Construction of an octosyl acid backbone catalyzed by a radical S-adenosylmethionine enzyme and a phosphatase in the biosynthesis of high-carbon sugar nucleoside antibiotics

This work provides, for the first time, significant in vitro evidence for the biosynthetic origins of octosyl acid through free radical and dephosphorylation enzymatic reactions.


Introduction
Nucleoside antibiotics constitute more than 200 structurally unique natural products with a wide range of bioactivities, such as antibacterial, antifungal, antitumor, and antiviral activities. 1 This family of antibiotics is usually biosynthesized by sequential modications of the nucleoside core scaffold. 1 The unique bicyclic uronic acid nucleosides, which include ezomycin, 2a malayamycin, 2b and octosyl acid (OA), 2c are structurally characterized by OA, an unusual 8-carbon furanosyl nucleoside skeleton (Fig. 1). The formation of the high carbon sugar (8C's) nucleoside scaffold, OA, has been proposed as a key step in the biosynthesis of polyoxin and nikkomycin and as the precursor to the second key intermediate, aminohexuronic acid (AHA, uracil polyoxin C) ( Fig. 1 and Scheme 1A). 1e, 3,4 Polyoxin, a group of structurally-related peptidyl nucleoside antibiotics, is produced by Streptomyces cacaoi var. asoensis and Streptomyces aureochromogenes. 4,5 The structure of polyoxin resembles UDP-N-acetylglucosamine, a building block for fungal chitin biosynthesis, and it thus acts as a potent chitin synthetase inhibitor by targeting fungal cell wall biosynthesis. 6a Polyoxin has been extensively used as an important fungicide in agriculture and as a potential lead for mammalian antifungal drug development. 6b Structurally, polyoxin consists of three building blocks including a nucleoside skeleton and two nonproteinogenic amino acids involving polyoximic acid and carbamoylpolyoxamic acid. 1a Besides the OA skeleton derived natural products including polyoxin, nikkomycin, 7a malayamycin, 2b and ezomycin, 2a only a few high-carbon (>6C's) sugar nucleoside antibiotics have been identied, such as liposidomycin, 7b caprazamycin, 7c and tunicamycin 7d (ESI, Fig. S1 †). However, little is known about how the high-carbon sugar scaffolds are biosynthesized, except for 5 0 -C-glycyluridine, which is a key intermediate in the biosynthesis of liposidomycin and has been characterized and shown to be generated from uridine-5 0 -monophosphate (UMP), consecutively catalyzed by LipL and LipK, a non-heme iron(II)dependent dioxygenase and a PLP-dependent transaldolase, respectively (Scheme 1B). 8 In the biosynthesis of tunicamycin, TunB, a putative radical SAM-dependent protein with its partner TunM, is predicted to catalyze the formation of the C-C bond between the C-5 of the uridyl moiety and the C-6 of the central sugar, exo-glycal, which is enzymatically produced by TunA and TunF in a recent revised biosynthesis pathway (Scheme 1C). 9 Since the identication of OA as the metabolite from the polyoxin producer, Streptomyces cacaoi var. asoensis, in the 1970s, 2c discoveries of this unique family of natural products or derivatives have been continuously reported. 1 To our knowledge, how OA is constructed by nature has remained elusive for almost half a century. Herein, we report the genetic evidence and in vitro functional characterization of two enzymes, PolH and PolJ, involved in the early stage of polyoxin biosynthesis. On the basis of the detailed investigation and biochemical analysis of these two consecutive enzymatic reactions, we revised the pathway by identifying OA as a key intermediate.

In vitro characterization of PolJ and PolH
We investigated the proposed pathway with in vitro characterization of the reactions catalyzed by PolH and PolJ. Based on bioinformatic analysis, PolJ (ESI, Fig. S10 †) and PolH (ESI, Fig. S11 †) had been assigned as a phosphatase and a radical SAM protein, respectively. 4 The polJ and polH genes were individually cloned into pET28a with N-terminal His6 tags followed by protein over-expression in E. coli. PolJ was puried to near homogeneity with partial protein contamination due to issues with poor solubility and weak expression (ESI, Fig. S12 †). PolH, Scheme 1 Proposed or characterized biosynthetic pathway for building the high-carbon sugar nucleoside skeleton. with the characteristic C 225 XXXC 229 XXC 232 radical SAM protein motif, 12 was puried anaerobically and contained a low concentration of iron/sulfur clusters, as evident from the UV-Vis spectrum ( Fig. 2A and B). The PolH was reconstituted with enhanced iron/sulfur content and contained one [4Fe-4S] cluster per protein based on the iron 13a and sulfur 13b analyses, which indicated that the reconstituted PolH has 3.9 iron ions and 3.2 sulfur atoms per PolH monomer compared to 1.1 iron ion and 1.0 sulfur atom per as-isolated PolH (ESI, Fig. S13 †). Electron paramagnetic resonance (EPR) characterization of the as-isolated PolH displayed a typical signal with a g value of 2.00 for a [3Fe-4S] + cluster (ESI, Fig. S14A †). Aer reconstitution and reduction, a typical [4Fe-4S] + cluster feature was observed with an axial EPR signal with g k ¼ 2.03, and g t ¼ 1.92 (Fig. 2C). Moreover, the same g-values were also determined with less intensity from the as-isolated PolH in the presence of a dithionite reductant (ESI, Fig. S14B †). These EPR signals are in agreement with those of other radical-AdoMet [4Fe-4S] + clusters. 14

The biochemical activity of PolH and PolJ
Initially, we tried to optimize a variety of reaction conditions for PolH using compound X as a substrate, but unfortunately, our test of the PolH activity yielded negative results, suggesting that it's not a correct prime substrate for PolH. It is probable that compound X is originally synthesized from the very substrate of PolH via a pathway independent of polyoxin biosynthesis. We then were motivated to reinvestigate the initial steps leading to the nucleoside skeleton of polyoxin, which are proposed to be governed by PolJ and PolH in our previous report. 4 As a consequence, the key substrate 3 0 -EUMP was prepared chemo-enzymatically by PolA's homolog, NikO, and puried by HPLC (ESI, Fig. S15 and Methods S4 †). 3 0 -EUMP was rst incubated with PolJ, and the reaction mixtures were analyzed by LC-MS/MS. 3 0 -Enolpyruvyl-uridine (3 0 -EU) was determined by tandem MS analysis as the actual dephosphorylation product instead of the proposed oxidative dephosphorylation product, OFUA (ESI, Fig. S16 †). Chemically, the C-5 0 of 3 0 -EU(MP) has to be activated as well as dephosphorylation for conversion to OFUA. Enzymatically, the activation of the C-5 0 of uridine analogues could be achieved either by an oxidase, like LipL, 8a with the formation of the aldehyde in the biosynthesis of liposidomycin, or through a radical mechanism, like TunB, 9 predicted in the biosynthesis of tunicamycin (Scheme 1B and C). Considering the observed net dephosphorylation function and bioinformatic analysis of PolJ as a putative phosphatase, 4 the proposed transformation of 3 0 -EUMP to OFUA catalyzed by PolJ would require another enzyme. To test this hypothesis, a PolJ-PolH coupled enzymatic assay was performed either to initiate the reaction with the radical or to push the reaction forward by consumption of the potential intermediates generated by PolJ. The reconstituted PolH plus PolJ were incubated anaerobically with 3 0 -EUMP with the presence of SAM and dithionite as the general radical SAM enzymatic assay conditions (ESI, Methods S7 and S9 †). 14 The reaction mixture was analyzed by LC-MS/MS, and compared with the purely PolJ catalyzed reaction; the substrate 3 0 -EUMP with a retention time of 10.2 min was consumed completely and the dephosphorylated 3 0 -EU was observed at t R ¼ 4.05 min. In addition, two new signals at t R ¼ 9.89 min and t R ¼ 3.67 min with the same molecular mass of 393.03 (designated as an isomer of 3 0 -EUMP) and 313.06 (designated as an isomer of 3 0 -EU) as 3 0 -EUMP and 3 0 -UM, respectively, were detected (ESI,  (78.957, 96.967, 111.018, 295.057, 296.062, 350.028, 351.033, 393.034, and 394.039 (from left to right)).
Scheme 2 The proposed reaction mechanism catalyzed by PolH.  S17A and B †). Further comparative tandem MS/MS analysis of these two peaks indicated that the two isomers shared a common uridine motif as 3 0 -EUMP and 3 0 -EU with the same ion peaks at m/z 111.02 (ESI, Fig. S17C and D †). By omitting PolJ in the coupling assay, only the product, the isomer of 3 0 -EUMP, was detected in the presence of PolH as shown by the LC-MS/MS spectrum. From the above studies, it's most likely that 3 0 -EUMP was converted by PolH to its isomer.
To elucidate the isomer of 3 0 -EUMP catalyzed by PolH, a HPLC assay with different control experiments was performed (ESI, Methods S9 †). The new product was only generated in the presence of both SAM and sodium dithionite in comparison with the control and standard traces ( Fig. 3A1-8, ESI and Fig. S18 †). The product was further puried by HPLC and fully characterized by HR-MS and NMR ( Fig. 3B and C, ESI and Fig. S18-S23 †). The m/z 393.03 ratio of the puried product determined by ESI-MS/MS analysis was the same as 3 0 -EUMP with a common ion peak at m/z 111.02, but different ion peaks at m/z 295.05 to 305.02 of 3 0 -EUMP, which may be due to a cyclization reaction of a pyruvate side chain to the nucleoside sugar group (Fig. 3B, ESI and Fig. S18B †). There are 10 and 12 signals in the 1 H and 13 C-NMR spectrum, respectively, except for the extra signals from glycerol contamination (ESI, Fig. S19 and S20 †). The majority of the peaks in the 1 H-NMR spectrum were highly consistent with the chemically synthesized standard octosyl acid A (5-carboxylic octosyl acid). 15 1 H-1 H COSY Scheme 3 The proposed biosynthetic pathway for polyoxin. R ¼ H, CH 3 , CH 2 OH or COOH. Fig. 5 Target-directed genome mining of the potential nucleoside antibiotics containing AHA using PolH and PolJ as probes. The genome sequence data was obtained from GenBank, and the related ORFs associated with aminohexuronic acid (uracil polyoxin C) are linked together and probably encode potential nucleoside antibiotics related to polyoxin and nikkomycin.

The PolH reaction follows a radical reaction mechanism
During the cyclization reaction catalyzed by PolH, the amount of OAP and 5 0 -dA was increased simultaneously based on the time course monitoring experiments, indicating that SAM was being used as a co-substrate (ESI, Fig. S25 †). Therefore, 5 0 -dA was a co-product instead of a catalytic source for the regeneration of 5 0 -dA radicals like the well characterized lysine-2,3-aminomutase (LAM). 16 Presumably, the substrate radical intermediate generated from 3 0 -EUMP abstracted a hydrogen atom from an amino acid residue with solvent exchangeable ability, as is the case with NeoN as opposed to 5 0 -dA in the biosynthesis of neomycin. 17 To verify this hypothesis, the PolH assay was carried out in a 60% deuterium oxide buffer with the clear detection of m/z 396.05 against m/z 395.05 by LC-MS/MS for the unlabeled OAP (Fig. 4, ESI and Fig. S26 †).
Given the above results, the PolH reaction mechanism is described in Scheme 2 which aligns well with the previous isotopic feeding experiments about the carbon sources of C-5 0 and C-6 0 . 3, 10 We thus propose that the 5 0 -dA radical, generated from SAM and catalyzed by the reduced [4Fe-4S] + , abstracts the hydrogen at the C-5 0 position followed by cyclization to form a C-7 0 radical intermediate. The substrate based radical intermediate continually abstracts a hydrogen atom from the protein to give OAP. Additional detailed enzymatic mechanism studies including on which residue is the hydrogen atom source are currently being pursued in our laboratory.
PolH and PolJ are capable of being utilized as potential probes for the discovery of new natural products Although the OA moiety is a common structure shared by several high-carbon sugar nucleoside natural products including polyoxin and nikkomycin, the puzzle of its biosynthesis remained unsolved for over half a century. With the functional roles of PolH and PolJ distinctly dissected in this study, we could utilize both of them as probes for the targetdirected genome mining of potential novel OA-or AHA-containing compounds with antifungal bioactivity, and we are gratied to have identied several gene clusters that also contain the conserved 6 genes including polH and polJ, suggesting that the OA, as an unusual building block for polyoxin, is diversely distributed in nature (Fig. 4, ESI and Table S3 †). Furthermore, using polH and polJ as probes not only offers great potential for natural product discovery, in particular for those containing the OA moiety, but also provides new directions for future natural product development.

Conclusions
In summary, we have dissected the initial two crucial enzymatic steps by PolH and PolJ leading to OA biosynthesis by a combination of in vivo and in vitro investigations. A revised biosynthetic pathway for polyoxin was proposed (Scheme 3), and the second key intermediate, AHA, is most likely generated from OA with C6 0 -C7 0 bond cleavage to give ketohexuronic acid (KHA) by two putative a-KG dependent dioxygenases, PolD and PolK, followed by a PLP dependent aminotransferase, PolI, whose characterized homolog is NikK. 18 Our studies revealed that construction of the OA backbone highlights a free radical strategy to rst form OAP followed by dephosphorylation which agrees well with the labelling feeding experiments. More importantly, we unveiled several potential gene clusters coding for OA-or AHA-containing high-carbon sugar nucleoside antibiotics biosynthesized from diverse strains, by target-directed genome mining using PolH and PolJ as probes (Fig. 5, ESI and Table S3 †). We anticipate that the deciphering of OA biosynthetic puzzles will contribute to polyoxin pathway engineering, and also enrich the toolbox for novel enzymatic mechanisms.