PtmC catalyzes the final step of thioplatensimycin, thioplatencin, and thioplatensilin biosynthesis and expands the scope of arylamine N-acetyltransferases

Abstract

The arylamine N-acetyltransferase (NAT) family of enzymes are important for their many roles in xenobiotic detoxification in bacteria and humans. However, very little is known about their roles outside of detoxification or their specificities for acyl donors larger than acetyl-CoA. Herein, we report the detailed study of PtmC, an unusual NAT homologue encoded in the biosynthetic gene cluster for thioplatensimycin, thioplatencin, and a newly reported scaffold, thioplatensilin, thioacid-containing diterpenoids and highly potent inhibitors of bacterial and mammalian fatty acid synthases. As the final enzyme of the pathway, PtmC is responsible for the selection of a thioacid arylamine over its cognate carboxylic acid and coupling to at least three large, 17-carbon ketolide-CoA substrates. Therefore, this study uses a combined approach of enzymology and molecular modeling to reveal how PtmC has evolved from the canonical NAT scaffold into a key part of a natural combinatorial biosynthetic pathway. Additionally, genome mining has revealed the presence of other related NATs located within natural product biosynthetic gene clusters. Thus, findings from this study are expected to expand our knowledge of how enzymes evolve for expanded substrate diversity and enable additional predictions about the activities of NATs involved in natural product biosynthesis and xenobiotic detoxification.

Graphical Abstract

INTRODUCTION

Platensimycin (PTM, 1) and platencin (PTN, 2), originally isolated from Streptomyces platensis MA7327 and MA7339 by an antisense differential sensitivity whole-cell diffusion assay,^1–2 are selective and potent inhibitors of bacterial and mammalian fatty acid synthases. They have been pursued extensively as promising drug leads for both antibacterial and antidiabetic therapies.^1–3 Structurally, 1 and 2 are comprised of two distinct moieties, a polar 3-amino-2,4-dihydroxybenzoic acid and a lipophilic ketolide of diterpenoid origin, linked by a non-peptidyl amide bond (Figure 1A). While the ptm biosynthetic gene cluster (BGC) from S. platensis MA7327 encodes the biosynthesis of both 1 and 2, the ptn BGC from S. platensis MA7339 encodes the biosynthesis of 2 only (Figure 1B).^1–3

Figure 1.

PtmC catalyzes the final step of thioPTM (4), thioPTN (5), and thioPTL (6) biosynthesis, exhibiting substrate specificities unprecedented for NATs. (A) The structures of 4–6, as well as their non-enzymatic hydrolysis congeners PTM (1), PTN (2), and PTL (3), featuring the varying arylamine (red) and ketolide (blue) moieties that are linked by a non-peptidyl amide bond. (B) Genetic organizations of the ptm BGC, encoding 4–6 biosynthesis, and the ptn BGC, encoding 5 biosynthesis, from S. platensis strains. The ptm BGC differs from the ptn BGC by the presence of a five-gene “PTM cassette” within the ptm BGC. (C) A convergent biosynthetic pathway for 4–6, featuring the PtmC-catalyzed amidation of the arylamine thioacid (9) with the corresponding ketolide-CoAs (16–18). Enzymes catalyzing the biosynthesis of the arylamines (7–9) (PtmB1-3, PtmA3, PtmU4), the partition of ent-copalyl diphosphate (ent-CPP) to the ent-atiserene scaffold (PtmT1), and to the ent-kauranol and ent-beyerene scaffolds (PtmT3), respectively, and the final step coupling 9 with 16–18 (PtmC), are highlighted (also see Figure S1 for the complete pathway for 4–6 biosynthesis).

Six additional Streptomyces platensis species were subsequenctly discovered by genome mining for bacterial diterpene synthase genes, all of which were founded to harbor ptm BGCs encoding both 1 and 2 biosynthesis.⁴ Significantly, inactivation of the pathway-specific negative regulator ptmR1 yielded several overproducers of 1 and 2, which, under optimized fermentation conditions, were capable of producing 1 at titers of >2.0 g L⁻¹.⁴ This overproducing phenotype enabled the discovery of thioplatensimycin (thioPTM) (4) and thioplatencin (thioPTN) (5) as the nascent products encoded by the ptm BGC, with 1 and 2 established as the non-enzymatic hydrolysis products of 4 and 5, respectively, during the fermentation and isolation processes (Figure 1A).⁵

One of the overproducers, SB12029, engineered from Streptomyces platensis CB00739, has been developed into a model system to study the biosynthesis of 1, 2, 4, and 5.⁶ On the basis of in vivo and in vitro studies, a convergent biosynthetic pathway has been proposed, featuring (i) one set of enzymes for the 3-amino-2,4-dihydroxybenzoic thioacid (ADHBSH) (9) moiety, (ii) another set of enzymes for the diterpene-derived penultimate ketolide intermediates platensicyl-CoA (16) and platencinyl-CoA (18), and (iii) PtmC for the final step coupling 9 with 16 and 18 to afford 4 and 5, respectively (Figures 1C, S1).^7–8

PtmC belongs to the arylamine N-acetyltransferase (NAT) family of enzymes, members of which are widespread across all branches of life.⁹ The NAT family typically catalyzes the N- and O-acetylation of arylamines, arylhydrazines, arylhydroxylamines, and related metabolites (Figure S2).^10–11 These enzymes have been well studied for their role in xenobiotic detoxification in bacterial pathogens, especially regarding metabolism of the anti-tuberculosis drug isoniazid, and in humans, including as a cancer target.^12–13 While NATs often have some promiscuity towards their arylamine substrates, they are highly specific for the acetyl donor, i.e. acetyl-CoA, with only rare exceptions.^14–17 Thus, all NATs share a well-conserved structural fold that enables the transient formation of an acetyl-thioester species at the catalytic cysteine residue before transfering the acetyl group to the arylamine substrates (Figure S3).^{10–11, 18} To date, no NAT has been characterized from natural product biosynthetic pathways. Furthermore, with only five thioacid-containing natural products known to date (Figure S4),^{5, 19–21} little is known about their biosynthesis; no enzyme that uses a thioacid substrate has been characterized.⁸ PtmC, predicted to catalyze the amidation of 9 with 16 and 18 in the biosynthesis of 4 and 5, respectively, therefore represents an outstanding opportunity to investigate the roles NATs may play in natural product biosynthesis, providing insights into the molecular mechanism of how NATs have evolved to accommodate ketolide-CoAs, much larger than acetyl-CoA, as substrates in catalysis.

Herein, we first report the fortuitous discovery of platensilin (PTL) (3) and thioplatensilin (thioPTL) (6) as the third scaffold encoded by the ptm BGC, highlighting the ingenuity of nature in generating natural product diversity. We then report the in vitro characterization of PtmC as an NAT that specifies large ketolide-CoAs and prefers an arylamine thioacid, significantly expanding the substrate scope of NATs, and molecular modeling of PtmC that reveals a conserved NAT architecture with distinct features, laying out a molecular basis for its unprecedented substrate specificity. We finally report a bioinformatics analysis that shows a wide distribution of PtmC homologues, demonstrating outstanding prospects for future investigation of the roles of NATs in natural product biosynthesis and genome mining for natural product discovery.

RESULTS AND DISCUSSION

Discovery of PTL (3) and thioPTL (6) reveals an extraordinary pathway plasticity for the PTM-PTN-PTL biosynthetic machinery.

Previously, we inactivated ptmC in vivo, and the resultant ΔptmC mutant abolished production of 1, 2, 4, and 5, and instead accumulated 3-amino-2,4-dihydroxybenzoic acid (ADHBA) (8), platensic acid (13), and platencinic acid (15), a phenotype that supported the functional assignment of PtmC to catalyze the last step of 4 and 5 biosynthesis⁷ (Figure 1C). We carried out large scale fermentation of SB12035 (i.e., the ΔptmR1/ΔptmC mutant of S. sp. CB00739) to isolate sufficient quantities of ketolide acids 13 and 15 for semisynthesis of 16 and 18 as CoA-linked substrates for in vitro characterization of PtmC (SI Materials and Methods). In our effort to isolate 13 and 15, we noticed that 15 was co-eluted with an additional metabolite (14) with an identical mass of 274 (Figure 2A), the identical masses and retention time of which led to the speculation of an additional ketolide acid with a similar structure like 15. Thus, SB12029 (i.e., the ΔptmR1 mutant of S. sp. CB00739), the engineered overproducer of 1 and 2,⁶ was re-fermented (SI Materials and Methods). Close examination of the SB12029 metabolite profile indeed revealed two distinct peaks, 2 and 3, with identical masses of 424 (Figure 2B), and both peaks, corresponding to 2 and 3, were abolished in SB12035 (Figure 2A). A large-scale fermentation of SB12029 allowed the isolation, and thereby fortuitous discovery, of 3, together with the nascent thioacid congener 6 (SI Materials and Methods). The structures of 3 and 6 as the ent-beyerene-derived analogues, named platensilin (PTL) (3) and thioplatensilin (thioPTL) (6), respectively (Figure 1A), together with 14 isolated from SB12035, named platensilic acid (Figure 1C), were fully characterized by 1D and 2D NMR spectroscopy (Figures S4–S17 and Table S3).

Figure 2.

Discovery of PTL (3) and thioPTL (6) as the third diterpenoid scaffold encoded by the ptm BGC. (A) HPLC analysis of production of platensilic acid (14), in addition to platensic acid (13) and platenicinic acid (15), by the ΔptmR1/ΔptmC mutant strain SB12035. (B) HPLC analysis of production of 3 and 6, in addition to PTM (1), thioPTM (4), PTN (2), and thioPTN (5) by the ΔptmR1 mutant overproducer SB12029. (C) HPLC analysis of production of 2 and 5 only by the ΔptmR1/ΔptmT3 mutant strain SB12008. Panel I, HPLC analysis with UV detection at 254 nm; Panel II, LC-MS analysis with the extracted ion chromatogram (EIC) for m/z at 424 for 2 and 3.

Discovery of the ent-beyerene-derived 6 (and 3) as the third scaffold reveals extraordinary pathway plasticity for the PTM-PTN-PTL biosynthetic machinery, but the origin of this scaffold remained unknown. The two diterpene synthases within the ptm BGC, PtmT3 and PtmT1, which, prior to discovery of 6 (and 3), are proposed to partition the common precursor ent-copalyl diphosphate (ent-CPP) specifically to the ent-atiserene (PtmT1) and ent-kauranol scaffolds (PtmT3) for the eventual biosynthesis of the thioacids 4 and 5, respectively.²² To determine the biosynthetic origin of 6 (and 3), the previously engineered SB12008 strain (i.e., the ΔptmR1/ΔptmT3 mutant of S. platensis sp. CB00739)²² was re-fermented. Close re-examination of the metabolite profile revealed the complete abolishment of both 4 (and 1) and 6 (and 3) production, with the production of 5 (and 2) unaffected (Figure 2C), indicating that PtmT3 is responsible for the biosynthesis of both ent-beyerene and ent-kauranol scaffolds (Figure 1C). Importantly, re-analysis of SB12035 fermentation revealed no production of the thioacids 4–6, or their non-enzymatic hydrolysis congeners 1–3 (Figure 2A), confirming that the biosynthesis of all three scaffolds is encoded by the ptm BGC (Figure 1). The PTM-PTN-PTL machinery, capably of parallelly processing three distinct scaffolds across such a long and intricate biosynthetic pathway, is unprecedented in natural product biosynthesis, showcasing once again Nature’s ingenuity in combinatorial biosynthesis (Figure S1).²³

As both 1 and 2, as well as their thioacid congeners 4 and 5, are known as potent bacterial fatty acid synthase inhibitors (targeting the FabF and FabH subunits),^1–3 3 and 6 were also expected to share their antibacterial activities (Table 1). Both 3 and 6 retain a comparable level of bioactivity as 1 and 2, with 6 showing slightly weaker activity as expected based on previous results with 4 and 5 (4–8 fold higher MICs).⁸ Despite higher MICs, previous studies have also shown that 4 and 5 bind FabF tighter than 1 and 2, indicating that this difference between 1–3, and their thioacid congeners 4–6 may be due to bioavailability.⁸ Future comparative studies among the three scaffolds may inform new efforts to optimize the PTM-PTN-PTL family of natural products for further clinical development, including opening up the possibility of combinatorial biosynthesis using a variety of diterpene cyclases in an engineered strain for additional structural diversity.

Table 1.

Antibacterial activities of PTL (3) and thioPTL (6) in comparison with PTM (1) and PTN (2).

Compounds	MIC (μg mL−1)
	B. subtilis	S. aureus	M. smegmatis
PTM (1)	1	1	2
PTN (2)	0.125	0.25	0.5
PTL (3)	2	2	8
ThioPTL (6) (6)	4	8	32

PtmC exhibits an unprecedented acyl-CoA specificity for an NAT.

With the three ketolide acids 13–15 isolated from the ΔptmR1/ΔptmC mutant strain SB12035, we first prepared the corresponding ketolide-CoAs 16–18 semisynthetically by following literature procedures (SI Materials and Methods).²⁴ The arylamines 8 and 9 were also isolated and enzymatically synthesized according to our previously published methods (SI Materials and Methods).⁸ To overproduce the PtmC protein for in vitro characterization, the ptmC genes from the seven ptm BGCs and one ptn BGC were all cloned (pBS12117-pBS12124) and heterologously expressed in E. coli (SI Materials and Methods).^25–26 Among the eight overproduced PtmC homologues, five were soluble, but surprisingly considering their shared 97% amino acid identity, the remaining three were insoluble (Figure S68). Notably, previous studies have identified key residues involved in NAT family stability;^27–29 however, none of the corresponding residues differed between the eight PtmC homologues studied (Figure S69). Of the five soluble homologues, ptmC from S. sp. CB02304 (i.e., pBS12117) was selected for its high level of expression in E. coli BL21(DE3), and the overproduced PtmC protein was purified to homogeneity (Figure S70A). PtmC was determined to be a monomer in solution by size-exclusion chromatography (Figure S70A), consistent with known NATs of both prokaryotic and eukaryotic origins.¹⁰

PtmC was assayed in vitro to investigate its activities between the selected arylamines 8 and 9 and the selected ketolide-CoAs 16–18 (Figure 3A), and all assays were carried out in 50 mM MOPS at pH 6.5 (Figure S71) (SI Materials and Methods). In the presence of PtmC, new peaks representing the products 1–3 were readily observed after reactions of carboxylic acid 8 with 16, 17, or 18, respectively, within 15 minutes at 30 °C, with boiled PtmC used as negative controls (Figure 3B). Additionally, thioacids 4–6 were also readily generated by PtmC under the same assay conditions upon replacing 8 with 9 (Figure 3C). Conspicuously, and despite the relatively rapid non-enzymatic hydrolysis of 9 to 8, none of 1–3 were seen in the reactions forming 4–6, respectively, suggesting that thioacid 9 was taken preferentially over carboxylic acid 8 and that the thioacid products 4–6 apparently did not undergo non-enzymatic hydrolysis under the assayed conditions and time frame.

Figure 3.

PtmC catalyzes amidation of selected arylamines with selected ketolide-CoAs in vitro. (A) Reaction schemes for PtmC-catalyzed amindation of ADHBA (8) or ADHBSH (9) with platensicyl-CoA (16), platensilyl-CoA (17), or platencinyl-CoA (18). (B) HPLC analysis of PtmC-catalyzed amidation of 8 with 16, 17, or 18. Panels I-III, authentic standards of substrates 8 and 16–18, and products 4–6 and CoA; panels IV and V, 8 + 16; panels VI and VII, 8 + 18; panels VIII and IX, 8 + 17. (C) HPLC analysis of PtmC-catalyzed amidation of 9 with 16, 17, or 18. Panels I-III, authentic standards of substrates 9 and 16–18, and products 4–6 and CoA; panels IV and V, 9 + 16; panels VI and VII, 9 + 18; panels VIII and IX, 9 + 17. Under the assay conditions, 9 undergoes non-enzymatic hydrolysis to 8, which co-eluted with CoA under the HPLC conditions analyzed.

While members of the NAT family are known for their arylamine substrate promiscuity, they are highly specific for their acyl-CoA substrate, typically acetyl-CoA,¹⁰ with only rare instances of alternative CoA- or pantetheine-linked substrates.^14–17 As PtmC apparently utilizes all three ketolide-CoAs 16–18 efficiently (Figure 3), the kinetic parameters of PtmC for 16–18 were determined, with carboxylic acid 8 at 750 mM and varying concentrations of 16, 17, or 18 (from 5 to 120 μM) (SI Materials and Methods). PtmC-catalyzed formation of 1–3 followed Michaelis-Menten kinetics, with K_m values of 24.0 ± 7.0 mM (for 16), 6.3 ± 1.3 mM (for 17), and 9.9 ± 3.8 mM (for 18), and k_cat values of 73.4 ± 8.1 min⁻¹ (for 16), 38.5 ± 2.1 min⁻¹ (for 17), and 46.8 ± 6.2 min⁻¹ (for 18), respectively (Figure S72B). While the three ketolide-CoAs were utilized with nearly equal catalytic efficiencies (k_cat/K_m), 17 and 18 exhibited lower K_m values than the PTM congener 16, which was compensated for with a higher k_cat. This result is reminiscent of the PtmA2 acyl-CoA ligase, which exhibits similar catalytic efficiencies towards both ent-kauranol- and ent-atiserene-derived substrates,³⁰ but in contrast to the PtmO3 and PtmO6 hydroxylases, which display an eight-fold higher catalytic efficiency towards the ent-atiserene-derived substrate over the ent-kauranol-derived substrate.³¹ As the CoA moiety likely contributes the most to substrate-enzyme interactions, less difference in catalytic efficiency would be expected among the three ketolide-CoAs. Significantly, PtmC has evolved to exhibit some of the lowest K_m values for its acyl-CoA substrates (< 25 μM) in comparison to any NATs characterized to date (with K_m values reaching into the mM range).^32–33

Inspired by the relaxed substrate specificity towards the three distinct ketolide-CoA scaffolds (Figures 1C, 3), we set out to investigate if PtmC could also utilize other acyl-CoAs, particularly the varying CoA-activated intermediates on the PTM-PTN-PTL biosynthetic pathway (Figures 1C, S1), as well as acetyl-CoA, the canonical substrate for the NAT family of enzymes.¹¹ A panel of seven additional ketolide-CoAs, with varying acyl moieties of ent-kauranol, ent-atiserene, and ent-beyerene origins, was selected (Figure S73A).^{6, 34} Under the identical assay conditions with 8 as the arylamine substrate, PtmC showed no activity towards either acetyl-CoA or the selected seven ketolide-CoAs with varying moieties of diterpenoid origin (Figure S73C). This is striking, considering the high structure similarity between 10 and 16, 11 and 17, and 12 and 18, which each differ only in the side chain, with an extra methyl-branched two-carbon unit between the CoA and the ketolide moieties (Figure 1C). PtmC therefore must have evolved specifically to accommodate structural diversity far from the CoA moiety yet to restrict the distance between CoA and the varying ketolide moieties, as exemplified by 16–18, so that only the precisely processed penultimate intermediates on the PTM-PTN-PTL pathway are specifically recognized by PtmC, thereby catalyzing their efficient coupling with thioacid 9 to afford 4–6 as the final natural products (Figure 1C).

PtmC prefers an arylamine thioacid over its carboxylic acid congener as a substrate.

Since NATs are known for their arylamine substrate promiscuity,^{27–29, 35} we next investigated the substrate scope of PtmC towards a selected panel of arylamines. Previously, we demonstrated the production of a total of 18 PTM and PTN analogues, with varying benzoic acid moieties, by feeding a panel of 34 arylamines and congeners to SB12032 (i.e., theΔptmR1/ΔptmB1 mutant of S. sp. CB00739, in which the biosynthesis of arylamines 7–9 is abolished). These findings are indicative of some degree of arylamine substrate promiscuity of PtmC, at least in vivo under the fermentation conditions examined.⁷ Based on their relative incorporation efficiencies in vivo, three representative arylamines, including 3-amino-4-hydroxybenzoic acid (AHBA) (7), the biosynthetic precursor of 8 and 9 (Figure 1C), were selected, with 8 and 9 as positive controls (Figure S73B). However, under the identical assay conditions with 16 as the ketolide-CoA substrate, PtmC only showed trace activity towards 7, and no activity was detected for the other two arylamines tested in vitro (Figure S73C), despite the fact that all three arylamines were incorporated with varying efficiencies in vivo.⁷ In contrast to known NATs, PtmC therefore must have evolved to be unusually specific for its arylamine substrate. In vivo incorporation of the varying arylamines by SB12032, when supplemented exogenously in high concentrations in the production medium, most likely resulted from the extended time of fermention.⁷

While PtmC has evolved a high substrate specificity towards its arylamine substrate, any enzymatic preference for a thioacid substrate over a carboxylic acid is unprecedented. Therefore, we set out to determine the kinetic parameters of PtmC for 8 and 9; the poor activity of 7 in vitro prevented us from determining its kinetic parameters under the assay conditions (SI Materials and Methods). With saturating 16 at 250 μM and varying concentrations of 8 or 9 (from 25 to 1500 μM), PtmC-catalyzed the formation of 1 or 4 followed Michaelis-Menten kinetics, with K_m values of 372 ± 49 μM (for 8) and 142 ± 54 μM (for 9), and k_cat values of 92.6 ± 4.5 min⁻¹ (for 8) and 282 ± 4 min⁻¹ (for 9), respectively (Figure S72A). Steady-state kinetics of PtmC revealed an eight-fold higher catalytic efficiency (k_cat/K_m) for thioacid 9 over carboxylic acid 8. While an eight-fold difference in catalytic efficiencies between 8 and 9 is notable, it is insufficient to account for the exclusive formation of 4–6 as the nascent products of the PTM-PTN-PTL biosynthetic machinery in vivo.⁵ Instead, it appears that PtmC lacks the necessary evolutionary pressure to further distinguish between 8 and 9 as all available 8 is already being sequestered before its conversion into 9. Previously determined kinetic parameters of PtmA3 reveal a K_m value of only 22.9 ± 3.0 μM for 8, more than 16-fold lower than PtmC.⁸ Thus, in the two biosynthetic steps prior to PtmC, PtmA3 ligates 8 to CoA before its conversion to 9 by PtmU4 (Figure 1C). Taken together, the kinetic and metabolic flux advantages for 9 over 8 represent the first significant evidence of how a biosynthetic pathway efficiently incorporates a thioacid intermediate in direct competition with its more stable corresponding carboxylic acid congener in natural product biosynthesis (Figure 1C).

PtmC shares a conserved NAT structural fold with distinct features to accommodate the large ketolide-CoA substrates.

After several failed attempts to solve the PtmC structure by X-ray crystallography, we built a homology model of PtmC, based on an NAT from Salmonella typhimurium (PDB: 1E2T),³⁶ to provide some evolutionary clues for its divergent substrate selectivity (SI Materials and Methods). The three catalytic residues, Cys69, His107, and Asp121, are conserved in PtmC (Figure 4A), and site-directed mutagenesis was conducted to confirm their roles in catalysis. As expected, the C69A mutant lost its activity (Figure S74); however, both the H107A and D121A mutants were insoluble, precluding experimental validation of their catalytic roles in vitro (Figure S70B–C). The PtmC model shows the conserved overall NAT structural fold (Figures 4A, S75), but instead of the C-terminal helix found in prototypical prokaryotic NATs (blue, Figure 4B),^36–37 PtmC has a slightly longer coil prior to the helix and then a 20-residue coil at the C-terminus (blue and red, Figure 4A). This feature is reminiscent of the eukaryotic NAT structures, which lack the C-terminal helix and instead have a single 19-residue C-terminal tail (red, Figure 4C).¹¹ In the eukaryotic NATs, this coil extends into the active site and is proposed to maintain the size and shape of the active site.²⁹ The eukaryotic NATs have a second feature, termed the domain II loop (purple, Figure 4C), that is a 17-residue insertion (residues 164 to 186 in the human NAT1) and is proposed to provide structural stability as well as limit active site access.^{29, 35} While this feature is not found in any known prokaryotic NATs, PtmC lacks an additional three residues at this site compared to other prokaryotic NATs.

Figure 4.

Structural modeling of PtmC. (A) A homology model of PtmC was subjected to a MD simulation and docked with CoA (green). The three catalytic residues are depicted as sticks (coral). The C-terminal coils are highlighted in red, with the prokaryotic C-terminal helix highlighted in blue. The loop most impacted in the truncated mutant PtmC^t is shown in gold. (B) The M. loti NAT1 structure (PDB: 4NV7) is an example of a prototypical prokaryotic NAT with CoA (green) co-crystallized. (C) The structure of human NAT2 (PDB: 2PFR) co-crystallized with CoA (green) shows the insertion of the C-terminal coil (red) into the active site and the domain II loop (purple) preventing the same CoA orientation as that of prokaryotic NATs. (D) Docking of the covalently-bound platensicyl group via a thioester to Cys69 (dark green) shows the importance of key residues whose identities are unique or rare in other NATs (magenta). The residues implicated in binding of the aromatic substrates (teal) remain conserved.

The C-terminal tail of PtmC is unprecedented among the NAT family members. To probe its role, we truncated the 12 C-terminal residues from PtmC to yield PtmC^t (pBS12128) (SI Materials and Methods). Upon expression in E. coli BL21 (DE3), the overproduced PtmC^t retained the similar solubility as the wild-type PtmC and was purified; however, PtmC^t retained significantly less catalytic activity than PtmC (Figure S74). The kinetic parameters of PtmC^t were then similarly determined for comparison (SI Materials and Methods). With 8 at 750 μM and varying concentrations of 16 (from 10 to 200 μM), or 16 at 250 μM and varying concentrations of 8 (from 100 to 1000 μM), PtmC^t-catalyzed formation of 1 followed Michaelis-Menten kinetics, affording K_m values of 375 ± 119 μM (for 8) and 23.7 ± 6.1 μM (for 16), respectively, with the k_cat value estimated between 8.3 ± 0.7 min⁻¹ (determined under partial saturation of 8) and 19.1 ± 2.5 min⁻¹ (determined under full saturation of 16) (Figure S72C). PtmC^t retained similar K_ms as PtmC, and the significant loss in catalytic efficiency of PtmC^t resulted mainly from the five-fold decrease in k_cat (Figures S72C, S74). We finally calculated the melting temperatures (T_m) for both PtmC and PtmC^t to estimate their relative stability (SI Materials and Methods). While the wild-type PtmC had a higher T_m than PtmC^t (41.22 ± 0.27 °C and 39.53 ± 0.30 °C, respectively, Figure S76), the difference seems to be insufficient to account for a five-fold decrease in activity.¹²Without much difference in overall stability or substrate recognition, other structural roles for the C-terminal tail needed to be considered. Molecular dynamics (MD) simulations (50 ns) of PtmC and PtmC^t models were used to probe the role of the C-terminal tail in catalysis. In the simulations, the overall structures do not differ significantly (RMSD of 0.92 Å, Figures S77–S78). However, comparison of the root-mean-square fluctuation (RMSF) values showed significant differences in the dynamics for the loop proximal to the β-strand containing the catalytic His107 (gold, Figure 4A,D), shifting the position of the β-strand itself (Figure S78). Any change in the position of His107, whose position relative to Cys69 is very well conserved in NATs, would be expected to have a negative effect on k_cat. Therefore, we proposed that the C-terminal coil functions to stabilize this loop in order to maintain the integrity of the catalytic triad.

An examination of sequence alignments of PtmC with other NATs reveals several key differences in addition to those readily apparent from the structures (Figure S79), especially at two residues near the catalytic Cys69 – Thr68 and Ile70. The two residues flanking Cys69 are often aromatic; however, both are smaller in PtmC, likely to provide additional space for the bulky ketolide acyl groups. Covalent docking of the platensicyl group of 16 via a thioester to Cys69 (dark green, Figure 4D) shows interactions with, in addition to Thr68 and Ile70, four additional active site residues that diverge from the consensus sequence (magenta, Figure 4D) – Val106 (usually hydrophilic), Leu125 (Gly), Trp186 (Ala or Ser), Leu191 (Ser), and Met193 (Phe/His/Pro). Together, these seven residues most likely make up the large and hydrophobic ketolide acyl group binding pocket. Identification of these residues may be important for future studies of other NATs with unusual acyl specificities. In contrast, residues involved in binding of the arylamine substrate appears to be mostly unchanged from similar NATs, such as the human NAT1.²⁷ These residues include Phe124, whose cognate residue participates in π-π stacking in human NAT1, and Arg126, which should form a salt bridge with the carboxylate and thiocarboxylate of 8 and 9, respectively (teal, Figure 4D).²⁷

PtmC homologues are widespread but underappreciated in natural product biosynthesis.

As a new member of the NAT family, the unique chemistry, enzymology, and structure of PtmC prompted a broader look into homologues outside of the ptm BGC. Remarkably, of more than 5,700 members of the NAT Pfam identified across almost all branches of life, with more than 1,500 (26%) from Actinobacteria (Figures 5A, S80), which are the most prolific producers of natural products.^{9, 38–40} Importantly, while members of the NAT Pfam appear to cluster by taxonomic origin, they are also likely to cluster by predicted functions. PtmC is weakly associated with the second largest cluster in the sequence similarity network (SSN) (Figure 5A). To determine if those enzymes most closely related to PtmC are likely to utilize unusual acyl-CoA substrates, their encoding genes and genomic neighbors were manually analyzed. Surprisingly, considering the dearth of NAT enzymes being reported in natural product biosynthesis,¹⁴ the top five homologues of PtmC all appear to be in natural product BGCs, often surrounded by genes encoding for CoA ligases and/or polyketide synthases and non-ribosomal peptide synthetases (Figure S81).⁴¹

Figure 5.

Broad distribution of PtmC-like NATs in nature. (A) A sequence similarity network (SSN) of NATs in different organisms shows grouping by taxonomic relationships. A BLAST e-value threshold of 10⁻⁵⁴ was employed. Colors represent different phylogenetic classes. PtmC is indicated in red. (B) The Actinobacterial NATs from the cluster containing PtmC were subjected to a more stringent e-value threshold of 10⁻⁷⁷. Proteins were colored by the identity of residues flanking the catalytic cysteine. Sequence logos created for the residues surrounding the catalytic cysteine for (C) the entire NAT family and (D) the 80 closest homologues to PtmC, indicating that these enzymes likely specify larger acyl-CoA substrates.

Gratifyingly, these homologues contain similar amino acid sequence signatures as PtmC, particularly at the residues flanking the catalytic cysteine (Figure S82). A broader analysis of 750 Actinobacterial NATs revealed >15% deviating from the two flanking aromatic residues found in NATs specific for acetyl-CoA, indicative of the high potential to further expand the acyl-CoA substrate scope of NATs (Table S8). This can also be visualized by examining the Actinobacterial NATs from the same cluster as PtmC at a higher resolution (Figure 5B) and comparing the residues flanking the catalytic cysteine from those more closely related to PtmC to the entire NAT family (Figure 5C–D). The predicted novelty within the NAT family indicates that PtmC may be just one of many non-canonical NATs and the prevalence of NATs in natural product biosynthesis has not been appreciated to date. With the expanding substrate scope of alternative acyl-CoAs, new chemical scaffolds, containing non-peptidyl amide bonds, are likely waiting to be discovered. As highlighted by the representative set of BGCs encoding PtmC homologues (Figure S82), this study should inspire future utilization of these novel NATs as genetic beacons for genome mining for novel natural products.

CONCLUSIONS

The PTM-PTN-PTL biosynthetic machinery is an extraordinary example of Nature’s combinatorial biosynthesis at its best, affording three scaffolds as final natural products, with at least 14 enzymes acting on each, the last of which being PtmC (Figures 1C, S1). Each of these enzymes has evolved to be specific to its substrates while still being permissible enough to accept three varying diterpenoid scaffolds partitioned by two diterpene cyclases from the common precursor ent-CPP. Terpenoid biosynthetic pathways are often either highly promiscuous with few tailoring enzymes or highly specific with more intricate tailoring steps.^42–43 With the discovery of 3 and 6, the first new scaffold in the PTM-PTN class of antibiotics and antidiabetics since 2007,¹ the PTM-PTN-PLL biosynthetic machinery elegantly highlights Nature’s ingenuity in fine-turning pathway plasticity between the two extremes. The molecular basis of these enzymes is critical to future understanding, thereby engineering designer pathways, for both natural product structural diversity and targeted analogues.

PtmC has been biochemically characterized as a novel NAT with unique specificities for both its large ketolide-CoAs and arylamine thioacid as substrates. Cumulatively, PtmC distinguishes itself from other NAT family members by eschewing the typical high levels of arylamine promiscuity in favor of a narrow substrate scope,^14–16 including an unprecedented preference for the thioacid 9 over its carboxylic acid congener 8. Structurally, PtmC has several unprecedented features for a NAT family member, containing characteristics of both eukaryotic and prokaryotic NATs. Molecular modeling and MD simulations revealed the typical C-terminal helix found in prokaryotic NATs, but PtmC also has a 20-residue C-terminal coil that may provide additional stability for the catalytic triad contained within an enzyme scaffold known to be unstable.^{12, 44} PtmC also contains several active site residue changes, some unprecedented, relative to other NAT family members, likely to accommodate the large ketolide-CoA substrates relative to acetyl-CoA for most NATs known to date.^9–10 The identity of these residues should enable further identification of NATs with unusual acyl-CoA specificities. While most NATs, which span nearly all branches of life, are currently known mainly for their roles in xenobiotic detoxification, bioinformatics has revealed that PtmC may be a member of a significant, yet underappreciated, subgroup of NATs that are responsible for natural product biosynthesis. Thus, PtmC may be used as a genetic beacon to mine genomes for new enzymology and novel classes of natural products containing non-peptidyl amide bonds. Together, PtmC provides a wealth of knowledge for NATs involved in bacterial natural product biosynthesis and xenobiotic metabolism in humans and bacterial pathogens.