New Glycosylated Polyene Macrolides: Refining the Ore from Genome Mining

Abstract

Glycosylated polyene macrolides include effective antifungal agents, such as pimaricin, nystatin, candicidin, and amphotericin B. For the treatment of systemic mycoses, amphotericin B has been described as a gold-standard antibiotic because of its potent activity against a broad spectrum of fungal pathogens, which do not readily become resistant. However, amphotericin B has severe toxic side effects, and the development of safer alternatives remains an important objective. One approach towards obtaining such compounds is to discover new related natural products. Advances in next-generation sequencing have delivered a wealth of microbial genome sequences containing polyene biosynthetic gene clusters. These typically encode a modular polyketide synthase that catalyzes the assembly of the aglycone core, a cytochrome P450 that oxidizes a methyl branch to a carboxyl group, and additional enzymes for synthesis and attachment of a single mycosamine sugar residue. In some cases, further P450s catalyze epoxide formation or hydroxylation within the macrolactone. Bioinformatic analyses have identified over 250 of these clusters. Some are predicted to encode potentially valuable new polyenes that have not been uncovered by traditional screening methods. Recent experimental studies have characterized polyenes with new polyketide backbones, previously unknown late oxygenations, and additional sugar residues that increase water-solubility and reduce hemolytic activity. Here we review these studies and assess how this new knowledge can help to prioritize silent polyene clusters for further investigation. This approach should improve the chances of discovering better antifungal antibiotics.

Keywords: antifungal antibiotics; biosynthetic gene clusters; genome mining; glycosylated polyene macrolide.

Figure 31

Proposed biosynthesis of d and l sugars in Actino. algeriensis and Actino. xanthii. In Actino. algeriensis, the 2,3-DH (MBB4911558.1) is 61% identical to SelSVII, the 3KR (MBB4911557.1) is 54% identical to SelVI, the 4AT (MBB4911553.1) is 61% identical to SpnR, the methylase (MBB4911552.1) is 50% identical to SpnS, the EPIM (MBB4911555.1) is 57% identical to StaE [126], and the 4KR (MBB4911556.1) is 50 % identical to Nbc15 [122]. In Actino. xanthii, the 2,3-DH (WP_075125785.1) is 46% identical to EryBVI, the 3KR (WP_075125784.1) is 59% identical to SelSVI, the 4AT (WP_075125780.1) is 62% identical to SpnR, the methylase (WP_075125779.1) is 52% identical to SpnS, the EPIM (WP_075125782.1) is 54% identical to EryBVII, and the 4KR (WP_075125783.1) is 49% identical to Nbc15 [122].

Figure 1

Structures of polyene macrolides.

Figure 2

Conserved hemiketal structure that forms in polyene macrolides. (a) A hexamodular polyketide synthase protein synthesizes this region of the polyketide chain from malonyl-derived acetate (red arrows) and methylmalonyl-derived propionate (blue arrow) building blocks. (b) Exocyclic carboxyl group formation and glycosylation with mycosamine occur in this region. X represents the primer end of the polyketide chain, and Y represents the carboxy-terminal end.

Figure 3

Structures of aromatic heptaenes. Candicidin D = ascocin A2 = levorin A2; candidicin A1 = ascosin A1 = levorin A1; candicidin A3 = ascocin A3 = levorin A3. Partricin B (not shown) is the same as partricin A, except that it lacks an N-methyl group on the p-aminobenzoyl moiety. Partricin A and gedamycin are identical, and partricin B and vacidin are identical. Candicidins are synthesized by Streptomyces albidoflavus (formerly Streptomyces griseus); ascocins are preoduced by Streptomyces canescus; levorins are made by Actinomyces levoris. Partricins, gedamycin, and vacidin are synthesized by different isolates of Streptomyces aureofaciens. Some carbon atoms are numbered to highlight structural differences between the polyenes.

Figure 4

Conflict between structures determined by NMR spectroscopy and by bioinformatic predictions from aromatic heptaene PKS KR2 sequence motifs.

Figure 5

Structures of ECO-02031, clethramycin, and mediomycin.

Figure 6

Formation of primers for biosynthesis of linear polyene polyols ECO-020301, clethramycin, and mediomycin.

Figure 7

Predicted partial structure of the A. saalfeldensis polyene. The structure-prediction method used is given in the 2017 paper by Sheehan et al. [43].

Figure 8

Predicted partial structure of Amyc. albispora heptaene (see also Supplementary Materials Figure S4). The BGC contains a gene for an AmphL cytochrome P450 homologue that may hydroxylate the polyol chain between C2 and C14. No prediction is made for this modification.

Figure 9

Conversion of chorismate to 3-hydroxybenzoate catalyzed by Hyg5. Homologues of Hyg5 occur in the polyene BGCs of Sacc. flava (WP_093416095.1), Sacc. dendrathemae (WP_145741796.1), Lentz. Waywayandensis (WP_093588251), and Lentz. Xingjiangensis (WP_089951167.1).

Figure 10

Predicted partial structures of heptaenes made by Sacc. dendrathemae and Sacc. flava. The two clusters each contain genes for AmphN homologues that form the exocyclic carboxyl group and two further cytochrome P450s for which functions cannot be predicted. The Sacc. dendrathemae polyene structure was predicted as detailed in Supplementary Materials Figures S6 and S7. The Sacc. flava structure was predicted by Usachova [61].

Figure 11

Ps2 phylotype polyene. This structure is the same as that predicted by Holmes and co-workers [46], except for the geometry of the C26–C27 and C28–C29 double bonds. In the PKS extension modules 3 and 4, DH domains are paired with A-type KR domains, indicating that cis alkenes are formed. The genome used for this structure prediction has the accession number MCIQ00000000.

Figure 12

Structure of meijiemycin.

Figure 13

Predicted partial structure of Amyc. lexingtonensis polyene. Methyl branches are shown at C2 and C44 as predicted by antiSMASH, although the relevant domains AT29 and AT8 have FAAH motifs, such as the methoxymalonate-specific AT24, not YASH motifs, such as methylmalonate-specific ATs. AT17 has YASH, and so module 17 is predicted to introduce the methyl branch at C26.

Figure 14

Kineosporicin/actinospene from A. spheciospongiae.

Figure 15

Biosynthetic pathways for GDP-α-d-perosamine and GDP-α-d-mycosamine.

Figure 16

Structure of 67-121C.

Figure 17

Disaccharide-modified Pseudonocardia polyenes.

Figure 18

Predicted structure for octaene from Cryptosporangium arvum.

Figure 19

Partial structure predicted for Amyc. suaedae tetraene. The BGC contains genes for a AmphN P450 homologue (63% identity) that forms the exocyclic carboxyl group and genes for two further cytochrome P450 enzymes (WP_130475496.1 and WP_130475500.1) homologous to SelL (47% identity) and Lcm10 (47% identity). The modifications catalyzed by these enzymes cannot be predicted. Lcm10 catalyzes epoxide formation in lucensomycin. There are also genes for biosynthesis of additional deoxyhexoses and two further GT genes (see Section 10).

Figure 20

Structures of nystatin A1 and candidin. Modification of nystatin A1 and candidin at C35 with l-digitoxose gives nystatin A3 and candidinin, respectively [76]. Modification of candidin with l-cinerulose gives candidoin. Nystatin analogues modified with l-mycarose have also been identified [98].

Figure 21

Structure of selvamicin.

Figure 22

Biosynthesis of dTDP-β-l-4-O-methyl-digitoxose in selvamicin-producing Pseudonocardia species. It is possible that methylation occurs after transfer of the l-digitoxosyl residue to the macrolactone. The abbreviations used are as follows: 2,3-DH, dTDP-4-keto-6-deoxyhexose 2,3-dehydratase; 3 KR, dTDP-3,4-diketo-2,6-dideoxyhexose 3-ketoreductase; 5 EPIM, dTDP-4-keto-2,6-dideoxyhexose C5 epimerase; 4KR, dTDP-4-keto-2,6-dideoxyhexose 4 ketoreductase.

Figure 23

Proposed pathway for biosynthesis of NDP-α-d-2,6-dideoxyhexose in Sacc. gloriosae. The combination of genes present could give any one of the four possibilities shown in panel (a). NDP-α-d-olivose is predicted in (b), because the 2,3-DH (MBB5070950.1) is 67% identical to SelSVII, the 3KR (MBB5070951.1) is 58% identical to SelSVI, and the 4KR (MBB5070952.1) is 47.6 % identical to LanR (AAD13548) [108,112,113].

Figure 24

Predicted partial structure for Sacc. gloriosae pentaene. The site for glycosylation with d-olivose is predicted to be C27 because the homology between SelSV and MBB5070953.1 is 60% identity. Sacc. gloriosae has a SelL homologue (58% identity) that is likely to hydroxylate the polyol chain. This modification is not shown.

Figure 25

Proposed biosynthesis of dTDP-4-amino-2,3,4,6-tetradeoxy-α-d-glucose in Amyc. suaedae. The enzymes encoded could give unmethylated forms of either NDP-α-d-ossamine or NDP-α-d-forosamine (a). NDP-α-d-forosamine is predicted because the 4-aminotransferase is 52% identical to VinF, which gives the stereochemistry shown in (b). The 2,3-DH (WP_130478887.1) is 62% identical to SelSVII, the 3 KR (WP_165436470.1) is 53% identical to SelSVI, the 3 DEOX (WP_130478889.1) is 70% identical to UrdQ [116], and the 4-AT (WP130479013.1) is 52% identical to VinF [115]. The new abbreviations are as follows: 3 DEOX, NDP-4-keto-2,6-dideoxyhexose 3-deoxygenase; 4-AT, NDP-4-keto-2,3,6-trideoxyhexose 4 aminotransferase.

Figure 26

Proposed pathway for biosynthesis of NDP-N-demethyl-α-d-vicenisamine in Amycolatopsis cihanbeyliensis. The 2,3-DH (WP_141995336.1) is 59% identical to ScaDH1 [118], the 3KR (WP_141995334.1) is 67% identical to VinE [115] and the 4AT (WP_141995335.1) is 66% identical to VinF [115].

Figure 27

Predicted partial structure for Crossiella pentaene.

Figure 28

Proposed pathway for biosynthesis of d-vicenisamine in Crossiella cryophila. The 2,3-DH (WP_185005856.1) is 62% identical to SelSVII, the 3KR (WP_185005854.1) is 66% identical to VinE, the 4AT (WP_185005851.1) is 62 % identical to SpnR, and the methylase (WP_185005852.1) is 47% identical to SpnS [115,120].

Figure 29

Proposed biosynthesis of NDP-α-d-olivose and NDP-didemethyl-α-d-forosamine in Amyc. antarctica. The 2,3-DH (OZM74152.1) is 61% identical to SelSVII, the 3KR (OZM74153.1) is 51% identical to SelVI, the 4KR (OZM74154.1) is 53% identical to Nbc15 [122], the 3DEOX (OZM74156.1) is 74 % identical to IdnS12 [123], and the 4AT (OZM74157.1) is 67% identical to IdnS13 [123]. IdnS12 and IdnS13 function in biosynthesis of dTDP-N-demethylforosamine in Streptomyces ML694-90F3, producer of the macrolactam incednine.

Figure 30

Partial structure predicted for Actino. algeriensis tetraene. The cluster includes genes for homologues of AmphL and Lcm10 cytochrome P450 enzymes, the exact functions of which cannot be predicted.