Identification of the Early Steps in Herbicidin Biosynthesis Reveals an Atypical Mechanism of C-Glycosylation

Abstract

Herbicidins are adenosine-derived nucleoside antibiotics with an unusual tricyclic core structure. Deletion of the genes responsible for formation of the tricyclic skeleton in Streptomyces sp. L-9-10 reveals the in vivo importance of Her4, Her5, and Her6 in the early stages of herbicidin biosynthesis. In vitro characterization of Her4 and Her5 demonstrates their involvement in an initial, two-stage C–C coupling reaction that results in net C5′-glycosylation of ADP/ATP by UDP/TDP–glucuronic acid. Biochemical analyses and intermediate trapping experiments imply a noncanonical mechanism of C-glycosylation reminiscent of NAD-dependent S-adenosylhomocysteine (SAH)-hydrolase catalysis. Structural characterization of the isolated metabolites suggests possible reactions catalyzed by Her6 and Her7. An overall herbicidin biosynthetic pathway is proposed based on these observations.

Graphical Abstract

INTRODUCTION

Herbicidins (1–7, Figure 1) are adenosine-derived, undecose (C₁₁) nucleoside antibiotics isolated from several strains of Streptomyces.^1-12 Herbicidins show selective herbicidal activity against dicotyledonous plants and can protect rice plants from bacterial leaf blight.² They are also known to inhibit proliferation of the human parasite C. parvum¹⁰ and exhibit antialgal³ as well as antifungal⁴ activities. The biosynthesis of herbicidins has attracted particular attention on account of their tricyclic structures, which consist of a pyran (ring C) connected to a furan (ring A) via a methylene bridge and a hemiketal linkage thereby forming the bridging six-membered ring B. The herbicidins differ with respect to methylation at C2′-OH and C11′-OH, acylation at C8′-OH, and the stereochemistry at C8′ (see Figure S1). Consequently, aureonuclemycin (2), which was first isolated from Streptomyces aureus var. suzhoueusis and is characterized by an undecorated furano-pyrano-pyran glycosyl core alone,¹² is believed to be the principal intermediate during the biosynthesis of all herbicidins and their derivatives.

Figure 1.

(A) Organization of the herbicidin gene cluster (her) and part of the aureonuclemycin gene cluster (anm). (B) Proposed herbicidin biosynthetic pathway and representatives of isolated herbicidin derivatives. The reactions catalyzed by Her8/9/10/11 have been verified in vitro.¹³ (C) Potential mechanism of bond formation between C5′ and C6′ if catalyzed by Her6.

Rings C and A of the herbicidins have been demonstrated to originate biosynthetically from glucose and ribose, respectively, as indicated by the in vivo incorporation of ¹³C-labeled glucose and ribose in 1.¹⁴ Therefore, C-glycosylation appears to be required for the introduction of the C–C bond between rings C and A thereby facilitating construction of ring B during herbicidin biosynthesis. Natural products possessing O-, N-, and S-glycosidic linkages are ubiquitous, and the underlying glycosylation reactions generally involve glycosyltransferases that catalyze the addition of a nucleophilic secondary metabolite, which can be structurally quite diverse, to an activated carbohydrate donor often in the form of a nucleotidyl diphosphate (NDP)-linked sugar (Figure 2).¹⁵ The biosynthesis of C-glycosylated natural products follows a similar paradigm requiring activation of a nucleophilic carbon on the acceptor fragment before addition to the electrophilic NDP–sugar donor. For example, aromatic aglycones often possess a phenolic hydroxyl group ortho or para to the nucleophilic carbon facilitating electrophilic aromatic substitution of the aglycone by the sugar donor.^15,16 An analogous mechanism has also been noted to construct the C-glycosidic bond in the biosynthesis of pyrazofurin¹⁷ and several other C-nucleosides.^18-20 In contrast, the structures of herbicidins show little resemblance to the typical C-glycosylated natural products. Hence, ring B assembly in herbicidin biosynthesis may proceed along a unique course of C–C bond formation.

Figure 2.

O-Glycosylation with an alcohol acceptor and C-glycosylation with a phenolic aromatic aglycone.

An herbicidin biosynthetic gene cluster (her) has been identified in Streptomyces sp. L-9-10 (Figure 1A) and is similar to the corresponding gene clusters found in other herbicidin-producing strains.¹⁴ The aureonuclemycin (2) gene cluster (anm) has also been cloned from S. aureus, which is not known to produce any other herbicidin besides 2 (Figure 1A).¹³ Genes her4/5/6/7 in the her cluster show sequence similarity to anmB/C/E/D (56/47/56/62% amino acid identity, respectively) in the anm cluster and have been shown by a heterologous expression experiment to be essential for the production of 2.¹³ Therefore, her4/5/6/7 and anmB/C/D/E are predicted to fulfill similar biosynthetic roles in the respective biosynthetic pathways. In contrast, the anm cluster lacks homologs of her8/9/10/11, which were recently established to encode enzymes catalyzing the tailoring reactions that transform 2 to 1 during the biosynthesis of herbicidin (see Figure 1B).¹³ These observations are consistent with the hypothesis that her4/5/6/7 and anmB/C/D/E participate in the construction of the tricyclic undecose core; however, the catalytic properties of the encoded gene products have yet to be elucidated. Reported herein is the characterization of the early steps during the biosynthesis of herbicidin including a novel C-glycosidation reaction. These results thus suggest a likely pathway for assembly of the tricyclic undecose core.

Sequence alignment revealed that the four her genes (4, 5, 6, and 7), respectively, encode an NAD-dependent S-adenosylhomocysteine (SAH) hydrolase (Her4), two NAD-binding oxidoreductases (Her5 and Her7), and a B₁₂-dependent radical S-adenosyl-l-methionine (SAM, 10) enzyme (Her6). Since none of the encoded gene products was annotated as a glycosyltransferase, how the C–C bond between C5′ and C6′ is formed was not apparent. However, the presence of her6 in the cluster was initially of particular interest, because a number of radical SAM enzymes (e.g., MqnE, HpnH, NosL, and NikJ/PolH) have been documented to catalyze C–C bond formation between C5′ of SAM and the nominal substrate.^21-25 Thus, Her6-catalyzed reductive cleavage of SAM (10) could yield methionine (12) and a 5′-deoxyadenosyl radical (11) equivalent prior to radical addition of 11 to an electrophilic pyranosyl precursor (e.g., 13) resulting in the C5′–C6′ covalent linkage (Figure 1C).

RESULTS AND DISCUSSION

To explore the speculated function of Her6, the her6 gene was heterologously expressed in Streptomyces lividans and isolated as an N-His₆/C-strepII doubly tagged protein (Figure S2). The reconstituted Her6 exhibited a UV–vis absorption around 420 nm (Figure S3) and contained 4.44 ± 0.16 Fe per mole of the enzyme monomer. However, anaerobic incubation of 5 μM Her6 with 1 mM UDP–d-glucuronic acid (15, see Figure 4 for the structure) and 2 mM SAM (10) in the presence of 4 mM DTT, 0.1 mM HO-Cbl, 0.25 mM methyl viologen, and 2 mM NADPH in 25 mM HEPES buffer (pH 7.5) at room temperature for 16 h did not yield any observable reaction except the uncoupled reduction of SAM to 5′dAdo even in the absence of 15. Altering the incubation conditions including the use of several different reducing systems similarly did not result in any detectable reaction.

Figure 4.

(A) Sequential Her4/5 reactions with substrate UDP-GA (15). (B) Extracted-ion chromatograms (EICs) of Her4/5 reaction products corresponding to [M – H]⁻ signals with m/z = 440. Peak a likely corresponds to 22. (C) EIC of Her5 reaction product 23 corresponding to [M – H]⁻ signals with m/z = 173. (D) Nonenzymatic benzilic acid rearrangement converting 22 to 18. (E) EIC of the (D) reaction corresponding to [M – H]⁻ signals with m/z = 440 at different time points. (F) Proposed mechanism of Her5-catalyzed conversion of 15 to 23. The β-elimination may not be concerted.

Efforts were thus turned to investigate the biosynthesis of 2 beginning with the in vivo functions of Her4/5/6/7. Accordingly, the her4/5/6/7 genes from the producing strain were each in-frame deleted to obtain the four Δher4, Δher5, Δher6, and Δher7 mutants. The S. sp. L-9-10 wild-type (WT) strain and mutants were separately grown under the same conditions, and the fermentation broth was analyzed by high-performance liquid chromatography (HPLC) with detection by both UV–vis absorbance and high-resolution electrospray ionization mass spectrometry (HR-ESI-MS). Four herbicidin congeners, herbicidin A (1), herbicidin G (3), herbicidin B (6), and herbicidin H (7), which had been isolated and characterized in aprevious study,¹³ were all detected in the WT growth culture (Figure 3A) and identified based on co-elution with the compound standards and high-resolution mass spectrometry (HR-MS, Figure S4). In contrast, none of the above four herbicidin derivatives were produced by any of the Δher4, Δher5, and Δher6 mutants. However, it was possible to detect production of 1 in the Δher7 mutant fermentation media albeit at significantly reduced levels versus wild-type. These observations suggested that the her4/5/6 genes are all needed for construction of the tricyclic ring, whereas her7 does not appear to be essential for the biosynthesis of herbicidins.

Figure 3.

(A) HPLC analysis of the fermentation culture of Streptomyces sp. L-9-10 wild-type (WT) and mutant strains. (B) Extracted-ion chromatograms (EICs) of the fermentations corresponding to [M + H]⁺ signals with m/z = 426. (C) EIC of the fermentations corresponding to [M + H]⁺ signals with m/z = 442. (D) Structures of 17, 18, and 19.

A search for other possible biosynthetic intermediates was carried out by LC-ESI-MS analyses of the mutant cultures using extracted-ion chromatograms (EICs) at the predicted m/z for protonated aureonuclemycin (2) and its oxidized/reduced derivatives (m/z ± 2). In doing so, a new metabolite (proposed as 17, calcd for C₁₆H₂₀N₅O₉⁺ m/z 426.1256, obs. 426.1253) having the same molecular weight as 2 (calcd for C₁₆H₂₀N₅O₉⁺ m/z 426.1256, obs. 426.1362) was found to accumulate in the growth culture of Δher7 (Figure 3B); however, this new compound failed to co-elute with 2 on co-injection. Furthermore, EICs based on compounds carrying one more oxygen atom than 2 led to the identification of a second metabolite (proposed as 18, calcd for C₁₆H₂₀N₅O₁₀⁺ m/z 442.1205, obs. 442.1239), which was detected in the wild-type, Δher6 and Δher7 cultures but absent in the Δher4 and Δher5 cultures (Figure 3C). Heterologous in vivo expression of her4/5/6/7 in different combinations in the nonproducing strain S. lividans further supports the assignment of 17 and 18 as metabolites related to the her4/5/6 genes (Figure S5). Moreover, a third metabolite (proposed as 19, calcd for C₁₆H₂₀N₅O₉⁺ m/z 426.1256, obs. 426.1292) could also be detected in the cultures of S. lividans-her45, S. lividans-her4S6, and S. lividans-her4567 when the EICs were acquired at m/z 426 (Figure S5C).

Compounds 17 (1.1 mg), 18 (1.3 mg), and 19 (0.72 mg) were isolated by HPLC from 1.5 L fermentations with their molecular formulas determined by HR-ESI-MS (Figure S6) and structures proposed (Figure 3D) based on ¹H-NMR, ¹³C-NMR, and 2D NMR spectroscopy characterizations (see Tables S3-S5 and Figures S9-S24). The HR-ESI-MS and NMR data indicated that all three compounds consist of an adenine and an 11-carbon moiety. In compounds 18 and 19, the J_1′,_2′ values of the H1′ doublet are both 5.5 Hz, which is comparable to the coupling constant of the adenosine H1′ signal (about 6.1 Hz, see the adenosine structure (9) in Figure 1B). However, the H1′ J value of 17 is around 1.0 Hz, which is similar to that reported for other herbicidins that have a tricyclic ring structure.^7,10 The configuration at C8′ of 17 was confirmed by the observation of NOESY signals between H6′/H8′ and H8′/H8 indicating R-stereochemistry at C8′. Based on these analyses and the observation that 17 has the same molecular weight as aureonucleomycin (2), 17 was assigned as the new herbicidin congener 8′-epi-aureonuclemycin. Compound 17 accumulated in the Δher7 mutant and was not processed by the tailoring enzymes (i.e., Her8/9/10/11) in vivo, implying that 17 is likely a dead-end product.

Compounds 18 and 19 were not reduced when treated with NaBH₄, suggesting that there is no ketone nor other moieties susceptible to NaBH₄ reduction in either compound. Analysis of the ¹H-NMR spectra of isolated 18 and 19 revealed the absence of a C7′–H proton resonance. The ¹³C-NMR spectra showed two signals at δ 176 ppm and δ 83 ppm in addition to the C10′–COOH resonance (δ 178 ppm). The chemical shifts of these signals are different from those of a typical ketone carbon (ca. δ 185–220 ppm) or a hemiketal carbon (ca. δ 90–100 ppm). Instead, these ¹³C-NMR data suggested the presence of a hydroxylated quaternary carbon at C7′ that is conceivably also linked to a carboxyl group. Therefore, 18 and 19 were proposed to have a furanose C-ring resulting from a putative ring contraction process (Figure 4D). Furthermore, compound 19 also lacks a C8′ hydroxyl group (corresponding to the C9′ hydroxyl in 2) compared with 18. While it remains unclear how the deoxygenation occurs, herbicidin derivatives without a C9′ hydroxyl group have nevertheless been isolated as minor products from the producing strain.⁷

The above gene deletion experiments are inconsistent with the initially proposed function for Her6 (see Figure 1C) and instead imply that C5′–C6′ bond formation is likely catalyzed by Her4 or Her5. To investigate the catalytic activities of these enzymes, Her4 was heterologously expressed in Escherichia coli and purified as an N-His₆ tagged protein. Her5 was expressed in S. lividans and purified as an N-His₆/C-strepII doubly tagged protein (Figure S2). A reaction was observed when 1 mM uridine diphosphate glucuronic acid (UDP–GA, 15) and 1 mM ADP (20) were used as substrates in the presence of 1 mM NAD, 2 μM Her4, and 2 μM Her5 at room temperature for 18 h, with the formation of compound 18 detected by LC-ESI-MS (Figure S29). Subsequent screening of potential substrates that are possible glucuronic acid moiety donors and adenosine moiety donors (Table S6) in the putative reactions catalyzed by Her4 and Her5 led to the identification of either uridine diphosphate glucuronic acid (UDP–GA, 15) or thymidine diphosphate glucuronic acid (TDP-GA, 16, see Section S2.7 for preparation) as possible precursors to the glucuronic acid moiety and ADP (20) or ATP (21) as donors for the adenosine unit (Figures S29 and S30).

The Her4/5 reactions were also carried out in the presence of one enzyme (Her4 or Her5) first at room temperature for 18 h, and this enzyme was subsequently filtered out followed by addition of the second enzyme (Her5 or Her4); the reaction mixtures were then incubated at room temperature for an additional 18 h. LC-ESI-MS analysis revealed the formation of 18 only if the incubation was conducted by using Her5 first and then Her4 in sequence (Figure 4A route a). Reversal of the reaction sequence abolished the production of 18 (Figure 4A route b). The identity of 18 was confirmed by MS analysis and co-elution with the 18 standard (Figures 4B, S29, and S30B). An additional product was also noted (Figure 4B, peak a) exhibiting the same molecular weight as 18 suggesting an isomer. While this compound could not be isolated due to its scarcity and instability, time course analysis showed that it was slowly converted to 18 nonenzymatically at room temperature with a half-life of ca. 24 h (Figure 4D/E). This observation led to the proposed assignment of the new product as 22. When the Her4/5 reaction mixture was treated with NaBH₄ or NaBD₄, this compound disappeared, and four reduced product peaks formed (Figure S31). This observation is consistent with its assignment as 22, which is a diketone derivative with a gem-diol at either C7′ or C8′. As 22 has a bulky moiety at C6′, a gem-diol at C7′ is expected to be more sterically hindered than a gem-diol at C8′, and thus, the gem-diol is proposed to be at C8′ in 22 (Figure 4A). The fact that 22 can be nonenzymatically transformed into 18 (Figure 4D) further supports this structural assignment.

The above results established the reaction sequence in which the Her5 reaction proceeds prior to Her4 catalysis (Figure 4A, route a). Since Her5 is annotated as a NAD-dependent oxidase and the conversion of 15/16 and 20/21 to 22/18 involves a two-electron oxidation process, it is thus conceivable that the Her5 reaction may be initiated by C3″ oxidation of 15/16 followed by β-elimination of UDP or TDP to give 23 (Figure 4F). This putative Her5 product may therefore add to a 5′-deoxyadenosine derivative to generate 22/18. Examination of the EIC for the proposed Her5 product 23 (calcd for C₆H₅O₆⁻ m/z 173.0092, obs. 173.0084) in the incubation with Her5 alone led to the detection of a peak at 3.3 min (Figure S32). Thus, compound 23 may indeed be the product of Her5 reaction (Figures 4C, S32, and S33). The fact that C3″ deuterium labeling was not incorporated into 23 but transferred to NADH when [3"-²H₁]-15 was employed as the substrate in Her5 reaction (Figure S34) is consistent with the proposed structure of 23 and supports the initial oxidation of 15 occurring at C3″ (Figure 4F). The k_cat/K_M values of Her5 against 15 and 16 were determined to be 0.091 ± 0.008 and 0.007 ± 0.001 μM⁻¹ min⁻¹, respectively (Figure S35). The catalytic efficiency for 15 was about 10 times higher than that for 16, meaning that 15 is the more suitable substrate of Her5.

Despite low sequence identity (18% identity and 30% similarity versus human SAH hydrolase AHCYL1), Her4 is nevertheless annotated as an NAD-dependent SAH hydrolase that is commonly found in eukaryotes and catalyzes the conversion of SAH (24) to l-homocysteine and adenosine (9).²⁶ The overall transformation involves oxidation of the C3′ carbon so as to facilitate elimination of homocysteine to yield an enone intermediate (i.e., 25) prior to reductive hydration to give 9.^27,28 The reaction catalyzed by Her4 may thus proceed via a similar mechanism using ADP or ATP instead of SAH. Addition of 23 instead of water to form the furo-C-linked-pyran structure 26 followed by reduction would then yield the product 27 (Figure 5A), which then undergoes nonenzymatic conversion to 22/18. Indeed, in the absence of 23, the Her4-catalyzed reaction generated 4′,5′-didehydro-5′-deoxyadenosine (28), which presumably forms via reduction of the putative enone 25 instead of 26 (Figure S36A).

Figure 5.

(A) Proposed mechanism of Her4-catalyzed reaction. (B) Her4 reactions in buffered H₂O or D₂O with substrate 29. Incubation in the absence of 29 led to 28. Compound 24 was the product when homocysteine was used instead of 29. (C) Extracted-ion chromatograms (EICs) at m/z 404.085 ± 0.005, corresponding to [M – H]⁻ for 30, for Her4 reactions in buffered H₂O (1) without Her4, (2) without NAD, (3) full reaction, and (4) without NAD but with 1 mM NADP. (D) MS analysis of the Her4 reaction product 30 in buffered (1) H₂O and (2) D₂O.

To test this hypothesis, a 4-pyrone analog (29) of 23 was used as an alternative substrate for Her4 (Figure 5B) and the reaction was carried out with 1 mM 29, 1 mM NAD, 1 mM ATP (21), and 2 μM Her4 at room temperature for 18 h. Formation of a new compound was observed (Figure 5C, peak b), which was absent in the control reactions. This compound could be detected by LC-MS in negative mode at m/z = 404.0863 (Figure 5C,D). When the reaction was carried out in D₂O, an LC-MS peak at the same retention time with m/z = 405.0934 in negative mode was observed, indicating incorporation of a single deuteron (Figure 5D). ¹H-NMR analyses of the isolated compound and its deuterated counterpart confirmed its identity as 30 and the incorporation of a deuterium at C4′ (30D) when the reaction was conducted in buffer prepared with D₂O (Figures S37-S42). Likewise, one deuterium incorporation was also noted using unlabeled substrates with Her4/5 in buffered D₂O. The products, 18D and 22D, are thus believed to be deuterated at C4′ as well (Figure S43), suggesting that solvent hydron exchange at C6′ through tautomerization did not occur. When l-homocysteine was used instead of 29, SAH (24) was produced (Figures 5B and S36B). The formation of 24 is consistent with the intermediacy of 25 during the Her4 catalytic cycle similar to that of SAH hydrolase (Figures 5A and S36C). These results also revealed that the C3′ stereochemistry of ATP remains unchanged in 28 as well as the adducts formed from 29 (i.e., 30) and l-homocysteine (i.e., 24), which implies retention of C3′ stereochemistry during the Her4/Her5-catalyzed C–C coupling reaction.

Taken together, the enzymes responsible for the initial coupling of rings A and C during biosynthesis of the herbicidin core have been identified. This reaction is catalyzed by two NAD-dependent enzymes, Her4 and Her5, and proceeds in two stages to affect net C5′-glycosylation of ADP/ATP (20/21) with UDP/TDP–glucuronic acid (15/16). The first stage involves Her5-catalyzed C3″ dehydrogenation of NDP–glucuronic acid to facilitate μ-elimination of NDP and thus formation of the activated glycosyl donor 23. Meanwhile, C3′ dehydrogenation of ADP/ATP catalyzed by Her4 facilitates the elimination of the C5′ phosphate group to give 25, which can subsequently undergo C–C bond formation with 23 to yield 26. The initially formed NADH can then reduce the C3′ keto group of 26 to afford 27 such that NAD serves as a catalytic redox cofactor being regenerated upon each turnover (Figure 5A). The resulting product of the Her4-catalyzed reaction (27) is susceptible to hydration of the C8′ keto group to give 22, which may then undergo benzilic acid rearrangement to produce 18 in the absence of the next enzyme in the pathway.

Three possible mechanisms can be envisioned for C–C bond formation between the activated species 23 and 25 (Figure 6). Mechanism A proceeds according to canonical glycosylation biochemistry involving a nucleophilic acceptor (25) that undergoes Michael addition to the electrophilic sugar donor (23). This generates a tricyclic oxonium intermediate 31 that can transform via retroaldol elimination and subsequent tautomerization to yield 26. In contrast, the more plausible mechanism B proceeds counter to the typical glycosylation mechanism with the sugar donor (23) serving instead as the nucleophile that adds to the electrophilic acceptor (25). This mechanism leads directly to 26 without the requirement of an oxonium intermediate and essentially mirrors the currently accepted mechanism of NAD-dependent SAH hydrolase catalysis.^27,28 Finally, mechanism C involves a hetero-Diels–Alder reaction between 23 and 25 to form the hemiacetal 32 followed by a ring-opening reaction. Similar intermolecular Diels–Alder reactions forming heterocycles have been reported in other biological systems^29-32 and utilized in synthetic applications.³³ Although distinguishing between these potential mechanisms of catalysis must await further experimentation, the Her4/Her5-catalyzed reaction appears to be distinct from the typical mechanisms of C-glycosylation observed in natural product biosynthesis.

Figure 6.

Proposed C–C bond formation mechanisms catalyzed by Her4.

Identification of the reactions catalyzed by Her4 and Her5 also suggests a possible pathway for herbicidin biosynthesis (Figure 7). Both 17 and 2 have the same R-configuration at C3′ and were indiscernible in the S. lividans-her45 culture but could be detected in the culture with S. lividans-her456. Therefore, the B₁₂-dependent radical SAM enzyme Her6 is proposed to catalyze the necessary epimerization of C3′–OH likely with 22 (with a 3′S configuration) serving as the substrate. To date, no B₁₂-dependent radical SAM enzyme has been reported to catalyze an epimerization reaction; however, radical SAM enzymes have been reported that do not depend on B₁₂ and are known to function as epimerases.^34-39 Subsequent cyclization to form ring B of herbicidin after C3′ epimerization need not be enzyme-catalyzed. The final step in the biosynthesis of 2 is then proposed to be the reduction of 33, which is hypothesized to be catalyzed by Her7. As shown by the gene deletion experiments, Her7 is not essential for the biosynthesis of 2 and reduction of C8′ in 33 may also be catalyzed in vivo by nonspecific reductases to generate both 17 along with 2 (Figure 7).

Figure 7.

Proposed biosynthetic pathway for aueronuclemycin (2).

CONCLUSIONS

In summary, a number of mechanistically important features have been identified in the early stages of herbicidin biosynthesis. Of particular interest is the two-stage glycosylation reaction catalyzed by Her4/5, during which C–C bond formation appears to proceed in a manner reminiscent of NAD-dependent SAH hydrolase catalysis. Consequently, this enzyme-catalyzed glycosyl transfer differs from canonical C-glycosylation reactions in that the NDP–sugar donor appears to act as the nucleophile rather than an electrophile. The reaction, once verified, may thus represent an example of an inverse-demand glycosylation. While Her6 is not involved in the C-glycosylation as previously proposed, it may instead contribute to C3′ epimerization. Additional studies of herbicidin biosynthesis including the reaction catalyzed by Her6, the mechanism of Her4-catalyzed C–C bond formation, and the structural characterization of the key biosynthetic enzymes are currently underway.