C—N-Coupled Metabolites Yield Insights into Dynemicin A Biosynthesis

Abstract

The biosynthesis of the three structural subclasses of enediyne antitumor antibiotics remains largely unknown beyond a common C₁₆-hexaene precursor. For the anthraquinone-fused subtype, however, an unexpected iodoanthracene γ-thiolactone was established to be a mid-pathway intermediate to dynemicin A. Having deleted a putative flavin-dependent oxidoreductase from the dynemicin biosynthetic gene cluster, we can now report four metabolites that incorporate the iodoanthracene and reveal the formation of the C—N bond linking the anthraquinone and enediyne halves emblematic of this structural subclass. The coupling of an aryl iodide and an amine is familiar from organometallic chemistry, but has little or no precedent in natural product biosynthesis. These metabolites suggest further that enediyne formation occurs early in the overall biosynthesis, and that even earlier events might convert the C₁₆-hexaene to a common C₁₅ intermediate that partitions to enediyne and anthraquinone building blocks for the heterodimerization.

The 9- and 10-membered enediyne antitumor antibiotics share their origins from acetate units,^[1–3] but their intermediate biosynthetic steps remain confounding, unsolved problems among natural products. Of the three main structural subclasses, exemplified by the 9-membered neocarzinostatin (NCS, 1; Figure 1) and the 10-membered calicheamicin (CLM, 2) and dynemicin A (DYN, 3), insights have been gained recently that are believed to be general for the anthraquinone-fused family represented by DYN. It was known from the acetate labeling experiments above that the enediyne and anthraquinone halves of DYN arise from two separate polyketide chains.^[3] Despite great structural dissimilarity and presumed different biosynthetic origins of the two halves, however, they both derive from DynE8, the single polyketide synthase encoded in the DYN biosynthetic gene cluster (BGC).^[4] DynE8 is an iterative, highly-reducing polyketide synthase (HR-PKS) that is phylogenetically related to other enediyne HR-PKSs, which collectively constitute a distinct group in the polyketide synthase superfamily.^[5] The highly homologous CLM PKS, CalE8, has been studied in detail. Careful reconstitution experiments point to acyl-carrier protein-bound β-hydroxyhexaene 4 (Figure 2) as the on-pathway intermediate synthesized by this PKS.^[6,7] In the presence of its genomically-adjacent thioesterase, CalE7, this reactive PKS-bound intermediate undergoes hydrolysis with concomitant decarboxylation and dehydration to afford the yellow heptaene 5.^[8] It is believed that heptaene 5 is a common shunt product of all enediyne biosynthetic pathways,^[8,9] supporting a role for 4 as the programmed product of all enediyne PKS subtypes.

Figure 1.

Representative structural types of enediyne antitumor antibiotics: Neocarzinostatin (NCS, 1), calicheamicin (CLM, 2), and dynemicin A (DYN, 3). The dashed line in 3 represents the junction of the two halves of DYN, each of separate polyketide origin.

Figure 2.

Biogenetic proposal for coupling the enediyne and anthraquinone precursors of DYN. DynE8 assembles β-hydroxyhexaene 4, which is processed by the enediyne TE to generate heptaene 5. The C₁₆-hexaene 4 is also believed to be an intermediate en route to both the enediyne and anthraquinone halves of DYN (X and iodoanthracene 6, respectively), which are coupled to generate the heterodimeric natural product.

The force of the evidence brought by these findings dictates that the 16-carbon β-hydroxyhexaene 4 must partition into two distinct, but parallel pathways to prepare precursors of the enediyne and anthraquinone halves of DYN (each 14 carbons in the final natural product). Assuming that the synthetic flux through these two channels is not perfectly matched, the HPLC profiles of the wild-type DYN producer, Micromonospora chersina, and a CRISPR-Cas9 knockout strain of DynE8 were compared, which led to the identification of an accumulated iodoanthracene 6 bearing a γ-thiolactone ring (Figure 2).^[10] This metabolite subsequently proved to be the first intermediate identified mid-pathway to an enediyne natural product, which was a surprising discovery as iodine and sulfur occur rarely in natural products, especially aromatic thiols.^[11,12] Furthermore, the DYN BGC encodes no apparent halogenase or sulfur-insertion enzyme, and neither atom appears in the final product.^[4] The fact that it accumulates in readily detectable amounts, however, suggests – but does not prove – that 6 could couple with an unknown X from the “enediyne” arm of the pathway (Figure 2) to form a heterodimeric intermediate en route to DYN. Mechanisms to account for how such a process might occur have been discussed and point to a unimolecular radical process (S_RN1) rather than a nucleophilic aromatic substitution, where the former would involve single-electron transfer followed by loss of iodide and C–N bond formation.^[10]

Further CRISPR-Cas9 genetic deletions in the DYN BGC demonstrated that a pair of putative cytochromes P450, E10 and orf19, specifically hydroxylate C15 and C18 of the anthraquinone, respectively.^[13] Bearing in mind the global elevation of oxidation state inherent to every enediyne natural product from β-hydroxyhexaene 4, we sought to further delete oxidative biosynthetic proteins from the DYN BGC. Bioinformatics analysis identified E13, a putative flavin-dependent oxidoreductase, as a candidate for study, which has homologs in all other anthraquinone-enediyne BGCs sequenced to- date,^[14,15] but which is notably absent from the CLM BGC.^[16] In view of this difference, it was proposed that E13 could potentially play a unique role in the biosynthesis of the anthraquinone-containing subclass of 10-membered enediynes, as we report in the following.

An overlay of the HPLC traces of the ΔE13 mutant and wild-type M. chersina is presented in Figure 3a. DYN production was completely abolished in the mutant, but could be restored by introducing a copy of E13 on a separate plasmid under the control of the constitutive ermEp* promoter.^[17] Importantly, production of the iodoanthracene 6 appeared unaffected by the removal of E13, and four new major metabolites 7–10 were observed. Their UV-vis spectra are displayed in Figure 3b along with exact mass data (UPLC-ESI-MS) and their predicted elemental compositions, which differ only slightly in oxygen and hydrogen count. Especially striking were molecular masses consistent with late-pathway, post-heterodimerization structures, which were further supported by the predicted presence of nitrogen and the absence of iodine, but retention of sulfur in all four molecules.

Figure 3.

a) HPLC comparison of fermentation extracts from ΔE13 and wild-type M. chersina, and a ΔE13 reconstitution strain expressing C-terminally hexahistidine-tagged E13. b) UV-vis spectra, solution images, masses, and predicted molecular formulae of 7-10

To structurally characterize 7–10, the metabolites were isolated from 6 to 12 L fermentations of the ΔE13 mutant; when necessary, [1-¹³C]- and [2-¹³C]sodium acetate were administered to the fermentations to increase the sensitivity of downstream NMR experiments. Purification of the ΔE13 products was aided by their bright neon hues (Figure 3b), which eased visualization by standard silica gel flash chromatography. Owing to its higher yield, we first focused on characterization of 9 (Figure 4a). The ¹H NMR spectrum of 9 (Figure S2) immediately revealed a substructure corresponding to 6, as the characteristic signals of the anthracene were readily apparent.^[10] Chemical shift changes for these peaks were all less than 0.2 ppm except for a hydrogen in the C-ring, which was shifted dramatically upfield by about 1.5 ppm. We reasoned that substituting the iodine of 6 with the strong electron donor nitrogen (in accordance with the anthraquinone-enediyne C—N linkage of DYN) would have this effect and, thus, attributed a thiolactone-fused aminoanthracene partial structure to 9.

Figure 4.

Rationale for the production of 7–10 by M. chersina ΔE13. a) β-hydroxyhexaene 4 is proposed to partition to enediyne precursor X and iodoanthracene 6 (through a possibly common C_l5-intermediate Z), which are coupled to generate the on-pathway anthracene-enediyne Y. b) In the presence of E13, Y is processed to DYN; in its absence, this mutant would be expected to accumulate 13, its possible substrate, which would rapidly isomerize to 10. Carbons of 13 indicated with a dot (•) would be lost en route to DYN. c) Alternatively, epoxidation could occur just after the double bond is formed at the D/F-ring fusion in an enediyne precursor, and in its absence 10 could be formed directly in the heterodimerization event.

Accounting for the aminoanthracene left C₁₅H₁₄O to be assigned to the “enediyne” half of the molecule, although the diagnostic enediyne hydrogen resonances were conspicuously absent from the ¹H NMR spectrum. Instead, we identified four coupled hydrogens (δ = 7.21, 7.17, 7.05, 6.83) characteristic of a 1,2-disubstituted benzene ring, which indicated that 9 is the Bergman rearrangement product of an enediyne precursor. Assuming that the core connectivity of DYN was retained, we then identified a methyl group at δ = 2.14, which we assigned to the F-ring. The methyl group was coupled to two alkene carbons by HMBC (Figure S6) and split into a doublet (J = 1.4 Hz) by a vinyl hydrogen (δ = 5.80) as determined by COSY (Figure S4). These observations led to a proposed 1-methyl-cyclohexene substructure for the F-ring. The chemical shift of the vinyl hydrogen, however, seemed unlikely without an adjacent electron-withdrawing group (the analogous hydrogen of 1-methyl-cyclohexene appears upfield of our hydrogen at δ = 5.38).^[18] The ¹³C NMR spectrum revealed an additional carbonyl resonance at δ = 198.61 (Figure S3). We positioned this carbonyl, therefore, next to the vinyl hydrogen; comparison of the ¹H NMR spectrum of 9 to that of 3-methyl-2-cyclohexen-1-one (where the vinyl hydrogen appears at δ = 5.88)^[19] supported this assignment. To finally piece together the heptaketide core of DYN, we linked four contiguous methine carbons (δ = 5.64, 3.36, 3.48, 3.67) bridging the D- and F-rings using a combination of COSY and HMBC. Their orientation was anchored by the terminal hydrogens at δ = 5.64 and 3.67, which coupled to a carbon of the aromatized enediyne and the F-ring ketone, respectively.

Despite assigning the full polyketide backbone of the DYN enediyne to 9, our structure was incomplete due to the presence of an additional methylene in the enediyne half of the molecule that DYN lacks (determined by HSQC, Figure S5). A crosspeak was identified between this methylene and the methine carbon in the β-position to the F-ring carbonyl by HMBC (Figure S6), so we appended the methylene to this carbon, thereby extending the polyketide chain by one carbon. Further distinguishing 9 from the final natural product was the absence of a nitrogen-associated hydrogen. To account for this distinction, we linked the methylene carbon to the nitrogen to generate a pyrrolidine ring, and thus arrived at the complete structure for 9 (Figure 4a). Owing to an extra carbon in each half of the molecule (the methylene and thiolactone carbonyl), 9 was given a distinct numbering system from that of DYN (Figure 1).^[20] This numbering system was extended to metabolites 7, 8, and 10 in the descriptions below.

Following the assignment of 9, the structure elucidation of compounds 7 and 8 was straightforward since comparison of their NMR spectra revealed conserved core structures. Notably, the longest wavelength λ_max of 8 (Figure 3b) was greater than that of 9 by about 50 nm, implying an apparent extension of the conjugated anthracene chromophore. However, the predicted formula of 8, C₃₀H₁₉NO₂S, only differed from that of 9 by two fewer mass units (Figure 3b), suggesting that a simple structural alteration prompted this sizable red shift. In agreement with the MS data, two hydrogens were missing from the ¹H NMR spectrum of 8: the diagnostic B-ring singlet of the anthracene, and one “enediyne” hydrogen, leaving a 1,2,3-trisubstituted ring (Figure S7). We postulated, therefore, that a C–C bond formed between the enediyne and the proximal anthracene by reaction of a Bergman rearrangement radical with the anthracene at its most reactive position.^[21] Substantiating this proposal was the observation that the hydrogen at C10 was shifted downfield by about 0.7 ppm, which could result from a deshielding aromatic ring current by fixing the cycloaromatized enediyne over the anthracene A-ring. Nuclear Overhauser Effect Spectroscopy (NOESY) validated this assignment; a crosspeak was detected between the C10 and C24 hydrogens (Figure S12), thus confirming the structure of 8 (Figure 4a).

Like compound 8, the ¹H NMR spectrum of 7 revealed a 1,2,3-trisubstituted enediyne-derived benzene ring (Figure S13). However, the molecule still possessed its anthracene B-ring hydrogen and one more oxygen than 9. To account for the mass data and ¹H NMR spectrum, we positioned a hydroxyl group on the cycloaromatized enediyne. This assignment was supported by the appearance of a new carbon resonance between 160 and 150 ppm in the ¹³C NMR spectrum (Figure S14). The hydrogen at C21 (δ = 6.17) was coupled to this hydroxylated carbon by HMBC (Figure S17), so we assigned the hydroxyl to C23 and thereby completed the structure of 7 (Figure 4a). A similarly oxygenated enediyne byproduct was reported from the uncialamycin producer,^[22] thus suggesting a common mechanism of oxygenation following Bergman rearrangement.

Finally arriving at compound 10, we identified two characteristic enediyne hydrogens (δ = 5.83 and 5.69) in its ¹H NMR spectrum (Figure S18), which were complemented by a simplified aromatic region in comparison to the cycloaromatized metabolites just described. The calculated mass of this putative enediyne – that is, 9 pre-Bergman cyclization – was two mass units fewer than the measured mass of 10. Additionally, the ¹³C NMR spectrum of 10 had only one carbonyl signal (Figure S20), although the overall oxygen count matched those of 9 and 10 (Figure 3b). We reasoned that modifying the oxidation state at C29 from a ketone to an alcohol would account for these differences. Closer inspection of the ¹H NMR spectrum revealed a doublet at δ = 1.80, which disappeared upon the addition of D₂O, confirming the presence of a secondary alcohol (Figure S19). Moreover, the vinyl hydrogen at C30 in 7–9 was replaced by a new methylene (δ = 2.79 and 2.00; Figure S22), and the multiplicity of the hydrogen at C21 revealed loss of a vicinal coupling partner (Figure S18). We rationalized these changes by migrating the double bond from C30-C31 in 7–9 to between C20 and C31 in 10 (Figure 4a; ¹H and ¹³C NMR data for 7–10 are in Table S2 and S3).

Heartened by the identification of the enediyne-containing heterodimer 10, we sought to test its intermediacy in the biosynthesis of DYN. Following a similar protocol that established iodoanthracene 6 as on-pathway,^[10] [1-¹³C]- and [2-¹³C] sodium acetate were administered to the ΔE13 mutant to enrich the carbon framework of isolated 10. This material in turn was supplied to a fermentation of wild-type M. chersina without NaI, as the absence of iodide in the fermentation prevents accumulation of DYN.^[10] In contrast to our previous work, which demonstrated that supplementation of iodoanthracene 6 rescued wild-type production of the final natural product, repeated trials with 10 saw no return of DYN biosynthesis (Figure S24).

The failure of 10 to support DYN biosynthesis is perhaps to be expected – the position of the C—C double bond in the F-ring is incorrect. Spanning the fusion of the D- and F-rings would lead to the proper regiochemistry of DYN epoxidation, which is essential to its mechanism of DNA cleavage.^[23] Accompanying recovered ¹³C-labeled 10 (Figure S24), however, were barely detectable amounts of several minor metabolites. Their low concentrations impeded confident mass assignments, but one species observed just short of the retention time of DYN produced a mass of m/z 448.1542. The molecular composition was determined to be C₂₉H₂₁NO₄, which corresponds to the loss of CS from 10 and the gain of O₂. We propose that the most probable downstream modifications to account for these mass changes are loss of the thiolactone and oxidation to hydroxyanthraquinone 11. Although too small an amount was isolated for NMR analysis, the unusual aminohydroxyanthraquinone chromophore of 11 is highly similar to that of dideoxydynemicin A (12), which we previously characterized from a double cytochrome P450 mutant (ΔE10Δorfl9) of M. chersina (Figure 5).^[13] The absence of late-stage A-ring hydroxylations in 11, like 12, would be consistent with the structures of 7–10. Lastly, the appearance of the isotopic cluster of the parent ion made clear its derivation from labeled 10, and confirmed entry of the latter into the cell (Figure S24).

Figure 5.

Structure of putative anthraquinone-enediyne 11. Reincorporation of shunt metabolite 10 by wild-type M. chersina yielded several minor products, including compound 11, whose proposed structure is based on exact mass determination and UV-vis comparison to dideoxydynemicin A (12).

In considering the potential function of E13 in DYN biosynthesis, a BLAST search revealed it is homologous to the hydroxylase that installs one of the quinone oxygens in ubiquinone.^[24] Similar chemistry mediated by E13 could lead to anthraquinone formation, providing a plausible rationale for the production of 11 from the incorporation of 10 by wild-type M. chersina without NaI. Alternatively, flavin-dependent enzymes are known to carry out Baeyer-Villiger reactions and could cleave the thiolactone ring to introduce hydroxyl at C11 in DYN (C3 in 7–10) from molecular oxygen.^[13,25] Lastly, E13 is homologous to epoxidases, for example, that involved in the biosynthesis of the quinoline alkaloid aurachin.^[26] One can appreciate that a C19-C20 double bond at the D/F-ring junction such as that in hypothetical intermediate 13 (Figure 4b) would be planar and appreciably strained, especially in a tricyclic system bearing the enediyne bridge. Epoxidation would reduce overall strain by taking carbon hybridizations from sp² to sp³, but retain sufficient rigidity to prevent Bergman rearrangement. While diligent search for 13 was unsuccessful, its isomerization into the F-ring to produce 10 (Figure 4b) would be a rapid and thermodynamically-favored event, and a logical outcome of the ΔE13 mutant accumulating such a chemically unstable substrate. The role of E13 may be to trap this labile bond before it migrates. Alternatively, rather than E13 intercepting the highly-strained alkene 13, epoxidation at the D/F-ring junction might occur much earlier and be already present in enediyne intermediate X before heterodimerization (Figures 4a and c). This epoxidized enediyne precursor could then undergo a coupling reaction with iodoanthracene 6. Although efforts to identify this enediyne precursor have been unsuccessful, the accumulation of 6 in both the wild-type and ΔE13 mutant fermentations supports its placement at the end of the “anthraquinone” pathway, and, therefore, solidifies its role as a partner in the heterodimerization. The experiments reported here do not directly address the mechanism of this coupling, however, the data suggest linkage of 6 with a 2°-amine. Such a reaction is precedented classically in the copper-mediated Ullmann-Goldberg reaction,^[27–29] but as far as we are aware, is unknown in natural product biosynthesis. Suggesting potential mechanistic similarities, an aryliodide is the optimal halide substrate to undergo the Ullmann-Goldberg reaction.^[30,31]

Further deductions from the data can be briefly made. First, installation of the enediyne, which one might envision to be a very late stage process in the overall pathway to allow for stepwise introduction of increasing ring strain and delayed formation of a reactive structural element, appears to be a much earlier event, likely pre-heterodimerization. Second, the F-ring carboxylate at C5 in DYN is absent in 7–10 and poses a mysterious carbon branching step late in the overall biosynthetic pathway known to be derived from the C2 of acetate.^[3] Finally, the structures of 7–10 all show coupling of two C₁₅ intermediates. One implication of this observation is that the C₁₆ β-hydroxyhexaene 4 found on the PKS DynE8 is carried forward in shared steps to a common C₁₅ intermediate Z (Figure 4c) of unknown structure before partitioning and differentiating to the “enediyne” and “anthraquinone” component pathways. In the hypothetical intermediate 13, two carbons [labeled (•) in Figure 4b] each at the carboxy terminus of its C₁₅ polyketide half, would then have to be lost in the subsequent unknown steps to DYN. These structural mileposts, and the insights they give, both order major synthetic events and guide further experiments to decrypt still obscure chemistry of the DYN pathway.