Formicamycin biosynthesis involves a unique reductive ring contraction

Using a combination of biomimetic chemistry and molecular genetics we demonstrate that formicamycin biosynthesis proceeds via reductive Favorskii-like reaction.


Introduction
Bacterial aromatic polyketide natural products comprise an extensive family of biosynthetically related molecules that encompass great structural variety and oen display potent biological activity. 1 They include several important clinical agents, for example the tetracycline antibiotics and anthracycline antineoplastic drugs. The carbon skeletons of aromatic polyketides are invariably derived from the function of dissociated type II polyketide synthases (PKSs), and additional structural diversity is introduced by tailoring enzymes following polyketide biosynthesis. 2,3 Such post-PKS transformations include acylation, alkylation, glycosylation and halogenation, as well as redox changes. Further structural complexity can be introduced via carbon-carbon bond cleavage and rearrangements in processes commonly catalyzed by oxidative enzymes. Illustrative examples include the reactions catalyzed by avindependent Baeyer-Villiger (BV) monooxygenase enzymes 4 during the biosynthesis of the jadomycins, 5 gilvocarcins 5 and chartreusins 6 amongst others.
We recently described the formicamycins, polyhalogenated tridecaketides that are co-products, along with the fasamycins, of the Streptomyces formicae KY5 for biosynthetic gene cluster (BGC; Fig. 1 and S1 †) encoding a type II PKS. 7 The fasamycins exhibit potent antibacterial activity and congeners were reported from heterologous expression of environmental DNA and were encoded by a clone expressing a type II PKS BGC, but no co-produced formicamycins were reported. 8 Despite their common biosynthetic origin and carbon backbone connectivity, the fasamycins and formicamycins have very different three-dimensional structures as determined from NOESY NMR, ECD spectroscopy and computational modelling. 7 Moreover, the formicamycins displayed more potent antibacterial activity and a barrier for the selection of resistant isolates.
Mutational analysis of the formicamycin (for) BGC identi-ed a single avin dependent halogenase (ForV) that can presumably catalyze multiple halogenation events. 7 Deletion of forV abolished the production of all halogenated congeners and the DforV mutant produced only a single compound fasamycin C 1 but no formicamycins (Fig. 2b); the deletion could be complemented by ectopic expression of forV. 7 This suggests a gatekeeper function for ForV, meaning the enzymes which convert fasamycins to formicamycins operate on halogenated intermediates, conrming that the fasamycins are biosynthetic precursors of the formicamycins. Structural comparison of the two skeletons indicates that unusual biosynthetic transformations are required to enable their interconversion, involving a dearomatization of fasamycin ring C via a formal tautomerization and two-electron reduction, plus the introduction of a tertiary hydroxyl group at C10. This suggests the requirement for a reductive enzyme in addition to a monooxygenase. These intriguing biosynthetic observations, in concert with their antibacterial properties, prompted us to study the biosynthetic steps linking the two compound families.

Bioinformatics analysis
The for BGC was sequenced previously 7 and comparison to the originally reported fasamycin BGC 8 did not identify any additional genes obviously responsible for formicamycin biosynthesis. However, comparison to a for-like BGC identi-ed in a 140 kbp region of the Streptomyces kanamyceticus genome that is syntenic with S. formicae KY5 (Genbank ID LIQU00000000.1), identied a set of four grouped genes forX-forAA which are not present in S. kanamyceticus (Fig. 3). A similar observation was made when we compared the for BGC to the recently reported BGC encoding production of accramycin A, a new fasamycin congener. 9 Three of these gene products (ForY, ForZ and ForAA) show high homology to sequences from Actinomadura species, suggesting they may have been acquired via horizontal gene transfer. The gene product ForX is a putative avin-dependent monooxygenase, and both sequence homology and predicted structural homology 10 searches indicated the most closely related characterized homologue is aklavinone 12-hydroxylase (RdmE) ( Table 1), a Group A avin-dependent monooxygenase. 11,12 Group A monooxygenases are reported to catalyze hydroxylation or sulfoxidation. Sequence alignments of ForX and RdmE showed an overall identity of 41%, with three noticeable regions of low homology between residues 70-119, 212-245 and 365-393 in the RdmE structure (31HG), corresponding to the substrate binding pocket. Structural homology searches predicted ForY to be a TIM beta/alpha barrel oxidoreductase ( Table 1), homologues of which typically utilize avin cofactors to perform redox chemistry. [13][14][15][16][17] The genes forZ and forAA encode a MarR family transcriptional regulator and an MFS family transporter respectively.

Mutational analysis identies a two-step biosynthetic transformation
Guided by our bioinformatic observations we made individual in-frame deletions in forX and forY using Cas9 mediated gene editing. The resulting mutants failed to produce formicamycins, but this was restored aer ectopic expression of the deleted genes ( Fig. S2 †). Despite producing no formicamycins, S. formicae DforX accumulated the previously isolated fasamycin E (3) (Fig. 2c) which was conrmed by isolation and NMR characterization ( Fig. S3 and S4 †). This is consistent with the hypothesis that ForX catalyzes an oxidation required for formicamycin biosynthesis, and that it functions on halogenated fasamycin substrates. In contrast, S. formicae DforY accumulated six new compounds 17-22 (Fig. 2d). Aer scaling up fermentation (4 L) and solvent extraction, a small amount of each of the congeners 17-21 (5-13 mg) was isolated and subjected to structure determination using 2D NMR and MS. Compound 22 could not be isolated but analytical MS and UV data were consistent with it being a related biosynthetic congener. We rst determined the structure of 18 because high-resolution ESI-MS predicted a molecular formula of C 28 H 22 O 8 Cl 2 indicating one more oxygen atom compared with 3, but with a distinct UV spectrum (l max 248 and 300 nm vs. 249, 289, 355, and 415 nm for 3) indicating a major structural change. 1D and 2D NMR experiments showed a striking low-to high-eld chemical shi for the carbonyl group (C11) in comparison to 3, in addition to the appearance of new methylene and lactone carbonyl moieties. Based on these observations, and other connectivity, we assigned the structure of 18 as a Baeyer-Villiger lactone derivative of 3 (Fig. 1). Further analysis identied 19 as the C5-O-methyl congener of 18, and 17 as lacking a chlorine atom at C2 when compared to 18. The remaining two congeners 20 and 21 were determined to contain further variations in chlorination and O-methylation (Fig. S1 †).
These results are consistent with a two-step ring expansion-ring contraction pathway for the conversion of fasamycins into formicamycins with ForX catalyzing a Baeyer-Villiger reaction on fasamycin substrates, and ForY catalyzing the two-electron reduction of the Baeyer-Villiger intermediates with concomitant ring contraction via hydride addition at C19. To provide support for our hypothesis we attempted to express and isolate both proteins for biochemical analysis but, despite extensive experimentation, we were unable to obtain soluble protein. Undeterred, we instead utilized biomimetic chemical synthesis to investigate the biosynthetic pathway (see below), but rst veried the intermediacy of the isolated lactones through a cross feeding experiment. Exogenous 18 (isolated from S. formicae DforY) was added to growing cultures of the S. formicae DforX biosynthetic mutant and aer nine days incubation LCMS analysis of an ethyl acetate extract of the culture broth indicated that 18 was converted efficiently into the previously isolated formicamycin congeners 5, 8, 9, and 13 plus the O-methylated lactone 19 (Fig. S5 †). This is consistent with 18 being a direct biosynthetic intermediate of the formicamycins and a putative substrate of ForY.

Biomimetic semi-synthesis of the formicamycins
To mimic the ForY catalyzed biosynthetic transformation we subjected an analytical sample ($0.1 mg) of the lactone 19 to sodium borohydride reduction (Fig. 4a). This yielded three new products with a mass increase of 2 Da, plus an additional compound with a mass decrease of 14 Da (Fig. 4b).
Notably, and in contrast to 19, the products 23 and 24 (0.97 : 1 ratio) displayed formicamycin like UV spectra (Fig. 4c). The reaction was then scaled up (using 4 mg of 19) and the products 23 ($0.4 mg) and 24 ($0.2 mg) were puri-ed by HPLC. 2D NMR experiments indicated that these both contained the formicamycin ring structure, consistent with the proposed reduction and ring contraction mechanism (Fig. S59 †). Chemical shi comparisons to biosynthetic congeners, in addition to 1D NOE data, conrmed that 23 corresponds to the natural (10RS,19RS) diastereoisomer and 24 to the (10RS,19SR) diastereoisomer. The product 25 ($0.4 mg) was isolated with 24 (1 : 1 mixture) and NMR analysis clearly showed 25 to be a lactone in which the C10-C19 double bond was reduced but for which rearrangement via ring contraction had not occurred. NMR analysis then showed that 25 corresponds to a single diastereoisomer and the NOESY spectrum showed a strong correlation between H-10 and H-19 indicating the (10RS,19RS) conguration. The NMR and MS data for the remaining À14 Da component is consistent with deoxygenation at C11 as shown for 26 (Fig. 4). The biomimetic reaction was then repeated using lactone congener 17 (5.4 mg) as the starting material and analysis of the products showed an almost identical outcome to that for sodium borohydride reduction of 19 (Fig. S6 †).
To better understand their stereochemical composition, we determined the [a] D 20 values of 23 and 24 to be +68 and À73 respectively indicating that the reductive transformation had occurred with induction of chirality, most probably due to the S a -atropoisomeric axis of the C6-C7 bond deriving from the fasamycin precursors. The xed atropoisomeric axis was conrmed by determining the rotational energy barrier for the C6-C7 bond of fasamycin C (1) using DFT calculations (see ESI †). This yielded a value of DE ¼ 156.1 kJ mol À1 indicating that little to no axial rotation should be expected about this bond, with the frequency of rotation on the order of years. 18 The positive sign of the [a] D 20 for 23 is consistent with all previous naturally occurring formicamycins which have [a] D 20 values in the range +162 to +469 . Unfortunately, we had not isolated the formicamycin congener 23 in our previous work and therefore could not assign the enantiomeric excess of this molecule based on optical rotation comparison to a standard. While the isolated sample of 23 is a mixture of two enantiomers at C10/C19, given the essentially xed atropoisomeric axis at C6-C7, these two molecules should be diastereotopic with respect to each other. However, the 1 H NMR spectrum for this sample did not indicate split chemical shis at any position and the initial HPLC analysis showed a single peak. We thus repeated the chromatographic analysis using the higher resolving power of UPLC, and varied the elution conditions, and were able to resolve the isolated sample of 23 into two peaks of the same mass and with the same UV spectra (Fig. 5) consistent with their diastereotopic relationship. Integration of these two peaks indicated a diastereomeric excess (% de) of 41.3%. The equivalent analysis of isolated 24 indicated a % de of 40.5%. Mixing the isolated samples of 23 and 24 and repeating the UPLC analysis conrmed the presence of 4 peaks, whereas reanalysis of this mixture using the original HPLC conditions showed only 2 peaks, consistent with no change to the sample following purication. UPLC analysis of formicamycins isolated from S. formicae KY5 indicated all congeners were single diastereoisomers.

Conclusions
The two-step pathway described here explains the signicant structural changes that must occur to transform the almost at fasamycin skeleton into the twisted chair-like conguration of the formicamycins. The dearomatization of aromatic systems as a prelude to structural diversication has been described for several natural products including, for example, recent mutational analysis of the agnestins 19 and cryptosporioptides 20 where deletion of genes encoding putative Baeyer-Villigerases led to accumulation of the anticipated aromatic precursors. For the Baeyer-Villiger activity of ForX, two alternative mechanisms can be invoked for reaction between the fasamycin substrate, reduced avin cofactor and oxygen. In the rst pathway (Pathway I) the peroxyavin species acts as an electrophile to catalyze oxygen introduction as a tertiary hydroxyl group at C10 following deprotonation of the adjacent phenol (Fig. 3). Subsequent rearrangement then occurs with bond migration and protonation at C20. Alternatively, (Pathway II) ForX could rst induce tautomerization of ring C, possibly through substrate binding, followed by donation of a proton from the peroxyavin cofactor to the resulting C9 ketone generating an oxonium intermediate. The resulting avoperoxy anion can then act as a nucleophile in a canonical BV-like reaction to generate the observed lactone intermediate. The tautomerization required in the rst step of Pathway II has precedent in the reaction of dehydrorabelomycin and congeners during the biosynthesis of gilvocarcins and jadomycins. 5 However, we favor Pathway I which is reminiscent of that by which a nucleophilic chloride ion is converted into a highly reactive electrophilic species via the intermediacy of hypochlorous acid by avindependent halogenases. 21 This is consistent with our observation that ForX groups phylogenetically with the Group A monooxygenases that typically catalyze hydroxylation. We propose that ForY then acts as a avin-dependent reductase catalyzing two-electron reduction of the intermediate lactone. This reaction most likely proceeds via 1,4-conjugate attack of a hydride ion at C19 to form the corresponding enolate 27. Collapse of the enolate can then occur with formation of the new C9-C10 sigma bond and formation of an epoxide intermediate with subsequent rearrangement to yield the C10-tertiary hydroxyl group with overall ring contraction in a process reminiscent of a Favorskii reaction. This pathway (Fig. 3) is consistent with the side products observed in the sodium borohydride biomimetic reduction, and with the steric requirements for orbital overlap during C9-C10 bond formation. Although rare in nature a Favorskii-type reaction has been demonstrated to occur during biosynthesis of the polyketide natural product enterocin produced by Streptomyces maritimus. 22 This pathway involves oxidative activation of the preceding intermediate via a dual 4-electron oxidation. The reaction is catalyzed by the unique avin dependent monooxygenase EncM which utilizes a avin-N5-oxide, rather than a peroxyavin intermediate, to catalyze substrate oxidation and trigger a Favorskii-type reaction. 23 Oxidation promotes further aldol and heterocylization reactions which are also mediated by EncM. 22,23 Based on the outcome of careful stable isotope labeling experiments, a Favorskii-like reaction has also been proposed to occur during biosynthesis of the fungal polyketide metabolite roussoellatide. 24 The production of a single steric isomer during the biosynthesis of formicamycins and the lack of any conjugate reduction side products as are observed during the sodium borohydride biomimetic reduction of isolated lactone biosynthetic intermediates indicates that ForY must control both the stereochemical progress of this reaction and limit alternative reaction coordinates. As described, the proposed reductive ring contraction catalyzed by ForY is unique in polyketide biosynthesis and future work will seek to understand this novel biochemical process in greater detail. A search of all available microbial genome sequences did not identify any additional close homologues. Further searches using MultiGeneBlast 25 with sequences for the minimal PKS genes plus forX, with and without forY, failed to identify any additional candidate BGCs that might encode pathways that operate as described here suggesting that this pathway is unique to formicamycin biosynthesis.

Materials and methods
For details regarding experimental procedures, microbiology and molecular biology procedures, spectroscopic data, see the ESI. †

Conflicts of interest
The authors declare no competing interest.