Introduction
Recently, development of E. coli as a vehicle for heterologous metabolite biosynthesis has seen considerable progress. The greatest advantages of using E. coli are the wealth of knowledge available about its metabolic pathways and genetic makeup and the availability of well-established techniques for its genetic manipulation. Additionally, the ease of E. coli fermentation makes this organism particularly suitable for metabolite overproduction. Its tolerance toward heterologous protein production and its short doubling time also facilitate its use for metabolic engineering. However, so far, no biologically active complex natural product synthesized by heterologous polyketide synthase (PKS), NRP synthetase (NRPS) or mixed PKS/NRPS has been obtained de novo from E. coli. Therefore, we aimed to establish an E. coli system capable of total biosynthesis of biologically active forms of NRPs. To this end, we chose echinomycin (Compound 1) as our target (Scheme 1).
Scheme 1: The quinoxaline- and quinoline-type antibiotics and the proposed mechanism for echinomycin biosynthesis.
(a) Proposed pathway for QC biosynthesis (yellow). (b) Proposed pathway for octadepsipeptide core-structure biosynthesis (green). (c) Proposed mechanism for peptide chain homodimerization and cyclorelease from Ecm7 (green) and subsequent modifications for the formation of the quinoxaline antibiotics (pink). (d) Chemical structures of quinoline-type antibiotics. Enzymes are represented by pentagonal boxes, and the catalytic domains found within the enzymes are represented by A, adenylation; C, condensation; T, thiolation; E, epimerization; M, methyltransferase; and TE, thioesterase. TANDEM (Compound 11)15 is a synthetic analog of triostin A (Compound 2).
Full size image (119 KB)Compound Compound 1 is an NRP isolated from various bacteria, including Streptomyces lasaliensis4, that belongs to a large family of quinoxaline antibiotics, which have quinoxaline chromophores attached to the C2-symmetric cyclic depsipeptide core structure. The great interest in this group of compounds stems from their potent antibacterial, anticancer and antiviral activities. Many, including Compound 1 and triostin A (Compound 2), have nanomolar potency2. Also, Compound 1 contains unique chemical structures, including the quinoxaline-2-carboxylic acid (QC) moiety and the thioacetal bridge, whose biosynthetic mechanisms remain unknown.
We have isolated the echinomycin biosynthetic gene cluster from the S. lasaliensis linear plasmid (Supplementary Methods online). DNA sequence analysis of the 36-kilobase-long cluster (Fig. 1) revealed the presence of eight genes that seemed responsible for QC biosynthesis (ecm2, ecm3, ecm4, ecm8, ecm11, ecm12, ecm13 and ecm14), five genes for peptide backbone formation and modifications (ecm1, ecm6, ecm7, ecm17 and ecm18) and a resistance gene (ecm16). Based on the previous finding that L-tryptophan is the precursor for QC5 and on the predicted functions of the proteins Ecm2, Ecm3, Ecm4, Ecm8 Ecm11, Ecm12, Ecm13 and Ecm14, we hypothesized that QC biosynthesis parallels the first stage of nikkomycin6 biosynthesis, in which Ecm12 hydroxylates the L-tryptophan bound to the thiolation domain of Ecm13 (Scheme 1a). The product (2S,3S)-
-hydroxytryptophan (Compound 3), determined to be an intermediate by substrate feeding experiments, is released from Ecm13 by the thioesterase activity of Ecm2. Then, as in the first two steps of kynurenine biosynthesis7, oxidative ring opening of Compound 3 by Ecm11 and subsequent hydrolysis by Ecm14 can yield -hydroxykynurenine (Compound 4). Subsequently, oxidative cyclization and hydrolysis of 4 by Ecm4 to form N-(2'-aminophenyl)--hydroxyaspartic acid (Compound 5) and oxidation of Compound 5 by Ecm3 to form N-(2'-aminophenyl)--ketoaspartic acid (Compound 6) can follow. Finally, Compound 6 can undergo spontaneous decarboxylation, cyclic imine formation and oxidative aromatization to yield QC.
Figure 1: The echinomycin biosynthetic cluster from S. lasaliensis.
(a) The organization of the echinomycin biosynthetic gene cluster isolated from S. lasaliensis. (b) Predicted fatty acid synthase gene organization in S. lasaliensis. (c) Deduced functions of the ORFs of the S. lasaliensis echinomycin biosynthetic gene cluster and fatty acid synthase acyl carrier protein. Expected value is the number of matches expected to be found purely by chance for a given query sequence in a database29; n.a., not applicable.
Full size image (90 KB)Curiously, the aryl carrier protein (ArCP) required for incorporating QC into Compound 1 was absent from the cluster. However, as in the proposed triostin A biosynthesis8, we expected the adenylation domain–containing Ecm1 to activate and transfer QC to the phosphopantetheine arm of FabC, the fatty acid biosynthesis acyl carrier protein (ACP). The first module of the bimodular NRPS Ecm6 can accept QC-S-FabC as the starter unit. Overall, Ecm6 and the NRPS Ecm7 catalyze 17 chemical reactions for peptide core formation (Scheme 1b). Ecm7 contains a terminal thioesterase domain that seems to homodimerize and cyclorelease the peptide chain (Scheme 1c)9. The cyclized product can then become the substrate for the oxidoreductase Ecm17, which can catalyze an oxidation reaction within the reducing cytoplasmic environment to generate the disulfide bond in Compound 2.
The last step of echinomycin biosynthesis involves an unusual transformation of the disulfide bridge of Compound 2 into a thioacetal bridge10. Ecm18, which is highly homologous to a known S-adenosyl-L-methionine (SAM)-dependent methyltransferase, is thought to be responsible for this transformation (Fig. 2a). To verify this idea, we have demonstrated in vitro that purified Ecm18 catalyzes the transformation of Compound 2 to Compound 1 in the presence of SAM (Fig. 2b–d). This is the first finding of a single methyltransferase's being responsible for the biotransformation of a disulfide bridge into a thioacetal bond. Notably, a group of compounds, SW (Compound 7,Compound 9)11 and UK (Compound 8,Compound 10)12 (isolated from Streptomyces sp. SNA15896 and S. braegensis, respectively), were found to have an identical backbone while having different modifications at the interbackbone bridge (Scheme 1d). Whereas the conversion of Compound 7 to Compound 8 can proceed the same way as the bioconversion of Compound 2 to Compound 1, the formation of Compound 9 and Compound 10 demands attaching alkyl substituents of differing lengths onto the disulfide bridge. Nonetheless, the reaction mechanism, which involves formation of a sulfonium ylide through deprotonation followed by methyl transfer13, 14, is similar to the proposed mechanism for the Ecm18-catalyzed reaction, and it can explain the way in which a single SAM-dependent methyltransferase can perform an iterative methylation of a disulfide bridge followed by deprotonation and rearrangement to yield Compound 9 and Compound 10 (Fig. 2a). Given that Ecm18 mediates neither further methylation of Compound 1 nor installation of a thioacetal bridge in TANDEM (Compound 11, a synthetic des-N-methyl derivative of Compound 2; ref. 15), we suspect the presence of another related methyltransferase that can iteratively methylate a disulfide bond. As methyltransferases capable of performing multiple rounds of methylation have not been reported so far, we are currently searching the SNA15896 genome for the biosynthetic cluster involved in the biosynthesis of Compound 7 and Compound 9 and their associated methyltransferase(s) to obtain further insight into this unique enzymatic mechanism of thioacetal bridge formation.
Figure 2: LC-MS analyses of Ecm18-catalyzed thioacetal formation.
(a) Proposed mechanism for the formation of thioacetal bridges with different alkyl substituents. Black arrows represent the proposed steps associated with Ecm18-catalyzed conversion of Compound 2 to Compound 1 in the echinomycin biosynthetic pathway, whereas red and green arrows represent the proposed pathways for the biosynthesis of the differently alkylated thioacetal bridge in the antibiotics from Streptomyces sp. SNA15896 and S. braegensis, respectively. Ha, methylene proton located on the C of the N-methylcysteine residue; Hb, proton located on the methyl group donated by SAM. (b) LC-MS analysis of the crude extract of the reaction mixture for the Ecm18-catalyzed conversion of Compound 2 to Compound 1. (c) LC-MS analysis of the product Compound 1 isolated from the crude extract of the reaction mixture. (d) LC-MS analysis of the substrate Compound 2 isolated from the crude extract of the reaction mixture. In b–d, the UV trace (
= 254 nm) is shown on top and the MS spectrum is shown on the bottom.
For the E. coli production of Compound 1, after confirming the feasibility of expressing each of the 15 S. lasaliensis genes (ecm1, ecm2, ecm3, ecm4, ecm6, ecm7, ecm8, ecm11, ecm12, ecm13, ecm14, ecm16, ecm17, ecm18 and fabC) in E. coli (Supplementary Fig. 1 online), we assembled the 15 genes, along with the Bacillus subtilis phosphopantetheine transferase gene sfp (which is known to efficiently phosphopantetheinylate heterologous ACPs and thiolation domains16), into three separate plasmids (Supplementary Fig. 2 online), with each gene carrying its own T7 promoter, ribosome-binding site and T7 transcriptional terminator. We chose this multimonocistronic arrangement for our multigene assembly not only to simplify the assembly process but also to minimize both potential premature terminations and mRNA degradation17 in transcribing excessively long polycistronic gene assemblies. Moreover, we used orthogonal origins of replication and antibiotic resistance genes to ensure the stable retention of all three plasmids in E. coli18 (Supplementary Fig. 3 online). The E. coli strain BL21 (DE3) that we transformed with the three plasmids was subjected to 8-day-long fed-batch fermentation in minimal medium. We purified the culture extract for Compound 1 through a series of chromatographic steps to give the final yield of 0.3 mg of Compound 1 per liter of culture. When we analyzed the purified Compound 1 by ESI-MS (Fig. 3a) and MS/MS (Fig. 3b), the data were consistent with the chemical structure of Compound 1. Also, the 1H NMR spectrum (Fig. 3c) showed the presence of characteristic resonances at 
9.66 (s, 1 H, QC H-3), 9.64 (s, 1 H, QC H-3) and 2.10 (s, 3 H, -SCH3). Additional TOCSY NMR data (Fig. 3d) confirmed unambiguously that our engineered E. coli produced Compound 1. These findings establish for the first time the identity of a set of genes sufficient for the biosynthesis of Compound 1. More importantly, the result serves as the first example of de novo production of a bioactive form of a heterologous NRP in E. coli. Furthermore, to demonstrate the ease and effectiveness of modifying the E. coli–based heterologous biosynthetic system, we chose to convert the echinomycin biosynthetic pathway into a triostin A biosynthetic pathway. We modified the plasmid by simply removing ecm18. The resulting strain produced the expected compound Compound 2 at a yield of 0.6 mg per liter of culture, as confirmed by LC-MS, MS/MS, 1H NMR and TOCSY 1H NMR analyses (Fig. 4).
Figure 3: Chemical characteristic spectra of compound 1 produced by the engineered E. coli strain.
(a) LC-MS spectrum of Compound 1. (b) MS/MS spectrum of Compound 1 collected at the collision energy of 4.0 eV. Important fragment ions30 that are assigned are shown in red. (c) 1H NMR spectrum of Compound 1. (d) 1H NMR TOCSY spectrum of Compound 1 collected at the mixing time of 100 ms, showing important correlations (numbered in red). Correlation 1: 0.95–0.75 (Val–CH3) and 5.15 (Val–H
); 2: 1.15–1.05 (Val–CH3) and 5.21 (Val–H); 3: 1.45–1.35 (Ala–CH3) and 5.00–4.90 (Ala–H); 4: 1.45–1.35 (Ala–CH3) and 6.96 (Ala–NH), and 1.45–1.35 (Ala–CH3) and 6.81 (Ala–NH).
Figure 4: Chemical characteristic spectra of compound 2 produced by the engineered E. coli strain.
(a) LC-MS spectrum of Compound 2. (b) MS/MS spectra of Compound 2 collected at the collision energy of 4.5 eV, (c) 4.0 eV and (d) 3.0 eV. Important fragment ions30 in the MS/MS spectra that are assigned are shown in red. (e) 1H NMR spectrum of Compound 2. (f) 1H NMR TOCSY spectrum of Compound 2 collected at the mixing time of 100 ms, showing important correlations (numbered in red). Correlation 1: 0.72 (Ala–CH3) and 5.15–4.90 (Ala–H); 2: 0.72 (Ala–CH3) and 8.39 (Ala–NH); 3: 1.46 (Ala–CH3) and 4.85–4.70 (Ala–H); 4: 1.46 (Ala–CH3) and 7.21 (Ala–NH); 5: 0.87 (Val–CH3) and 4.27 (Val–H); 6: 0.87 (Val–CH3) and 5.22 (Val–H); 7: 1.06 (Val–CH3) and 4.27 (Val–H); 8: 1.06 (Val–CH3) and 5.22 (Val–H); 9: 1.10 (Val–CH3) and 4.27 (Val–H), and 1.12 (Val–CH3) and 4.27 (Val–H); 10: 1.10 (Val–CH3) and 5.22 (Val–H), and 1.12 (Val–CH3) and 5.22 (Val–H).
When introducing any exogenous biosynthetic pathway into E. coli, the toxicity of the biosynthetic product can impair the host. This problem can be circumvented by introducing a self-resistance mechanism into E. coli that confers resistance to the host without destroying the product. For echinomycin biosynthesis, the homology between Ecm16 and daunorubicin resistance–conferring factor (DrrC)19 and the similarity of the mode of action between Compound 1 and daunorubicin20 suggest that Ecm16 achieves nondestructive resistance against Compound 1 in S. lasaliensis. Indeed, we were able to demonstrate that ecm16 confers echinomycin resistance to BL21 (DE3). Also, when ecm16 was absent from our system, the growth of the host was hampered, suggesting that sufficient yields of Compound 1 and Compound 2 would have been unattainable without the self-resistance mechanism in place.
In conclusion, our study has demonstrated for the first time the viability of E. coli–based total biosynthesis of a bioactive form of heterologous, complex NRPs from simple carbon and nitrogen sources, paving the way to the development of an economical, general platform for one-pot mass production of natural products and their analogs. Our system shows that using a multiplasmid, multimonocistronic gene assembly is a straightforward, highly stable and easily modifiable approach for establishing and engineering exogenous biosynthetic pathways in E. coli. With the use of appropriate orthogonal selection markers and origins of replication, in combination with other potential approaches such as chromosome integration21, the introducion of even larger, more complex biosynthetic pathways seems possible. Combining our current efforts with the successes in introducing other PKS22, 23 and mixed PKS-NRPS24 pathways into E. coli and in engineering of PKS25 and NRPS26 should help broaden the scope of E. coli–based heterologous mass production of a wide range of natural products and their analogs.
Methods
Assembly of the plasmid-based echinomycin biosynthetic gene cluster.
Initially, we cloned each open reading frame (ORF) individually from cosmids harboring the echinomycin biosynthetic cluster into the pET28b (Novagen)-derived plasmid pKW409 as either an NdeI-EcoRI or an NdeI-XhoI fragment prepared by PCR (Supplementary Methods online). In pKW409, we moved the single XbaI recognition site to the 5' side of the T7 promoter and created an SpeI recognition site at the 3' side of the T7 terminator. The assembly formation exploited the compatibility of the cohesive ends generated by XbaI and SpeI digestion. The cassette arrangement not only facilitated evaluation of the expression level of each gene individually but also was necessary for rapid construction of the multimonocistronic gene assemblies (Supplementary Fig. 2).
Thioacetal formation assay.
We preincubated the assay mixture containing 10 
M Ecm18 and 10 mM SAM in 0.1 M Tris-HCl at pH 7.2 and 30 °C for 5 min. After adding Compound 2 to the final concentration of 1 mM, we incubated the reaction at 30 °C for 5 min before terminating it by additing 10% (w/v) SDS. We extracted the reaction mixture with ethyl acetate and concentrated the extract in vacuo to yield a white residue. We isolated a mixture of the substrate Compound 2 and the product Compound 1 from the extracts using preparative TLC (5% MeOH/CHCl3). We analyzed the isolated samples of Compound 1 and the unreacted substrate Compound 2, along with the mixture of the two, independently by LC-MS using Alltech 2.1 
100 mm C18 reverse-phase column. We subjected samples to a linear gradient of 5 to 95% CH3CN (v/v) in H2O supplemented with 0.05% (v/v) formic acid over 60 min at a flow rate of 0.15 ml min-1 at room temperature (25 °C). The LC-MS analysis of the mixture showed the presence of both compounds Compound 1 and Compound 2 (Fig. 2b). Compound Compound 1 eluted at 41.65 min under the conditions described above and yielded characteristic ions at m/z = 1,139.24 [M+K]+, 1,123.34 [M+Na]+, 1,101.29 [M+H]+ and 1,053.38 [M–SCH3]+ (Fig. 2c). The result matched well with the analysis of the authentic sample of Compound 1 (Supplementary Fig. 4 online). Similarly, Compound 2 eluted at 41.15 min and yielded characteristic ions at m/z = 1,125.30 [M+K]+, 1,109.44 [M+Na]+ and 1,087.49 [M+H]+, as expected (Fig. 2d and Supplementary Fig. 5 online).
E. coli production of 1 and 2.
We incubated BL21 (DE3) transformed with pKW532/pKW538/pKW541 and pKW532/pKW539/pKW541 for the production of Compound 1 and Compound 2, respectively (Supplementary Fig. 2), at 37 °C overnight in 2 ml LB medium and subsequently in 100 ml M9 minimal medium. We used the entire culture to inoculate 1.5 liters of M9 minimal medium kept at 37 °C, pH 7.0 by the BioFlo110 fermentor system (New Brunswick Scientific). Once the glucose was exhausted from the medium, as indicated by a sudden increase of the dissolved oxygen concentration, feeding of the feed media24 was initiated. When the culture reached an optical density at 600 nm of 11, we reduced the temperature to 15 °C and added IPTG to the final concentration of 200 M. After 8 d of incubation, we centrifuged the culture to separate the supernatant and the cells. We extracted the supernatant and the cell pellet with ethyl acetate and acetone, respectively. We combined the extracts and concentrated them in vacuo to yield an oily residue, which we fractionated by silica gel flash column chromatography with 50% MeOH/CHCl3. We collected the fractions containing the target compound and further purified them through a series of preparative TLC (i, 50% EtOAc/hexane; ii, 2-butanone; iii, 5% MeOH/CHCl3) to afford purified samples of Compound 127 and Compound 228. We fully characterized the isolated compounds using LC-MS, MS/MS, 1H NMR and 1H NMR TOCSY. The MS and 1H NMR spectra of the authentic reference of Compound 1 (Supplementary Fig. 4) and Compound 2 (Supplementary Fig. 5) are provided for comparison.
Echinomycin resistance assay.
We determined E. coli echinomycin resistance using BL21 (DE3) transformed with pKW409 carrying ecm16. We incubated the transformant in 3 ml LB medium at 37 °C for 5 h. We spread the culture on LB agar plates containing two different concentrations of echinomycin (10 and 100 g ml-1) supplemented either with or without 300 M IPTG. We incubated the plates at 37 °C overnight to determine colony formation.
Accession codes.
The echinomycin biosynthetic gene cluster sequence has been deposited in the DNA Data Bank of Japan (DDBJ: http://www.ddbj.nig.ac.jp/) with the accession numbers AB211309 and AB211310.
Note: Supplementary information is available on the Nature Chemical Biology website.
