴⁹  Moreover, the series of reactions from 5mC is suggested to play a role in active demethylation in DNA, either via (i) excision of 5fC or 5caC by thymine-DNA glycosylase, (ii) deamination of 5hmC followed by glycosylase removal, (iii) retro-aldol chemistry of 5hmC or 5fC with release of formaldehyde or formate, or (iv) decarboxylation of 5caC.¹⁵⁰  The structures of human TET2 in complex with DNA and a TET family protein from Naegleria gruberi in complex with 5meC-containing DNA are reported.^151,152  Additional information on the biology of the TET proteins is provided in Chapter 11 and in reviews.^153,154 

Nuclear genomes of kinetoplastid flagellates and some unicellular flagellates contain β-D-glucopyranosyloxymethyluracil in their telomeric repeats.¹⁵⁵  This modified base, first identified in 1993, is also called β-d-glucosyl-hydroxymethyluracil or base J.¹⁵⁶  Synthesis of base J involves two steps, hydroxylation of thymidine to form hydroxymethyluracil followed by glucosylation (Figure 1.4D).^157,158  JBP1, a protein that binds to and enhances the levels of base J,^159,160  was shown to belong to the 2OG-dependent dioxygenase family and is related to the TET proteins.^161,162  JBP1 along with the related JBP2, which does not bind to base J in DNA,¹⁶³  were directly demonstrated to catalyse thymidine hydroxylation.¹⁶⁴  Further information about the role of 2OG-dependent dioxygenases in base J biosynthesis is provided in Chapter 12.

Several 2OG-dependent hydroxylase reactions are used in lipid biosynthesis (Figure 1.5).

Carnitine or γ-trimethyl-hydroxybutyrobetaine is a small molecule that becomes linked to fatty acids, allowing for their transport across the inner mitochondrial membrane and degradation in the matrix by the β-oxidation pathway.¹⁶⁵  The synthesis of carnitine initiates with protein-derived ε-N-trimethyllysine that undergoes β-hydroxylation, aldolase cleavage to glycine plus trimethylaminobutyraldehyde, dehydrogenation to γ-butyrobetaine, and another hydroxylation (Figure 1.5A). The ε-N-trimethyllysine hydroxylase activity was shown to require 2OG in 1978,¹⁶⁶  and the enzyme was purified and characterized in 2001.¹⁶⁷  The hydroxylation of γ-butyrobetaine was found to require 2OG in 1968,¹⁶⁸  and the enzyme was purified from Pseudomonas spp. AK1 in 1977 and from calf liver in 1981.^169,170  A detailed comparison of substrate specificities for the human and bacterial enzymes revealed surprisingly large differences in reactivity for several γ-butyrobetaine analogues.¹⁷¹  The structure of human γ-butyrobetaine hydroxylase has been reported.^172,173  Further discussion of the enzymes involved in carnitine metabolism is provided in Chapter 13.

Phytanic acid (2,6,10,14-tetramethylhexadecanoic acid), a polyisoprenoid derived from the phytol moiety of chlorophyll, cannot be directly degraded via the β-oxidation pathway due to its 3-methyl group. To overcome this hurdle, the C-1 carbon is removed by action of a series of four enzymes: a ligase converts the molecule to phytanoyl-coenzyme A (CoA),¹⁷⁴  hydroxylation occurs at the C-2 position,¹⁷⁵  a lyase releases formyl-CoA and methylpentadecanal (pristanal),^176,177  and a dehydrogenase forms pristanic acid (2,6,10,14-tetramethylpentadecanoic acid)¹⁷⁸  which is a substrate for β-oxidation. Interruption of this pathway results in Refsum disease, which is associated with dysfunctions including retinitis pigmentosa, polyneuropathy and ataxia.¹⁷⁹  Phytanoyl-CoA hydroxylase (Figure 1.5B), shown to be a 2OG-dependent enzyme,¹⁷⁵  was purified from rat liver and as the recombinant human protein.^180,181  The human enzyme was structurally characterized, providing an understanding of mutations associated with Refsum disease.¹⁸²  Further description of phytanic acid metabolism is found in Chapter 14.

Much like phytanoyl-CoA hydroxylase, OlsD from Burkholderia cenocepacia (and probably from other Burkholderia and Serratia species) catalyses the hydroxylation of an acyl chain (Figure 1.5C) as part of a lipid metabolism pathway.¹⁸³  In this case, however, the pathway is biosynthetic rather than degradative. The first step of the pathway involves OlsB, an N-acyl transferase that transfers a 3-hydroxy fatty acyl group from the acyl carrier protein to the α-amino group of ornithine. The resulting lyso-ornithine lipid is the substrate of OlsA, an O-acyltransferase that transfers a fatty acyl group from the acyl carrier protein to the C-3 hydroxyl group. This ornithine-containing lipid, or its C-2 hydroxylated derivative (not shown), is the substrate of OlsD that hydroxylates somewhere on the amide-linked fatty acyl chain.

Two 2OG-dependent enzymes have been shown to modify lipid A (Figure 1.5D), with one targeting a specific acyl side chain and the other acting on the 3-deoxy-d-manno-oct-2-ulosonic (Kdo) moiety. Salmonella typhimurium, previously known to synthesize lipid A containing secondary S-2-hydroxyacyl chains, was demonstrated to use the O₂- and 2OG-dependent enzyme LpxO to achieve this modification.¹⁸⁴  LpxO was the first integral membrane protein representative of this enzyme family.¹⁸⁵  A Kdo 3-hydroxylase (KdoO) that adds an oxygen atom to the deoxysugar was identified in Burkholderia ambifaria and Yersinia pestis.¹⁸⁶  Homologues to the genes encoding these proteins are found in other Gram-negative bacteria.

Plants possess many of the 2OG-dependent oxygenases already mentioned, but in addition they utilize members of this enzyme family for some of their unique biosynthetic needs.¹⁸⁷  Here we illustrate how plants (and, in a few cases, other organisms) use these enzymes for generating flavonoids, gibberellins, alkaloids and other predominantly plant-specific products.

The flavonoids are polyphenolic compounds that include flavanones, flavones, isoflavones, flavonols and anthocyanins with more than 9000 such compounds known (see examples in Figure 1.6). These substances are used by plants for defence against pathogens, as signalling molecules in plant–microbe interactions, to minimize photodamage, in flower colour, and other roles.¹⁸⁸  When consumed by humans they function as antioxidants, antimalarials and potential anticancer agents.¹⁸⁹  The entrance to the main pathway shown in Figure 1.6 (indicated by the green arrow) is the flavanone naringenin, the substrate for two distinct 2OG-dependent oxygenases. Flavone synthase I (FNS) catalyses a desaturation reaction at the C-2/C-3 position forming trans-dihydroflavonol,¹⁹⁰  while flavanone 3β-hydroxylase (FHT) adds a hydroxyl group to C-3.¹⁹¹  The cytochrome P450 enzyme flavanone 3′-hydroxylase (F3′H) uses naringenin (or its 3β-hydroxy derivative) and converts the appended phenol to a catechol, with this species also being a substrate for FNS and FHT. The FHT-derived trans-dihydroflavonol products are substrates for another 2OG-dependent desaturase called flavonol synthase (FLS).^192,193  Alternatively, the trans-dihydroflavonols can be reduced via non-2OG-dependent enzymes to provide leucoanthocyanidin substrates for anthocyanidin synthase (ANS).¹⁹⁴  The structure of ANS from Arabidopsis thaliana has been elucidated with the bound substrate analogue dihydroquercetin and the unnatural substrate naringenin.^192,195  Two other flavonoid-related reactions catalysed by 2OG-dependent hydroxylases are 2,4-dihydroxy-2H-1,4-benzoxazin-3(4H)-one (DIBOA) 7-hydroxylase and methylated flavonol 6-hydroxylase (see boxes in Figure 1.6).^196–200  Flavonoid biosynthesis is discussed in much greater detail in Chapter 15.²⁰¹ 

Tetracyclic diterpenoids known as gibberellins (GAs) are widely distributed in higher plants and are also found in some lower plants, bacteria and fungi.²⁰²  At least 136 distinct GA structures are reported (commonly referred to as GA₁-GA₁₃₆; see http://www.plant-hormones.info/gibberellins.htm). A small sampling of such structures is shown in Figure 1.7, which depicts selected 2OG-dependent transformations of these molecules. Many GAs possess a carboxylate at C-7, introduced by oxidation of the GA₁₂-aldehyde to form GA₁₂, with this product undergoing partial oxidization at C-7 by a hydroxylase to form GA₅₃. Depending on their biological sources these reactions can be catalysed by either cytochrome P450 or 2OG-dependent oxygenases (Figure 1.7A).²⁰³  The GA C20 oxidase can catalyse sequential reactions that convert the C-20 methyl group (e.g. GA₁₂/GA₅₃) to the alcohol (GA₁₅/GA₄₄), aldehyde (GA₂₄/GA₁₉), and in some cases the carboxylate (GA₂₅/GA₁₇) (Figure 1.7B).^204,205  Alternatively, the same enzyme can catalyse an oxidative transformation that eliminates the carboxylate as CO₂ while forming a γ-lactone (GA₉/GA₂₀).²⁰⁶  GA 3β oxidase converts these products to the corresponding hydroxylated species (GA₄/GA₁).²⁰⁷  GA 2β oxidase acts on the same substrates (producing GA₅₁/GA₂₉) or on the products of the prior reaction (producing GA₃₄/GA₈).²⁰⁸  In addition, other 2OG-dependent enzymes can catalyse several types of desaturation reactions (not shown),^202,209  such as a recently characterized fungal GA₄ desaturase that introduces a double bond between C-1 and C-2.²¹⁰  Further information on this remarkable family of enzymes is available in Chapter 16.

Four examples are depicted to illustrate how plants use 2OG-dependent oxygenases for alkaloid biosynthesis. Scopolamine is a hallucinogenic tropane alkaloid produced by Hyoscyamus niger (henbane). The last two steps in its synthesis are catalysed by hyoscyamine 6β-hydroxylase that carries out both hydroxylation and epoxidation steps (Figure 1.8A).^211,212  Vinblastine and vincristine are alkaloids produced by Caranthus roseus (periwinkle), which have been used for treatment of Hodgkin’s lymphoma and acute leukaemia. These compounds are synthesized by a complex biosynthetic pathway that utilizes a 2OG-dependent hydroxylase to generate deacetylvindoline (Figure 1.8B), which is then further modified.^213,214  Codeine and morphine are important pharmaceuticals obtained from Papaver somniferum (opium poppy). Their complex biosynthetic pathway includes an intermediate named thebaine, which can be demethylated at one site by thebaine 6-O-demethylase to produce codeinone (Figure 1.8C) or demethylated at a second site by codeine O-demethylase to form oripavine (Figure 1.8D).²¹⁵  These two reactions plus that catalysed by codeinone reductase yields morphine. A series of additional reactions (not shown) involving alkaloid metabolism in opium poppy are catalysed by these enzymes and the 2OG-dependent dioxygenase protopine O-dealkylase, PODA.²¹⁶  In the fungus Claviceps purpurea, a cyclization reaction (not shown) in the synthesis of d-lysergic acid alkaloid peptides is catalysed by a member of this enzyme family, and the 2OG-bound holoprotein, EasH, was structurally characterized.²¹⁷  Another fungal example is found in the synthesis of loline alkaloids by Epichloë species, where the 2OG-dependent oxygenase LolO introduces an ether bridge into a pyrrolizidine ring system (Figure 1.8E).³⁷⁸ 

Additional representatives of the 2OG-dependent oxygenases that are primarily restricted to plants include enzymes involved in the biosynthesis of phytosiderophores, coumarins and glucosinolates, or the degradation of hormonal compounds.

Under iron-deficient conditions, some grasses, cereals and rice produce and secrete iron-binding compounds such as the mugineic acid-related species made by Hordeum vulgare (barley). This plant contains two 2OG-dependent dioxygenases, IDS2 and IDS3, that act on 2′-deoxymugineic acid to form 3-epihydroxy-2′-deoxymugineic acid and mugineic acid, respectively (Figure 1.8F).²¹⁸  2′-Deoxymugineic acid is converted to 3-epihydroxymugineic by combined actions of the two enzymes.

Coumarins (1,2-benzopyrones) such as scopoletin and umbelliferone are synthesized by many higher plants where they are used for defence against phytopathogens. A key enzyme in the pathway used by Arabidopsis thaliana for generating scopoletin is a 2OG-dependent enzyme that hydroxylates the ortho position of feruloyl-CoA,²¹⁹  with subsequent isomerization, hydrolysis and lactonization steps providing the product (Figure 1.8G). The same activity was detected using two recombinant proteins from Ipomoea batatas (sweet potato), one of which also used p-coumaroyl-CoA to form umbelliferone (where H is present at the 3′ position in Figure 1.8G).²²⁰  An enzyme with the latter dual activity has also been characterized from Ruta graveolens (the common rue).²²¹ 

Glucosinolates (over 130 are known) are predominantly associated with the Brassicaceae family of plants where, following tissue disruption, they are decomposed to form compounds with diverse protective roles against herbivores and pathogens.²²²  Two sequence-related 2OG-dependent enzymes in A. thaliana, named AOP2 and AOP3, have been shown to participate in glucosinolate biosynthesis by converting methylsulfinylalkyl glucosinolate to the alkenyl or hydroxyalkyl species (Figure 1.8H).²²³  The chemistry of these reactions has not been well defined.

Indole-3-acetic acid and salicylic acid are plant hormones that play important roles in growth, development, disease resistance or other functions, but until recently their degradation pathways within plants were unclear. Using rice, a dioxygenase for auxin oxidation (DAO) has now been identified and shown to transform this substrate into 2-oxoindole-3-acetic acid (Figure 1.8I).²²⁴  Mutants affecting the corresponding gene display male sterility and produce infertile seeds. Similarly, a salicylic acid 3-hydroxylase (Figure 1.8J) of Arabidopsis was identified and mutants in the corresponding s3h gene were found to accumulate salicylate species and exhibit early senescence.²²⁵  When assayed in vitro, both 2,3- and 2,5-dihydroxybenzoate are formed; however, only the former appears to be made in vivo.

The reactions described in this section are diverse, but generally involve rather small-sized molecules. Several products from these reactions are precursors that become incorporated into antibiotics, whereas 2OG-dependent tailoring enzymes that act directly during antibiotic synthesis are described in Section 1.7.

2OG-Dependent hydroxylases acting on free amino acids (i.e. not as a side chain of a protein) are known. In the case of l-Pro, different enzymes exhibit each of four distinct specificities. Pro 4R-hydroxylase (Figure 1.9A, left) from Streptomyces griseoviridus P8648 produces the trans-isomer of 4-hydroxyproline that is subsequently utilized for etamycin synthesis.^226,227  This enzyme activity is also found in other bacterial strains of Streptomyces, Dactylosporangium and Amycolatopsis,²²⁸  and in the pneumocandin-producing fungus Glarea lozoyensis.²²⁹  l-Pro 4S-hydroxylase (Figure 1.9A, right), producing the cis isomer, has been identified in Mesorhizobium loti and Sinorhizobium meliloti where it was shown to require 2OG.²³⁰  l-Pro 3S-hydroxylase (Figure 1.9B, left) has been reported in the G. lozoyensis fungus mentioned above.²²⁹  Enzymes from two actinomycetes were shown to exhibit both 3S- and 4S-hydroxylase activities using l-Pro.²³¹  The best studied enzyme using this substrate is l-Pro 3R-hydroxylase (Figure 1.9B, right), an activity identified in strains of Streptomyces and Bacillus.²³²  The enzyme was purified from Streptomyces sp. strain TH1²³³  and the structure of the apoprotein was elucidated.²³⁴ 

Several 2OG-dependent hydroxylases act on free l-amino acids with polar side chains. A 3S-hydroxylase of l-Asn (creating the threo isomer, Figure 1.9C) was characterized from recombinant cells containing asnO from Streptomyces coelicolor.²³⁵  High-resolution crystal structures of AsnO reveal the basis of this substrate specificity. The product of the reaction is subsequently incorporated (at position nine) into a daptomycin-type lipopeptide called the calcium-dependent antibiotic. A single amino acid substitution (D241N) of AsnO led to use of l-Asp as substrate, forming the threo isomer of hydroxyaspartic acid (not shown).²³⁶  By contrast, AspH from Pseudomonas syringae catalyses 3R-hydroxylation of l-Asp (and l-Asp thioesters, not shown) to form the erythro isomers (Figure 1.9D).²³⁷  VioC, a 2OG-dependent Arg 3S-hydroxylase (Figure 1.9E), was shown to be used by Streptomyces vinaceus to provide a precursor for viomycin biosynthesis.²³⁸  The VioC structure was elucidated and explains the hydroxylation specificity.²³⁹  Enduracididine is an Arg-derived amino acid that undergoes 3S-hydroxylation by MppO (Figure 1.9F) to provide β-hydroxyenduracididine, which is incorporated into mannopeptimycins, glycopeptide antibiotics produced by Streptomyces hydrogroscopicus NRRL 30439.²⁴⁰ 

Hydroxylases of l-Ile and l-Leu have been widely studied since the first report of 2OG-dependent 4-hydroxyisoleucine synthesis in Trigonella foenum-graecum (fenugreek).²⁴¹  The recombinant enzyme (IDO) from Bacillus thuringiensis was shown to exhibit l-Ile 4S-hydroxylase activity,²⁴²  and the enzyme was later shown to also form 2-amino-3-methyl-4-ketopentanoate, probably arising from dihydroxylation of the same carbon position to produce a gem diol that loses water (Figure 1.9G).²⁴³  The products of two adjacent genes in Pantoea ananatis, HilA and HilB, catalyse sequential hydroxylation reactions at the 4′ and 4 positions, respectively, of l-Ile (Figure 1.9H).²⁴⁴  Related family members were identified in a range of other bacteria, in several cases using l-Leu as the preferred substrate and yielding 4-hydroxyleucine (Figure 1.9I).²⁴⁵  Alternatively, l-Leu 5-hydroxylase (LdoA, Figure 1.9J) was characterized from Nostoc punctiforme.²⁴⁶  All of these enzymes also act on other substrates with reduced catalytic efficiencies (including Leu/Ile substitution), in various cases forming Met sulfoxide, 4-hydroxyvaline and 4-hydroxythreonine (not shown).^243–245 

Four other enzymes are mentioned in this section because they modify amino acid-like substrates or combine another reaction with amino acid hydroxylation. SadA from Burkholderia ambifaria acts on several N-substituted amino acids with hydrophobic side chains, notably catalysing 3R hydroxylation of N-succinyl-l-Leu (Figure 1.10A).²⁴⁶  The structure of SadA has been reported,²⁴⁷  and variants with altered specificity have been studied.²⁴⁸  PvcB of Pseudomonas aeruginosa functions in the biosynthesis of the siderophore pyoverdine by participating in the formation of 2-isocyano-6,7-dihydroxycoumarin (reminiscent of the coumarin product shown in Figure 1.8G). The P. aeruginosa enzyme catalyses a cyclization reaction (Figure 1.10B),²⁴⁹  although the mechanistic details remain obscure, and the structure of the protein has been characterized in the absence of ligands.²⁴⁹  Ectoine is an acid with a secondary amine, so the associated enzyme is described here although this tetrahydropyrimidine could alternatively be included in the next section with nucleobases and nucleosides. Bacteria synthesize ectoine and a variety of other compatible solutes to prevent excessive loss of water when grown in high salinity conditions, and some species convert ectoine to 5-hydroxyectoine during stationary growth. EctD is a 2OG-dependent enzyme responsible for this ectoine 5-hydroxylase activity (Figure 1.10C).²⁵⁰  The properties of this protein have been examined from a wide distribution of microorganisms, including extremophiles.²⁵¹  Of particular interest, the structure of EctD from Virgibacillus (formerly Salibacillus) salexigens in the absence of substrate has been characterized,²⁵²  and a model of the holoprotein with ectoine and 2OG has been simulated.²⁵³  Finally, Pseudomonas syringae pv. phaseolicola contains the ethylene-forming enzyme (EFE) which is reported to convert 2OG to ethylene and three molecules of CO₂ while simultaneously converting Arg into guanidine and Δ¹-pyrroline-5-carboxylate.²⁵⁴  The presumed Arg 5-hydroxylase activity (Figure 1.10D) and formation of the resulting degradation products have not been clearly documented. 2OG-dependent ethylene formation has been observed in other microorganisms, including a mushroom,²⁵⁵  and the gene encoding EFE is functional when inserted into other hosts,^256,257  including cyanobacteria,^257,258  with the latter microorganisms offering the potential for generating a biofuel from CO₂. The mechanism of 2OG conversion to ethylene by EFE is not understood.

Two distinct 2OG-dependent oxygenases are known to act on nucleobases. Thymine 7-hydroxylase from the fungus Neurospora crassa was among the first 2OG-dependent oxygenases to be characterized.^259,260  The enzyme catalyses sequential oxygen additions to the methyl group of the free base, forming 5-hydroxymethyluracil, 5-formyluracil and 5-carboxyuracil (Figure 1.11A).^261,262  Additional studies have focused on the purified fungal enzyme from Rhodotorula glutinis,²⁶³  including analysis of its broad substrate specificity.^264,265  Xanthine hydroxylase, XanA, is a fungal enzyme that catalyses the reaction of Figure 1.11B.²⁶⁶  This enzyme was discovered by studies involving a mutant strain of Aspergillus nidulans that was defective in xanthine dehydrogenase, a widely distributed molybdopterin-containing enzyme, yet was able to grow on xanthine as a nitrogen source. XanA was purified both from the fungal host and as a recombinant protein from E. coli; although differing in various types of post-translational modifications, both forms exhibited the activity shown.²⁶⁷  Although no crystal structure is available for XanA, the likely active site residues were identified from mutagenesis studies and shown to support results of a homology model.²⁶⁸ 

Three types of 2OG-dependent oxygenases have been reported to catalyse reactions with the sugar components of nucleosides. Several fungi contain pyrimidine deoxyribonucleoside 2′-hydroxylases that form the corresponding ribonucleosides (Figure 1.11C).^269,270  Furthermore, R. glutinis contains a deoxyuridine or uridine 1′-hydroxylase that forms an unstable intermediate which decomposes with release of the nucleobase and formation of a lactone (Figure 1.11D).²⁷¹  Finally, a uridine-5′-monophosphate 5′-hydroxylase (Figure 1.11E) named LipL has been purified from a Streptomyces species where it provides a precursor for incorporation into antibiotic A-90289.²⁷² 

Although the enzyme has not been characterized, it appears that a thymine 7-hydroxylase-like activity might be used for nucleoside modification during polyoxin biosynthesis in Streptomyces avermitilis.²⁷³  The cells produce 14 distinct forms of polyoxins that possess a common nucleoside core containing 1-(5′-amino-5′-deoxy-β-d-allofurauronosyl)pyrimidine. The SAV_4805 open reading frame of this microorganism was shown to be associated with enhancement of structural diversity at the C-5 position of the pyrimidine ring, consistent with hydroxylation of this site by the encoded protein (Figure 1.11F).

Phenoxyalkanoic acids are widely used as herbicides for selective control of broad-leaf weeds. An important example is 2,4-dichlorophenoxyacetic acid (2,4-D), for which the biodegradative process has been extensively studied.²⁷⁴  Bacteria that are capable of growth on 2,4-D as sole carbon source often possess a series of enzymes: TfdA (initially referred to incorrectly as 2,4-D monooxygenase), 2,4-chlorophenol hydroxylase (TfdB), 3,5-dichloro–catechol dioxygenase (TfdC), dichloromuconate cycloisomerase (TfdD), dienelactone hydrolase (TfdE) and maleylacetate reductase (TfdF) yielding 3-oxoadipate which enters intermediary metabolism. Using the enzyme from Cupriavidus necator (formerly Ralstonia eutropha), the reaction of TfdA was revised to be that of a 2OG-dependent hydroxylase (Figure 1.12A).²⁷⁵  The properties of TfdA have been extensively characterized, including the specificity towards a diversity of phenoxyalkanoic acids, several spectroscopic features of the protein, and its ability to catalyse self-hydroxylation.^64,276,277  This was the first example of a 2OG-dependent dioxygenase being used for biodegradation of a xenobiotic compound. By incorporating tfdA-like genes into transgenic plants, enhanced herbicide tolerance has been obtained.²⁷⁸ 

2-Phenoxypropionic acids differ from phenoxyacetic acids by a single methyl group, resulting in two enantiomeric forms of these compounds. Only the (R) enantiomer of 2-(2,4-dichlorophenoxy)propionic acid (dichlorprop) or 2-(4-chloro-2-methyl-phenoxy)propionic acid (mecoprop) are active as herbicides. Sphingomonas herbicidovorans MG was shown to possess two proteins, RdpA and SdpA, specific for hydroxylating the (R) and (S) enantiomers, respectively (Figure 1.12B).²⁷⁹  The structural basis of the distinct enantiospecificities for these proteins has been assessed by homology modelling, substrate docking and mutagenesis.²⁸⁰ 

The mechanism of herbicidal action of the above aryloxyalkanoic acids depends on the interaction of the compounds with the auxin receptor of plants. It is thus of some interest that rice possesses DAO (see Section 1.5.4), which converts indole-3-acetic acid into 2-oxoindole-3-acetic acid (Figure 1.8I).²²⁴  The distribution of this activity in other plants remains to be defined.

TauD is the best characterized 2OG-dependent oxygenase in terms of understanding its catalytic mechanism.^9,281  This E. coli enzyme hydroxylates 2-aminoethanesulfonate (taurine), with the hydroxylated sulfonate intermediate spontaneously decomposing to aminoacetaldehyde and sulfite (Figure 1.12C), which is used as a sulfur source by the cells.²⁸²  Crystal structures are available for the E. coli protein as well as that from Pseudomonas putida KT2440.^283–285  Several of the intermediate states of catalysis have been examined spectroscopically,^286,287  including the Fe(iv)-oxo^288–291  and later species.¹⁰  In addition, at least two distinct types of self-hydroxylation chemistries have been studied with TauD, with one case involving a transient tyrosyl radical.^66–68  The enzyme decomposes a variety of other sulfonates (not depicted), including the widely used buffer 3-(N-morpholino)propanesulfonic acid, MOPS.²⁸²  A yeast homologue is most active with MOPS among the potential substrates tested, but physiologically it functions to degrade taurocholate – the amide formed between taurine and cholic acid.²⁹² 

Similar to the sulfonate-degrading enzymes discussed above, 2OG-dependent hydroxylases are also used to decompose alkylsulfates (Figure 1.12D). The first such enzyme discovered was AtsK from P. putida S-313; this protein is 38% identical to TauD, but it exhibits no activity with taurine.²⁹³  A related enzyme is associated with the Rv3406 locus of Mycobacterium tuberculosis.²⁹⁴  Crystal structures have been obtained with AtsK proteins from both sources.^294–296 

Several members of the 2OG oxygenase enzyme family have been shown to act on various types of phosphonate compounds. The gene encoding PhnY was identified from a screen of ocean-derived metagenomic DNA that allowed E. coli cells deficient in C–P lyase (Δphn) to grow on 2-aminoethylphosphonic acid as the sole source of phosphorus. The substrate, a close analogue of taurine, is hydroxylated by PhnY (Figure 1.12E)²⁹⁷  much like the reaction of TauD; however, the hydroxylated phosphonate is stable – unlike the hydroxylated sulfonate. DhpA catalyses a very similar reaction, oxygen addition to hydroxyethylphosphonate (Figure 1.12F) or its O-phosphomonomethyl ester (not shown), in the pathway for biosynthesis of dehydrophos, a vinyl phosphonate tripeptide antibiotic of Streptomyces luridus.²⁹⁸  A hydroxylation reaction is also catalysed by FrbJ of Streptomyces rubellomurinus, in this case using the antibiotic FR-900098 as substrate (Figure 1.12G).²⁹⁹  Rather than hydroxylation, EpoA of Penicillium decumbens catalyses epoxidation using cis-propenylphosphonic acid (Figure 1.12H) during synthesis of the antibiotic fosfomycin.³⁰⁰  Another enzyme participating in the synthesis of dehydrophos by S. luridus is DhpJ, which catalyses a desaturation reaction of the monomethyl diester of l-Leu-l-1-aminoethylphosphonic acid (Figure 1.12I).³⁰¹  While the substrate is not a phosphonate, it is also appropriate to mention here the phosphite-producing hypophosphite hydroxylase activity (Figure 1.12J) of HtxA from Pseudomonas stutzeri WM88.³⁰² 

Several 2OG-dependent oxygenases described earlier function in the synthesis of antibiotics, but their actions typically involve provision of small precursor molecules that become incorporated into the final compounds. This section focuses on other family members that function in antibiotic biosynthesis by forming bicyclic β-lactams, tailoring terpenoids, modifying protein-bound S-pantetheinyl thioesters compounds, and other roles.

This section covers several 2OG-dependent oxygenases that participate in the formation of clinically important bicyclic β-lactam antibiotics.³⁰³  Also of interest for this topic, Section 1.8.1 describes isopenicillin N synthase, a structurally-related enzyme that does not use 2OG.

Clavaminate synthase (CAS) is a remarkable trifunctional 2OG-dependent enzyme in the pathway for synthesis of clavulanic acid in Streptomyces clavuligerus.³⁰⁴  Starting with l-Arg and glyceraldehyde-3-phosphate, the cells utilize a thiamine diphosphate-dependent enzyme to form N²-(2-carboxyethyl)arginine, which cyclizes via an ATP-dependent ligase reaction to generate deoxyguanidinoproclavaminate. This compound is the substrate for CAS, which catalyses a 2OG-dependent hydroxylation reaction (Figure 1.13A, left reaction). The guanidine group is removed from the resulting guanidinoproclavaminic acid by a separate hydrolase to yield proclavaminate. This substrate is used in a 2OG-dependent cyclization reaction catalysed by CAS (Figure 1.13A, middle reaction) to form dihydroclavaminic acid, which undergoes desaturation to clavaminic acid in the third 2OG-dependent reaction of CAS (Figure 1.13A, right reaction). Subsequent amino transferase and reductase reactions provide the final clavulanic acid (with an alcohol replacing the amine in the last structure shown). Purified CAS in its various states was studied by several spectroscopic methods, which provided evidence that the binding of both substrates (2OG and the antibiotic precursors) led to the loss of all metal-bound water, thus creating an oxygen binding site on the metal.^305,306  Structures of CAS have been solved in the presence of Fe(ii), 2OG and either N-α-acetyl-l-Arg, proclavaminic acid, or deoxyguanidinoproclavaminate (with the oxygen analogue NO bound to the metal).^307,308 

Cephalosporin biosynthesis starts with the formation of isopenicillin (see Section 1.8.1), which is epimerized to penicillin, modified by two 2OG-dependent enzymes in prokaryotes or a single dual-function enzyme in eukaryotes, and additionally tailored by other reactions, sometimes including another 2OG family member. The first two 2OG-dependent oxygenases in S. clavuligerus are deacetoxycephalosporin C synthase (DAOCS), which catalyses a ring expansion reaction (Figure 1.13B, left), and deacetylcephalosporin C synthase (DACS; Figure 1.13B, right), a hydroxylase.^309–311  In contrast, both reactions are catalysed by the same DAOCS/DACS enzyme from Cephalosporium acremonium.^312,313  Several structural studies of DAOCS included the first structure for any 2OG-dependent oxygenase³¹⁴  and reveal the modes of substrate and product binding – with the surprising finding of overlapping binding sites of 2OG and penicillin substrates.^315–319  Recent pre-steady state kinetics and binding studies have called into question the interpretations from earlier results and conclude that a ternary complex does form in the protein.³²⁰  Modelling combined with mutagenesis studies suggest that a single residue controls whether the enzyme catalyses ring expansion or hydroxylation, and offers insight into how the C. acremonium DAOCS/DACS enzyme catalyses both reactions.³²¹  The deacetylcephalosporin C resulting from these reactions undergoes further transformations including acylation of the newly introduced hydroxyl group, 2OG-dependent hydroxylation at the 7α position (Figure 1.13C),³²²  and methylation of the latter site. These enzymes are described in greater detail in Chapter 17.

Carbapenems are a third grouping of the bicyclic β-lactams that utilize a 2OG-dependent oxygenase in their synthetic pathway. Pectobacterium carotovorum (formerly Erwinia carotovora) contains the enzyme (2S,5S)-carboxymethylproline synthase (CarB) that uses glutamate semi-aldehyde and malonyl-CoA to produce the carboxymethylproline derivative. This substrate is subjected to a ligase reaction by (3S,5S)-carbapenam synthetase (CarA), forming carbapenam-3-carboxylate with its β-lactam ring. Carbapenem synthase (CarC) is a 2OG-dependent oxygenase proposed to catalyse two sequential non-hydroxylase reactions: epimerization to (3S,5R)-carbapenam-3-carboxylate and desaturation to yield (5R)-carbapenem-3-carboxylate (Figure 1.13D).³²³  The crystal structure of CarC with bound Fe(ii) and 2OG has been elucidated.³²⁴  The stereoinversion reaction appears to involve ferryl abstraction of a substrate hydrogen atom followed by hydrogen atom donation from a tyrosyl side chain, limiting the enzyme to a single turnover unless external electrons are provided.³²⁵ 

The rich biochemistry of terpenoid metabolism includes several examples of reactions catalysed by 2OG-dependent oxygenases. The synthesis of pentalenolactone, a sesquiterpenoid produced by dozens of strains of Streptomyces, nicely illustrates this point. A hydroxylase reaction, the conversion of 1-deoxypentalenic acid to 11β-hydroxy-1-deoxypentalenic acid (Figure 1.14A), is catalysed by S. avermitilis PltH.³²⁶  The structure of this protein in complex with Fe(ii), 2OG and ent-1-deoxypentalenic acid (a non-reactive enantiomer), has been defined.³²⁷  Analogous proteins appear to be present in S. exfolatus UC5319 and S. arenae TU469, where they are named PenH and PntH, respectively.³²⁸  These three strains each possess a second 2OG-dependent oxygenase (PtlD, PenD and PntD, respectively) capable of desaturating pentalenolactone D to pentalenolactone E, with PenD and PntD subsequently forming the epoxide pentalenolactone F (Figure 1.14B). In addition, these same three enzymes can transform neopentalenolactone D to an unstable enollactone desaturation product which is hydrolysed (Figure 1.14C).³²⁸ 

Two 2OG-dependent oxygenases utilize the same substrate, fusicocca-2,10(14)-diene-8β,16-diol, to produce distinct diterpene phytohormone-like compounds in fungi.³²⁹  An enzyme from Alternaria brassicicola first abstracts a hydrogen atom from the eight-membered ring, the carbon-centered radical migrates, and hydroxyl radical rebound yields the distal hydroxylated product (Figure 1.14D, left) during a step in the synthesis of cotylenin A or brassicicene C. In contrast, an enzyme from Phomopsis amygdali catalyses oxidation at C-16 to yield the aldehyde 8β-hydroxyfusicocca-1,10(14)-diene-16-al (Figure 1.14D, right) during fusicoccin A biosynthesis.³²⁹ 

A final example of a 2OG-dependent oxygenase involved in synthesis of a terpene glycoside antibiotic is PlaO1 from Streptomyces sp. Tü6071.³³⁰  PlaO1 catalyses the critical formation of a γ-butyrolactone during synthesis of phenalinolactone (Figure 1.14E). The chemical mechanism associated with this remarkable reaction has not been detailed, but a cogent hypothesis has been proposed.³³¹ 

Several 2OG-dependent oxygenases act on protein-tethered S-pantetheinyl thioesters of amino acids which undergo non-ribosomal incorporation into peptide-related antibiotics. For example, Pseudomonas syringae pv. syringae B301D produces the non-ribosomal peptide phytotoxin called syringomycin E by assembling components on a multimodular megasynthetase, SyrE. SyrP is a 2OG-dependent hydroxylase that acts on l-Asp bound to the eighth module of SyrE to yield the threo isomer of 3-hydroxyaspartic acid which is inserted into the eighth position of the final phytotoxin (Figure 1.15A).²³⁷ 

Two other proteins involved in syringomycin E biosynthesis, SyrB1 and SyrB2, are required for synthesis of the 4-chloro-l-Thr located at position nine of the phytotoxin. Remarkably, SyrB2 was shown to be a 2OG-dependent oxygenase that catalyses a halogenation reaction using l-Thr tethered to SyrB1 (Figure 1.15B).³³²  The enzyme also catalyses bromination, dichlorination, nitration and azidation reactions (not depicted).^333,334  The structure of SyrB2 reveals a metallocentre in which Fe(ii) is bound only by two histidyl residues (with the typical carboxylate ligand replaced by an alanyl side chain), and coordinated to 2OG and chloride ion.³³⁵  Spectroscopic studies have revealed evidence for a chloroferryl intermediate in catalysis,³³⁶  and studies with substrate analogues suggest that substrate positioning determines whether halogenation or hydroxylation take place.³³⁷  Streptomyces sp. OH-5093 produces free 4-chlorothreonine by a similar pathway, in which Thr3 catalyses the halogenation of the tethered amino acid, followed by thioester hydrolysis.³³⁸ 

Much like the SyrB1/SyrB2 system just discussed, a cytotrienin-producing Streptomyces sp. uses CytC3 to catalyse halogenation of l-2-aminobutyric acid (or l-Val) tethered to CytC2 (Figure 1.15C).³³⁹  Tandem chlorinations followed by thioesterase activity lead to the γ,γ-dichloraminobutyrate antibiotic that is released into the soil. Two interconverting Fe(iv) intermediates were detected in that study, and later work provided evidence for the formation of a bromoferryl intermediate when bromide was used to replace chloride.³⁴⁰  The structure of CytC3 exhibits the same general architecture as seen for SyrB2, including an Ala residue replacing the typical aspartyl metal ligand.³⁴¹ 

Five other examples illustrate additional interesting features of 2OG-dependent halogenation enzymes. The marine cyanobacterium Lyngbya majuscula produces barbamide, an antibiotic containing trichlorinated leucine that is active against molluscs. BarB2 converts BarA-tethered l-Leu or 4-chloro-l-Leu to the dichlorinated species and BarB1 halogenates the mono- or dichlorinated species to the trichlorinated species (Figure 1.15D).³⁴²  Pseudomonas syringae pv. tomato DC3000 synthesizes the phytotoxin coronatine which contains 1-amino-1-carboxy-2-ethylcyclopropane (coronamic acid). Synthesis of coronamic acid derives from l-allo-isoleucine which is tethered to CmaD, chlorinated by CmaB (Figure 1.15E), and converted to the cyclopropane species by CmaC with elimination of chloride in this cryptic chlorination pathway.³⁴³  An analogous system is present in the ascomycete Kutzneria, where KtzD chlorinates l-Ile bound to KtzC, with the product cyclized to the bound (1S,2R)-allocoronamic acid by KtzA.³⁴⁴  Another example of cryptic chlorination occurs during curacin A synthesis by L. majuscula, the marine bacterium mentioned above. In this case, acyl carrier protein-bound (S)-3-hydroxy-3-methylglutarate is chlorinated by CurA (Figure 1.15F), followed by dehydration using CurC, decarboxylation via a CurF domain, and reductive dechlorination by another CurF domain to yield the protein-bound (1R,2S)-2-methylcyclopropane-1-carboxylate which is incorporated into curacin A.³⁴⁵  The structures of CurA in five ligand states have been elucidated.³⁴⁶  Kutzneria sp. 744 chlorinates a protein-bound piperazate residue by using the KthP halogenase (Figure 1.15G) during biosynthesis of a family of antifungal kutznerides.³⁴⁷  Finally, Lyngbya majuscula contains the three-domain protein HctB that catalyses dichlorination (along with hydroxylation and introduction of a vinyl chloride group) on a fatty acyl group attached to its acyl carrier domain region (not depicted).³⁴⁸  Interestingly, the oxidative reactions of this enzyme are stimulated more than 200-fold by the presence of saturating concentrations of chloride, thus providing an explanation for why hydroxylation does not dominate in the absence of halide salts.³⁴⁹  Further discussion of 2OG-dependent halogenases can be found in Chapter 18.

Final examples for this section relate to hydroxylation reactions during synthesis of four antitumour natural products made by non-ribosomal peptide synthesis pathways: bleomycin, tallysomycin, zorbamycin and spliceostatin made by Streptomyces verticillus ATCC15003, Streptoalloteichus hindustanus E465-94, Streptomyces flavoviridis ATCC21892 and Burkholderia spp., respectively. A 2OG-dependent hydroxylase, potentially associated with open reading frames 1, 10 and 30 of the gene clusters in the first three microbes, is thought to catalyse histidyl group hydroxylation of a large precursor of the final antibiotic species bound to carrier proteins (Figure 1.15H), rather than hydroxylating just the tethered amino acid.³⁵⁰  The 3-hydroxyhistidyl group is the site of glycation and additional tailoring reactions are used to generate the final antibiotics. The Burkholderia enzyme catalyses the synthesis of a hemiketal group (not shown) using the 2OG-dependent mechanism.³⁵¹ 

Other 2OG-dependent oxygenases serve important roles in antibiotic biosynthesis, but don’t fit into the above categories. These examples are not meant to be comprehensive, but provide useful illustrations of the diversity of reactions catalysed by these enzymes. The just mentioned antibiotic tallysomycin requires TmlH for its synthesis. This enzyme catalyses two 2OG-dependent hydroxylations, at positions C-41 and C-42 (Figure 1.16A).³⁵²  The carbanolamide produced by hydroxylation of C-41 is unstable, but appears to be immediately modified further by TlmK to produce a stable species. It remains unclear whether TlmH acts on the protein-free species as shown or if the modification takes place on the protein-tethered substrate.³⁵⁰  The fumonisins are mycotoxins produced by Fusarium verticillioides and several other filamentous fungi via polyketide synthetic routes. Hydroxylation at the C-5 position (Figure 1.16B) is carried out by the 2OG-dependent enzyme Fum3p (formerly associated with the FUM9 locus).^353,354  Another mycotoxin, verruculogen, is produced by Aspergillus fumigatus. The novel creation of an endoperoxide within fumitremorgin B is catalysed by the 2OG-dependent FtmOx1 protein (Figure 1.16C). Notably, this reaction requires two molecules of oxygen.³⁵⁵  Several Oscillatoria and other cyanobacterial species produce the cyanotoxins cylindrospermopsin and 7-epi-cylindrospermopsin. The final step in the synthesis of these compounds is catalysed by the 2OG-dependent hydroxylase Cyrl (Figure 1.16D).³⁵⁶  Streptomyces hydrogscopicus subsp. Limoneus produces a family of validamycin compounds with antifungal activity. The main species, validamycin A, is converted to the less effective validamycin B by VldW, a 2OG-dependent hydroxylase (Figure 1.16E).³⁵⁷  Finally, the penultimate step in kanamycin biosynthesis by Streptomyces kanamyceticus is carried out by KanJ. This protein catalyses the hydroxylation of an amino sugar (Figure 1.16F) to yield an unstable hemiaminal which decomposes with release of ammonia to yield 2′-oxokanamycin; the latter compound is the substrate for an NADPH-dependent reductase, KanK, to produce the final antibiotic.

The enzyme isopenicillin N synthase (IPNS) catalyses the fascinating transformation of a linear tripeptide, l-δ-(α-aminoadipoyl)-l-cysteinyl-d-valine, into a bicyclic structure, as shown in Figure 1.17A. The resulting product is subsequently metabolized into penicillins and cephalosporins as described earlier in Section 1.7.1.³⁰³  Although IPNS does not utilize 2OG as a cosubstrate, it is related in sequence and structure to the 2OG-dependent oxygenases; indeed, it was the first family member to be structurally characterized.³⁵⁸  Like the 2OG-dependent enzymes, IPNS coordinates Fe(ii) via a 2-His-1-carboxylate motif and binds substrate within a double-stranded β helix fold.³⁵⁹  The thiolate sulfur atom of the substrate forms a ligand to the metallocentre. Snapshots of the reaction were structurally visualized after brief exposure of anaerobic crystals to high pressures of oxygen.³⁶⁰  This remarkable enzyme is described further in Chapter 19.

Another enzyme that does not utilize 2OG, yet is related by sequence and structure to the 2OG-dependent oxygenases, is 1-aminocyclopropane-1-carboxylate oxidase (ACCO).^361,362  This plant enzyme catalyses the synthesis of ethylene (Figure 1.17B) in a reaction that is distinct from that described in Section 1.6.1. The gaseous product is a phytohormone that functions in germination, fruit ripening and senescence. The cyclopropane-containing substrate of ACCO binds to the Fe(ii) site via its α-amino and α-carboxylate groups according to spectroscopic analyses with isotopically-labelled substrates and substrate analogues;^363,364  such coordination is reminiscent of the bidentate binding of 2OG to the metallocentre of related family members. Ascorbate is required for multiple turnovers of ACCO, but the reductant is not needed for a single turnover.³⁶⁵  Surprisingly, carbon dioxide is essential for catalysis in addition to being a product, and this molecule is proposed to stabilize the enzyme from undergoing inactivation reactions.³⁶⁶  Oxidative inaction of ACCO occurs rapidly and can result in cleavage of the peptide backbone.^367,368  Crystal structures have been resolved for the apoprotein and holoprotein forms of ACCO,³⁶⁹  and modelling studies have led to proposals for the binding sites of substrate, bicarbonate and ascorbate.³⁷⁰  This enzyme is described in further detail in Chapter 20.

Hydroxymandelate synthase (HMS) and 4-hydroxyphenylpyruvate dioxygenase (HPPD) are of special interest because they transform the same substrate into distinct products (Figure 1.17C) by reactions that are mechanistically related to those carried out by the 2OG-dependent oxygenases; however, HMS and HPPD are unrelated to the latter enzymes while being closely related to each other.³⁷¹  The substrate for both enzymes is 4-hydroxyphenylpyruvate, a 2-oxo acid, and like 2OG it undergoes oxidative decarboxylation with formation of carbon dioxide and 4-hydroxyphenylacetate. The enzyme intermediate resulting from the C–C cleavage reaction is used either to hydroxylate the methylene group of 4-hydroxyphenylacetate to form hydroxymandelate (in HMS) or it catalyses the ‘NIH shift’ in which substituent migration and ring hydroxylation produce homogentisate (in HPPD). Isotope effect studies have provided keen insights into how the intermediate partitions to form the two products.³⁷²  Four key residues in the active sites of these enzymes are critical for defining product specificity, and an HPPD has been engineered to exhibit HMS activity.³⁷³  Several catalytic intermediates have been detected by using spectroscopic approaches with these enzymes.^374,375  The structure of HMS from Amycolatopsis orientalis³⁷⁶  exhibits the same fold as found in HPPD, first reported for the enzyme from Pseudomonas fluorescens;³⁷⁷  however, their folds are unrelated to the typical 2OG oxygenase fold. Further information on this pair of enzymes is provided in Chapter 21.

Conspectus: The 2OG-dependent oxygenases catalyze a diverse array of reactions that have profound implications in biology. The structures and mechanisms of these fascinating enzymes are discussed in Chapters 2–4 and representative topics are detailed more fully in Chapters 5–21.

A large number of publications related to this topic have appeared since submission of this chapter, only three of which are cited here. A prolyl residue in prokaryotic elongation factor Tu is hydroxylated by a 2OG-dependent oxygenase, perhaps serving as an evolutionary precursor to prolyl hydroxylases used in oxygen sensing.³⁷⁹  A review describing these enzymes in coumarin synthesis has appeared.³⁸⁰  Finally, Wel05 was shown to be a 2OG-dependent halogenase acting on a free substrate (i.e. not tethered to a peptidyl carrier protein).³⁸¹ 

2OG-dependent oxygenase work in the author’s laboratory was supported by the National Institutes of Health (GM063584).