The shikimate pathway: gateway to metabolic diversity

Abstract

The shikimate pathway is the metabolic process responsible for the biosynthesis of the aromatic amino acids phenylalanine, tyrosine, and tryptophan. Seven metabolic steps convert phosphoenolypyruvate (PEP) and erythrose 4-phosphate (E4P) into shikimate and ultimately chorismate, which serves as the branch point for dedicated aromatic amino acid biosynthesis. Bacteria, fungi, algae, and plants (yet not animals) biosynthesize chorismate and exploit its intermediates in their specialized metabolism. This review highlights the metabolic diversity derived from intermediates of the shikimate pathway along the seven steps from PEP and E4P to chorismate, as well as additional sections on compounds derived from prephenate, anthranilate and the synonymous aminoshikimate pathway. We discuss the genomic basis and biochemical support leading to shikimate-derived antibiotics, lipids, pigments, cofactors, and other metabolites across the tree of life.

Graphical Abstract

1. Introduction

In plants and microorganisms, the 20 common proteinogenic amino acids can be organized into six families based on the shared metabolites that serve as their biosynthetic origin.¹ One of these biosynthetic families gives rise to the three aromatic amino acids (AAAs): phenylalanine (Phe, 1), tyrosine (Tyr, 2), and tryptophan (Trp, 3). The biosynthesis of these amino acids proceeds through a metabolic route called the “shikimate pathway”, named after the toxic Japanese star anise flower, Illicium anisatum, known as shikimi in Japanese, from which the key metabolite shikimate (4) was first isolated.²

The pathway begins with the intersection of two important primary metabolic processes, namely glycolysis and the pentose phosphate pathway, whereupon phosphoenolypyruvate (PEP, 5) and erythrose 4-phosphate (E4P, 6) combine to yield 3-deoxy-d-arabino-heptulosonate 7-phosphate (DAHP, 7) (Figure 1).³ A multi-step oxidation, elimination of phosphate, and reduction cascade then transforms pyranose DAHP into the first carbocyclic intermediate, 3-dehydroquinate (DHQ, 8). Dehydration of DHQ produces 3-dehydroshikimic acid (DHS, 9), the third intermediate in the pathway. Reduction of DHS forms shikimate (4), the pathway namesake and key metabolic intermediate. Phosphorylation via shikimate kinase generates shikimate-3-phosphate (10), the fifth shikimate pathway intermediate, which subsequently reacts with an additional unit of PEP to give 5-enolpyruvylshikimate-3-phosphate (EPSP, 11). Finally, the elimination of phosphate from EPSP yields chorismate (12), the last common intermediate for aromatic amino acid biosynthesis and the recognized endpoint of the shikimate pathway. The biosynthesis of Phe and Tyr proceeds through prephenate (13), via a enzymatic Claisen rearrangement catalyzed by chorismate mutase, while Trp is biosynthesized from anthranilate (14) generated by anthranilate synthase. In microbes, the shikimate pathway is regulated in part by AAA inhibition of DAHP synthase,⁴ while in higher organisms the complex regulation of this pathway has yet to be fully characterized. Allosteric regulation of enzymes downstream of the canonical shikimate pathway, including chorismate mutases and anthranilate synthases, have been identified as key regulatory pathways across kingdoms to control AAA flux.^3,5

Figure 1.

The shikimate pathway consists of seven biosynthetic steps beginning with the convergence of PEP and E4P and ending in the formation of chorismate. Steps converting chorismate to prephenate and anthranilate traditionally are not included in the shikimate pathway but are covered in this review as they are fundamental building blocks of natural products and specialized metabolites. Enzyme abbreviations in order of biosynthetic step: DAHPS, 3-deoxy-d-arabino-heptulosonate 7-phosphate (DAHP) synthase; DHQS, 3-dehydroquinate synthase; DHQD, 3-dehydroquinate dehydratase; SDH, shikimate dehydrogenase; SK, shikimate kinase; EPSPS, 5-enolpyruvylshikimate 3-phosphate synthase; CS, chorismate synthase; CM, chorismate mutase; and AS, anthranilate synthase.

Comparing the shikimate pathway enzymes between plants, fungi, and bacteria reveals remarkable harmony in reaction mechanisms across kingdoms.⁶ However, the molecular architecture of these biocatalysts varies dramatically (Figure 2). In bacteria, the shikimate pathway is encoded by discrete monofunctional enzymes that catalyze each biosynthetic step, often referred to as aro homologs, after the aro naming system adopted in Escherichia coli for their role in aromatic amino acid production.³ In plants, these seven reactions are performed by six enzymes, including a bifunctional DHQ dehydratase / shikimate dehydrogenase that leads to the formation of shikimate.^7,8 Fungi and protists, however, have evolved the AROM complex, a pentafunctional enzyme that performs the five consecutive reactions from DAHP to EPSP.⁹

Figure 2.

Crystal structures of shikimate pathway enzymes from bacteria, fungi, and plants. PDB codes for bacterial structures: AroG (phenylalanine regulated) – 1QR7; AroB – 3CLH; AroD (type I) – 1QFE; AroD (type II) – 1GU0; AroE – 1NYT; AroK – 1KAG; AroA – 1G6S; and AroC – 1Q1L. Fungal structure PDB codes: ARO4 – 1HFB; AROM – 6HQV; and CS – 1CSM. Plant PDB codes: DHQ-SDH – 2GPT; and SK2 – 3NWJ. Structural domains are color coded according to function regardless of organism of origin.

Notably, the shikimate pathway is found across domains of life in bacteria, archaea, fungi, algae, some protozoans, and plants, but not in animals. Instead, animals must retrieve aromatic amino acids from their diet, meaning that phenylalanine and tryptophan are considered “essential” amino acids. Tyrosine remains an exception as it can be produced in animals via oxidation of dietary phenylalanine by the biopterin-dependent phenylalanine hydroxylase. Because of this metabolic difference, the shikimate pathway has long been recognized as an attractive target for herbicides, antibiotics, and antiparasitics due to the inherent lack of human cross toxicity. For example, the widely used herbicide glyphosate (known as the Monsanto product Roundup^®), inhibits the sixth step of the shikimate pathway, EPSP synthase. Transgenic Roundup^® Ready plants carry a form of the EPSP synthase gene that is not sensitive to glyphosate, conferring resistance to the herbicide.¹⁰

While the amino acid endpoints of the shikimate pathway, Phe, Tyr, and Trp, are commonly incorporated into peptides, cofactors, pigments, organic polymers, alkaloids, and a bevy of other molecules, each of the pathway intermediates en route to these AAAs can serve as a branchpoint for specialized metabolite biosynthesis as well. The highly oxidized and densely functionalized scaffolds of shikimate pathway intermediates make them perfectly poised to veer off their primary metabolic pathway midpoint and into secondary metabolism instead. Even the multidomained AROM complex, encoded by the gene ARO1, is notoriously “leaky,” allowing for derivatization of pathway intermediates.^8,9,11 This review focuses on these cases, where secondary metabolism hijacks intermediates from the shikimate pathway and uses them instead in the biosynthetic pathways of specialized metabolites in bacteria, fungi, and plants. We highlight natural products derived from intermediates of the shikimate pathway along the seven steps from PEP and E4P to chorismate, as well as sections on compounds originating from prephenate and anthranilate. Additionally, we have included a section on the aminoshikimate pathway, as many compounds originally thought to be of shikimate origin instead arise from this analogous pathway. The examples highlighted here are not meant to be comprehensive, but rather to demonstrate the breadth of molecular diversity generated from shikimate intermediates and to highlight topics for future discovery.

We intend for this article to update Professor Heinz Floss’s 1997 review on the same topic¹² for the genomic age by focusing on the genomic basis for shikimate-derived natural products and connecting molecular products to their encoding genes. Unfortunately, a systematic naming convention has not been adopted for either the shikimate pathway nor products thereof, and this is reflected in the literature with compounds sometimes named as acids and sometimes as their conjugate bases. To enable readers to easily find additional articles on topics and molecules discussed herein, we have intentionally selected the name most widely used in the literature for each compound. We discuss the biochemical and genetic experiments used to link shikimate pathway-derived secondary metabolites to their biosynthetic genes, as well as point out research areas for further investigation.

2. 3-Deoxy-d-arabinoheptulosonate 7-phosphate (DAHP)

The convergence of phosphoenolpyruvate (PEP, 5) and D-erythrose 4-phosphate (E4P, 6) to yield 3-deoxy-d-arabinoheptulosonate 7-phosphate (DAHP, 7) is commonly recognized as the first committed step in the shikimate pathway across taxa (Figure 3).¹³ This reaction is catalyzed by DAHP synthase¹⁴ (DAHPS, PFAM: PF00793), an aldolase whose catalytic function is often dependent on divalent cations.¹⁵ The first high-resolution X-ray crystal structure was that of the phenylalanine-regulated DAHPS from E. coli (AroG).¹⁶ Since then, crystal structures of DAHPS from across the kingdoms of life have been solved, many in complex with substrates, metal cofactors, and inhibitors, illuminating the catalytic mechanism of this enzyme.¹⁷

Figure 3.

Aldol reaction between PEP (5) and E4P (6) catalyzed by DAHP synthase.

In addition to catalyzing the first step in the shikimate pathway, DAHPS controls the flux of carbon into the shikimate pathway through complex and well-studied regulatory mechanisms. The genetic and feedback regulation of DAHPS has been extensively studied and summarized.⁴ Briefly, this control is accomplished, in part, by plants, fungi, bacteria, archaea, and protists encoding multiple DAHPS orthologs in their genome, each uniquely sensitive to a specific downstream product of the shikimate pathway (i.e., Phe, Tyr, Trp). DAHPSs are classified into two phylogenetically distinct groups, the smaller (~39 kDa) “bacterial type” AroA_I class, which supports primary metabolic processes as epitomized by the three paralogous DAHPSs in E. coli,⁴ and the larger (~54 kDa) “plant type” AroA_II class associated with the production of specialized metabolites.¹⁸

There are no examples yet reported in the literature wherein DAHP (7) is shunted directly into a non-shikimate biosynthetic intermediate. However, AroA_II-class DAHPSs are often embedded in secondary metabolite biosynthetic gene clusters (BGCs)^19,20 and have been used as gene hooks to query for orphan BGCs that utilize a shikimate pathway-derived building block.²¹

3. 3-Dehydroquinate (DHQ)

3-Dehydroquinate (8, DHQ) is the second intermediate in the shikimate biosynthetic pathway and the first carbocyclic product of the pathway (Figure 4). The mechanism for conversion of the pyranose precursor DAHP to carbocyclic DHQ has been informed by high resolution crystal structures of dehydroquinate synthase (DHQS) in complex with metals, nicotinamide cofactors, and substrate-based inhibitors, and is proposed to proceed via a complex cascade of oxidation, elimination, reduction, ring opening, and intramolecular aldol cyclization reactions.^22–24 Despite the intricate nature of this multistep transformation, DHQS rarely generates shunt products, demonstrating the ability of this enzyme to stabilize intermediates and disfavor alternative modes of reactivity throughout the cascade.²⁴

Figure 4.

Conversion of DAHP (7) to DHQ (8) by DAHPS and its similarity to the transformation of sedoheptulose-7-phosphate to desmethyl-4-deoxygadusol (16) by Ava_3858 from A. variabilis en route to mycosporines such as shinorine (15).

Structural similarity between the widely distributed mycosporines, mycosporine-like amino acids (MLAAs), and shikimate pathway intermediates garnered early hypotheses that DHQ may be an intermediate in the biosynthesis of MLAAs.²⁵ Isotope feeding studies in a variety of mycosporine/MLAA producers demonstrated the likely origin of the cyclohexane core to be derived from the shikimate pathway rather than polyketide origin.²⁵ To determine the source of the mycosporine/MLAA carbocycle, Walsh and coworkers identified and characterized the complete biosynthetic gene cluster for shinorine (15) from the cyanobacterium Anabaena variabilis. They demonstrated that the conserved mycosporine precursor, desmethyl-4-deoxygadusol (16), is derived from an alternative seven-carbon pyranose, sedoheptulose-7-phosphate (17) (Figure 4).²⁶ These 3-DHQS-like sugar phosphate cyclases are implicated in the biosynthesis of a large and diverse family of shikimate-resembling natural products and have been the subject of a comprehensive review.²⁷

The paucity of microbial specialized metabolites directly derived from DHQ is curious and may reflect an underrepresented and yet to be discovered group of microbial metabolites and thus a target for future genome mining campaigns. This dearth of microbial compounds of DHQ origin is particularly notable given the large family of acyl-quinic acids isolated and characterized from plants.²⁸ Numbering over 300 examples to date, acyl-quinic acids such as 1-o-feruloyl-1,5-lactone (18) and 1,2-o-dicaffeoylquinic acid (19), also known as chlorogenic acids, play fundamental roles in plant maturation and lignin biosynthesis. The general composition of these compounds consists of two units derived from the shikimate pathway: 1) the coumaryl-(20), caffeoyl-(21) or feruloyl-CoAs (22) derived from phenylalanine, and 2) the quinate (23) core which is acylated by recently characterized classes of plant acyltransferases (Figure 5).²⁹ Notably, a recent phylogeny driven investigation into the genetic origin of quinate in plants by Ehlting and coworkers demonstrated that seed plants evolved an offshoot from the shikimate pathway to generate the quinate utilized in these specialized metabolites.³⁰ A combination of bioinformatics and in vitro biochemistry data strongly support a gene duplication of DHQ dehydrogenase ~300 million years ago generated a phylogenetically distinct clade of paralogs, quinate dehydrogenases, which selectively convert DHQ (8) to quinate (23) rather than dehydroshikimate. Using ancestral sequence reconstruction, the molecular basis for this switch in selectivity between converting DHQ to DHS or quinate was pinpointed to a single active site variation.

Figure 5.

General biosynthesis of acyl-quinic acids and representative examples of known acyl-quinic acids from plants.

4. 3-Dehydroshikimate (DHS)

The dehydration of DHQ by 3-dehydroquinate dehydratase (DHQD) introduces the double bond into the six-membered carbocycle and yields the third intermediate of the shikimate pathway, 3-dehydroshikimate (DHS, 9) (Figure 6). Two classes of DHQD enzymes exist: type I and type II, both found in bacteria.³¹ These proteins share little sequence or structural homology and catalyze dehydration via opposite stereochemical routes. Type I DHQD enzymes catalyze dehydration using an active site lysine residue that forms a transient Schiff base intermediate, which enables the net syn elimination of water via deprotonation of the C-2 pro-R hydrogen.³² Type II DHQDs lack this catalytic Lys and instead catalyze dehydration via deprotonation of the C-2 pro-S hydrogen to give an enolate intermediate followed by E1cB-type elimination to give the net anti-elimination of water.³³ Interestingly, the distribution of type I vs type II DHQDs in bacteria does not appear to follow species phylogeny, although the presence of type II DHQHs in many pathogenic organisms has fueled interest in developing selective inhibitors.³⁴ In fungi, DHQ dehydration is catalyzed by a DHQD type I domain embedded in the multifunctional AROM protein complex, while in plants, bifunctional didomain enzymes catalyze dehydration to DHS and subsequent reduction to shikimate.³⁵ DHS serves as a starting point for the biosynthesis of several metabolic building blocks in microbes and plants, namely 3,4-dihydroxybenzoic acid (24) as found in the natural product petrobactin (25), ^36–43 the unusual 3-aminobenzoic acid (26) substituent in pactamycin (27), ^44–50 and gallic acid (28), a key component of plant tannins and the structurally diverse ellagitannins such as tellimagrandin II (29).^51–57

Figure 6.

The dehydration reaction catalyzed by 3-dehydroquinate dehydratase to generate 3-dehydroshikimate (9) with H_R and H_S representative of pro-R and pro-S hydrogens, respectively, and enzymes and products responsible for trafficking DHS into 3,4-DHBA, 3-ABA, and gallic acid.

4.1. 3,4-Dihydroxybenzoic acid (3,4-DHBA)

Bacillus anthracis, the causative agent of anthrax, produces the siderophore petrobactin (25), which is considered an essential virulence factor in B. anthracis infections.³⁸ The catecholic 3,4-dihydroxybenzoic acid (3,4-DHBA, 24) termini of petrobactin are requisite for iron chelation and allow for petrobactin to evade sequestration by the mammalian immunoprotein siderocalin, which sequesters 2,3-dihydroxybenzoic acid containing siderophores as part of an antibacterial iron-depletion defense system.^39,40 Discovery of the petrobactin BGC in Bacillus and sequential deletion of every gene in the cluster revealed that asbF, first annotated as a hypothetical protein,^41,42 was essential for 3,4-DHBA (24) production. Stable-isotope experiments suggested 3,4-DHBA was derived from the shikimate pathway,⁴³ with the Koppisch and Sherman groups independently revealing that AsbF is a DHS dehydratase.^36,37 AsbF catalyzes the dehydration and aromatization of DHS into 3,4-DHBA via a proposed metal-dependent E1cB reaction reminiscent of type-II DHQDs (Figure 7).^36,37

Figure 7.

Aromatizing dehydration reaction catalyzed by AsbF to yield 3,4-DHBA (24) from DHS (9) en route to biosynthesis of stealth siderophore petrobactin (25).

Recent work has demonstrated that the production of petrobactin by Alteromonas species is critical for mediating iron acquisition across ocean environments, further demonstrating the utility of this shikimate-derived moiety, as well as its widespread distribution across bacteria.⁵⁸ To further validate this observation, we performed a cblaster analysis⁵⁹ using the Bacillus anthracis asb cluster as a search query, requiring resultant hits to contain asbF as well as one other gene from the asb cluster. Both the minimum identity and minimum query coverage were set to 20%, and the maximum number of BLAST hits was increased to 10,000, allowing us to visualize distribution of asbF-containing gene clusters across prokaryotes. Our search returned 5,309 gene clusters in 4,125 organisms. Representative examples of gene clusters containing asbF from diverse bacteria are displayed in Figure 8. This initial analysis demonstrates that asbF, as well as petrobactin biosynthesis, is widely distributed across bacteria. Moreover, it revealed that poaching DHS from shikimate biosynthesis for conversion to 3,4-DHBA is similarly a conserved biosynthetic strategy despite the existence of a variety of alternative biosynthetic routes to generate 3,4-DHBA, including tyrosine degradation,⁶⁰ oxidation of hydroxybenzoic acids,⁶¹ and demethylation of vanillic acid.⁶²

Figure 8.

Representative examples of asb-type clusters distributed across diverse bacterial classes. Figure made with cblaster⁵⁹ and clinker⁶³ via CAGECAT.⁶⁴

4.2. 3-Aminobenzoic acid (3-ABA)

Pactamycin (27) is a cytotoxic natural product produced by Streptomyces pactum comprised of a distinctive aminocyclopentitol core, decorated with 3-aminoacetophenone (3-AAP), 6-methylsalicyclic acid, and N,N-dimethylurea moieties (Figure 9).⁴⁷ Despite the potent bioactivity of pactamycin, broad-spectrum toxicity limited its development as a therapeutic, and interest in the development of analogs with improved selectivity was a key driver to discover the pactamycin biosynthetic genes.^48,49 Elegant stable-isotope experiments were key in determining the shikimate pathway origin of 3-AAP,⁵⁰ and revealing 3-ABA (26) as a 3-AAP precursor.^45,46 The enzymes transforming a shikimate intermediate to 3-ABA, however, remained elusive.

Figure 9.

Characterized reactions in pactamycin biosynthesis include the conversion of DHS (9) to 3-ABA (26) and subsequent thiotemplated elongation and glycosylation.

A putative pactamycin BGC (pct) was identified independently by the Eguchi and Mahmud groups using a characteristic methyltransferase as a genome mining hook,^48,65 and gene inactivation and heterologous expression experiments linked this putative pct BGC to pactamycin production. In vitro biochemistry identified the PLP-dependent aminotransferase, PctV, as the sole enzyme responsible for both transamination and double dehydration of DHS with strict substrate specificity (Figure 9).⁴⁶ Recently, the Eguchi and Mahmud groups also independently demonstrated that the conversion of 3-ABA into 3-AAP as well as glycosylation are both carrier protein dependent, where 3-ABA is adenylated to 3-ABA-AMP (30) by PtmS and loaded onto carrier protein PtmI to give thiotemplated intermediate 3-ABA-PtmI (31). The two-carbon elongation of 31 via 32 to 33 is catalyzed by β-ketoacyl-ACP synthase, PtmK, revealing the origin of the methyl substituent in 3-AAP. Glycosyltransferase PtmJ couples N-acetylglucosamine to carrier protein-bound intermediate 33 to give the glycosylated product 34, which must still undergo multiple, yet to be described, complex transformations and rearrangements before arriving at pactamycin (27).⁴⁴

These newly characterized biosynthetic steps in pactamycin assembly are mirrored in the biosynthesis of the aminoshikimate pathway derived natural products, the ansamycins and mitosanes. The 3,5-aminohydroxybenzoic acid (3,5-AHBA) starter unit of the mitomycins and rifamycins is synthesized from amino-dehydroshikimate by so-called “3,5-AHBA synthases” that catalyze an analogous reaction to PctV. In further reflection of pactamycin biosynthesis, glycosyltransfer as well as additional oxidative chemistry have also recently been demonstrated to be carrier protein-dependent in mitomycin biosynthesis (Section 11).

4.3. Gallic acid

Dehydroshikimate is the direct precursor to gallic acid (28), a key component of plant tannins and recognized on its own for potent antioxidant and anti-inflammatory activity.⁶⁶ To determine if gallic acid was derived from an early shikimate precursor or phenylalanine catabolism, feeding experiments with stable-isotope labeled glucose were conducted using a prolific gallic acid producing fungus, Phycomyces blakesleeanus, which revealed the direct conversion of DHS to gallic acid.⁶⁷ Follow up studies in the plant Rhus typhina used natural oxygen isotope abundance to demonstrate the same early shikimate origin, DHS.⁶⁸ Ultimately, a shikimate dehydrogenase (SDH) enzyme from Juglans regia (English walnut) was identified that catalyzed the formation of both shikimate and gallic acid.⁶⁹ The plant SDH enzyme was able to complement shikimate deficient E. coli, restoring shikimate production and enabling gallic acid biosynthesis. It also produced gallic acid in vitro. Additionally, transgenic expression of the J. regia SDH enabled gallic acid production in Nicotiana tabacum. Gallic acid is thus an unusual case in which both a shikimate pathway intermediate (DHS in this case) and the shikimate pathway enzyme itself are appropriated to play a role in secondary metabolite production.

Gallic acid is a key component of the ellagitannins, an extremely large and structurally diverse family of hydrolysable tannin natural products produced by plants and generally derivatized from 1,2,3,4,6-penta-O-galloyl-β-d-glucopyranose (35). Since the early 1980’s, Gross and coworkers have investigated the biosynthesis of the ellagitannins and through activity guided fractionation experiments unveiled novel and unexpected enzymology, which has been reviewed.⁵⁵ However, as gallic acid-based reviews are often not written in the context of the shikimate pathway, and often reviews on the shikimate pathway and natural products thereof are focused on microbial biochemistry, we briefly highlight ellagitannin biosynthesis here to demonstrate the true diversity of natural products derived from shikimate intermediates across taxa.

Using ¹⁴C-radiolabeled gallic acid and UDP-glucose, Gross used cell-free extracts⁵⁶ and partially purified enzyme⁵⁷ from oak leaves (Quercus rober) to identify a glycosyltransferase that generates β-glucogallin (βG, 36), a proposed theoretical precursor to the ellagitannins. In an unusual set of reactions catalyzed by acyltransferases isolated from oak and sumac, βG acts as both galloyl-donor and acceptor,⁷⁰ with “galloylation” of βG proceeding in a highly specific order.^51–54 This reaction mimics the order of substitution of glucose with chemical electrophiles. βG (and not a gallate-CoA thioester) is also the gallate donor in the biosynthesis of other gallotannins and flavonoid hydrolysable tannins.⁷¹ To identify enzymes involved in the oxidative coupling of 35 to complex ellagitannins, Gross and Rausch developed an elegant approach to generate [U-¹⁴C]1,2,3,4,6-pentagalloylglucose through photoassimilation of ¹⁴CO₂,⁷² which enabled highly sensitive activity guided fractionation campaigns that identified a novel laccase from the prolific ellagitannin producer Tellima grandiflora (fringe cups, Saxifragaceae), which selectively oxidized 35 to tellimagrandin II (37).^73,74 In cell free extracts from T. grandiflora, a selective dimerization of 37 to cornusiin E (38) was also observed, and subsequent fractionation identified a second laccase, which catalyzed the dimerization of 37.⁷⁵ These studies laid the groundwork for the current understanding of ellagitannin biosynthesis, and defied expectations that dimerizations would be non-selective. Despite this work, few additional ellagitannin biosynthetic enzymes and pathways have been characterized and represent a vast unexplored chemical space for genome mining.

5. Shikimic acid

Shikimate dehydrogenase catalyzes the stereoselective reduction of 3-dehydroshikimate (DHS, 9) to shikimate (4), the namesake molecule and fourth intermediate of the pathway (Figure 11). The archetypal shikimate dehydrogenase, AroE from E. coli, has been the subject of extensive structural and biochemical characterization.⁷⁶ Bacterial shikimate dehydrogenases are monofunctional enzymes, whereas shikimate dehydrogenases are discrete domains in eukaryotic multifunctional enzymes that catalyze multiple steps in shikimate biosynthesis. Bacterial shikimate dehydrogenases can be classified into four groups (AroE, YdiB, SDL, RifI) based on their differential enzymatic properties,⁷⁷ where AroE is selective for both dehydroshikimate and NADPH as a reducing agent. Homologs demonstrate relaxed substrate specificity for both hydride donor (NADH or PQQ) as well as acceptor (quinate), including E. coli paralog YdiB, which can utilize both NADH and NADPH.⁷⁸ To date, the processing of shikimate into reduced and dehydrated cyclohexanes and cyclohexenes are the known routes to repurpose shikimate into natural product scaffolds, however recent investigations have demonstrated the flexibility of shikimate dehydrogenases in the biocatalytic synthesis of iminosugars.⁷⁹

Figure 11.

Conversion of 3-dehydroshikimic acid (9) to shikimic acid (4) by shikimate dehydrogenase.

Shikimate itself is rarely directly incorporated into natural product scaffolds, with few notable exceptions such as the chlorogenic acids found in plants, as discussed in Section 3 (Figure 5). Recently, the fungal natural product (–)-maximiscin (39) was discovered as a cell-cycle inhibitor from Tolypocladium sp. Salcha MEA-2 and suggested to originate from shikimate based on isotope experiments. While the biosynthetic pathway for this compound remains to be fully elucidated, the cyclohexyl portion of this molecule may alternatively originate from sedoheptulose-7-phosphate. To date, the characterized pathway that utilizes shikimate as a dedicated starting material is the biosynthesis of cyclohexanecarboxyl-CoA (CHC-CoA, 40), a specialized precursor to fatty acids and antibiotics (Figure 12).

Figure 12.

Isotopic feeding studies demonstrated that the cyclohexane carbocycle in fatty acids 41 and 42 is derived from shikimate (4) and CHC (43) as a viable late-stage intermediate. Cyclohexane-containing secondary metabolites derived from CHC-CoA (40) are shown in which the cyclohexyl feature is colored red.

5.1. Cyclohexanecarboxylic acid (CHC)

Monosubstituted cyclohexanes are unusual motifs in metabolism, and the first insights into their molecular origin came from investigations using the thermoacidophilic bacterium Alicyclobacillus acidocaldarius, which produce ω-cyclohexyl fatty acids (41, 42) as a major constituent (73–94%) of their total lipid content.⁸⁰ Through isotope-labeled precursors, Oshima and Ariga were able to demonstrate the cyclohexyl moiety of these fatty acids was derived from shikimate.⁸¹ Floss and coworkers also used isotope tracing to demonstrate that the cyclohexanecarboxylic acid (CHC, 43) moiety in the antibiotic ansatrienin A (44) produced by Streptomyces collinus was similarly derived from all seven carbons of shikimate (Figure 12).⁸² Follow up work by Floss utilized detailed stable isotope feeding experiments in both A. acidocaldarius and S. collinus with stereoselectively labeled isotopologues of shikimate to establish the putative intermediates from shikimate to CHC via a multistep pathway wherein no intermediates are aromatic, and substrates are likely activated as CoA-thioesters rather than free acids (Figure 13).^83–87

Figure 13.

Zhang and coworkers⁹⁶ revised biosynthetic hypothesis for the CHC pathway (biochemically characterized steps from their work are shown in red).

Contemporaneously, Reynolds and coworkers identified a novel enoyl coenzyme A reductase (ChcA) from S. collinus, which catalyzed the final step in CHC-CoA biosynthesis, the reduction of 1-cyclohexenylcarboxyl-CoA (45) to cyclohexanecarboxyl-CoA (CHC-CoA, 40) (Figure 13).⁸⁸ Examination of the genetic context of chcA in the S. collinus genome revealed additional putative CHC biosynthetic enzymes upstream and downstream of chcA. Plasmid-based expression of the five-gene cassette in the heterologous host Streptomyces lividans resulted in dramatically increased production of cyclohexyl fatty acids (from <1% to 49% total fatty acid content), demonstrating that these five genes represent the minimal suite of enzymes required for the conversion of shikimate to CHC-CoA.⁸⁹ Transformation of this cassette into a mutant strain of Streptomyces avermitilis also resulted in a fermentation strain capable of the production of doramectin (46), a non-natural analogue of ivermectin still used in veterinary medicine.^90–94 This work laid the foundation for the characterization of additional cyclohexane-containing natural products such as asukamycin (47) and phoslactomycins A (48) and B (49) and has been summarized in depth previously.^12,95 This section will focus instead on recently published studies that have continued to refine our understanding of the biosynthesis of CHC-CoA from shikimate and its unusual processing in polyketide synthase pathways.

To resolve some of the remaining steps in CHC-CoA biosynthesis, Zhang and coworkers used in vitro enzymology to assign functional roles to AsuB1 and AsuB2, enzymes proposed to be involved in the early stages of CHC-CoA biosynthesis from the asukamycin producer, Streptomyces nodosus subsp. asukaensis.⁹⁶ AsuB1, a didomain protein with an N-terminal domain homologous to EPSP synthase and C-terminal domain homologous to coumarate-CoA ligase, was recalcitrant to functional expression in E. coli. However, the C-terminal domain, AsuB1_C, was successfully expressed and shown to catalyze the ATP-dependent thioester bond formation between shikimate and coenzyme A to give shikimate-CoA (50, Figure 13). Assays using the proposed alternative substrate shikimate-3-phosphate (10) resulted in no product formation, but AsuB1_C did show remarkable substrate flexibility amongst alternative substrates. Based on similarities to the ansatrienin homologue AsnJ, the flavin-dependent dehydratase AsuB2 was proposed to catalyze an unconventional dehydration of 55 to give known intermediate 45. ⁸⁹ Assays with recombinant AsuB2 and 55 indeed resulted in dehydration with 1-cyclohexenecarboxyl-CoA (45) as the major product. Finally, ChcA homolog, AsuB3, was confirmed to catalyze the reduction of 45 to 40. Through this initial biochemical analysis, a revised CHC biosynthetic hypothesis was proposed which excluded shikimate-3-phosphate as an intermediate (Figure 13). Details of the full biosynthetic conversion of shikimate to CHC-CoA are however still incomplete.

Qu and coworkers recently investigated the unusual PKS machinery that elongates the CHC-CoA starter unit into the polyene “upper chain” portion of asukamycin.⁹⁷ β-ketoacyl-ACP synthase III (KAS III) enzymes are canonically associated with the initiation of fatty acid and polyketide biosynthesis, wherein they catalyze the condensation of a malonyl-ACP and an acyl-CoA to generate a β-ketoacyl-ACP product.⁹⁸ Notably, the asukamycin BGC contains genes encoding two KAS III enzymes, AsuC3 and AsuC4 (29% AA sequence ID), adjacent to an acyl carrier protein (AsuC5), ketoreductase (AsuC7), and two dehydratases (AsuC8 and AsuC9). These enzymes were individually cloned and heterologously expressed in E. coli to investigate their roles in the elongation of CHC-CoA. Through in vitro reconstitution of the biocatalytic cascade performed by these enzymes, it was discovered they worked together to iteratively elongate CHC-CoA into the asukamycin upper chain (Figure 14). In addition to CHC-CoA, 19 additional acyl-CoAs were evaluated as alternative starter units that could be iteratively elongated with broad ranging flexibility observed across the system. The iterative activity of a KAS III system is unusual, and when AsuC3/AsuC4 were used as bioinformatic hooks, co-occurrence of two KAS III genes was observed across a variety of uncharacterized clusters of genes, intimating possibly overlooked BGCs with unique product profiles.

Figure 14.

Proposed elongation of CHC-CoA (40) into the upper triene chain of asukamycin (47).

6. Shikimate 3-phosphate (S3P)

Shikimate kinase, the fifth enzyme in the pathway, catalyzes the ATP-dependent phosphorylation of the 3-hydroxyl group of 4 to yield shikimate-3-phosphate (S3P, 10) (Figure 15). This enzyme requires a divalent cation (either Mg²⁺ or Mn²⁺) for activity and is a member of the nucleoside monophosphate kinase family, containing the characteristic domains for substrate binding, ATP binding, and the lid.⁹⁹ Both substrate and ATP binding induce massive conformational changes, lowering the lid domain and forcing ATP and the substrate into close proximity, resulting in transfer of the phosphate group. Two isozymes of shikimate kinase exist in E. coli (AroL and AroK), while the number of isozymes in plants is species specific.¹⁰⁰ No specialized metabolites have yet been reported to arise from a metabolic offshoot at S3P.

Figure 15.

The ATP-dependent phosphorylation of shikimate (4) to shikimate 3-phosphate (10) catalyzed by shikimate kinase in the fifth step of the shikimate pathway.

7. 5-Enolpyruvylshikimate 3-phosphate (EPSP)

In the penultimate step of the shikimate pathway, a second unit of PEP reacts with S3P, resulting in the loss of inorganic phosphate and generating 5-enolpyruvylshikimate-3-phosphate (EPSP, 11) (Figure 16). This reaction is performed by an alkyl transferase-type enzyme referred to as EPSP synthase and proceeds via a tetrahedral intermediate with the net result of C–O bond cleavage of PEP (5) rather than the traditional P–O bond cleavage performed by most PEP-utilizing enzymes.¹⁰¹

Figure 16.

The condensation of PEP and shikimate 3-phosphate to 5-enolpyruvylshikimate 3-phosphate is catalyzed by EPSP synthase.

EPSP synthase is of particular note as it is the target of herbicides such as glyphosate, which occupies the PEP binding site of EPSP synthase, acting as a competitive inhibitor (Figure 17).¹⁰² EPSP synthase enzymes are organized in three classes: Class I enzymes are sensitive to glyphosate and typically found in plants or bacteria; Class II enzymes are microbial in origin and vary in glyphosate sensitivity without compromising PEP binding; and finally Class III enzymes are also microbial in origin but share less than 25% identity to E. coli EPSPS.¹⁰³ Glyphosate tolerant versions of EPSP synthase have naturally been identified in some organisms, which inspired the development of transgenic Roundup Ready^® crop lines that express a resistant form of EPSP synthase to enable glyphosate herbicidal treatment.¹⁰ Because of the success of this system, EPSP synthase inhibitors are of tremendous interest in agricultural research. Despite the focus on EPSP synthase, there are no natural products derived from EPSP itself. However, this may be a promising area for genome mining as EPSP mimics may serve as potential herbicides or antibiotics through inhibition of EPSPS.

Figure 17.

E. coli EPSPS crystal structure with glyphosate and S3P in the active site (PDB: 1G6S). Glyphosate occupies the PEP binding site, interacting with residues: Lys-22, Arg-124, Asp-313, Arg-344, Arg-386, and Lys-411.

8. Chorismic acid

The final step in the shikimate pathway is the conversion of EPSP (11) to chorismate (12) catalyzed by chorismate synthase (Figure 18). This enzyme catalyzes the 1,4-trans elimination of phosphate from 11, adding a second double bond to the six-membered ring.¹⁰⁰ Chorismate synthase requires reduced flavin mononucleotide (FMNH₂) despite the fact that the reaction involves no net redox change.^104,105 The mechanism of chorismate synthase has been interrogated using a variety of elegant kinetic isotope and structure biology driven studies whereupon chorismate synthases can be grouped into two classes: those found in E. coli and higher plants are considered monofunctional and are unable to reduce FMN, while chorismate synthases from fungi and some protozoans are considered bifunctional and possess the ability to reduce FMN with NADPH.^105,106

Figure 18.

The reaction catalyzed by chorismate synthase (CS).

Chorismate has a tremendous number of metabolic fates that are particularly well studied in microorganisms (Figure 19).^107–109 It can undergo a Claisen rearrangement to yield prephenate (13) or be transformed to anthranilate (14), both reactions leading to canonical AAA biosynthesis. However, chorismate can also be directed away from amino acid biosynthesis entirely by chorismate-utilizing enzymes and converted into a variety of diverse scaffolds, which act as building blocks for cofactors and specialized metabolites. Chorismate-utilizing enzymes and their explicit mechanisms have been recently reviewed,¹⁰⁷ which demonstrated that a remarkably small number of enzyme folds are responsible for the diversity of metabolic products originating from chorismate. In this section, we touch briefly on the pathways that led to the discovery of these novel branchpoints and focus on the range of unique biosynthetic pathways that utilize chorismate for the construction of structurally distinct metabolites. Much like chorismate itself, many of these chorismate-derived metabolites also serve as “branchpoints” and give rise to a variety of structurally diverse products, many of which are highlighted in their respective subsections below. Specialized pathways deviating from prephenate and anthranilate, for instance, are not discussed here but rather separately in Sections 9 and 10, respectively.

Figure 19.

Metabolic fates of chorismate in microorganisms.

8.1. Isochorismic acid

One of the alternative metabolic routes of chorismate beyond AAA biosynthesis is the formation of isochorismate (56) by isochorismate synthase in bacteria, fungi, and plants. The reaction proceeds through the stereo- and regiospecific displacement of the 4-hydroxyl group in chorismate by addition of water at C-2. Here, we highlight three biosynthetic outcomes that originate from isochorismate leading to the 2,3-dihdyroxybenzoic acid moiety of bacterial siderophores like enterobactin (57), the 2-hydroxybenzoic acid (salicylic acid, 58) motif in microbes and plants, and the vitamin menaquinone (59, Figure 20).

Figure 20.

Natural products derived from isochorismate (56).

8.1.1. 2,3-Dihydroxybenzoic acid

The bacterial siderophore enterobactin (57), also known as enterochelin, is one of the most powerful natural chelators known, with an iron binding affinity of 10⁵² M⁻¹.¹¹⁰ Enterobactin is found widely in Gram-negative bacteria but is predominantly associated with E. coli and Salmonella strains. As a natural trimer, enterobactin possesses C3 symmetry and is composed of three 2,3-dihydroxybenzoyl-serine units wherein iron is chelated via the catechol functionalities. To identify the genetic basis for production of the enterobactin virulence factor, the enterobactin (ent) gene cluster was located through sequencing E. coli mutants unable to produce enterobactin.^111,112 Chorismate was identified as an intermediate en route to enterobactin, as crude enzymatic preparations showed the conversion of chorismate to 2,3-dihdyroxybenzoic acid (DHBA, 60),¹¹³ while the isolation of isochorismate from cell extracts began to establish a biosynthetic route to DHBA production.¹¹⁴ Expression and purification of each of six enzymes, EntA-F, as recombinant proteins revealed that EntC functions as an isochorismate synthase, and initiates enterobactin biosynthesis through the transformation of chorismate into isochorismate (56).¹¹⁵ The next enzyme, the lyase EntB, catalyzes the loss of pyruvate to 2,3-trans-CHD (61), which is then aromatized by EntA to yield DHBA (60) (Figure 21).¹¹⁶ EntDEF are then responsible for the downstream construction of mature enterobactin from three molecules of serine and three DHBA units. The elucidation of enterobactin biosynthesis revealed a new metabolic fate for chorismate in specialized metabolism, and EntC remains archetypical for the identification of isochorismate-derived natural products.

Figure 21.

Conversion of chorismate (12) to DHBA (60) catalyzed by enterobactin biosynthetic enzymes.

Beyond siderophores such as enterobactin, DHBA is a metabolic building block in diverse specialized metabolites. A contemporary example includes the β-lactone antibiotic, obafluorin (62) from Pseudomonas fluorescens wherein NRPS-type machinery couple DHBA with a nitrated non-proteinogenic amino acid, (2S,3R)-2-amino-3-hydroxy-4-(4-nitrophenyl)butanoate (also derived from chorismate), followed by thioesterase-mediated cyclization to the β-lactone core (Figure 22).

Figure 22.

Abbreviated biosynthesis of β-lactone natural product, obafluorin (62), constructed from two chorismate-derived building blocks.

8.1.2. Salicylic acid

Salicylic acid (58) plays key biological functions across the tree of life, including acting as an integral building block in bacterial siderophores, as a defensive response chemical in plants, and as the active ingredient in aspirin. Identifying the biosynthetic pathway for salicylic acid biosynthesis was driven, in part, by the identification of the salicylic acid-derived siderophore, mycobactin (63), as an essential virulence factor of Mycobacterium tuberculosis. Feeding radioisotopes of shikimate pathway intermediates by Ratledge, Bentley, and coworkers revealed salicylic acid was derived from isochorismate.^117–119 From salicylic acid-producing Pseudomonas aeruginosa PAO1, Haas and coworkers identified a two-gene operon composed of a putative isochorismate synthase (pchA) and a gene of unknown function (pchB), which were able to complement salicylic acid biosynthesis in vivo as well as generate salicylic acid from chorismate in vitro.¹²⁰ Further in vitro characterization validated the function of PchB as an “isochorismate pyruvate-lyase” (IPL) with PchB also having the surprising capacity to act as a chorismate mutase to generate prephenate from chorismate.¹²¹ In addition to this two-component system, Abell and coworkers identified a standalone “salicylate synthase,” irp9, from Yersinia enterocolitica, which directly transformed chorismate to salicylic acid (Figure 23).¹²² Both pchA/pchB and irp9 homologs are often found co-localized with bacterial siderophore BGCs.

Figure 23.

Orthogonal salicylic acid biosynthetic pathways from bacteria (top) and plants (bottom).

In plants, radioisotope labeling revealed two operative pathways for the biosynthesis of salicylic acid: degradation of phenylalanine, and a bacterial-like isochorismate pathway.¹²³ The isochorismate-dependent pathway remained a mystery as no homologues for ipr9 nor pchB bacterial-type salicylic acid synthases could be identified in plant genomes/transcriptomes. Through a combination of genetic manipulations in Arabidopsis, untargeted metabolomics, and in vitro enzymology, Weng and coworkers¹²⁴ and Feussner and coworkers¹²⁵ independently identified a novel plant salicylic acid pathway. In short, plastid-associated isochorismate synthase, ICS1, converts chorismate to isochorismate, which is shuttled to the cytosol by transporter EDS5 where acyl adenylase PBS3 generates isochorismate-9-glutamate (64) (Figure 23). Final conversion to salicylic acid can occur spontaneously or is accelerated by BAHD-type acyltransferase EPS1. The elucidation of this plant-specific pathway may enable the development of engineered salicylic acid pathways for improved resistance against plant pathogens.

8.1.3. Menaquinone

The biologically important metabolite menaquinone (59), also known as vitamin K2, is also of note as it is biosynthesized from isochorismate by a dedicated isochorismate synthase, MenF.¹²⁶ The successive action of pathway enzymes MenD, MenH, and MenC yields o-succinylbenzoate (65), which is further modified to the 1,4-dihydroxy 2-naphthoic acid core (66) before it is ultimately prenylated to yield menaquinone. This biosynthetic pathway was thought to be the exclusive pathway to this critically important vitamin¹²⁷ until Dairi and coworkers discovered an alternative menaquinone biosynthetic pathway in microbes.¹²⁸ The alternative futalosine pathway, named after key intermediate futalosine (67), is generated directly from chorismate rather than isochorismate via chorismate dehydratase MqnA which generates 3-[(1-carboxyvinyl)oxy]-benzoic acid (68).¹²⁹ Radical SAM enzyme, MqnE, generates aminofutalosine from 68,¹³⁰ which undergoes MqnF-catalyzed deamination to generate futalosine (67).¹³¹ Futalosine then undergoes a series of steps (some of which have yet to be characterized biochemically) to yield 59. Notably, pathogenic bacteria such as Heliobacter pylori are dependent on the futalosine pathway, and the promise of narrow-spectrum antibiotic treatments via targeting this pathway has spurred identification of selective futalosine pathway inhibitors.^132–134

8.2. 3,4-trans-Dihydroxycyclohexa-1,5-dienecarboxylic acid (3,4-trans-CHD)

The diagnostic 3,4-dioxocyclohexyl ring of the powerful immunosuppressant polyketides rapamycin (known clinically as Sirolimus, 70), FK520 (known as ascomycin, 71), and FK506 (known clinically as Tacrolimus, 72)¹³⁵ is also derived from chorismate (Figure 25), although early isotope feeding experiments first suggested that it originates from a shunt in the CHC pathway. The isolation, biosynthesis, and bioactivity of these lifesaving Streptomyces macrolides have been the subject of numerous reviews,^136,135 and as such we only briefly discuss the chorismatase enzymes which transform chorismate into their characteristic starter units.

Figure 25.

Natural products derived from 3,4-trans-CHD (69) include the immunosuppressants rapamycin (70), ascomycin/FK520 (71), and FK506 (72).

Feeding experiments demonstrated that ¹³C-labeled shikimate was incorporated into the cyclohexane moiety of both rapamycin¹³⁷ and ascomycin.¹³⁸ As such, it was initially proposed that an early shunt in the CHC pathway (Section 5, Figure 13) was involved. Ultimately, 4,5-dihydroxycyclohex-1-enecarboxylic acid (DHCHC, 73) was identified as the PKS starter unit for rapamycin biosynthesis.¹³⁹ Notably, throughout the literature “DHCHC” is used to describe both the fully saturated cyclohexane as well as the unsaturated cyclohexene (73); for the purposes of this article, DHCHC will refer to the unsaturated compound 73 (Figure 26). Through gene disruption and complementation studies, rapK/fkbO (homologous genes from the rapamycin and FK520/FK506 pathways, respectively) were identified to perform key, identical roles in the biosynthesis of DHCHC,¹⁴⁰ while rapK/fkbO knockout strains played a pivotal role in the mutasynthetic production of rapamycin analogs or “rapalogs.”¹⁴¹ In vitro assays with recombinant FkbO₅₂₀ revealed this enzyme to be a founding member of a new class of chorismate-utilizing enzymes that hydrolyze pyruvate from chorismate, with FkbO₅₂₀ efficiently and selectively converting chorismate to 3,4-trans-dihydroxycyclohexa-1,5-dienecarboxylic acid (3,4-trans-CHD, 69), the first committed step towards 70 biosynthesis (Figure 26).¹⁴² Notably, the enzyme catalyzing the final reduction of DHCHC (73) to the saturated starter unit has yet to be elucidated, and given the resemblance of these intermediates to those from the CHC pathway, presents a tantalizing possibility that this reduction may proceed via a CoA intermediate.

Figure 26.

Biosynthesis rapamycin biosynthetic starter unit DHCHC (73) from chorismate. Notably, the enzyme catalyzing DHCHC reduction is yet to be reported but is presumed to be an enoyl reductase present in the fkb BGC.

Deciphering the biosynthesis of rapamycin and FK506 has enabled the targeted discovery of notable bioactive analogs of this family, with a recent example from Warp Drive Bio wherein lysine cyclodeaminase gene rapL was used as a genome mining hook and discovered seven BGCs predicted to encode the biosynthesis of novel rapamycin analogs.¹⁴³ They ultimately isolated and characterized a new natural product, WDB002 (74) that lacks the trans-CHD feature, and in early bioactivity testing suggests WDB002 is a potent inhibitor of SARS-CoV-2.¹⁴³

8.3. 3-Hydroxybenozoic acid (3-HBA)

In the course of identifying and characterizing the chorismatases Rapk/FkbO involved in rapamycin, FK506, and FK520 biosynthesis as discussed in the previous section, Leadley, Wilkinson, and coworkers noticed that the Δrapk knockout strain of S. hygroscopicus, BIOT-4010, consistently produced a rapalog, BC325, wherein the DHCHC starter unit was replaced with 3-hydroxybenzoic acid (3-HBA, 75).¹⁴² Stable isotope feeding of BIOT-4010 with [1,7-¹³C₂]shikimate demonstrated 3-HBA was of shikimate origin, and the retention of both ¹³C labels indicated the branch point must be prior to prephenate. Genome mining of S. hygroscopicus identified hyg5 as a putative chorismatase, and its function as a candidate 3-HBA synthase was bolstered by phylogenetic analysis that revealed a hyg5 homolog, bra8, in the BGC of brasilicardin A (76),^144,145 another 3-HBA containing natural product. In vitro assays with chorismate and recombinant Hyg5 and Bra8 confirmed their role as 3-HBA synthases and defined a new offshoot pathway from chorismate (Figure 27). Chorismatases capable of biosynthesizing solely 3-HBA are sometimes referred to as type II chorismatases,¹⁴⁶ in contrast with the type I family that includes FkbO.

Figure 27.

Natural products derived from 3-HBA.

The discovery of this family of 3-HBA synthases opened the door for targeted genome mining efforts to discover new 3-HBA containing natural products, with the biosynthetic gene clusters for the structurally dissimilar cuevaene A (77)¹⁴⁷ and unantimycin B (78)¹⁴⁸ both discovered via this strategy. Notably, the molecular basis for the divergent chemical reactivity displayed between the 3-HBA forming chorismatases (type II) and 3,4-trans-CHD forming FkbO-type chorismatases (type I) has been pinpointed to subtle variations at two active site residues (Gly-> Ala, Ala-> Cys) (Figure 28), demonstrating how little perturbation is required to accomplish fundamentally different chemistry via different mechanisms and yield profoundly different natural products with diverse bioactivities.¹⁴⁶

Figure 28.

Alignment of key residues from FkbO (FK506), RapK (rapamycin), Bra8 (brasilicardin), Hyg5 (Bra8 homolog), Cuv10 (cuevaene), and Nat-Hyg5 (unantimycin).

8.4. 4-Hydroxybenzoic acid (4-HBA)

4-Hydroxybenzoic acid (4-HBA, 79) is another important building block of life that has two metabolic origins. In non-yeast eukaryotes, it is biosynthesized from tyrosine via a pathway that is yet to be fully described,¹⁴⁹ while in bacteria, 4-HBA is synthesized directly from chorismate via chorismatase enzymes.¹⁵⁰ The first chorismatase biochemically verified to generate 4-HBA from chorismate was UbiC, a chorismate lyase involved in the construction of the essential redox-active ubiquinones (also known as coenzyme Q, 80).¹⁵¹ Later the characterization of specialized biosynthetic pathways led to the identification of non-UbiC chorismate-utilizing enzymes that also generate 4-HBA, which include both type-III and type-IV chorismatases.¹⁵² In this section, we highlight three prokaryotic examples leading to the xanthomonadin arylpolyenes (xanthomonadin A, 81), polybrominated phenols like pentabromopseudilin (82), the polythioamide antibiotic, closthioamide (83), as well as the basidiomycete β-lactone natural product vibralactone (84) whose biosynthesis has recently been demonstrated to originate with 4-HBA (Figure 29).

Figure 29.

Compounds derived from 4-HBA (79) include ubiquinones (80) and the halogenated natural products xanthomonadin A (81), pentabromopseudilin (82), closthioamide (83), and vibralactone (84).

8.4.1. Xanthomonadins

The xanthomonadins (81) are a series of halogenated, aryl-polyene, membrane-bound, yellow pigments produced by phytopathogenic Xanthomonas bacteria. These compounds are critical for bacterial survival as they protect the pathogen from UV damage and reactive oxygen species.¹⁵³ The biosynthesis of these compounds has been shown to be regulated by the presence of 3-HBA so a link between an early intermediate of the shikimate pathway and xanthomonadin biosynthesis was established.¹⁵⁴ The discovery of Hyg5 and Bra8 as chorismatases responsible for generating 3-HBA for use in specialized metabolism, as discussed above, prompted the examination of the xanthomonadin BGC for a similar chorismate-utilizing enzyme.

Gene disruption experiments revealed that a functioning shikimate pathway was required for xanthomonadin production, identifying that the gene xanB2 was involved in 3-HBA synthesis in xanthomonadin producing strains.¹⁵⁵ Unexpectedly, in vitro experiments with purified XanB2 and chorismate revealed two products, 3-HBA (75) and 4-HBA (79). 3- and 4-HBAs can both be incorporated into xanthomonadins, with 4-HBA generated by XanB2 also possibly supporting an alternative route to ubiquinone production.¹⁵⁶

Both structurally and mechanistically, XanB2 is an unusual bifunctional chorismatase capable of producing both 3-HBA and 4-HBA simultaneously. This reaction is proposed to proceed through an arene oxide intermediate, which in XanB2 can open unselectively, resulting in both 3- and 4-HBA products.¹⁴⁶ This ability to biosynthesize both products led to a new classification of XanB2 as a type-III chorismatase.¹⁵² In a bioinformatic study based on the active site sequence motif differences noted between type I and type II chorismatases, Grüninger et al. identified additional XanB2/type-III chorismatases and characterized a new class of type-IV chorismatases that led solely to the formation of 4-HBA.¹⁵² Biochemical validation of type-IV chorismatase enzymes confirmed this exclusive 4-HBA biosynthetic ability.

8.4.2. Pentabromopseudilin

Pentabromopseudilin (82) is a marine bacterial antibiotic notable for its high degree of bromination.¹⁵⁷ Feeding experiments by Laatsch, Floss, and coworkers indicated that the phenol ring of pentabromopseudilin was derived from a shikimate pathway intermediate, presumably via 4-HBA due to observed symmetrical isotope labeling patterns.¹⁵⁸ Using the characterized carrier-dependent pyrrole halogenase gene pltA from pyoluteorin biosynthesis as a query for genome mining,¹⁵⁹ Moore and coworkers were able to identify a candidate pentabromopseudilin BGC, bmp, from Pseudoalteromonas bacteria that notably contained a putative chorismate lyase, bmp6.¹⁶⁰ Extensive in vitro enzymology demonstrated 4-HBA generated from Bmp6 was converted to 2,4-dibromophenol (85) by the single-component, flavin-dependent decarboxylase-halogenase, Bmp5, via 3-bromo-4-hydroxybenzoic acid as an isolable, on pathway intermediate. Finally, 85 was oxidatively coupled by cytochrome P450, Bmp7, to either proline-derived tribromopyrrole^161,162 to give pentabromopseudilin (82), or a second molecule of 2,4-dibromophenol to give polybrominated diphenylethers, biphenyls, and dioxins (Figure 31).^160,163 Smaller bmp gene clusters were subsequently identified in Hormoscilla spongeliae cyanobacterial symbionts associated with marine tropical Dysidea sponges that contain high levels of polybrominated diphenyl ether molecules such as 2OH-BDE68 in excess of 10% of the sponge tissue by dry weight.^164,165 Functional characterization of these prokaryotic BGCs revealed that they were indeed responsible for production of sponge animal-associated polybrominated diphenyl ethers that proceeded through the same shikimate shunt.

Figure 31.

Polybrominated aryl natural products are biosynthesized from chorismate via 4-HBA (79) by the bmp biosynthetic pathway in Pseudoalteromonas marine bacteria.^160–163

8.4.3. Closthioamide

Closthioamide (83) is a thioamide-containing antibiotic that is notably biosynthesized by an anaerobic bacterium, Ruminiclostridium cellulolyticum (formerly Clostridium cellulolyticum DSM 5812).^166,167 Few biosynthetic pathways from anaerobic organisms have been elucidated, and 83 is a notable exception. 4-HBA was a presumed building block for 83, and using a genome mining approach with UbiC as a search hook, Hertweck and coworkers were able to identify a candidate gene, ctaA, with significant homology to chorismate lyase enzymes.¹⁶⁸ Inactivation of this gene abolished production of 4-HBA and 83, with production able to be complemented by the addition of 4-HBA to the cultured mutants. This work not only confirmed the role of 4-HBA in 83 biosynthesis and validated CtaA as the enzyme responsible for diverting chorismate towards antibiotic production, but also identified an unusual thiotemplated but NRPS-independent biosynthetic route to 83 production. Further biochemical work demonstrated 4-HBA is the starter unit for 83 biosynthesis, which is loaded onto carrier protein, CtaH, by acetyl-CoA synthetase-like enzyme (Figure 32).¹⁶⁹ CtaG catalyzes the transacylation of the thiotemplated 4-HBA (86) onto to the N-terminus of three amide-linked units of β-alanine thiotemplated to carrier protein CtaE (87) to give benzamide, 88. Iron-sulfur cluster enzyme CtaC then transforms amide bonds of 88 to thioamides resulting in 89.¹⁷⁰ 89 then acts as a branching intermediate wherein transglutaminase CtaJ cleaves 89 to give covalent intermediate 90 and a single unit of β-alanine still tethered to CtaE. CtaK catalyzes a reductive offloading to yield aldehyde 91. The second nitrogen in the central diamine linker is installed via transaminase, CtaB, and the final thioamide is forged from resulting amine 92 and 90, completing the “split-merge” biosynthesis of closthioamide (83).¹⁷¹

Figure 32.

“Split-merge” biosynthetic assembly of closthioamide (83) from 4-HBA (79).

8.4.4. Vibralactone

Vibralactone (83) is a rare bicyclic fused β-lactone terpenoid produced by the mushroom Boreostereum vibrans.¹⁷² The β-lactone warhead of this molecule serves as the covalent attachment site for a catalytic serine residue in the human pancreatic lipase, thereby inhibiting fat absorption.^{172 13}C-labeling studies surprisingly suggested a shikimate origin for the bicyclic skeleton of vibralactone, despite its non-aromatic structure. The labeling patterns revealed that both phenylalanine and an earlier shikimate pathway compound may contribute to this structure, and so the authors hypothesized 4-HBA as a possible intermediate.¹⁷³

In examining the biosynthetic diversity of 4-HBA derived metabolites from B. vibrans, Zeng, Liu and coworkers then identified an FAD-dependent monooxygenase, VibMO1, that catalyzes the conversion of prenyl-4-HBA (93) into structurally unrelated vibralactones, intimating a shikimate derived 4-HBA metabolite as a key intermediate in all vibralactone production.¹⁷⁴ The biosynthesis of the β-lactam vibralactones were deciphered using activity guided fractionation coupled with proteomics, with Zheng, Liu, Huang and coworkers first identifying and characterizing VibC, an α/β hydrolase family enzyme which catalyzes the final biosynthetic step, formation of the vibralactone bicyclic warhead from 1,5-seco-vibralactone (94), with the crystal structure of VibC lending molecular detail into hypothesized cyclization mechanisms.¹⁷⁵ This strategy enabled the recent discovery of the full suite of enzymes catalyzing vibralactone biosynthesis from 4-HBA by Zheng, Pan and coworkers (Figure 33), with the pathway fully reconstituted both in vivo and in vitro and a crystal structure of flavin-dependent oxidase, VibO.¹⁷⁶ However, the molecular origin of 4-HBA and genetic organization of these enzymes have both yet to be fully described.

Figure 33.

Vibralactone biosynthesis as characterized through activity guided fractionation.

8.5. 4-Amino-4-deoxychorismic acid (ADC)

Chorismate also undergoes transformations to introduce amino groups to the cyclohexyl ring, the first of which reported here is in the conversion to 4-amino-4-deoxychorismic acid (ADC) by ADC synthase. Arguably the most universal metabolite derived from ADC (95) is para-aminobenzoic acid (PABA, 96) (Figure 34), a precursor for critical folate cofactors as well as natural products.^177,178 ADC can also be converted into 4-amino-4-deoxyprephenate (97), which is a precursor to the important antibiotics stravidin S2 (98)^179–182 and chloramphenicol (99)^183,184 as highlighted below.

Figure 34.

Natural products derived from chorismate via ADC (95), including para-aminobenzoic acid (PABA, 96), stravidin S2 (98), and chloramphenicol (99).

8.5.1. para-Aminobenzoic acid

The conversion of chorismic acid to ADC was originally characterized in folic acid biosynthesis in E. coli, where two proteins, PabA and PabB, form a heterodimeric complex to convert chorismic acid and glutamine to ADC and glutamate.^185–187 ADC can then be subsequently transformed by a lyase (PabC) to yield PABA (Figure 35). In fungi, some bacteria, Plasmodium, and plants, ADC is formed by fused protein domains homologous to PabA and PabB, referred to as PABA synthases, despite the fact that an additional lyase protein is still required for the ADC to PABA conversion.^188,189

Figure 35.

The conversion of chorismate to ADC is the branch point for PABA (96) and 4-amino-4-deoxyprephenate (97) biosynthesis. 97 further diverges through either aromatic or non-aromatic decarboxylation reactions to chloramphenicol (99) and stravidin S2, respectively (98).

PABA is also broadly incorporated into natural product biosynthetic pathways. It is used as an assembly line primer such as in candicidin^190,191, and assembly line extender as in albicidin (100).¹⁹² In the case of aureothin (101),¹⁹³ the amine of PABA is oxidized to a nitro group prior to initiating polyketide assembly, while in xanthomonic acid (102),¹⁹⁴ PABA serves as a prenylation substrate.

8.5.2. 4-Amino-4-deoxyprephenic acid

The stravidins are a family of antibiotic dipeptides produced by Streptomyces avidinii and are known inhibitors of biotin biosynthesis. These small molecules are isolated along with the biotin binding protein streptavidin, resulting in a potent synergistic cocktail that causes biotin deficiency in microbes.^180,181 Stravidins, such as stravidin S2 (98), contain the nonproteinogenic amino acid amiclenomycin (Acm) that selectively binds and inhibits mycobacterial biotin biosynthesis protein, BioA, by forming an irreversible, covalent adduct with BioA’s PLP cofactor. The characteristic amino-substituted 1,4-cyclohexadiene moiety of Acm is critical for this activity, however the molecular origin of this amino acid, as well as the biosynthesis of stravidins, remained unknown for decades. Kelleher and Montaser were able to identify a candidate stravidin biosynthetic gene cluster¹⁷⁹ through the reported “metabologenomics” platform,¹⁸² wherein related BGCs from across a wide range of bacteria are grouped into gene cluster families that are then scored against LC-MS data from those same strains to determine the likelihood of a particular metabolite originating from a gene cluster family. Heterologous expression of the putative stravidin (svn) BGC validated this assignment. Like other BGCs that utilize shikimate pathway intermediates, the svn BGC contains a dedicated DAHP synthase to ensure ample precursor production and avoid negative feedback regulation by aromatic amino acids, suggesting a shikimate origin for the Acm amino acid. Based on the gene cluster organization, Acm (and therefore stravidin) biosynthesis was proposed to diverge from the shikimate pathway at the point of chorismate via ADC (95), as precedented in the biosynthesis of chloramphenicol (99) (Figure 35).^183,184

In chloramphenicol biosynthesis, ADC is converted by a mutase (CmlD) to 4-amino-4-deoxyprephenate (97), followed by aromatic decarboxylation by CmlC towards mature chloramphenicol synthesis.¹⁸³ The stravidin gene cluster contains both an ADC synthase homolog (svnN) as well as a ADC mutase homolog (svnK), however, Acm retains the 1,4-cyclohexadiene feature of the prephenate ring, suggesting a novel nonaromatic decarboxylation step as also seen in prephenate derived natural products (Section 9). The enzyme responsible for this reaction is proposed, but not yet biochemically validated, leaving the door open for future mechanistic evaluation.

8.6. 2-Amino-2-deoxyisochorismic acid (ADIC)

A second amino-containing chorismate branch product is 2-amino-2-deoxyisochorismic acid (ADIC, 103). This versatile intermediate is transformed into a large diversity of specialized metabolites including benzoxazolinates, phenazines, and benzodiazepines (Figure 36). There are notable similarities between ADIC synthase and anthranilate synthase, which is discussed separately in Section 10. In both cases, chorismate undergoes an amine transfer at C-2 resulting in the elimination of the C-4 hydroxyl to give ADIC. In anthranilate synthase, a second irreversible reaction takes places to eliminate pyruvate to give anthranilate en route to tryptophan biosynthesis. Notably, a mutant form of anthranilate synthase from Salmonella typhimurium was able to decouple these reactions and accumulate ADIC, revealing the enzymatic similarities.¹⁹⁵ As such, many ADIC-derived specialized products were once thought to derive from anthranilate given their related chemistry and protein sequence.

Figure 36.

Natural products derived from chorismate via ADIC (103), with ADIC-derived portions of each molecule highlighted in red.

8.6.1. Benzoxazolinates

ADIC synthase was first biochemically validated by Shen and coworkers in the biosynthesis of the potent antitumor antibiotic, C-1027 (104).¹⁹⁶ C-1027 is constructed of four structurally distinct biosynthetic units, namely the reactive enediyne pharmacophore, a β-amino acid, a deoxy aminosugar, and notably the shikimate-derived benzoxazolinate moiety (105). The biosynthesis of enediyne natural products has been reviewed extensively,^197–199 and while to date there is no comprehensive understanding of their assembly, the molecular and genetic origin of the benzoxazolinate moiety has been characterized. In short, the benzoxazolinate was hypothesized to arise from anthranilate due to the presence of two anthranilate synthase homologs, SgcD and SgcG, encoded in the C-1027 BGC.²⁰⁰ Notably, neither enzyme generated anthranilate from chorismate. Rather SgcD was revealed to exclusively generate ADIC with no pyruvate lyase activity, a previously unknown biochemical transformation.¹⁹⁶ SgcG was then found to act directly after SgcD as an “ADIC dehydrogenase” generating 3-enolpyruvoylanthranilate (OPA, 106) (Figure 37).

Figure 37.

Biosynthesis of benzoxazinolate moiety (105) from chorismate (12).

In a recent global analysis of BGCs from nematode-associated symbiotic bacteria, Bode and coworkers identified and characterized an orphan BGC, xvb, from Xenorhabdus vietnamensis due its unusual combination of tailoring enzymes and an adenylase with unpredictable amino acid specificity.²⁰¹ Insertion of a promoter upstream of the xvb BGC enabled the isolation of the sole example of a non-Streptomyces derived benzoxazinolate natural product, benzobactin A (107). Follow up work capitalizing on ScgD/SgcG gene probes identified additional benzoxazinolate pathways, including xsb from X. szentirrmaii DSM 16338 that yields benzoxazinolate itself (105). In vitro biochemistry established the function of the CoA ligase XsbC as the previously unidentified “cyclase” catalyzing ATP-dependent intramolecular cyclization of OPA to benzoxazinolate (Figure 37). With the activity of XsbC identified, it was then used as a bioinformatic hook to demonstrate benzoxazinolate BGCs are widespread beyond actinomycete bacteria and are in fact prevalent in Proteobacteria and Firmicutes.

8.6.2. Phenazines

While SgcD from the C-1027 biosynthetic pathway was the first reported ADIC synthase enzyme, ADIC was originally identified as an intermediate in the biosynthesis of phenazines.^202,203 Phenazines are a large class of heterocyclic, redox-active secondary metabolites produced by both Gram-positive and Gram-negative bacteria, most commonly associated with Pseudomonas.^204–206 The myriad ecological roles, impacts on human health, and the structural diversity of the phenazine class of compounds drove investigations into their molecular origin, and the biosynthesis of the phenazine core has been the subject of thorough reviews.^207–210 Briefly, a conserved set of five genes, phzBDEFG, make up the phz operon (Figure 38), starting with ADIC synthase, PhzE, converting chorismate to ADIC (103). ADIC is converted to trans-2,3-dihydro-3-hydroxyanthranilic acid (DHHA, 108) by isochorismatase homolog, PhzD, and isomerase PhzF catalyzes the conversion of DHHA to 6-amino-5-oxocyclohex-2-ene-1-carboxylic acid (AOCHC, 109).²¹¹ Two AOCHC molecules can spontaneously condense to yield the tricyclic phenazine core, but multiple studies have shown this reaction is greatly facilitated by PhzA/B.^212,213 The resulting intermediate, hexahydrophenazine-1,6-dicarboxylic acid (HHPDC, 110), is the branch point for phenazine biosynthesis and can undergo spontaneous decarboxylation prior to oxidation by PhzG to generate phenazine-1-carboyxlic acid (PCA, 111), or PhzG can directly oxidize 110 to generate the phenazine-1,6-dicarboyxlic acid (PDC, 112).²¹⁴

Figure 38.

The biosynthesis of phenazine PDC (112) and PCA (111) cores from chorismate (12).

The biosynthetic route for formation of the phenazine PDC and PCA cores from chorismate is highly conserved in Streptomyces bacteria, where the phenazine molecular family is diversified via enzymatic tailoring, particularly enzymes incorporating polyketide²¹⁵ or terpene structural features.²¹⁶ Early isotope experimental work with the antimicrobial saphenic acid (113) and antitumor esmeraldin B (114) revealed the biosynthetic connection between the shikimate pathway and phenazine biosynthesis in these colorful molecules.²¹⁷ Characterization of the plasmid-born esmeraldin BGC from Streptomyces antibioticus Tü 2706 established that the one-carbon extension of the PDC core of the saphenic acid monomer of esmeraldin was derived from a PKS extension with malonyl-CoA followed by a decarboxylation reaction.²¹⁵

A more elaborate modification takes place in the biosynthesis of the heavily tailored streptophanzines (streptophenazine R, 115). In exploring the metabolic potential of marine actinomycete Streptomyces sp. CNB-091, Moore and coworkers detected a suite of streptophenazine-like compounds produced intermittently and in very low titer. Targeted mining of the CNB-091 genome revealed a candidate BGC, spz, which contained a phenazine operon colocalized with both PKS and NRPS machinery. To characterize how this “cluster of clusters” functioned together to assemble the streptophenazines, the 49.5 kb spz BGC was captured using transformation-associated recombination cloning and transformed and expressed in heterologous host S. coelicolor M1146. Initial transformants failed to reconstitute production and drove efforts for targeted refactoring of the spz BGC. Fermentation of M1146 with refactored spz resulted in the production of >100 streptophenazine analogs at dramatically increased titer relative to the CNB-091. Gene deletions of this refactored pathway enabled a proposed biosynthetic route to the streptophenazines (Figure 39), as well as the identification of phylogenetically distinct adenylation domain, spz15, as involved in the installation of the rare N-formylglycine unit.

Figure 39.

Proposed streptophenazine biosynthetic pathway from Streptomyces sp. CNB-091 with representative streptophenazines depicted for the structurally diverse product types.

8.6.3. Bacterial benzodiazepines

Benzodiazepines are a structurally diverse class of molecules minimally defined by a benzene ring fused to a seven-membered diazepine ring. Diazepinomicin (116) is a potent antibiotic benzodiazepine isolated from terrestrial and marine strains of Micromonospora bacteria and is comprised of a tricyclic dibenzodiazepinone ring system wherein the substitution pattern of both benzene rings intimates a shikimate origin.²¹⁸ Stable isotope feeding experiments identified 3-hydroxyanthranilic acid (3-HA, 117) as a precursor for the dibenzodiazepinone core (Figure 40). Genomic analysis revealed a candidate diazepinomicin gene cluster which contained isoprenoid biosynthesis genes colocalized with shikimate pathway genes.²¹⁹ ORF19 was of particular interest due to its homology to phenazine ADIC synthase, phzE. Two additional genes in the diazepinomicin BGC were predicted to catalyze the downstream conversion of ADIC to 3-hydroxyanthranilic acid via pyruvate lyase and aromatization reactions. An adenylate ligase, ORF24, was also proposed to adenylate 3-hydroxanthranilic acid for a subsequent construction of the dibenzodiazepinone core. This putative pathway was validated biochemically by Najmanova and coworkers in their work characterizing the biosynthesis of related bacterial benzodiazapines, the limazepines.²²⁰ The identified limazepine BGC also contained a gene coding for a PhzE homolog, lim6, again suggestive of a biosynthetic route proceeding through to 3-hydroxyanthranilic acid via ADIC.²²⁰ In vitro assays with Lim6 confirmed its role as an ADIC synthase, as well as identified the activity of downstream enzymes, Lim5 and Lim4, in converting ADIC to 3-hydroxyanthranilic acid, which is then elaborated into the pyrrolobenzodiazepine core limazepine C (118) via an NRPS (Figure 39). Notably, while the molecular origin of anthranilate building blocks in bacterial benzodiazepine biosynthesis is often ADIC, fungi utilize an alternative route to pyrrolobenzodiazepines as discussed in Section 10.

Figure 40.

Biosynthesis of 3-hydroxyanthranilate (3-HA, 117) from ADIC (103) in actinobacteria to the benzodiazepine natural products, diazepinomicin (116) and limazepine C (117).

9. Prephenic acid

The conversion of chorismate (12) to prephenate (13) by chorismate mutase (CM) is the first committed step towards phenylalanine and tyrosine biosynthesis (Figure 41).²²¹ Chorismate mutase catalyzes a [3,3]-sigmatropic rearrangement of chorismate, the first reported example of a naturally occurring pericyclic reaction.^222,223 Complex allosteric regulation of chorismate mutase by aromatic amino acids controls the flux of chorismate towards either dedicated phenylalanine and tyrosine biosynthesis via prephenate, or tryptophan biosynthesis via anthranilate. The metabolites and pathways discussed in this section are all bacterial and utilize prephenate as a substrate to generate non-Phe or Tyr based products that in turn get incorporated into structurally diverse natural products.

Figure 41.

Sigmatropic rearrangement of chorismate to prephenate by chorismate mutase and prephenate derived natural products.

9.1. Dihydro-4-hydroxyphenylpyruvate (H₂HPP)

Salinosporamide A (119) is an unusual hybrid polyketide/non-ribosomal peptide natural product originally isolated from the obligate marine actinomycete Salinispora tropica (Figure 41).²²⁴ Salinosporamide A acts as a selective and irreversible covalent inhibitor of the β-subunit of the 20S proteasome, and blocks a cell’s ability to control proteolysis.²²⁵ This activity propelled salinosporamide A into oncology clinical trials in 2006, just three years after its initial discovery²²⁶. Notably, despite the discovery of numerous natural analogs of salinosporamide A, as well as synthetic and mutasynthetic derivatives,²²⁷ it is the original natural product itself that has advanced through clinical trials, entering Phase III trials for the treatment of glioblastoma in late 2017.²²⁸

Work in the Moore group has focused on deciphering salinosporamide biosynthesis, beginning with stable isotope feeding experiments that revealed salinosporamide is comprised of three distinct building blocks that correlate to the side chains surrounding the biscyclized core.²²⁹ Assimilation of [U-¹³C₆]glucose demonstrated the cyclohexenyl-containing amino acid, H₄Phe (120), originated from shikimate via a novel shunt in aromatic amino acid biosynthesis prior to aromatization. Subsequent identification of the salinosporamide BGC^230,231 and gene disruption experiments established the key role of salX, a “prephenate dehydratase homolog” in the formation the “cyclohexenylalanine" amino acid and enabled the mutasynthetic production of non-natural salinosporamides.^232,233

Concomitant to work in the Moore lab, Walsh and coworkers investigated the biosynthesis of the dipeptide antibiotic, bacilysin (121), produced by Bacillus subtilis and recognized by its signature epoxycyclohexanone amino acid, anticapsin (122).²³⁴ Feeding experiments with radioisotope labeled precursors back in 1966 demonstrated bacilysin was derived from an intermediate of the shikimate pathway prior to phenylalanine,²³⁵ with subsequent gene knockouts identifying prephenate as the likely intermediate.²³⁶ Biochemical characterization of enzymes in the bacilysin gene cluster revealed another prephenate dehydratase homolog, BacA, as capable of performing a decarboxylation and protonation to yield the non-aromatic ketoacid 123.²³⁴

Collaboratively, the Moore and Walsh groups demonstrated much like BacA, SalX also utilizes prephenate as a substrate and performs a non-aromatizing decarboxylation, converting prephenate to endocyclic dihydro-4-hydroxyphenylpyruvate (H₂HPPen, 123, Figure 42).²³⁷ H₂HPP can undergo spontaneous, or in the case of bacilysin enzyme catalyzed, isomerization conversion to exocyclic H₂HPP (H₂HPPex, 124) en route to cyclohexenylalanine formation. Additionally, a third enzyme, AerD, was identified in the cyanobacterial BGC responsible for aeruginoside 126A (125), and in vitro characterization of AerD validated its prephenate decarboxylase activity to generate H₂HPP en route to the nonproteinogenic amino acid residue “choi” (126), which is found in a number of other cyanobacterial peptides.²³⁸

Figure 42.

Biosynthesis of H₂HPPen and H₂PHPPex en route to salinosporamide A (119), bacilysin (121), and aeruginoside 126A (125).

Identification of the “prephenate decarboxylases” has enabled continuing systematic evaluation of these biosynthetic pathways,²³⁹ however the final steps detailing the tailoring of non-aromatic intermediates into their fully elaborated products in bacilysin, salinosporamide, and aeruginosin biosynthesis have yet to be characterized.^240,241 Additionally, the general activity for these non-aromatizing prephenate decarboxylases can be seen in non-prephenate utilizing biosynthetic gene clusters such as the stravidin BGC seen in Section 8 as well as the recently reported BGC for radiosumin (127), wherein a putative pathway includes a non-aromatizing decarboxylation of an aminoprephenate.^222,223

9.2. Dihydrophenylalanine (H₂Phe)

The identification of non-aromatizing prephenate decarboxylases BacA, SalX, and AerD enabled a genome mining approach to search for orphan prephenate decarboxylases from other organisms. Walsh and coworkers targeted the insect pathogen, Photorhabdus luminescens, which they hypothesized might utilize a similar prephenate decarboxylation reaction in the biosynthesis of dihydrophenylalanine (H₂Phe, 128). This non-proteinogenic amino acid was a suspected intermediate en route to the antibiotic, 2,5-dihydrostilbene (129, Figure 43)^242,243 based in part on precursor feeding experiments that saw no incorporation of isotopically labeled phenylalanine.²⁴⁴

Figure 43.

H₂Tyr (128) pathway identified and characterized from P. luminescens genome mining campaign and proposed biosynthetic pathway to dehydrostillbene (129).

Using the newly established prephenate decarboxylases as a search query, an eight gene cluster was identified and confirmed through genetic knockout experiments to be responsible for H₂Phe production.¹⁶⁵ In vitro biochemical experiments verified Plu3043 as an additional prephenate decarboxylase capable of converting prephenate to non-aromatized H₂HPP. Aminotransferase Plu3042, also encoded in the cluster, catalyzes the transamination of H₂HPPen to generate H₂Tyr (130) using Gln as the amine donor. Feeding experiments with P. luminescens mutants validated that isotopically labeled H₂Phe as an on pathway intermediate to dehydrostilbene, however enzymes catalyzing the final conversion to H₂Phe have yet to be described (Figure 43).

10. Anthranilic acid

Chorismate (12) also directs the biosynthesis of tryptophan and indole-based metabolites via the dedicated intermediate anthranilate (14) (Figure 44). Anthranilate is a key building block to diverse specialized metabolites. Anthranilate synthase is composed of two subunits that assemble into a variety of organism-dependent oligomeric states and can be found in multi-subunit complexes with other tryptophan biosynthetic enzymes.²⁴⁵ The α-subunit catalyzes the aminotransfer from either glutamine or ammonia onto chorismate yielding ADIC (103), followed by the irreversible elimination of pyruvate by the β-subunit to give anthranilate (14).^246,247 While the direct use of anthranilate as a building block in metabolites is often associated with fungal alkaloids,²⁴⁸ bacteria and plants have also evolved specialized pathways that assimilate anthranilate. In this section we highlight recent examples of anthranilate-derived natural products across taxa.

Figure 44.

Aminotransfer and pyruvate lyase reaction catalyzed by anthranilate synthase, and representatives of each class of molecule discussed in this section.

10.1. Fungal anthranilate-containing alkaloids

Filamentous fungi are prolific producers of anthranilate containing alkaloids, such as fumiquinazoline A (131), acetylaszonalenin (132), and asperlicin (133).²⁴⁸ These compounds have been isolated from both marine and terrestrial environments and possess broad ranging bioactivities including cytotoxicity, toxicity, and cholecystokinin antagonism. Anthranilate alkaloid BGCs cover significant biosynthetic diversity, with anthranilate integrated into a variety of fused, heterocyclic “privileged scaffolds” (benzodiazapenes, quinazolines, pyrroloindolines) which are also present in many clinically relevant compounds.^249,250

Houck and coworkers at Merck originally observed the direct incorporation of anthranilate into fungal natural products via radioisotope feeding experiments.^251,252 The discovery and characterization of acetylaszonalenin (132) by Walsh and coworkers later revealed the earliest enzyme utilizing anthranilate as a secondary metabolite building block, with the NRPS AnaPS catalyzing the condensation of anthranilate and tryptophan to yield the benzodiazepinedione core (Figure 45).²⁵³

Figure 45.

Abbreviated biosyntheses of acetylaszonalenin (132) and nanangelenin A (134) highlighting their orthogonal assembly line cyclization modes.

In vitro characterization of AnaPS revealed the assembly line module that selectively activated, loaded, and elongated anthranilate, culminating in Ames and Walsh identifying a 10-residue sequence in the fungal adenylation domain that conferred specificity for anthranilate.²⁵⁴ The predictive power of this specificity code was further demonstrated through genome mining and heterologous expression of two previously unidentified anthranilate activating NRPS modules from Aspergillus fumigatus and Neosartorya fischeri. In all cases examined in this work, anthranilate served as the starter unit for assembly via a multimodular NRPS.^248,255

A recently characterized anthranilate activating NRPS was identified in the BGC for nanangelenin A (134), a 1-benzazepine containing natural product produced by the fungus Aspergillus nanangensis.²⁵⁶ The 1-benzazepine core was predicted to arise from condensation of anthranilate and the nonproteinogenic amino acid l-kynurenine, and based on this prediction, the presence of an anthranilate-activating AnaPS NRPS homolog helped identify the nanangelenin BGC. Co-expression of the NRPS nanA with the indolamine-2,3-dioxygenase nanC, which catalyzes the conversion of tryptophan to kynurenine, resulted in the production of the anthranilate-kynurenine cyclized dipeptide (Figure 45). This rare 1-benzazepindione moiety was ultimately revealed to be formed by a regioselective lactamization catalyzed by a terminal condensation domain, which displays orthogonal selectivity compared to previously characterized cyclization domains and uses the kynurenine aniline nitrogen as the cyclizing nucleophile instead of the anthranilate nitrogen. Following assembly of the benzazepindione core to yield nanangelenin B (135), the anthranilate nitrogen can then be subjected to a suite of tailoring enzymes, with NanD catalyzing prenyltransfer to give nanangelenin C (136), oxidation to hydroxylamine by flavin mono-oxygenase NanF, and O-acylation by NanB to give nanangelenin F (137). The pathway terminates in a methylation of the kynurenine nitrogen by NanE to give nanangelenin A (134).

10.2. Quinoline/Quinolone Meroterpenoids

Aurachin A (138) inhibits mitochondrial respiration and is a member of a family of quinoline alkaloids produced by the myxobacterium Stigmatella aurantiaca (Figure 45).²⁵⁷ Initial feeding experiments by Müller and coworkers with ¹³C and ¹⁸O-labeled anthranilate increased aurachin production, demonstrating the role of this compound as an early biosynthetic precursor. A putative BGC for aurachin production was identified using a transposon mutagenesis strategy, revealing a five-gene cluster encoding a type II PKS, the first such PKS identified from a Gram-negative bacterium at the time.²⁵⁸

However, none of the candidate genes appeared capable of activating anthranilate, the presumed first step in aurachin biosynthesis.²⁵⁹ A BLAST search using the benzoate-CoA ligase, auaE, from the aurachin BGC as a hook revealed a homologous gene located ~8 kb upstream of the gene cluster. Targeted disruption of this gene, auaEII, completely abolished aurachin production. Both AuaE and AuaEII were heterologously expressed for in vitro biochemical assays to evaluate their respective biosynthetic roles.²⁵⁹ This work demonstrated that they function in series; AuaEII activates anthranilate as an anthranilate-CoA ligase, proceeding through an AMP-activated intermediate, while AuaE is responsible for transferring anthranilate onto the AuaB PCP domain (Figure 46). The activation of anthranilate in aurachin biosynthesis thus proceeds via an unusual priming mechanism, involving two ligase enzymes encoded in different genomic regions.

Figure 46.

The biosynthesis of aurachin A (138) requires two activating enzymes to load anthranilate onto the AuaB-PCP domain.

Recently, a series of structurally related prenylated quinoline natural products, the marinoterpins (marinoterpin A, 139), were isolated from a marine actinomycete.²⁶⁰ By using the aua cluster as a genome mining hook, Fenical and coworkers identified a putative BGC with a similar CoA ligase, mrtE. Notably, the putative marinoterpin BGC contains only a single ligase, rather than the pair required for aurachin biosynthesis. The identification of this BGC implies that the activation of anthranilate as a starter unit for bacterial PKS biosynthesis may be more widely distributed and underrepresented from bacteria.

10.3. Pyrrolobenzodiazepines

Pyrrolobenzodiazepines (PBDs) are tricyclic natural products produced by both terrestrial and marine microorganisms. These molecules bind the minor groove of DNA, forming crosslinks that result in cell death and therefore potent antitumor activity (Figure 47).²⁶¹ However, despite this powerful bioactivity, the cardiotoxicity of this class of compounds has precluded their clinical development. The founding member of this class of natural products is anthramycin (140), isolated from the thermophilic Streptomyces refuineus subsp. thermotolerans,²⁶² and named for its characteristic anthranilate moiety. Additional members of this class include sibiromycin (141) and tomaymycin (142).

Figure 47.

Orthogonal routes to anthranilic acid incorporation in pyrrolobenzodiazepine natural products. Sibiromycin (141) derives its anthranilate unit via degradation of tryptophan (3), while tomaymycin (142) directly uses chorismate.

Isotope feeding studies demonstrated 3-hydroxyanthranilic acid (143) derived from tryptophan is incorporated into the anthranilate moiety of sibiromycin and anthramycin,²⁶³ and subsequent identification of the gene clusters responsible for sibiromycin and anthramycin biosynthesis revealed homologs of the kynurenine tryptophan degradation pathway enzymes (sibP, sibK, sibC, sibO), further implicating this metabolic route for anthranilate production and incorporation.^264,265 However, feeding experiments with labeled anthranilate showed incorporation in tomaymycin, and given the lack of substitution at C9 compared to sibiromycin, Hurley and coworkers proposed an alternative biosynthetic origin of the anthranilate moiety in tomaymycin.²⁶⁶ Analysis of the tomaymycin BGC by Gerratana and coworkers revealed a dedicated DAHP synthase and two anthranilate synthetase homologs (tomD, tomP), and no genes related to tryptophan catabolism.²⁶⁷ They proposed that the anthranilate (14) unit is directly hydroxylated and methylated by colocalized flavin mono-oxygenases (tomO, tomE/F) and a methyltransferase (tomG) prior to loading onto the NRPS assembly line enzymes TomA and TomB that catalyze diazepine ring formation. This work demonstrated that the pyrrolobenzodiazepine family of natural products has developed two distinct strategies to access anthranilate analogs; sibiromycin and anthramycin utilize tryptophan as an end point of the shikimate pathway, while tomaymycin exploits the on-pathway intermediate anthranilate.

10.4. Tasikamides

Tasikamide A (144) belongs to a series of cyclic peptides produced by the fresh water sediment-dwelling Streptomyces tasikensis P46 (Figure 48).²⁶⁸ The tasikamides are notable for a hydrazone feature²⁶⁹ that connects an alkyl 5-hydroxyanthranilate (AHA) fragment to a cyclic pentapeptide. Whole genome sequencing and subsequent genome mining enabled Ma and coworkers to identify a biosynthetic gene cluster, tsk, which contained a five module NRPS whose A-domain predicted amino acids matched the five amino acids present in the tasikamide core, as well as additional tailoring genes in line with those required for functionalization.²⁶⁸ While gene disruption experiments confirmed this cluster was involved in tasikamide construction, enzymes catalyzing the biosynthesis of the AHA portion of the molecule remained unaccounted for. Gene clusters distal to the tsk cluster were examined and prioritized based on the hypothesis that the AHA portion was derived from the shikimate pathway or tryptophan catabolism. This analysis resulted in the discovery of a 25 kb BGC, aha, that contained genes encoding a putative anthranilate synthase, a thioredoxin, and a flavin-dependent monooxygenase, all presumably involved in converting chorismate to 5-hydroxylanthranilate (Figure 48). The smaller aha cluster also contained three genes that shared high sequence similarity to genes from the cremeomycin biosynthetic pathway (aha1, aha2, aha11) responsible for the production of nitrite and subsequent enzymatic diazotization.²⁷⁰ In vitro assays with alkyl 5-hydroxylanthranilate resulted in the formation of diazo-AHA (145), which underwent a non-enzymatic Japp-Klingemann reaction with the peptidic aldehyde core (146), coupling the cyclic peptide and diazo-AHA to generate the mature tasikamides. The tasikamides harbor a wealth of structurally intriguing features, and discovery of their biosynthetic pathway revealed a novel β-amino acid, two separate BGCs that act in concert, and a new biosynthetic route to C-N bond formation.

Figure 48.

The biosynthesis of tasikamide A (144) A is encoded by two disparate biosynthetic gene clusters to the AHA moiety (145) and tasikamide core (146).

10.5. Acridone alkaloids and avenanthramides

The acridone (147) alkaloids are a large family of natural products often found in the Rutaceae family of flowering plants, particularly in citrus. Compounds like normelicopidine (148) and rutacridone (149) are characterized by a shared tricyclic 9-acridone core.²⁷¹ An early biosynthetic hypothesis proposed these compounds arise from the condensation of anthranilate with three acetate units, which was confirmed by Gröger and Johne via radioactive precursor feeding experiments wherein anthranilate was incorporated into only one half of the molecule.²⁷² In vitro biochemical experiments by Matern and coworkers using a recombinant anthranilate N-methyltransferase from Ruta graveolens demonstrated that anthranilate methylation is the first committed step towards acridone biosynthesis.^273,274 In 2019, acridone biosynthesis was fully reconstituted in E. coli by Ahn and coworkers.²⁷⁴

Recently, hydroxylated anthranilates have been validated as a precursors in the biosynthesis of the acyclic avenanthramides (avenanthramide A, 150), antioxidant natural products produced by oat plants characterized by anthranilate derivatives coupled to cinnamates via an amide bond. The biosynthetic pathway for avenanthramide biosynthesis has been characterized in detail (Figure 49). The key reaction is catalyzed by hydroxycinnamoyl-CoA:hydroxyanthranilate N-hydroxycinnamoyl transferase (HHT) that couples hydroxylated anthranilates to CoA-activated cinnamates.^275,276

Figure 49.

Biosynthesis of avenanthramide A (150) in oat, PAL = phenylalanine lyase, C4H = cinnamic acid 4-hydroxylase, 4CL = 4-coumarate-CoA ligase, HHT = hydroxycinnamoyl-CoA:hydroxyanthranilate N-hydroxycinnamoyl transferase.

11. Aminoshikimic acid (ASA) and 3-amino-5-hydroxybenzoic acid (AHBA)

Aminoshikimate (ASA, 151) and 3-amino-5-hydroxybenzoic acid (3,5-AHBA, 152) were once thought to originate from the shikimate pathway,²⁷⁷ however seminal work by the Floss and Frost groups established the alternative, but parallel, aminoshikimate (ASA) pathway. While its discovery and characterization has been the subject of previous reviews,^278,279,280 this section will briefly discuss key biosyntheses and enzymes known to date involved in the biosynthesis of ASA and 3,5-AHBA in the context of the archetype ansamycin molecule, the antibiotic rifamycin, from which many of the first biosynthetic enzymes were characterized (Figure 50).

Figure 50.

3,5-AHBA containing natural products with the AHBA core colored red.

11.1. Ansamycins

Biosynthetic logic suggested 3-amino-5-hydroxybenoic acid (3,5-AHBA, 152), a known naturally occurring amino acid from Streptomyces,²⁸¹ may be the source of the naphthalenic core of the ansamycins, including ansatrienin A (44), rifamycin S (153), and geldanamycin (154). This early hypothesis was confirmed in feeding studies²⁸² with stable isotope labeled 3,5-AHBA, and ultimately via gene knockout and chemical complementation studies. However, in contrast with biosynthetic studies of anthranilate, feeding experiments with stable isotope labeled shikimate (4), quinate (23), and DHQ (8) failed to isotope enrich rifamycins in Amycolatopsis mediterranei, suggesting that the branch point to produce the ‘mC₇N’ (meta-substituted 7-carbon nitrogen) core must occur prior to DHQ. These results initially led to hypotheses of crosstalk between observed aminoshikimate and shikimate intermediates and enzymes.²⁸⁰ Ultimately, additional studies with the rifamycin-producing strain A. mediterranei revealed a completely independent aminoshikimate biosynthetic pathway containing a parallel set of enzymes to those found in shikimate biosynthesis (Figure 51).²⁸⁰ This pathway, however, should not be mistaken as related to the pathway involved in the formation of the constitutional isomer 3-amino-4-hydroxybenzoic acid (3,4-AHBA). The 3,4-AHBA core gives rise to the similarly structured mC₇N moiety in a variety of natural products such as the manumycin-type and grixazone phenoxazinone antibiotics, however its biosynthesis is completely unrelated to the shikimate pathway.²⁸³

Figure 51.

The aminoshikimate/3,5-AHBA biosynthetic pathway with representative enzymes from the A. mediterranei rifamycin biosynthetic gene cluster.

With respect to the aminoshikimate pathway,²⁸⁴ the functional DAHPS homolog, aminoDAHP synthase (RifH), catalyzes the an imino-aldol reaction between PEP (5) and proposed substrate, iminoE4P (155), to give aminoDAHP (156) (Figure 51).^285,286 The instability of iminoE4P has precluded its direct observation as a substrate in the RifH catalyzed reaction, however Frost and coworkers designed and executed elegant experiments using transketolase enzymes to demonstrate aminosugars such as kanosamine (157), are the origin of nitrogen in the aminoshikimate pathway and to infer the structure of iminoE4P.^287,288 This initial biosynthetic reaction is followed by two additional steps that mirror those from shikimate biosynthesis with aminoDHQ synthase (RifG) converting 156 to aminoDHQ (158) followed by dehydration to aminoDHS (159) via aminoDHQ dehydratase (RifJ). AminoDHS can then either be trafficked to aminoshikimate (151) via aminoDHS dehydrogenase (RifI), again echoing shikimate biosynthesis, or can be converted to the mC₇N precursor 3,5-AHBA (152) via AHBA synthase (RifK), which has no representative homologue in shikimate assembly. Notably, aminoshikimate itself does not serve as an intermediate in the assembly of 3,5-AHBA, nor any known secondary metabolite, however this stereochemically rich scaffold has been proposed to be valuable as a precursor for the construction of antiviral pharmaceuticals such as oseltamivir phosphate (Tamiflu).²⁸⁹ Elucidation of the core 3,5-AHBA biosynthetic genes not only resolved the key components for the assembly of this scaffold prevalent in a wide variety of natural products, but also enabled their use as “hooks” for genome mining strategies to identify new 3,5-AHBA containing natural products.^290–292

Rifamycin molecules are also biosynthesized by the marine bacterium Salinispora arenicola that uniquely produce the related saliniketals A (160) and B (161) that feature a similar polyketide framework, but without the 3,5-AHBA starter unit (Figure 52). In addition to their potent bioactivity as ornithine decarboxylase inhibitors, the saliniketals posed a curious biosynthetic question about their origin as shunt or independent products. The S. arenicola genome revealed a rifamycin BGC (SA-rif) along with an additional 12 PKS-driven BGCs, however, none possessed a domain architecture consistent with the assembly of the saliniketals.²⁹⁴ The shared biosynthetic origin of the saliniketals and rifamycins was demonstrated through disruption of the 3,5-AHBA synthase gene in the SA-rif BGC, which abrogated the production of both metabolites. Chemical complementation with [¹⁵N]-3,5-AHBA resulted in isotopically enriched saliniketals, confirming the origin of the primary amide as the 3,5-AHBA nitrogen and that C–N bond cleavage of a rifamycin-type intermediate is an on-pathway event in the biosynthesis of the saliniketals. However, disruption of rif tailoring genes failed to identify a C–N bond cleaving enzyme, and Fenical and Moore noted “if this reaction is indeed catalyzed by a dedicated enzyme, it may be encoded outside of the rif cluster.”

Figure 52.

Structural similarities in the polyketide region between saliniketals A (160) and B (161) and rifamycin S (153), all produced by S. arenicola and atom from 3,5-AHBA (152) colored in red.

Recently, Wright and coworkers discovered a widespread mechanism in bacteria for resistance against rifamycin and its derivatives wherein the C–N bond is oxidatively cleaved by flavin monooxygenases dubbed “Rox” enzymes (Figure 53).²⁹⁵ Bioinformatic investigation into rifamycin producers revealed the presence of Rox enzymes outside the bounds of their respective rif clusters, and in vitro characterization of these enzymes revealed they are not only capable of deactivating rifamycins via oxidative C–N bond cleavage but are also able to convert 16-demethyl salinisporamycin (162) into saliniketals such as 16-demethylsaliniketal (163), demonstrating these enzymes play a dual role in self resistance against rifamycins and catalyzing the final step in saliniketal biosynthesis.²⁹⁶

Figure 53.

Rox enzymes catalyze C–N bond cleavage on rifamycins/salinisporamycins (16-demethylsalinisporamycin, 162) to generate the saliniketals (16-demethylsaliniketal A, 163).

The Brady lab capitalized on the discrete phylogenetic clustering of 3,5-AHBA synthase genes with their corresponding metabolites and leveraged this observation as a strategy to efficiently identify rifamycin analogues from complex samples.²⁹⁷ A BGC from an isolated strain of Amycolatopsis vancoresmycina was prioritized due to its diversity of tailoring enzymes and altered predicted selectivity of the acyltransferase domain in the eighth PKS module from methylmalonyl-CoA to ethylmalonyl-CoA. Fermentation of A. vancoresmycina resulted in the isolation of known rifamycin congener kanglemycin A (164) and new kanglemycins, V1 (165) and V2 (166). These new kanglemycins demonstrated improved activity in vivo and in vitro against pathogenic Gram-positive bacteria harboring clinically relevant RNA polymerase (RNAP) mutations associated with rifamycin resistance. High-resolution crystal structures of kanglemycins in complex with RNAP demonstrated the kanglemycins possessed additional points of contact with RNAP which may assist in overcoming the loss of contact caused by known mutations in rifamycin resistant RNAP. While significant effort in genome mining is centered on the discovery of structurally novel and uncharacterized classes of natural products, this work demonstrates the utility of using the core 3,5-AHBA biosynthetic genes as hooks to identify natural products of known structural classes with improved efficacy towards pathogenic bacteria with clinically relevant antibacterial resistance.

11.2. Mitomycins

The mitomycins are a family of potent DNA-alkylating antitumor natural products, with mitomycin C (167) approved by the FDA for the treatment of gastric and pancreatic carcinomas in 1974. The mitomycins are characterized by their densely functionalized, fused polycyclic quinoid scaffold and their reactive aziridine and carbamate functionalities which, upon reduction, generate electrophilic centers for the alkylation of DNA. Stable isotope feeding studies with Streptomyces lavendulae demonstrated the entire mitomycin tetracyclic core is derived solely from 3,5-AHBA and glucosamine with peripheral methyl and carbamate groups derived from methionine and citrulline, respectively. Through generating a comprehensive series of disruptions in S. lavendulae, Sherman and coworkers were able to identify the genetic loci and boundaries of the mitomycin BGC, as well as implicate key genes involved in mitomycin biosynthesis, resistance, regulation, and transport.²⁹⁸ Notably, disruption of a gene encoding a standalone carrier protein, mmcB, completely abrogated production of mitomycins despite the lack of obvious involvement of either PKS or NRPS biochemistry. While the role of enzymes involved in the mitosane warhead remained elusive, Sherman and coworkers were able to functionally characterize a variety of enzymes involved in the late-stage tailoring of mitomycins, particularly hydroquinone²⁹⁹ and aziridine^300,301 methylation.

The thiotemplated nature of the mitomycin biosynthetic pathway has recently been elucidated through in vitro characterization of recombinant proteins. In nearly simultaneous publications, Yokoyama and coworkers³⁰² and Dairi and coworkers³⁰³ demonstrated that the CoA ligase MitE loads 3,5-AHBA (152) onto MmcB to generate the 3,5-AHBA-MmcB adduct 168 (Figure 54). To demonstrate that the convergence of precursors 3,5-AHBA and glucosamine (in the form of UDP-N-acetylglucosamine, UDP-GlcNAc) occurs on the carrier protein, UDP-GlcNAc and 168 were incubated with the glycosyltransferase MitB, which resulted in the formation of glycosylated 3,5-AHBA intermediate (169). This pathway is an intriguing representation of an ACP-dependent reaction wherein a substrate is tailored on an ACP in the absence of a PKS or FAS.

Figure 54.

Characterized steps in the thiotemplated assembly of mitomycins.

Further tailoring of the thiotemplated glycosyl-AHBA towards the mitosane core was explored by Liu and coworkers;³⁰⁴ by replacing the native mmcB with a His-tagged complement in mitomycin producing Streptomyces caespitosus, they were able to purify and characterize ACP-tethered intermediates. Incubation of 169 with the deacetylase MitC resulted in deacetylation to 170 with notable selectivity as MitC was unable to deacetylate UDP-GlcNAc. Disruption of nearby genes mitD (radical SAM) and mitF (NADPH-dependent dehydrogenase) and the purification of ACP-tethered intermediates also resulted in 170, suggesting these enzymes may catalyze the next steps in mitosane core assembly. Characterization of these genes in vivo and in vitro demonstrated that MitF catalyzes the reduction of the imine form of 171 to amine 172, which is then oxidized by MitD to generate diastereomeric epoxides (173a/b) whose cyclization products (174a/b) are representative of the diastereomeric mitomycins A (175) and B (176).

12. Perspectives and outlook

Aromatic amino acid biosynthesis proceeds through the shikimate pathway, seven enzymatic steps that begins with the combination of PEP and E4P and culminates in the production of chorismate, the last common intermediate between phenylalanine, tyrosine, and tryptophan. The biosynthesis of the aromatic amino acids is extraordinarily expensive metabolically,³⁰⁵ more so than any of the other proteinogenic amino acids,³⁰⁶ and so therefore AAAs are used relatively less frequently in bacterial proteomes. Perhaps because of these high metabolic costs associated with their synthesis, aromatic amino acids are valuable for secondary metabolism as well. These amino acids are widely incorporated into a variety of peptides, phenylpropanoids, and alkaloids. Often, such as in the case of neuroactive tryptophan-derived alkaloids, it is these aromatic amino acid moieties themselves that are responsible for the bioactivity of the molecule.³⁰⁷

However, as the body of work highlighted here demonstrates, intermediates from the shikimate pathway can be major drivers of metabolic diversity for specialized metabolism as well in microbes and plants. The densely functionalized intermediates between DAHP and chorismate can be diverted away from the shikimate pathway, often through particularly interesting enzymatic reactions, to yield a variety of specialized metabolites, including cofactors, antibiotics, lipids, and other bioactive molecules. The tremendous diversity of natural product classes that utilize shikimate-derived intermediates is also noticeable, as reflected throughout this review.

Notably, for almost all the biosynthetic pathways described in this article, the shikimate-derived portion of the molecule serves as the starter unit or core upon which additional biosynthetic enzymes act to construct a mature molecule. This phenomenon embodies the complex interplay between primary and secondary metabolism and showcases the shikimate pathway’s central role in both types of metabolism. Often, shikimate pathway orthologs are encoded within natural products BGCs under separate regulatory control to permit an organism to overcome primary metabolic bottlenecks and allow for specialized metabolic needs. In fact, these compounds are so valuable in secondary metabolism that some microbes have developed a similar pathway, dubbed “the aminoshikimate pathway”, for the biosynthesis of aminovariants, all of which fuel specialized metabolic pathways while also sharing similar metabolic pathways as their shikimate congeners as seen in the biosynthesis of the mitomycins and pactamycin.

Because of how valuable and widely used shikimate intermediates are in specialized metabolism, they are worthy targets for modern genome mining efforts. Many of the biosynthetic enzymes that perform the first dedicated biosynthetic step on a shikimate intermediate bear structural and/or sequence similarity to the shikimate pathway enzyme that typically utilizes that substrate. We suggest that using canonical shikimate enzymes as biosynthetic search hooks may be useful as a genome mining informatic tool that could lead to the discovery of new pathways and/or products. Perhaps using DAHP synthase, shikimate kinase, and EPSP synthase enzymes as genome mining hooks might reveal new pathways with homologs of these enzymes to specialized metabolites. Ordinarily, the genes used as genome mining hooks are those particularly distinctive of secondary metabolism, i.e., PKS or NRPS encoding genes. However, this also means the class of molecule discovered via this approach is predetermined, inherently biasing the search away from chemical novelty. Instead, using an enzyme from primary metabolism as a search hook and loosening the stringency for homologs, could result in the discovery of unknown molecular classes that incorporates a shikimate derived feature.

Despite the diverse pathways and enzymatic reactions discussed here, we were unable to identify any validated biosynthetic pathways that hijack and utilize the shikimate intermediates DAHP, shikimate-3-phosphate, and EPSP. Considering the biosynthetic diversity produced from other intermediates, and the variety of enzymatic reactions at play, it is likely that these intermediates can also enter secondary metabolism. Chorismate, on the other hand, is by far the most versatile of the shikimate pathway metabolites presently reported, giving rise to many distinct metabolic building blocks that nourish a bevy of diverse compounds (Figure 19). It should be pointed out however that other branch point pathway intermediates may be just as versatile as chorismate but have yet to be studied in such detail.

In preparing this review, we noticed that, not as a rule but as general practice, it seems that bacteria tend to utilize shikimate intermediates from the beginning of the pathway, while fungi use latter intermediates. For example, anthranilate is a relatively common structure feature in fungal natural products but is used much less frequently in bacterial systems. We also observed that higher organisms seem to allow the biosynthetic pathway to proceed almost to completion, while bacteria tend to derive intermediates from earlier in the pathway, and with higher frequency.

In summary, the shikimate pathway is a key driver of metabolic diversity in plants, bacteria, and fungi, both from the main pathway itself via aromatic amino acids, and from intermediary branch points as highlighted here. With the development of homology-based sequence network tools for identifying evolutionarily related enzymes together with increased genomic information across life,³⁰⁸ we anticipate that new biosynthetic features of the shikimate pathway to specialized metabolites are awaiting discovery.

Acknowledgements

This work was supported by the National Institutes of Health (N.I.H.) grants F32-GM145146 to V.S., F31-HD101307 to K.D.B., and R01-AI127622 and R01-GM085770 to B.S.M. We dedicate this manuscript to the late Prof. Heinz G. Floss (1934–2022; University of Washington) in recognition of his seminal contributions on the biosynthesis of shikimate-based microbial metabolites.

Figure 10.

Biosynthesis of gallate-containing ellagitannins, including cornusiin E (38), as informed by formative works by Gross and coworkers.

Figure 24.

Canonical menaquinone biosynthetic pathway proceeding via isochorismate (56, top route) and alternative futalosine pathway diverging directly from chorismate (12, lower route).

Figure 30.

Chorismatase activities of UbiC and XanB2 and their role in the biosynthesis of ubiquinone coenzyme Q8 (80) and xanthomonadin A (81).