Biosynthesis of lactacystin as a proteasome inhibitor

Identification of the biosynthetic gene cluster of lactacystin

To identify the biosynthetic gene cluster of lactacystin, we conducted a draft genome analysis of its producer, S. lactacystinicus NBRC 110082. Based on the chemical structure of lactacystin and the results of previous tracer experiments, the core structure of lactacystin was suggested to be formed by a PKS-NPRS system. By searching for PKS-NRPS hybrid proteins in the draft genome, several candidate genes were found. Among them, one PKS-NRPS gene was surrounded by genes similar to those in the salinosporamide A biosynthetic gene cluster (sal)⁸. In particular, an ortholog of the salC gene responsible for the formation of the γ-lactam-β-lactone moiety, which is found in both lactacystin and salinosporamide A, existed in the cluster. We therefore speculated that this region is responsible for the biosynthesis of lactacystin. The putative cluster was composed of a gene coding for a major facilitator superfamily (MFS) transporter (lctT), a hypothetical gene (lctH), genes similar to sal genes (lctA–E), and a formyltransferase gene (lctF) (Fig. 2a and Table S4). To examine whether the cluster is involved in lactacystin biosynthesis, we in-frame deleted the lctD gene encoding cytochrome P450, which perhaps hydroxylates the leucine moiety (Fig. S4). The resultant disruptant strain (∆lctD) lost lactacystin productivity, indicating that LctD is involved in lactacystin biosynthesis (Fig. 2b). Moreover, we observed accumulation of a compound with m/z 361, which is 16 mass units less than that of lactacystin (Figs. 2b and S5). This suggested that the compound was dehydroxyl-lactacystin (7) but we were unable to determine its exact structure because of low productivity. To narrow down the essential genes for lactacystin production, we employed a heterologous expression experiment. Two DNA fragments possessing “lctT to lctF” and “lctA to lctF” were prepared by PCR and each was inserted into a shuttle vector. The constructed plasmids were introduced into Streptomyces lividans TK23. The strains possessing lctT–lctF (8 genes) and lctA–lctF (6 genes) produced significant and small amounts of lactacystin, respectively. In contrast, the strain harboring the empty vector produced no lactacystin (Fig. 2c). These findings suggested that LctA–LctF play a vital role in lactacystin biosynthesis and that LctT and LctH might be crucial for high productivity.

a Biosynthetic gene cluster of lactacystin. The cluster contains lctA, PKS-NRPS (ACP, acyl carrier protein; KS, ketosynthase; AT, acyltransferase; C, condensation domains); lctB, NRPS (A, adenylation; PCP, peptidyl carrier protein domains); lctC, cyclase; lctD, cytochrome P450; lctE, MbtH-like protein (MLP); and lctF, formyltransferase. b LC-ESI-MS analysis (positive ion mode, monitored at m/z 377 (black) and 361 (blue)) of (i) lactacystin standard and broths of (ii) the wild-type strain and (iii) the ∆lctD strain. Traces monitored at m/z 377 were enlarged 25-fold vertically. A peak (*) with mass of 361.1439 corresponding to 7 (calculated mass, 361.1428) was detected (Fig. S5). c LC-ESI-MS analysis (positive ion mode, monitored at m/z 377) of (i) lactacystin standard and broths of transformants harboring (ii) lctT to lctF, (iii) lctA to lctF, and (iv) the empty vector. d LC-ESI-MS analysis (positive ion mode, monitored at m/z 377) of (i) lactacystin standard and broths of transformants harboring (ii) lctT, B, C, D, E, F, and ACP1-inactivated lctA (S45A) and (iii) lctB, C, D, E, F, and AT1-inactivated lctA (S791A).

Putative Biosynthetic Pathway of Lactacystin

Based on the abovementioned results and in silico analysis, the biosynthetic pathway of lactacystin was estimated as follows (Fig. 3). A putative formyltransferase (LctF) transfers a formyl group into CoA from 10-N-formyl tetrahydrofolate (10N-fTHF) to form formyl-CoA, and then the formyl group is loaded onto ACP1 by LctA_AT1. Concurrently, LctA_AT2 loads a methylmalonyl moiety, which is supplied from methylmalonyl-CoA, onto ACP2; this possibility was predicted by multiple alignment analysis of the AT2 domain (Fig. S6). Then, the KS domain in LctA catalyzes a Claisen-type condensation to form ACP2-tethered methylmalonyl-semialdehyde (3). The A domain of LctB likely loads Leu or 3-hydroxyl-Leu (OH-Leu) formed by LctD onto PCP domain of LctB after activation by adenylation. The MbtH-like protein (LctE) perhaps interacts with and activates the A domain¹⁶. Next, compound 3 tethered to ACP2 and Leu or OH-Leu bound to PCP are condensed to form an intermediate, 4 or 5, by the C domain in LctA. The intermediate is then cyclized and off-loaded by a stand-alone KS-like cyclase, LctC, to give 6 or 2 in a manner similar to the biosynthesis of salinosporamide A. It is then converted into 7 or lactacystin (1) by spontaneous addition of N-acetyl cysteine⁷. We next examined the plausibility of these proposed pathways by in vivo and in vitro analysis.

a Domain organization of LctA and LctB and predicted substrate specificities. b The proposed biosynthetic pathway.

ACP1 and AT1 in LctA were essential for lactacystin biosynthesis

First, the functions of the ACP1 and AT1 domains in LctA (PKS-NRPS hybrid) were investigated because of its unique domain organization. LctA consisted of ACP1-KS-AT1-AT2-ACP2-C domains but the existence of an ACP domain as the first domain is rare in PKS-NRPS hybrid enzymes and the same is true for the last C domain¹⁷. Furthermore, it was impossible to predict the substrate of the AT1 domain from its amino acid sequence using programs such as Minowa through antiSMASH 7.0 analysis^18,19. Therefore, we investigated whether ACP1 and AT1 were necessary for lactacystin biosynthesis. We constructed two derivatives from the plasmid used for heterologous expression, in which ACP1 or AT1 was inactivated by replacing the active Ser residue with an Ala residue (ACP1 (S45A) and AT1 (S791A)). The plasmids were then introduced into S. lividans and each culture broth of the transformants was analyzed by LC-ESI-MS. As shown in Fig. 2d, both transformants lost lactacystin productivity, indicating that both AT1 and ACP1 were necessary for the biosynthesis of lactacystin.

Formyltransferase (LctF) catalyzed formation of formyl-CoA

We next performed in vitro analysis of LctF, which has 32% identity to methionyl-tRNA formyltransferases. This type of enzyme usually requires 10N-fTHF as a formyl group donor, which was chemoenzymatically prepared from commercially available 5N-fTHF (Fig. S7)^20,21. 5N-fTHF was anaerobically acidified with 1 M HCl in the presence of 1 M mercaptoethanol to form 5,10N-methylene THF²⁰. Recombinant 5,10N-methylene THF cyclohydrolase FolD was then added to the neutralized reaction mixture to synthesize 10N-fTHF. Because 10N-fTHF is unstable and sensitive to molecular oxygen²¹, it was directly used for the in vitro reaction of LctF without purification, and all solutions used for the preparation and the in vitro reaction were bubbled with N₂ before use. Recombinant LctF was incubated with CoA in the presence of 10N-fTHF under a nitrogen gas atmosphere. We detected a specific peak, which eluted at the same retention time and had the same m/z as the chemically synthesized formyl-CoA, by LC-ESI-MS analysis only in the presence of LctF and 10N-fTHF (Fig. 4a). These results clearly indicated that LctF catalyzed the transfer of a formyl group to CoA with 10N-fTHF as a cofactor. To the best of our knowledge, this is the novel route to biosynthesize formyl-CoA using formyltransferase^22,23,24.

a LC-ESI-MS analysis (negative ion mode, monitored at m/z 794) of (i) formyl-CoA standard and reaction mixtures containing (ii) LctF and 10N-fTHF, (iii) 10N-fTHF, and (iv) LctF. b LC-MS analysis of reaction mixtures of LctA_AT1 and formyl-CoA; (i) without LctA_AT1, (ii) with LctA_AT1 (Fig. S8). A mass of 12794.3 corresponding to formyl-ACP1 (calculated mass, 12793.5) was detected in (ii). c LC-MS analysis of reaction mixtures of a triple mutant of LctA_AT1 (M-LctA_AT1) with (i) acetyl-CoA and (ii) formyl-CoA (Fig. S9). A mass of 12807.9 corresponding to acetyl-ACP1 (calculated mass, 12807.5) was detected in (i). d Relative activities of LctA_AT1 and M-LctA_AT1 toward formyl- (For-), acetyl- (Ac-), and malonyl- (Mal-) CoA. The activities were quantified with Ellman’s reagent (DTNB) reacting with free thiol released from the substrates. Error bars indicate standard deviation (SD) (n = 3), and data are presented as mean values ± SD. Black dots denote individual data points.

LctA_AT1 recognized formyl-CoA and transferred the formyl group to ACP1

We next examined the functions of the LctA_AT1 and ACP1 domains in LctA by in vitro analysis. Truncated holo-ACP1 prepared with Sfp²⁵ and recombinant LctA_AT1 were incubated with formyl-CoA, and the reaction mixtures were analyzed by LC-ESI-MS. A specific peak corresponding to formyl-ACP1 was observed only in the presence of LctA_AT1, confirming that LctA_AT1 accepted formyl-CoA and formed formylated ACP1 (Figs. 4b, and S8). This is the first example of an AT domain recognizing formyl-CoA as a substrate.

Recognition mechanism of formyl-CoA by LctA_AT1

LctA_AT1 recognized formyl-CoA but its ortholog, SalA_AT1, for salinosporamide A biosynthesis utilizes acetyl-CoA (Fig. 5a). We then examined whether LctA_AT1 accepts acetyl-CoA. However, no specific products were detected by LC-ESI-MS analysis (Fig. S8). To understand the basis of the strict substrate specificity of LctA_AT1, we constructed a modeled structure with local ColabFold (Fig. 5b)²⁶ and compared it with the modeled structures of SalA_AT1 (Fig. 5c)²⁷. The substrate binding pockets of LctA_AT1 and SalA_AT1 were also estimated by comparison with the crystal structure of malonate-bound AT in FabD (Fig. 5d)²⁸. As shown in Fig. 5c, the substrate binding pocket of LctA_AT1 was smaller than that of SalA_AT1 and this observation was reasonable because formyl-CoA is smaller than acetyl-CoA. The residues near the pocket were highly conserved between the two ATs; however, Leu718, Val794, and Leu840 in LctA_AT1 are replaced with Val741, Met820, and Val866 in SalA_AT1 (Fig. 5b, c), suggesting that these residues determine the pocket size. This hypothesis was supported by a modeled structure of mutant LctA_AT1 (M-LctA_AT1) possessing L718V/V794M/L840V replacements. Its pocket size was suggested to be sufficiently larger to accept acetyl-CoA, similar to SalA_AT1 (Fig. 5e). We therefore constructed M-LctA_AT1 and used it for in vitro assay with formyl-CoA and acetyl-CoA as the substrates (Fig. 4c). The LC-ESI-MS analysis revealed that M-LctA_AT1 could accept both formyl-CoA and acetyl-CoA, and the latter was more efficiently accepted (Figs. 4c, d and S9), indicating the importance of the three residues for recognition of the substrate.

a Alignments of orthologs of LctA_AT1 and SalA_AT1. Through genome database search, SalA_AT1 orthologs (1 to 3) and LctA_AT1 orthologs (4 to 9), which constitute similar gene clusters to the salinosporamide A and lactacystin clusters, respectively, were selected (Fig. S10). Residue numbers are based on LctA_AT1 and each accession number is described in Table S5. The red circles indicate pocket-forming residues conserved within each group (LctA_AT1 and SalA_AT1) but that are different between the groups. The His residues shown by the blue circle are usually Arg residues in acyltransferases utilizing malonyl-CoA. b Modeled structure of LctA_AT1. The amino acid residues shown in red are different from those in SalA_AT1. c Modeled structure of SalA_AT1. The substrate-binding pocket is shown by the dashed black circle. d Crystal structure of the AT in FabD accepting malonyl-CoA. The malonate-bound conserved Ser residue is shown (PDB: 2G2Z)²⁸. e Modeled structure of M-LctA_AT1 (L718V/V794M/L840V).

AT domains usually accept malonyl-CoA. However, neither LctA_AT1 nor SalA_AT1 accepted it (Fig. 4d). Previously, an Arg residue in ATs of fatty acids synthase accepting a malonyl-CoA (labeled in blue in Fig. 5d) was reported to be important to recognize malonyl-CoA by making a salt bridge with the carboxylic acid of malonyl-CoA (Fig. 5d)^28,29,30. The replacement of the Arg residue to Ala or Lys residue switched the substrate selectivity from malonyl-CoA to acetyl-CoA³⁰. Both LctA_AT1 and SalA_AT1 have His residues instead of Arg at the corresponding position (His 714 in LctA_AT1 and His737 in SalA_AT1, Fig. 5a–c). Therefore, the loss of the salt bridge might be a reason for the unavailability of malonyl-CoA by LctA_AT1 and SalA_AT1.

In vitro Analysis of the Adenylation Domain of NRPS in LctB

Because the hydroxyl group of lactacystin was suggested to be important for biological activity by structure-activity relationship studies⁷, the hydroxylation timing is an important issue for the engineering of lactacystin derivatives. The hydroxylation enzyme, LctD, perhaps accepts either free Leu or PCP-tethered Leu. We therefore investigated the substrate specificities of the A domain of LctB. We first tried to obtain a full-length LctB, but it was expressed as an insoluble form. Therefore, a truncated A domain was prepared. Because MbtH-like proteins were reported to interact with A domains and to enhance the adenylation activity³¹, a MbtH-like protein (LctE) was also prepared as a recombinant protein. The recombinant A domain and LctE were incubated with Leu or OH-Leu in the presence of ATP, and then the adenylation activities were measured by detecting the side product, inorganic diphosphate, by colorimetric assay³². The activity with Leu was 10-fold higher than that with OH-Leu (Fig. 6a). This result showed that the hydroxylation would occur after loading of the Leu moiety to NRPS. The adenylation activity without LctE was 20-fold lower than that with LctE (Fig. 6a), showing the importance of LctE in the adenylation reaction catalyzed by the A domain of LctB. The importance of LctE was also postulated by the multimer analysis between A domain in LctB and LctE using ColabFold (Fig. S11)^26,31.

a Relative activities of the A domain in LctB with Leu or OH-Leu in the presence of LctE (left) and with Leu in the absence of LctE (right). Error bars indicate standard deviation (SD) (n = 3), and the data are presented as mean values ± SD. Black dots denote individual data points. b HPLC analysis (detected with absorption at 210 nm) of PCP domains in reaction mixtures (i) containing full components (LctB_A domain, LctE and Leu), (ii) without the LctB_A domain, (iii) without LctE, and (iv) with OH-Leu. A mass of 14533.71 corresponding to Leu-PCP (calculated mass, 14533.19) was detected (Figure S12). c The chemical structure of 4’ (left) and LC-ESI-MS (right, positive ion mode monitored at m/z 198.1) analysis of the reaction mixture using 4’ as the substrate (i) with LctC and (ii) without LctC. A peak with an m/z of 198.1130 (m/z, [M + H]⁺) corresponding to the exact mass of 6 (198.1125; m/z, [M + H]⁺) was observed in (i) (Fig. S13).

We next examined whether the adenylated Leu was transferred to the PCP domain. A truncated PCP domain was co-expressed with sfp in E. coli to obtain the holo form. The holo-PCP was reacted with the A domain, ATP, and Leu or OH-Leu. LC-ESI-MS analysis showed that only Leu was transferred to the PCP domain (Figs. 6b and S12). These results clearly demonstrated that Leu is the substrate of LctB (NRPS) and suggested that the hydroxyl group is introduced during a late step of the biosynthesis.

In vitro analysis of the ketosynthase-like cyclase LctC

To examine the last step of the enzymatic reaction of lactacystin biosynthesis, we investigated the function of the KS-like protein LctC, which has 66% identity with SalC responsible for biosynthesis of salinosporamide A⁸. SalC was reported to catalyze an intermolecular aldol reaction using ketone with concomitant off-loading of the cyclized product from the carrier protein⁸. Therefore, we hypothesized that LctC also catalyzes a similar reaction although LctC was suggested to use a substrate possessing an aldehyde moiety instead of the ketone moiety. To examine this hypothesis, we synthesized an N-acetyl cysteamine thioester of methylmalonyl semialdehyde-Leu (4’) (Fig. 6c) and incubated it with LctC. LC-MS analysis revealed that the cyclization reaction occurred with concomitant release of the product from PCP in the presence of LctC (Figs. 6c and S13). The proposed catalytic residues of SalC are conserved in LctC and the reaction mechanism would be similar to SalC although LctC uses an aldehyde as an electrophile (Fig. S14)⁸.