The LnmK Bifunctional Acyltransferase/Decarboxylase Specifying (2R)-Methylmalonyl-CoA and Employing Substrate-assisted Catalysis for Polyketide Biosynthesis

Abstract

We previously showed that the bifunctional LnmK acyltransferase/decarboxylase (AT/DC) catalyzed the formation of propionyl-S-acyl carrier protein (ACP) from methylmalonyl-CoA but its substrate specificity to (2S)-, (2R)-, or (2RS)-methylmalonyl CoA was not known. We subsequently revealed that LnmK AT and DC activities share the same active site, employing a Tyr as the catalytic residue for AT, but failed to identify a general acid or base within the vicinity of the active site for LnmK catalysis. We now show that (i) LnmK specifies (2R)-methylmalonyl-CoA and (ii) the AT and DC activities are coupled, featuring substrate-assisted catalysis via the enolate to account for the missing general acid or base within the LnmK active site. LnmK homologs are the only bifunctional AT/DC enzymes known to date and are widespread. These findings therefore enrich PKS chemistry and enzymology. Since only the (2S)-methylmalonyl-CoA enantiomer has been established previously as a substrate for polyketide biosynthesis by PKSs, we now establish a role for both (2R)- and (2S)-methylmalonyl-CoA in polyketide biosynthesis, and (2R)-methylmalonyl-CoA should be considered as a substrate in future efforts for engineered production of polyketides by combinatorial biosynthesis or synthetic biology strategies in model hosts.

Graphical Abstract

LnmK is a bifunctional acyltransferase/decarboxylase (AT/DC) originally discovered in the acyltransferase-less (AT-less) type I polyketide synthase (PKS) gene cluster for the biosynthesis of leinamycin, an antitumor natural product with potent DNA cleaving properties.¹ LnmK and known homologues introduce propionate groups into polyketide biosynthetic pathways by accepting methylmalonyl-CoA and an acyl carrier protein (ACP) to generate propionyl-S-ACP.^2–4 In leinamycin biosynthesis, the propionyl-S-ACP is used by β-alkylation enzymes to generate a propionyl group that forms a part of the key dithiolane moiety responsible for the potent DNA damaging activity (Figure 1A).¹ More recently, the LnmK homologue Lom62 was biochemically characterized from the type II polyketide synthase gene cluster for lomaiviticin. Lom62 initiates biosynthesis by providing propionyl-S-ACP, rather than the canonical acetyl-S-ACP substrate (Figure 1B).⁵ A brief search of the MiBig database⁶ reveals that a few other known polyketide biosynthetic gene clusters contain LnmK homologues, such as those for myxovirescin⁷, nenestatin A,⁸ and fogacin C⁹ (Figure 1C). A BLAST search for LnmK homologues and genome neighborhood analysis reveals that there are currently over 150 unique sequences for LnmK homologues in uncharacterized PKS-containing biosynthetic gene clusters (Figure S1). Furthermore, a distant homologue of LnmK (annotated by the TIGR family 04098) resides in the initiating module of the kirromycin polyketide synthase¹⁰ and in other orphan PKSs. Therefore, the LnmK biochemistry is used widely and divergently in polyketide biosynthesis, warranting further investigation of substrate specificity and catalytic mechanism.

Figure 1.

The bifunctional LnmK AT/DC enzyme deriving a propionyl-S-ACP from methylmalony-CoA for polyketide biosynthesis in the presence of an ACP. (A) LnmK-derived propionyl-S-ACP for polyketide biosynthesis by a noniterative type I PKS. (B) Lom62-derived propionyl-S-ACP for polyketide biosynthesis by a type II PKS. (C) Other known polyketides with ethyl groups highlighted in red derived by LnmK homologues from methylmalony-CoA via a propionyl-S-ACP intermediate.

We previously demonstrated that, in the presence of the LnmL ACP, LnmK catalyzes the formation of propionyl-S-ACP only with methylmalonyl-CoA, and not propionyl-CoA, concluding that the AT activity precedes the DC activity (Figure 1A).³ This conclusion was further supported by the finding that chemoenzymatically prepared methylmalonyl-S-ACP is decarboxylated by LnmK. Our subsequent structure-function studies demonstrated that LnmK was the first member of the double-hot-dog fold enzyme family with an AT activity and the first AT that employs a Tyr as the catalytic residue at position 62 (LnmK numbering).² While LnmK forms a transient acyl-O-LnmK intermediate, en route to the propionyl-S-ACP product from methylmalonyl-CoA, the LnmK (Y62F) mutant neither forms the acyl-O-LnmK intermediate nor catalyzes formation of propionyl-S-ACP.² These studies left a number of unanswered questions. While LnmK specifically accepts methylmalonyl-CoA, it is not known if (2S)- or (2R)-methylmalonyl-CoA, or both are accepted or preferred substrates. Our initial structure-function studies of LnmK revealed a shared active site for AT and DC catalysis. However, the LnmK structures failed to reveal a general acid or base within the vicinity of the catalytic residue Tyr62. Here we demonstrate (i) LnmK is specific for (2R)-methylmalonyl-CoA and (ii) acyltransfer is coupled with decarboxylation, featuring substrate-assisted catalysis via deprotonation of the catalytic residue Tyr62 by the enolate intermediate to account for the missing general base within the LnmK active site. These findings enrich PKS chemistry and enzymology. Since only (2S)-methylmalonyl-CoA is known to date as a substrate for polyketide biosynthesis by non-iterative type I PKSs,¹¹ (2R)-methylmalonyl-CoA should now be considered as a substrate in future efforts for engineered production of polyketides by combinatorial biosynthesis or synthetic biology strategies in model hosts.

To determine if LnmK prefers one methylmalonyl-CoA enantiomer, we first examined the activity of LnmK on commercially available (2RS)-methylmalonyl-CoA (Figure 2A). In the absence of ACP, both LnmK and LnmK (Y62F) are competent DC enzymes, catalyzing efficient decarboxylation of (2RS)-methylmalonyl-CoA to generate propionyl-CoA (Figure 2C, panels II, IV). This was the first hint that DC precedes AT activity, which was contrary to our previous hypothesis.³ However, only ~50% of the (2RS)-methylmalonyl-CoA is converted to propionyl-CoA, suggesting that only one enantiomer is the substrate. We confirmed that only one of the two (2RS)-methylmalonyl-CoA enantiomers serves as a substrate for LnmK by adding Streptomyces coelicolor methylmalonyl-CoA epimerase (MCE),¹² which catalyzes interconversion of (2S)- and (2R)-methylmalonyl-CoA (Figure 2A), affording 100% conversion of the (2RS)-methylmalonyl-CoA substrate into propionyl-CoA (Figure 2C, panels III, V). We prepared (2R)-methylmalonyl-CoA using MatB¹³ (Figure 2B) and showed that LnmK catalyzes complete conversion of (2R)-methylmalonyl-CoA into propionyl-CoA (Figure 2C, panels VI, VII), unambiguously establishing that LnmK specifically accepts (2R)-methylmalonyl-CoA.

Figure 2.

LnmK or LnmK (Y62F) mutant decarboxylates (2R)-methylmalonyl-CoA to propionyl-CoA in the absence of an ACP. (A) Decarboxylation of (2R)-methylmalonyl-CoA by LnmK or LnmK (Y62F), and interconversion of (2R)- and (2S)-methylmalonyl-CoA by MCE. (B) (2R)-methylmalonyl-CoA biosynthesis by MatB and decarboxylation by LnmK or LnmK (Y62F) mutant to propionyl-CoA. (C) HPLC analysis of propionyl-CoA formation by LnmK or LnmK (Y62F): (I) authentic standards of CoA, (2RS)-methylmalonyl-CoA, propionyl-CoA; (II) LnmK with (2RS)-methylmalonyl-CoA; (III) panel II conditions plus MCE; (IV) LnmK (Y62F) with (2RS)-methylmalonyl-CoA; (V) panel IV conditions plus MCE; (VI) (2R)-methylmalonyl-CoA synthesized by MatB; and (VII) panel VI conditions plus LnmK.

We next confirmed that the transient acyl-O-LnmK intermediate was detected only with (2R)-methylmalonyl-CoA as a substrate, further supporting that LnmK is specific to (2R)-methylmalonyl-CoA. Incubation of LnmK with (2RS)-[methyl-¹⁴C]methylmalonyl-CoA afforded an acyl-O-LnmK intermediate within the first time point (Figure 3A, lane 2), which, in the absence of an ACP, was rapidly turned over into propionyl-CoA as the incubation time prolonged (Figure 3A, lanes 3, 5). The formation of the acyl-O-LnmK intermediate could be restored upon addition of the MCE, which converts the remaining (2S)-[methyl-¹⁴C]methylmalonyl-CoA into (2R)-[methyl-¹⁴C]methylmalonyl-CoA (Figure 3A, lanes 4, 6) or fresh (2RS)-[methyl-¹⁴C]methylmalonyl-CoA (Figure 3A, lane 7). The (2R)-methylmalonyl-CoA substrate specificity of LnmK contrasts to non-iterative type I PKSs, the AT domains of which are known to accept (2S)-methylmalonyl-CoA only for polyketide biosynthesis, as exemplified by the 6-deoxyerythronolide B synthase (DEBS).¹¹ The KS-AT didomain of DEBS modules 3 and 5 (i.e., KS-AT3 and KS-AT5) were overproduced and compared to LnmK for their opposite substrate specificity. While incubation of the KS-AT3 and KS-AT5 proteins with (2R)-[1,3-¹⁴C]methylmalonyl-CoA resulted in little ¹⁴C-labeled KS-AT species (Figure 3B, lanes 1, 3), inclusion of MCE converts (2R)-[1,3-¹⁴C]methylmalonyl-CoA into (2S)-[1,3-¹⁴C]methylmalonyl-CoA, resulting in specific ¹⁴C-labeling of both KS-AT3 and KS-AT5 proteins and confirming (2S)-methylmalonyl-CoA as the preferred substrate for the DEBS (Figure 3B, lanes 2, 4).¹¹

Figure 3.

LnmK specifies (2R)-methylmalonyl-CoA and, in the absence of an ACP, catalyzes decarboxylation to afford a propionyl-CoA, featuring a propionyl-O-LnmK intermediate in catalysis, with DEBS KS-AT3 and KS-AT5 specifying for (2S)-methylmalonyl-CoA as controls, revealed by SDS-PAGE (4–15%) (top) and autoradiography (bottom) analysis. (A) LnmK with (2RS)-[methyl-¹⁴C]methylmalonyl-CoA after varying incubation times: lane 1, molecular weight markers; lane 2, 2 min; lane 3, 20 min; lane 4, 2 min after addition of MCE at 20 min; lane 5, 80 min; lane 6, 2 min after addition of MCE at 80 min; lane 7, 2 min after addition of fresh (2RS)-[methyl-¹⁴C]methylmalonyl-CoA at 80 min. (B) DEBS KS-AT3 or KS-AT5 with MatB generated (2R)-[methyl-¹⁴C]methylmalonyl-CoA after 2 min incubation: lane 1, KS-AT3; lane 2, KS-AT3 and MCE; lane 3, KS-AT5 and MCE; lane 4, KS-AT5 and MCE. (C) LnmK after 2 min incubation: lane 1, (2RS)-[1,3-¹⁴C]methylmalonyl-CoA; lane 2, (2RS)-[methyl-¹⁴C]methylmalonyl-CoA; lane 3, (2RS)-[1,3-¹⁴C]methylmalonyl-CoA and MCE; lane 4, (2RS)-[methyl-¹⁴C]methylmalonyl-CoA and MCE. Please see (A) for scheme corresponding to lanes 2 and 4.

We subsequently identified the transient acyl-O-LnmK intermediate as propionyl-O-LnmK by differentially labeling LnmK with (2RS)-[1,3-¹⁴C]methylmalonyl-CoA or (2RS)-[methyl-¹⁴C]methylmalonyl-CoA at equal input of radioactivity (Figure 3C). Thus, incubation of LnmK with (2RS)-[1,3-¹⁴C]methylmalonyl-CoA afforded a [1-¹⁴C]propionyl-O-LnmK intermediate, losing a half of ¹⁴C-activity as ¹⁴CO₂, while incubation of LnmK with (2RS)-[methyl-¹⁴C]methylmalonyl-CoA afforded a [3-¹⁴C]propionyl-O-LnmK intermediate, retaining the full ¹⁴C-activity (Figure 3C, lanes 1 vs 2). Inclusion of MCE in a parallel reaction, ensuring that the substrate concentration was not limiting by conversion of the corresponding (2S)-forms into the (2R)-[1,3-¹⁴C]methylmalonyl-CoA and (2R)-[methyl-¹⁴C]methylmalonyl-CoA, respectively (Figures 2A, 3A), afforded the same differential ¹⁴C-labeling patterns of the resultant propionyl-O-LnmK species (Figure 3C, lanes 3 vs 4). With or without MCE, LnmK would be labeled with the same amount of ¹⁴C-activity with (2RS)-[methyl-¹⁴C]methylmalonyl-CoA or (2RS)-[1,3-¹⁴C]methylmalonyl-CoA if the transient intermediate was (2R)-methylmalonyl-O-LnmK. These findings establish propionyl-O-LnmK as a key intermediate in LnmK catalysis and reveal that decarboxylation precedes acyltransfer, a conclusion that would be consistent with the finding that the LnmK (Y62F) mutant is competent in catalyzing efficient decarboxylation of (2R)-methylmalonyl-CoA to propionyl-CoA (Figure 2).

Finally, we are motivated to revise the previous mechanism based on our current findings, including that (i) LnmK is specific to (2R)-methylmalonyl-CoA as a substrate (Figures 2, 3), (ii) LnmK decarboxylates (2R)-methylmalonyl-CoA to afford propionyl-CoA in the absence of an ACP (Figure 2), (iii) DC can be independent of AT as both LnmK and LnmK (Y62F) are competent in catalyzing decarboxylation of (2R)-methylmalonyl-CoA to propionyl-CoA (Figure 2), but AT must couple with DC as only (2R)-methylmalonyl-CoA, not propionyl-CoA,³ is utilized as a substrate by LnmK in the presence of an ACP, to afford the propionyl-S-ACP product, and (iv) DC precedes AT with propionyl-O-LnmK as a key intermediate in LnmK catalysis (Figures 3A, 3C). Our revised mechanism for LnmK catalysis accounts for the missing general acid or base in the shared AT/DC active site of LnmK as the decarboxylated enolate intermediate can deprotonate the catalytic residue Tyr62, allowing it to become a good nucleophile for acyltransfer and making it a good example of substrate-assisted catalysis (Figure 4).

Figure 4.

A revised mechanism for LnmK-catalyzed generation of a propionyl-S-ACP intermediate from (2R)-methylmalonyl-CoA for polyketide biosynthesis: (A) LnmK-catalyzed sequential decarboxylation and acyltransfer of (2R)-methylmalonyl-CoA, employing substrate-assisted catalysis for the acyltransfer step and exploiting LnmK and LnmL interaction to promote propionyl-S-ACP formation and (B) LnmK (Y62F)-catalyzed decarboxylation of (2R)-methylmalonyl-CoA followed by quenching of the resultant enolate intermediate with H₂O to account for propionyl-CoA formation in the absence of the LnmL ACP.

In conclusion, we have provided new mechanistic details for LnmK catalysis, which prefers (2R)-methylmalonyl-CoA and shares the same catalytic site for a coupled DC→AT reaction sequence that employs a substrate-assisted catalysis mechanism to afford a propionyl-S-ACP intermediate for polyketide biosynthesis (Figure 4). Since LnmK homologues are widespread (Figure S1), these new insights should now be considered for their roles in natural product biosynthesis and structural diversity. While other bifunctional AT/DC enzyme have been reported in the literature, LnmK remains unique. The PKS GNAT-like domains were originally characterized to catalyze acyltransfer of malonyl-CoA onto a cognate ACP with subsequent decarboxylation of the malonyl-S-ACP to afford an acetyl-S-ACP intermediate to initiate polyketide biosynthesis by non-iterative type I PKSs.¹⁴ However, recent mechanistic and structural studies have re-classified the PKS GNAT-like domains as DCs, excluding the AT activity.¹⁵ The Mcd bifunctional AT/DC enzyme catalyzes decarboxylation of malonyl-CoA to acetyl-CoA with subsequent acyltransfer to form a δ-N-acetyl-δ-N-hydroxy-L-ornithine product.¹⁶ Mcd differs from LnmK, featuring distinct DC and AT active sites likely within two different domains, allowing the Mcd AT to bypass DC and accept acetyl-CoA as an alternative substrate. Both (2S)- and (2R)-methylmalonyl-CoA are known metabolites and interconvertible by an epimerase, but only (2S)-methylmalonyl-CoA has been unambiguously established previously as a substrate for polyketide biosynthesis by type I PKSs.¹¹ We now establish a role for (2R)-methylmalonyl-CoA in polyketide biosynthesis. Informing the (2S)- or (2R)-methylmalonyl-CoA substrate specificity of PKS ATs is critical for engineered production of polyketides. Both (2R)- and (2S)-methylmalonyl-CoAs should now be considered as substrates in future efforts for engineered production of polyketides by combinatorial biosynthesis or synthetic biology strategies in model hosts.