A Putative Glycosyltransferase Complex is Responsible for the Sugar Diversity of Saquayamycins Isolated from Streptomyces sp. KY 40-1

Abstract

Glycosyltransferases are key enzymes involved in the biosynthesis of valuable natural products providing an excellent drug-tailoring tool. Herein, we report the identification of two cooperative glycosyltransferases from the sqn gene cluster directing the biosynthesis of saquayamycins in Streptomyces sp. KY40-1: SqnG1 and SqnG2. Gene inactivation of sqnG1 leads to 50 folds decrease in saquayamycin production, while inactivation of sqnG2 leads to complete production loss suggesting that SqnG2 acts as dual O- and C-glycosyltransferase. Gene inactivation of a third putative glycosyltransferase-encoding gene, sqnG3, does not majorly affect saquayamycin production suggesting that SqnG3 has no or supportive role in glycosylation. The data indicate that SqnG1 and SqnG2 are solely and possibly cooperatively responsible for the sugar diversity observed in saquayamycins 1–7. This is the first evidence of a proposed glycosyltransferase complex with dual O- and C-glycosyltransferase activity, utilizing NDP-activated d-olivose, l-rhodinose as well as an unusual amino sugar, presumably 3,6-dideoxy-l-idosamine.

Graphical abstract

Glycosylated natural products are abundant in nature. It is estimated that about one fifth of total bacterial natural products are glycosides.¹ Chemically, there are three classes of glycosyl transferases (GTs): O-, C- and N-GTs, which utilize nucleoside-diphosphate (NDP)-activated sugars and catalyze the regio- and stereospecific transfer of the sugars to either an aglycone or to other sugar residues, usually in S_N2-fashion under inversion of configuration, considering the stereochemical attachment of the NDP group versus the resulting glycosidic linkage.^2,3 This transformation often drastically affects the biological activity of the resulting natural products, and therefore GTs are considered powerful tools for synthetic biology approaches in drug design.⁴

Saquayamycins are potent farnesyl transferase inhibitors that belong to the angucycline group of natural products that also contains urdamycins and landomycins, which both show potent antibiotic and anticancer activities.^5,6 The saquayamycins 1–7 (Figure 1) were isolated from Streptomyces KY40-1 and were found to possess cytotoxic activities with saquayamycin B 2 and H 4 displaying the strongest activities against the non-small lung cancer cell line H460.⁷ Saquayamycins 1–7 consist of a polyketide derived benz[a]anthracene core that is decorated with saccharide chains at the C(3)-OH and C-9 positions, with d-olivose always at the C-9 and l-rhodinose always at the C(3)-OH position. There are two routes for the biosynthesis of the aglycone core of angucyclines. The most frequently observed route I involves the cyclization of the nascent polyketide in an angular fashion with the aid of cyclases as observed, e.g., in vineomycin, urdamycin and landomycin biosynthesis, while route II initially biosynthesizes a linear tetracyclic anthracyclinone intermediate that is rearranged into the angucycline core via Baeyer-Villiger monooxygenases (BVMOs) as observed in BE-7585A and PD116198.^8–11. The glycosylation pattern and sugar diversity observed in saquayamycins 1–7 (Figure 1) suggests the presence of GTs with interesting substrate flexibilities. In our quest to enrich the drug diversification toolbox with permissive GTs, we identified the gene cluster encoding the biosynthesis of saquayamycins 1–7 (Figure 1).

Figure 1.

Saquayamycins isolated from Streptomyces KY40-1.

Whole genome sequencing of Streptomyces KY40-1 led to the identification of a putative sqn gene cluster (Figure 2) that shares over 80% identity to the urd minimal PKS genes directing the biosynthesis of urdamycin and a gene organization similar to that of gcn and saq gene clusters directing the biosynthesis of grincamycin and saquayamycin Z, respectively.^12–15 The putative sqn cluster encodes the minimal PKS enzymes (SqnH, I, and SqnJ) as well as post-PKS enzymes (SqnBB, K, L, M, A, C, T and SqnU) and a BVMO SqnF required for the biosynthesis of the angucycline core (Scheme 1). The putative cluster also encodes a dedicated phosphopantetheinyl transferase (SqnCC), eight deoxysugar biosynthesis enzymes (SqnS1–SqnS8), three GTs (SqnG1–G3) and a sugar-modifying dehydrogenase (SqnQ) homologous to AknOx and GcnQ (Table S1).^14,16 To probe the role of the identified cluster in the biosynthesis of saquayamycins 1–7, the sqnCC gene encoding a phosphopantetheinyl transferase (PPTase) was replaced with an apramycin resistance marker through homologous recombination (Figure S1). The resultant mutant, ΔCC does not produce saquayamycins either on solid media (no brown color characteristic of WT was observed, Figure S1) or liquid fermentation (Figures S2). Genetic complementation of ΔCC mutant with pGusCC encoding the expression of SqnCC under ermE promotor restores saquayamycins production proving the role of the identified cluster in biosynthesis (Figure S3). Complementation with blank pGus-ermE plasmid does not rescue saquayamycins production. The presence of SqnF, a BexE homolog in the sqn gene cluster (72% identity), suggests that the biosynthesis of the saquayamycins 1–7 may share an initial cyclization pattern similar to that observed in BE-7585A, where an anthracycline intermediate is initially formed and later rearranged into the angucycline core by the action of BexE.¹⁷ To test this hypothesis and to study the cyclization pattern of the polyketide-precursor of saquayamycins 1–7, sodium [1,2-¹³C₂] acetate was fed to the WT Streptomyces KY40-1 strain followed by isolation of ¹³C-enriched 2. The ¹³C-NMR spectrum of 2 (Figure S4) showed 18 signals where a doublet is flanking the natural abundance signal, and one enriched singlet signal at δ 50 ppm corresponding to C-2. This indicates the incorporation of 9 intact acetate units, which was further confirmed by the 2D-INADEQUATE spectrum of 2 (Figure S5) revealing the doublet-pairing pattern shown in Scheme 1. This is in agreement with a biosynthetic pathway of the benz[a]anthracene core of saquayamycins following the typical direct route I of angucycline cyclization that was first discovered in context with the vineomycin biosynthesis.⁸ Route I is followed in saquayamycins biosynthesis despite the presence of the BexE homolog SqnF, suggesting that the major factor deciding either route I or II for polyketide cyclization in angucyclines is likely determined by the shape of the binding pocket for the nascent polyketide formed by the minimal PKS enzymes. Further study of this enzyme is underway.

Figure 2.

Organization of the sqn gene cluster.

Scheme 1.

¹³C-labelling pattern observed in saquayamycin B (2) and the putative biosynthetic pathway of the saquayamycins emphasizing the role of a proposed SqnG2/SqnG1 complex.

BLAST analysis of the primary amino acid sequence of the putative GTs SqnG1, SqnG2 and SqnG3 suggests that SqnG1 and SqnG2 are O-GTs homologous to SaqGT2 and SaqGT4, while SqnG3 is homologous to C-GTs, such as UrdGT2.^15,18 To elucidate the role of the sqnG1, sqnG2, and sqnG3 encoded enzymes in the biosynthesis of saquayamycins 1–7, the aforementioned genes were replaced by the apramycin resistance marker via homologous recombination, to generate three mutant strains, Streptomyces KY40-1ΔG1, ΔG2 and ΔG3, respectively.

The LC-UV/MS profile of the acetone extract of the mutant ΔG1 shows one major peak at retention time 16.5 minutes that is also present in the WT – although in very low intensity (Figure 3) – in addition to trace amounts of two other peaks of retention times/UV profiles and masses matching those observed for 1 and 2 (retention times 20.6 and 21.5 minutes, respectively), the latter confirmed as 2 by ¹H-NMR. The isolation and purification of the major peak accumulating in the ΔG1 extract revealed that it was a mixture of two compounds (A and B), subsequently resolved by further HPLC purification. Compound A, with a molecular ion [M+H]⁺ of m/z 323.0761, and compound B, with [M+H]⁺ of m/z 339.0703, were identified through 1D- and 2D-NMR data as tetrangomycin 8 and rabelomycin 9, respectively (Scheme 1, Tables S2–S3, Figures S6–S17). Both 8 and 9 are common shunt products of angucycline biosynthesis resulting from dehydration of UWM6 10, a well-known intermediate (Scheme 1).

Figure 3.

Comparison between the HPLC profile of WT and the mutants; ΔG1, ΔG2 and ΔG3.

The isolation of 2 from the extract of the ΔG1 mutant suggests that SqnG1 is not absolutely required for the glycosylation of putative intermediate 11, however, its inactivation reduced the production level of 2 by about 50 folds (0.16 mg/L of pure 2 isolated from the ΔG1 mutant versus 8 mg/L isolated from the WT). To ensure that the trace amounts of 1 and 2 observed in the ΔG1 mutant were not caused by contamination with the WT, all fermentations were repeated in presence of 50 µg/ml apramycin in which only the mutant can grow. This did not change the results of the earlier findings, thus the possibility of a contamination by the WT was refuted. Genetic complementation of ΔG1 mutant with pGusG1, a derivative of pGus-ermE¹⁹ encoding the expression of SqnG1 under ermE promotor restored saquayamycins 1–7 production with slight accumulation of saquayamycin B1 and saprolmycin C (Figure S23) while complementation with blank pGus-ermE plasmid does not have any effect on production. These data suggest an O-GT activity of SqnG1 (Fig. S23) that can be substituted by either SqnG2 or SqnG3. The presence of SqnG1 seems to be essential to achieve normal production level.

The LC-UV/MS profile of the second mutant, ΔG2, in which the GT encoded by sqnG2 had been disrupted, led to the accumulation of only 8 and 9 (Figure 3), and no glycosylated saquayamycins were detected whatsoever. Again, genetic complementation of ΔG2 mutant with pGusG2 encoding the expression of SqnG2 under effect of ermE promotor restores saquayamycins 1–7 production with slight accumulation of 3 and saprolmycin C (Fig. S23) also suggesting an O-GT activity and confirming that SqnG2 is essential for the initial glycosylation, presumably of putative intermediate 11.

Unexpectedly, disruption of sqnG3 did not have significant effect on 1–7 production (5 mg/L of 2 was isolated from ΔG3 mutant vs. 8 mg/L from WT), except for the slight accumulation of 8 and 9 (Figure 3) which could be attributed to polar effect on downstream genes resulting from gene replacement as the accumulation of these metabolites was observed in all ΔG1– ΔG3 mutants. To confirm the identity of the peak with retention time 21.5 minutes in ΔG3 mutant as 2 and to eliminate any possibility that this peak could be a different saquayamycin with different glycosylation pattern or stereochemistry, we sought its isolation and structure elucidation relying on 1D and 2D-NMR data (Table S4) (Figures S18–22) which confirmed its identity to be 2. These data confirm that the ΔG3 mutant, in which only the sqnG1 and sqnG2 gene products are functional, is fully capable of biosynthesizing 2 and other saquayamycins at a level comparable to that of the WT where, SqnG2 is dual O- and C-GT whose efficiency is greatly enhanced by the presence of SqnG1.

Surprisingly, genetic complementation of ΔG3 mutant with pGusG3 encoding the expression of SqnG3 under ermE promoter lead to significant reduction in the accumulation of shunt metabolites 8 and 9 relative to ΔG3 mutant and almost complete restoration of WT LC profile (Figure S24). This indicates a possible role of SqnG3 in glycosylation however, no accumulation of aquayamycin 12 (scheme 1) or other short chain saquayamycins was observed as a result of complementation to help assign the function of SqnG3 in glycosylation of 11 (Figure S24). There are two possible hypotheses to explain these results: either that SqnG3 is indeed a C-GT (based on annotation) but its function can be replaced by a remarkable substrate flexibility of SqnG2/G1 or most likely that SqnG3 acts only as a chaperone to assist the glycosylating activity of SqnG2/G1 by stabilizing intermediate 11. Multiple attempts to generate sqnG1/G2 double mutant to unambiguously assign the function of SqnG3 in biosynthesis of saquayamycins using the same strategy employed earlier have unfortunately failed. Attempts to use pCRISPomyces-2 plasmid²⁰ which relies on CRISPR/Cas9-mediated genome editing to create the sqnG1/G2 double mutant proved to be toxic to our strain therefore a sqnG1/G2 double mutant could not be generated. However, it is clear from our data that SqnG2 together with SqnG1 are fully capable to act as both O- and C-GTs regardless of the presence or absence of SqnG3.

Collectively, our results strongly suggest a complex glycosylation scheme of saquayamycins in which all GTs appear to be required in order to achieve efficient glycosylation. In this scheme, it is evident that SqnG2 can be solely responsible for both C- and O-glycosylation of possibly intermediate 11 to produce 1–7 with strong dependence on SqnG1 to achieve normal production levels, and possibly SqnG3 to stabilize 11 (shunt metabolites 8 and 9 accumulating in all mutants were significantly reduced with complementation with SqnG3, Figure S24). This hypothesis is corroborated by two facts: The first is the lack of accumulation of aquayamycin 12 or of any other short chain saquayamycin as a result of a single GT knockout in any of the mutants generated throughout our study. Accumulation of saquayamycin B1, saprolmycin C and 3 (Fig. S23) was observed only after overexpression of either SqnG1 or SqnG2 under the ermE strong promoter with no clear role for either in the sequence of glycosylation events. The second fact supporting our hypothesis is that mutant ΔG3 in which both SqnG1 and SqnG2 are functional show near normal production levels except for the accumulation of 8 and 9 whereas absence of SqnG1 in ΔG1 mutant greatly impairs production. We are proposing that SqnG2 and SqnG1 form a complex that enhances the ability of SqnG2 to catalyze both O- and C-glycosylation (Scheme 1) and that SqnG3 either further enhances the efficacy of the aforementioned complex or its role as C-GT can be readily substituted by SqnG2/G1.

Co-dependence of GTs on other proteins for their function have been reported before. The GT complexes reported so far consist of a GT with either a second GT, as in the case of the proposed MtmGI/MtmGII complex; or a GT is dependent on other auxiliary proteins, such as P450 oxygenases or proteins of previously unknown function, as observed in the DesVII/ DesVIII, the AknS/AknT, TylM2/TylM3, MycB/MydC, and EryCIII/EryCII GT complexes involved in the biosynthesis of pikromycin, aclacinomycin A, tylosin, mycinamicin, and erythromycin, respectively.^21–26 Recently, MtmGIV was shown to be co-dependent on C-methyltransferase/ketoreductase MtmC to achieve its control of transferred sugar type and sugar positioning at the acceptor substrate.²⁷ However, in all of these complexes, the GTs act - at least initially - as O-glycosyltransferases, transferring a first sugar unit to a specific OH group of the corresponding aglycone. The MtmGI/MtmGII complex is proposed to catalyze the stepwise transfer of two d-olivose units, the first to the C(8)-OH group of premithramycin A3, and display high substrate specificity.²⁶ Bi-functional or iteratively acting GTs such as AknK, LanGT1, LanGT4 and MtmGIV involved in the biosynthesis of aclacinomycin A, landomycin A and mithramycin, respectively, are all O-GTs that accept either only NDP-activated d-sugars or only NDP-l-sugars as donor substrates.^27–29 ElmGT, a GT involved in the biosynthesis of elloramycin, and UrdGT2, involved in the biosynthesis of urdamycin, are the only GTs reported so far that showed broad donor substrate flexibility to both NDP-activated d- and l-sugars.^30,31 However, ElmGT is a strict-regioselectively operating O-GT, while UrdGT2 is a C-GT with some O-GT side activity, also highly regioselective in affecting the position ortho to an OH-group of the corresponding acceptor substrate, e.g., ortho to C(8)-OH of the angucyclinone framework.³² Thus, our findings described here present the first GT complex with dual O- and Cglycosyltransferase functionality, affecting initially two different positions of the polyketide derived acceptor substrate core as well as some of its initially added sugar moieties later. Moreover, the proposed SqnG2/SqnG1 complex also displays broad donor substrate flexibility, and can transfer both d- and l-sugars, and interestingly also an amino sugar, namely a precursor of the rare 2-aminodeoxysugar l-rednose, likely 3,6-dideoxy-l-idosamine (Scheme 1). This makes the SqnG2/G1 complex a valuable candidate for natural product drug diversification via synthetic biology approaches. In vitro biochemical characterization of the proposed SqnG2/SqnG1 complex as well as SqnG3 is currently under investigation in our laboratory. Finally, this initial evaluation of the sqn gene cluster also revealed that alignment studies, here regarding the putative framework modifying BVMO gene sqnF and possibly the sqnG3 gene can be highly misleading, and biosynthetic pathways solely drawn from bioinformatics analysis, cannot be trusted to provide biosynthetic or mechanistic insights unless sufficient experimental data is presented.

METHODS

For detailed methods, please refer to Supporting Information.