Reprogramming of the antimycin NRPS-PKS assembly lines inspired by gene evolution
Reprogramming of the NRPS/PKS assembly line is an attractive method for the production of new bioactive molecules. However, it is usually hampered by the loss of intimate domain/module interactions required for the precise control of chain transfer and elongation reactions. In this study, we first establish heterologous expression systems of the unique antimycin-type cyclic depsipeptides: JBIR-06 (tri-lactone) and neoantimycin (tetra-lactone), and engineer their biosyntheses by taking advantage of bioinformatic analyses and evolutionary insights. As a result, we successfully accomplish three manipulations: (i) ring contraction of neoantimycin (from tetra-lactone to tri-lactone), (ii) ring expansion of JBIR-06 (from tri-lactone to tetra-lactone), and (iii) alkyl chain diversification of JBIR-06 by the incorporation of various alkylmalonyl-CoA extender units, to generate a set of unnatural derivatives in practical yields. This study presents a useful strategy for engineering NRPS-PKS module enzymes, based on nature’s diversification of the domain and module organizations.
Polyketides and non-ribosomal peptide hybrids are an important class of natural products, and include a variety of bioactive compounds, exemplified by the anti-cancer agents, bleomycin, epothilone, and calyculin1,2,3,4. They are built up by a complex system composed of polyketide synthases (PKSs) and non-ribosomal peptide synthetases (NRPSs)5,6,7, which are thiotemplate module enzymes consisting of several catalytic domains. In the PKS system, the ketosynthase (KS) domain catalyzes decarboxylative Claisen condensation between a starter acyl group and an extender unit thioesterified to an acyl carrier protein (ACP) via the action of an acyltransferase (AT), during a single polyketide extension cycle. In the NRPS system, the amino acid building blocks are activated as adenylates and thioesterified to the thiolation (T) domain by the adenylation (A) domain. The condensation (C) domain forms a peptide bond between two acylated T domains during a single peptide extension cycle. In both systems, if the termination domain is a thioesterase (TE), then chain release typically occurs by hydrolysis or macrocyclization. A set of domain complexes responsible for one cycle of chain extension is called a module. The amino acid sequences within subunits that join domains and modules covalently are referred to as a linker, and those which are located on distinct subunits and mediate communication between modules are referred to as a docking domain8,9. In the NRPS–PKS complex systems, the product of the NRPS module is transferred onto the KS domain of PKS module and condensed with a polyketide extender unit by PKS, or the product of PKS module is condensed with an amino acid to form a peptide bond by NRPS. The substrate specificities of the KS and C domains are usually rigid, thus contributing to the maintenance of the precise chain transfer order. Several reports have clarified the reaction mechanism of each module and domain in the NRPS and PKS systems in terms of the protein structure6; however, the engineering of these systems usually accompanies significant loss of productivity due to the strict regulation of the module enzymes10.
To overcome such problems, the optimal joint connections for each catalytic domain and module should be identified. Recently, Bode and co-workers reported a new strategy for the manipulation of NRPSs by using exchange units that are sets of A-T-C domains that can be transplanted into active chimeric modules11. However, this strategy may not be necessarily applied to hybrid NRPS-PKS systems because the quaternary organization of NRPS-PKS is likely different from that of pure NRPS system. In fact, the quaternary structure of NRPS module is defined as a monomer in the early chromatographic12 and X-ray crystallographic studies13, but we cannot exclude that it works as a dimer in the NRPS-PKS system, as the docking domain from tubulysin system forms a homodimer in the solution9. It has thus remained challenging to establish a general rule for rational reconstruction of module assembly lines, and scientists still rely on trial-and-error strategies. One possible solution is to learn how nature diversifies the domain and module organizations by taking advantage of bioinformatics analyses, through comparisons of closely related enzymes. In nature, various module enzymes have evolved through horizontal gene transfer, gene deletion, mutation, and so on14,15,16,17 to form functional sets of modules. Thus, for cis-AT PKSs, the predominant evolutionary model is a repeated cycle of duplication of ancestral modules coupled with domain exchanges14,15,17, while for the distinct class of trans-AT PKSs, it is rather horizontal gene transfer that appears to dominate16,17. However, the specific evolutionary history of hybrid cis-AT PKS/NRPS system has yet to be rigorously investigated. It is a promising way to modify the module compositions according to the evolutionary course, because this method is effective to maintain the connectivity between multiple modules while minimizing the change of the overall structure and the protein–protein interactions which underlie their function. The knowledge on the constraint for NRPS-PKS engineering should be accumulated more to understand how this system has been evolved and can be manipulated. This could lead to computational platforms which clearly predict domain/linker/docking-domain boundaries as the one established in cis-PKS system for designing chimeric modules18.
Antimycins (1) are a group of di-lactone depsipeptides with various important biological properties, including anti-fungal, insecticidal, anti-cancer, and anti-inflammatory activities19. They are biosynthesized by the NRPS-PKS system from substrates including a unique 3-formamidosalicylate (3-FSA) starter unit, L-threonine, pyruvate, and alkylmalonyl-CoA (Fig. 1). Their structures are characterized by the depsipeptides constructed by two NRPS modules (AntC), including one module with an unusual ketoreductase (KR)-domain, one PKS (AntD) that generates alkyl group variations at C-9 by the incorporation of a broad range of extender substrates provided by a crotonyl-CoA reductase/carboxylase (CCR)20,21,22,23 (AntE), and a reductase (AntM). Zhang’s group and we have created ~100 antimycin derivatives, by employing the biosynthetic enzymes with the tolerant substrate specificities in the antimycin systems23,24,25,26,27. Interestingly, in nature, there are tri-lactone, tetra-lactone, and penta-lactone antimycin-type depsipeptides that also utilize 3-FSA as a common starter unit19, including the tri-lactone JBIR-06 (2)28 and the tetra-lactone neoantimycin A (3a)29 (Fig. 1), known downregulators of GRP78, a molecular chaperone related to resistance to chemotherapy and hypoglycemic stress30,31. In 2013, Magarvey’s group reported two NRPS-PKS gene clusters based on bioinformatic analyses32, and proposed that they are involved in the biosyntheses of 2 and 3a, respectively; however, their proposal has not been validated experimentally and the sequence information has not been deposited in GenBank. The NRPS-PKS assembly line organizations involved in the biosyntheses of 1, 2, and 3a are very similar, and the conservation of the two modules that incorporate 3-FSA and L-threonine in these biosynthetic systems implies that they are generated from the same ancestor. Very recently, Zhang’s group also reported the neoantimycin gene cluster and showed the extender substrate of its PKS module was malonyl-CoA33. Here, we control the size of lactone rings of JBIR-06 and neoantimycin, and introduce diverse alkyl groups into the JBIR-06 scaffold, by rationally modifying their assembly lines with the aid of the information how they are diversified, to yield various bioactive antimycin-type depsipeptides.
Organization of modular enzymes involved in the biosynthesis of antimycin-type depsipeptides. C condensation, A adenylation, T thiolation, KR ketoreductase, KS ketosynthase, AT acyltransferase, ACP acyl carrier protein, MT methyltransferase, TE thioesterase, CCR crotonyl-CoA reductase/carboxylase, R1 and R2 various alkyl groups, Rstarter the acyl moiety of 3-FSA. The module and domain organizations of the starter, the module 1, and the module 2 are identical among the all three systems. The starter and the module 1 uptake 3-FSA and L-threonine, respectively, common to all pathways, but the module 2 uptake pyruvate (AntC), isoleucic acid (SmlB), and valic acid (NatB), respectively. JBIR-06 and neoantimycin systems include an additional NRPS modules (module 3) after the module 2 in the same ORF. The module 3 in SmlB uptakes phenylpyruvic acid, and the module 3 in NatB uptakes leucic acid. The PKS modules of JBIR-06 and neoantimycin systems (module 4) contain MT domain between AT and ACP, and accept malonyl-CoA as an extender unit, to yield dimethyl group at the α-position of polyketide moiety, differently from antimycin system. Furthermore, neoantimycin system includes an extra NRPS module (module 5 as NatD) which uptakes isoleucic acid or valic acid
Full size image
JBIR-06 and neoantimycin biosynthetic gene clusters
In this work, we independently identified the biosynthetic gene clusters of the tri-lactone JBIR-06 and the tetra-lactone neoantimycin (the sml and nat clusters) from each producing strain (Supplementary Figs. 1–2), and heterologously expressed them in Streptomyces lividans TK21 to identify their products. The JBIR-06 and neoantimycin gene clusters were each cloned into pKU51834,35, a bacteriophage-based BAC vector, resulting in pKU518J06 and pKU518nant, respectively (Supplementary Note 1, Supplementary Tables 1–2). From the sequence data of these two vectors, we confirmed that the assembly line organizations of JBIR-06 and neoantimycin NRPS-PKSs were identical with those in the previous report32, except for NatD (Fig. 1). In contrast to the Magarvey’s report, where natD and natE encode A-KR and T-TE domains, respectively, natD encodes a whole single module consisted of C-A-KR-T-TE domains in our sequence data. As Zhang’s group also reported the same NatD domain organization as ours33, the current module organization of neoantimycin is more plausible. HPLC analyses of the culture extracts of the recombinants revealed that S. lividans/pKU518J06 and S. lividans/pKU518nant yielded 2 and 3a–3b, respectively (Supplementary Fig. 3). Through NMR and MS analyses, 2, 3a, and 3b were identified as JBIR-06, neoantimycin A, and neoantimycin F (Supplementary Figs. 4–7, Supplementary Table 3), by comparisons with data in the literature28,36,37. These results linked 2 to the product of sml cluster, and 3a–3b to those of nat cluster, and provided the platform for the engineered biosynthesis.
Bioinformatic analyses of the NRPS-PKS module structures
Sequence analyses revealed that sml and nat NRPS/PKS gene clusters encode four and five modules, respectively, to build up the molecular scaffold of the tri-lactone JBIR-06 and the tetra-lactone neoantimycin (Fig. 1). We conducted the DNA sequence alignment for A, C, T, and KR domains from ant, sml, and nat system, respectively, and calculated their identities (Supplementary Fig. 8). Due to the high amino acid sequence identities (i.e., the PKS modules AntD and SmlC share 51% amino acid sequence identity, while AntD and NatC share 53%), we could precisely identify functional domain boundaries, and thus intervening linkers or adjacent docking domains (Fig. 2a, Supplementary Fig. 9). It is critical to retain the intimate domain/module interactions that are required for precise control of the chain transfer and elongation reactions for successful engineering of the NRPS/PKS assembly lines. As a result, as shown in Fig. 2a, we annotated region 1 (1301–1371 in SmlC and 1320–1390 in NatC) as the ACP domain of PKS module, region 2 (1393–1647 in SmlC), which is only conserved in AntD and SmlC, as the TE domain, and the region surrounded by ACP and TE (1372–1392 in SmlC) as an interdomain linker (Fig. 2a). Homology modeling (SWISS-MODEL server, https://swissmodel.expasy.org) supports the hypothesis on its structure including region 1 forms four α-helices and two loops, which are involved in the interactions with the KS-AT domains38,39 (Fig. 2b, Supplementary Fig. 10). The C-terminal region of NatC (1391–1411) was predicted to be a docking domain between PKS and NRPS modules, as the one previously reported for the epothilone NRPS-PKS system40. Therefore, with the bioinformatic information in our hands, we anticipated that the deletion of the NatD NRPS module from the tetra-lactone neoantimycin assembly line would lead to ring contraction (Fig. 3a), whereas the addition of the NatD module to the tri-lactone JBIR-06 system would expand the ring structure (Fig. 4a). Furthermore, in order to maintain the intimate domain/module interactions, we employed the TE domain of the SmlC PKS module (TESmlC), as well as the predicted interdomain linker between the ACP and TE domains of SmlC, for the ring contraction, while the docking domain between the NatC PKS module and the NatD NRPS module was kept for the ring expansion experiment. We therefore took advantage of this information to redesign the chimeric NRPS-PKS assembly lines for the ring contraction and expansion, as described below.
Bioinformatic analyses of the antimycin PKS modules. a Alignment of the PKS modules, AntD, SmlC, and NatC. We annotated SmlC1301–1371 as an ACP domain (region 1), SmlC1372–1392 as an interdomain linker, SmlC1393–1647 as a TE domain (region 2), and NatC1391–1411 as a docking domain. b A homology model of ACPNatC (1320–1411), based on the reported crystal structure of the surfactin synthase ACP subunit (PBD: 2VSQ) as the template. c A homology model of the ACPNatC-TESmlC domains, the NRPS module from Acinetobacter baumannii (NCBI PBD ID 4ZXH_A, https://www.ncbi.nlm.nih.gov/protein/4ZXH_A) was used as the template
Full size image
Ring contraction of neoantimycin A. a The reconstructed module structure for ring contraction. b HPLC analyses of the metabolites from transformants harboring (i) pKU518nant and (ii) pKU518nantΔnatD::smlCTE. The chromatogram represents the UV absorbance at 320 nm. * and ** indicate unidentified analogs with m/z values that were 14 and 28 less than 3b, respectively. c Engineering scheme and structure of 4
Full size image
Ring expansion of JBIR-06. a The reconstructed module structure for ring expansion. b HPLC analyses of the metabolites from transformants harboring (i) pKU518J06 and (ii) pKU518J06ΔsmlCTE/pZH2-NatD. The chromatogram represents the UV absorbance at 320 nm. *** indicates a mixture of analogs with m/z values that are 14 less than 5. c Engineering scheme and structure of 5 and 6
Full size image
Notably, our sequence analysis of the C-terminal region of NatC revealed that NatC1391–1411 contains several positively charged residues while other C-terminal docking domains that interact with NRPS subunits contain a run of negatively charged residues (Supplementary Fig. 11A). Further, the three α-helices and two β-turns, annotated as a docking domain in the EpoB NRPS structure9, are uniquely replaced with a shorter 9 amino acid sequences in NatD (Supplementary Fig. 11B). This suggests that the docking between NatC and NatD is fundamentally different from other interfaces. Despite these dissimilarities, our engineering strategy with the docking domain of NatC worked (see the section of ring expansion).
Ring contraction of neoantimycin A
We first conducted the ring contraction of the tetra-lactone neoantimycin A (3a), by deleting the NatD NRPS module from nat cluster and connecting the linker-TESmlC domain right after the C-terminus of NatC, to form a NatC-SmlC chimeric PKS (Figs. 2c, 3a). TE domains usually exhibit tolerant substrate specificities, as exemplified by the early studies where the TE domain from erythromycin system was shown to accept polyketide chains of diverse lengths41, and recognize the functional group at the acyl terminus and adjacent to the attacking nucleophile42. Therefore, we expected that TESmlC domain, which recognizes the acyl group derived from polyketide and the hydroxyl group of threonine as a nucleophile, would accept the intermediate from the new construct and catalyze macrocyclization. The interdomain linker of SmlC was also employed, to connect the ACPNatC domain to the TESmlC domain (Fig. 2a). The S. lividans strain harboring pKU518nantΔnatD::smlCTE, accordingly engineered with the Red/ET system43 (Supplementary Note 2, Supplementary Tables 1–2), was incubated in A3M medium, and the extract of the transformant was analyzed by HPLC. As expected, the production of 3a–3b was completely abolished, and 4 newly appeared (Fig. 3b). LC-MS analyses showed that the m/z of 4 was 581, which is 116 less than 3a, and its UV max was around 315 nm, characteristic of the antimycin starter unit (Supplementary Figs. 4, 12, Supplementary Table 3), suggesting that 4 was generated from the ring contraction of 3a. The 1H and 13C NMR data defined the structure of the side chain, and the HMBC correlations from H-28, H-29, and H-30 to C-8 (Supplementary Figs. 13–18, Supplementary Table 4) clearly showed that the hydroxyl group of the L-threonine moiety was esterified with the carboxyl group of the polyketide moiety, indicating that 4 is a trilactone generated by the loss of isoleucic acid from 3a (Fig. 3c). 4 bears a keto group at C-1, indicating that the ketoreduction unexpectedly did not work in the recombinant strain. It might be caused by the polar effect on the expression of natF, a ketoreductase gene located downstream of natD, though we cannot eliminate the possibility that NatF no longer recognizes the substrate. The yield of 4 (3.9 ± 0.7 mg/L) was 3-fold reduced relative to that of 3a (12 ± 2.0 mg/L) (Table 1), but it did not drop significantly, verifying the accuracy of our domain structure prediction.
Table 1 The yields of isolated products
Full size table
Ring expansion of JBIR-06
We next enlarged the lactone size of the tri-lactone JBIR-06 (2) by creating an sml/nat chimera. To this end, we appended the NatD NRPS module to sml cluster and replaced the linker and TE domain in the SmlC PKS module with the docking domain of NatC (Fig. 4a, Supplementary Fig. 6A), to maintain the module/module interactions to yield the tetra-lactone product. The accordingly engineered BAC, pKU518J06ΔsmlCTE (Supplementary Note 3, Supplementary Tables 1–2), was introduced into S. lividans with the natD expression vector, based on the ϕC31 phage integration system (pZH2-NatD) (Supplementary Note 4, Supplementary Fig. 19). As a result, the original product 2 disappeared completely, and 5 and 6 newly appeared in the transformant (Fig. 4b). The planar structure of 5 was determined as the tetra-lactone, which contains one threonine, two isoleucic acids, and one leucic acid (Fig. 4c), through a comparison of the 1D NMR data with the literature (Supplementary Figs. 20–23, Supplementary Table 5)27. The HMBC correlations between H-9 and C-8/-10, and H-7 and C-8 and MS data indicated that the additional isoleucic acid was introduced into the C-terminus of the depsipeptide chain of 2, and lactonized with the hydroxyl group of threonine to yield 5 (Supplementary Figs. 4, 18, and 24, Supplementary Table 3). The stereochemistry of each building block was identified as L-threonine, L-isoleucic acid, and L-leucic acid with the chiral GC-MS analyses of the acid-hydrolyzed fragments of 5 and 6, and the linear depsipeptide consisted of 3-FSA starter unit, threonine, isoleucic acid, and leucic acid (Supplementary Figs. 25–30, Supplementary Table 6). The yield of 5 (5.9 ± 1.6 mg/L) was almost the same as that of 2 (5.9 ± 0.7 mg/L, from S. lividans/ pKU518J06) (Table 1), indicating that NatD accepted the intermediate from SmlC efficiently and esterified one more L-isoleucic acid with the intermediate, to produce the tetra-lactone 5. The unexpected accumulation of the acyclic 6 (9.7 ± 0.8 mg/L) might have been caused by the extended residence time of the intermediate on SmlB relative to the native system, leading to its spontaneous hydrolysis.
Alkyl chain diversification of JBIR-06
As described above, both SmlC and NatC PKS modules include an additional methyltransferase (MT) domain, which is not present in the antimycin system (Fig. 1). In addition, the sml and nat clusters both lack the CCR enzyme20,21,22,23, which supplies various alkylmalonyl-CoA extender substrates to generate the alkyl group variations. Unlike the antimycin system, the JBIR-06 and neoantimycin systems lost the alkyl chain variations and only form the dimethyl group at C-9 in JBIR-06 and at C-11 in the neoantimycins, respectively (Fig. 1). Therefore, in order to increase the structural variations of the medicinally important JBIR-06, we planned to utilize the previously reported engineered CCR, the structure-based mutant of AntE that can supply a wide range of longer and bulkier alkylmalonyl-CoAs22, which is not straightforward by organic synthesis. Importantly, the long-chain alkyl group in the antimycin-type depsipeptide scaffold is reportedly essential for the Bcl2/Bcl-xL inhibitory activity, which leads to cancer cell death32,44. At the same time, we also engineered SmlC PKS module by introducing mutations into the AT domain in order to broaden its substrate specificity toward that of the antimycin system (Fig. 5a), so that the engineered PKS can process the unnatural extender units. Thus, we introduced SmlCATant, in which the substrate recognition motif “LRIAPH” was substituted with MPAAAH from AntD (Supplementary Fig. 9, Supplementary Note 5, Supplementary Tables 1–2), into S. lividans harboring pKU518J06ΔSmlC, with smlC deleted (Fig. 5a, Supplementary Note 6). We focused on this motif since it has been reported that the mutation of this site VDYASH in the AT domain of the erythromycin PKS altered the substrate preference45,46,47,48.
Alkyl chain diversification of JBIR-06. a The reconstituted module structure for alkyl chain diversification. b HPLC analysis of the metabolites from transformants harboring (i) pKU518J06 and (ii) pKU518J06ΔSmlC/pZH2-SmlCATant-AntEV350G. The chromatogram represents the UV absorbance at 320 nm. c Engineering scheme and structure of the representative product 7f
Full size image
The introduction of AntEV350G and SmlCATant gave rise to the production of 7a–7f, with m/z values that are 28 (7a–b), 42 (7c–e), and 56 (7f) higher than 2 (Supplementary Fig. 4), suggesting that they possess longer alkyl chains (Fig. 5b). Using NMR and HR-MS data, we characterized 7a, 7c, and 7f as JBIR06-type tri-lactone depsipeptides with butyl (7a), pentyl (7c), and hexyl (7f) chains at the C-9 position (Fig. 5c, Supplementary Figs. 4, 31–45, Supplementary Tables 3 and 7–9). Compounds 7b, 7d, and 7e were also identified as their minor derivatives (Supplementary Figs. 46–51, Supplementary Table 10). Small amounts of 7a–7f were observed in the strain harboring pKU518J06 with AntEV350G, but these yields were 2–3 times lower than that of the ATSmlC-mutated construct (Table 1). These data suggested that the mutation into ATSmlC broadens the substrate specificity to accept a variety of extender substrates supplied by AntEV350G. Notably, the C-9 di-substituted pattern of 2 changed into a mono-substitution, suggesting that the MT domain of SmlC does not work when the longer alkyl chain has been incorporated. However, it should be noted that the yield of the native product 2 is still dominant (5.0 ± 1.7 mg/L) (Table 1), indicating further optimization is required. Nonetheless, it was remarkable that the total yield of the newly obtained JBIR-06 derivatives with different alkyl chain structures, 7a, 7c, 7e, and 7f (3.8 mg/L), is almost comparable with that of 2.
In this study, we accomplished three manipulations of the antimycin-type NRPS-PKS assembly lines and obtained nine depsipeptides (4, 5, 6, 7a–f) with different lactone ring sizes in substantial yields. It is quite remarkable that the yields of the compounds which we obtained in the engineered NRPS-PKS system are 5–10 times higher than those in the reported module assembly line engineering11,49,50,51. For the ring reduction approach, the linker between the ACP and TE domains smoothly connected the TESmlC domain to the ACPNatC, resulting in the production of the tri-lactone 4 without significant drop of yield. Since TESmlC macrocyclizes a tri-lactone compound which shares the same structures around the carboxyl and attacking hydroxyl groups as the original substrate, we anticipated that TESmlC would work efficiently in the engineered system. In this approach, we reversely traced the evolutionary flow from the Sml to Nat modules, which facilitated the efficient reconstruction without large changes in the intermediate structures.
In the ring expansion approach, the intersubunit docking domain at the C-terminus of the NatC module was added onto the SmlC domain, instead of the TE domain. Remarkably, the yield of the ring-expanded compound 5 was same degree as that of original products 2. This data demonstrated that our strategy was effective to newly generate the module interactions. The interactions between NatC and NatD are likely to be different from the known systems, because the N-terminus of NatD lacks a predicted secondary structure consistent with the crystal structure of the docking domain from the EpoB NRPS9. The interactions between PKS and NRPS modules are still elusive, and need to be clarified by X-ray crystallization or cryo-EM analyses for future engineering studies.
In our alkyl chain diversification approach, we introduced the broad substrate specificity of AntD into SmlC by mutating the AT substrate definition sequence. The substrate definition sequence “M701PAAAH706” in AntD is exceptionally simple, and it likely causes the expansion of substrate binding cavity. In fact, the two alanines corresponding to A703 and A705 were recently shown to be the residues that influence substrate specificity of SpnD-AT52. In the case of the systems with relaxed substrate specificity for acyl-CoAs, the CCR enzyme family20,21,22,23 could be used to diversify product structures in future. The butyl-, 3-methylbutyl, and hexylmalonyl-CoAs provided by AntEV350G for 7a, 7c, and 7f production were likely synthesized from 2,3-hexenoyl, 5-methyl-2,3-hexenoyl, and 2,3-octenoyl-CoAs. This side chain pattern was also observed in our previous engineering studies for 123, implying the wide distribution of 2,3-alkenoyl-CoA in Streptomyces. As the original product (2) was still dominant in our system, the expansion of extender unit specificity of KS domain should be effective to increase the yields. With expecting the higher yield, we also tested several chimeric PKS module constructs such as KSSmlC-ATAntD-MT ACP TESmlC with domain swapping strategy (Supplementary Fig. 52); however, none of them afforded any detectable products (data not shown). Further adoption of the recently identified “optimal” fusion junctions for AT-domain swaping53 may improve the product yields.
Consequently, this study paves the way for the rational engineering of NRPS-PKS assembly machineries, by following the evolution of enzymes in nature. Through sequence comparisons, we can learn how nature cut and paste module structures. In addition, by employing the enzymes to give the various functional groups, we can further increase the diversity of products, as we have done previously with CCR enzymes23. The Escherichia-Streptomyces shuttle BAC genetic platform that we used is also an important factor to readily build artificial biosynthetic modules. The productivity of the engineered modules would be improved by optimizing the interactions betwen domains and modules based on the structural analysis of the whole module complex, even though it still remains a technically challenging endeavor. This study confirmed that the strategies applied to pure PKS and NRPS systems can be productively used for hybrid PKS/NRPS system. Furthermore, this study serves as an exemplar for bioengineering productive biosynthetic assembly lines that produce unnatural polyketide-non-ribosomal peptides.
General experimental procedures
Solvents and chemicals were purchased from Wako Chemicals Ltd. (Tokyo, Japan) or Kanto Chemical Co., Inc. (Tokyo, Japan), unless noted otherwise. Oligonucleotide primers were purchased from Eurofins Genetics (Tokyo, Japan) and Sigma-Aldrich Japan (Tokyo, Japan). PCR was performed using a TaKaRa PCR Thermal Cycler Dice® Gradient (TaKaRa), with Prime Star Max (Takara). Sequence analyses were performed by Eurofins Genetics (Tokyo). Analytical and preparative HPLC were performed on a Shimadzu Prominence system. Silica gel column chromatography was performed using Wakogel C-200. NMR spectra were obtained at 500 MHz (1H) and 125 MHz (13C) with a JEOL ECX-500 or ECZ-500 spectrometer, and chemical shifts were recorded with reference to solvent signals (1H NMR: CDCl3 7.26 ppm; 13C NMR: CDCl3 77.0 ppm). All NMR spectra were measured by using CDCl3 as solvent. Samples for LC-MS analysis were injected into an Shimadzu Prominence system HPLC-MicroTOF mass spectrometer (Bruker Daltonics), using electrospray ionization with a COSMOSIL 2.5C18-MS-II column (2.0 i.d. × 75 mm; Nacalai Tesque, Inc.).
The complete genome sequence of Streptomyces sp. ML55 and Streptomyces orinoci NBRC13466 was determined by using a Miseq (Illumina, San Diego, CA, USA) and a PacBio RS II (Pacific Biosciences, Menlo Park, CA, USA). End sequencing was carried out using the BigDye terminator ver3.1 kit (Applied Biosystems). The obtained sequence data were assembled using HGAP2 (Pacific Biosciences).
S. lividans transformants were cultured at 30 °C for 4 days in 500 mL flasks containing 100 mL A3M medium (Glucose 0.5%, Glycerol 2%, Soluble Starch 2%, Pharmamedia 1.5%, Yeast extract 0.3%, HP20 1%, pH = 7.0). After filtering, the mycelia were lyophilized to dryness. Dried mycelia were extracted by MeOH:CHCl3 = 2:1. The extraction solution was concentrated in vacuo to remove the solvent. Products from S. lividans TK21 transformant were analyzed by HPLC equipped with a STAR Separar C18G 5 μm column (4.6 i.d. × 250 mm, Rikaken Co. Ltd., Nagoya, Japan), with a solvent system of 0.1% formic acid (solvent A) and acetonitrile containing 0.1% formic acid (solvent B), at a flow rate of 1.0 mL/min and a column temperature of 40 °C. Separation was performed with solvent B/solvent A (50:50), a linear gradient from 50:50 to 80:20 within the following 5 min, an isocratic elution with 80:20 within the following 25 min.
Isolation of the metabolites from S. lividans TK21
S. lividans/pKU518nantΔnatD::smlCTE was cultured at 30 °C for 4 days in 30 × 500 mL flasks containing 100 mL A3M medium (Glucose 0.5%, Glycerol 2%, Soluble Starch 2%, Pharmamedia 1.5%, Yeast extract 0.3%, HP20 1%, pH = 7.0). After filtering, the mycelia were lyophilized to dryness. Dried mycelia were extracted by MeOH:CHCl3 = 2:1. The extraction solution was concentrated in vacuo to remove the solvent. The extract was subjected to silica-gel column chromatography and eluted using chloroform 100%. S. lividans/pKU518J06ΔsmlCTE/pZH2-NatD was cultured at 30 °C for 4 days in 5.0 L A3M medium, and S. lividans/pKU518J06ΔsmlC/pZH2-SmlCATant was cultured at 30 °C for 4 days in 12 L A3M medium. The compounds were extracted and purified by silica-gel as described above.
Fractions containing 4 was further purified by reverse-phase preparative HPLC equipped with an 5C18-MS-II column (Nacalai Tesque, Kyoto, Japan, 5 Δm, 10 mm i.d. × 250 mm). Separation was performed with solvent B (CH3CN)/solvent A (0.05% formic acid) (70:30), a linear gradient from 70:30 to 85:15 within the following 10 min, a linear gradient from 85:15 to 100:0 within the following 0.5 min, 100:0 for 4.0 additional min, to yield 4 (1.1 mg). Fractions containing 5 and 6 were further purified by preparative HPLC equipped with an YMC-Triant C18 column (YMC, Kyoto, Japan, 5 μm, 10 mm i.d. × 250 mm) using acetonitrile-0.05% formic acid (80:20) as the eluting solvent (flow rate 1.0 mL/min), to yield 5 (2.1 mg) and 6 (2.2 mg). Fractions containing 7a–f were further purified by preparative HPLC equipped with an YMC-Triant C18 column (YMC, Kyoto, Japan, 10 μm, 10 mm i.d. × 250 mm) using acetonitrile-0.1% formic acid (80:20) as the eluting solvent (flow rate 3.0 mL/min), and further purified by preparative HPLC equipped with an X-select HSS T3 column (Waters, MA, USA, 10 μm, 10 mm i.d. × 250 mm) using acetonitrile-0.1% formic acid (70:30) as the eluting solvent (flow rate 3.0 mL/min) to yield 7a (2.4 mg), 7b (0.6 mg), acetonitrile-0.1% formic acid (72.5:27.5) to yield 7c (1.0 mg), 7d (0.2 mg), 7e (0.3 mg), and acetonitrile-0.1% formic acid (75:25) to yield 7f (2.9 mg).
Chemical synthesis of isoleucic/leucic acids
L-isoleucine (100 mg) was dissolved in 1.25 M H2SO4 (5 mL) and stirred on ice, and ice-cold aqueous solution of NaNO2 (0.2 g/mL, 4 mL) was slowly added and stirred for 2 h on ice. Then, the reaction mixture was moved to room temperature, and further stirred for 15 h at room temperature. Finally, the reaction mixture was directly extracted with diethyl ether (5 mL × 2), and organic layer was dried with Na2SO4. The organic solvent was removed under reduced pressure and L-isoleucic acid was obtained as colorless oil (56 mg). Other isoleucic/leucic acids were also prepared from corresponding amino acids by diazotization except for L-leucic acid (Wako Pure Chemical Industries, Ltd.).
Methyl esterification of isoleucic/leucic acids
A methanol solution of L-isoleucic acid (50 mg/10 mL) in eggplant flask was stirred at room temperature. Then, thionyl chloride (1.2 mL) was slowly added to the solution. The mixture was further stirred for 2 h at 100 °C (in reflux), and cooled to room temperature. The solvent was removed under reduced pressure and the residue was suspended in diethyl ether (10 mL). The suspension was washed with saturated sodium hydrogen carbonate (10 mL × 2), and the organic layer was dried with Na2SO4. Finally, the ether was removed by evaporation and the L-leucic acid methyl ester was obtained as colorless oil (42 mg). For other isoleucic/leucic acid methyl esters were also prepared in the same procedure.
Preparation of N-TFA-threonine methyl esters
L-threonine (1 mg) was dissolved in hydrogen chloride-methanol reagent (5–10%) (0.5 mL, purchased from Tokyo Chemical Industry Co., Ltd.), and the mixture was incubated for 30 min at 100 °C (in reflux). The solvent was removed under reduced pressure. To the residue, a 1:1 mixture of trifluoroacetic anhydride (TFAA)/dichloromethane (0.5 mL) was added and the mixture was incubated for 30 min at 100 °C (in reflux). After the reaction, the solvents were removed by argon gas injection. Finally, the residue was dissolved in acetone (0.5 mL) and subjected to GC-MS analysis as a standard. The other stereoisomers of threonine were also derivatized in the same procedure and subjected to GC-MS analysis.
Hydrolysis of 5 and 6
Compound 5 (0.1 mg) was added to 6 M hydrochloric acid (0.5 mL) and stirred for 24 h at 110 °C (in reflux). After the reaction, the mixture was cooled on ice to room temperature, and the solvent was removed by lyophilization. The resulting residue was used for further chemical derivatization as noted below. Compound 6 was also hydrolyzed by the same procedure.
Chemical derivatization of hydrolysates of 5 and 6
The hydrolysate of 5 (from 0.1 mg) was in hydrogen chloride-methanol reagent (5–10%) (0.5 mL), and the mixture was incubated for 30 min at 100 °C (in reflux). The reaction mixture was divided in two equal parts, and the solvent was removed under reduced pressure. The half part of the residue was dissolved in acetone (0.5 mL) and subjected GC-MS to analyze isoleucine/leucine acids as methyl ester. The remaining half part of the residue was dissolved in the mixture of TFAA/dichloromethane (0.5 mL), and the mixture was incubated for 30 min at 100 °C (in reflux). The solvents were removed by argon and the residue was dissolved in acetone (0.5 mL), which was subject to the GC-MS to analyze threonine moiety as the N-TFA methyl ester. The hydrolysate of 6 (from 0.1 mg) was also derivatized and subjected to the GC-MS analysis in the same procedure.
GC-MS analysis condition
CP-Chirasil Dex-CB column (Alltech, 0.25 mm × 25 m; He as the carrier gas; program rate: 50–150 °C at 5 °C/min, 150–200 °C at 50 °C/min).
GC-MS for threonine derivatives
CP-Chirasil Dex-CB column (Alltech, 0.25 mm × 25 m; He as the carrier gas; program rate: 50–200 °C at 6 °C/min).
The DNA sequences of JBIR-06 and neoantimycin biosynthetic gene clusters were registered as LC375135 and LC375136 in DDBJ, respectively.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Walsh, C. T. Polyketide and nonribosomal peptide antibiotics: modularity and versatility. Science 303, 1805–1810 (2004).
ADSArticlePubMed Google Scholar2.
Walsh, C. T., O’Connor, S. E. & Schneider, T. L. Polyketide-nonribosomal peptide epothilone antitumor agents: the EpoA, B, C subunits. J. Ind. Microbiol. Biotechnol. 30, 448–455 (2003).
ArticlePubMed Google Scholar3.
Wakimoto, T. et al. Calyculin biogenesis from a pyrophosphate protoxin produced by a sponge symbiont. Nat. Chem. Biol. 10, 648–655 (2014).
ArticlePubMed Google Scholar4.
Wakimoto, T., Egami, Y. & Abe, I. Calyculin: nature’s way of making the sponge-derived cytotoxin. Nat. Prod. Rep. 33, 751–760 (2016).
ArticlePubMed Google Scholar5.
Sattely, E. S., Fischbach, M. A. & Walsh, C. T. Total biosynthesis: in vitro reconstitution of polyketide and nonribosomal peptide pathways. Nat. Prod. Rep. 25, 757 (2008).
ArticlePubMed Google Scholar6.
Weissman, K. J. The structural biology of biosynthetic megaenzymes. Nat. Chem. Biol. 11, 660–670 (2015).
ArticlePubMed Google Scholar7.
Sundaram, S. & Hertweck, C. On-line enzymatic tailoring of polyketides and peptides in thiotemplate systems. Curr. Opin. Chem. Biol. 31, 82–94 (2016).
ArticlePubMed Google Scholar8.
Richter, C. D., Nietlispach, D., Broadhurst, R. W. & Weissman, K. J. Multienzyme docking in hybrid megasynthetases. Nat. Chem. Biol. 4, 75–81 (2008).
ArticlePubMed Google Scholar9.
Dowling, D. P. et al. Structural elements of an NRPS cyclization domain and its intermodule docking domain. Proc. Natl Acad. Sci. USA 113, 12432–12437 (2016).
ArticlePubMed Google Scholar10.
Williams, G. J. Engineering polyketide synthases and nonribosomal peptide synthetases. Curr. Opin. Struct. Biol. 23, 603–612 (2013).
ArticlePubMedPubMed Central Google Scholar11.
Bozhüyük, K. A. J. et al. De novo design and engineering of non-ribosomal peptide synthetases. Nat. Chem. 10, 275–281 (2018).
ArticlePubMed Google Scholar12.
Sieber, S. A. et al. Evidence for a monomeric structure of nonribosomal peptide synthetases. Chem. Biol. 9, 997–1008 (2002).
ArticlePubMed Google Scholar13.
Reimer, J. M., Haque, A. S., Tarry, M. J. & Schmeing, T. M. Piecing together nonribosomal peptide synthesis. Curr. Opin. Struct. Biol. 49, 104–113 (2018).
ArticlePubMed Google Scholar14.
Jenke-Kodama, H., Sandmann, A., Müller, R. & Dittmann, E. Evolutionary implications of bacterial polyketide synthases. Mol. Biol. Evol. 22, 2027–2039 (2005).
ArticlePubMed Google Scholar15.
Ridley, C. P., Lee, H. Y. & Khosla, C. Evolution of polyketide synthases in bacteria. Proc. Natl Acad. Sci. USA 105, 4595–4600 (2008).
ADSArticlePubMed Google Scholar16.
Nguyen, T. A. et al. Exploiting the mosaic structure of trans-acyltransferase polyketide synthases for natural product discovery and pathway dissection. Nat. Biotechnol. 26, 225–233 (2008).
ArticlePubMed Google Scholar17.
Medema, M. H., Cimermancic, P., Sali, A., Takano, E. & Fischbach, M. A. A systematic computational analysis of biosynthetic gene cluster evolution: lessons for engineering biosynthesis. PLoS Comput. Biol. 10, e1004016 (2014).
ADSArticlePubMedPubMed Central Google Scholar18.
Eng, C. H. et al. ClusterCAD: a computational platform for type I modular polyketide synthase design. Nucleic Acids Res. 46, D509–D515 (2018).
ArticlePubMed Google Scholar19.
Liu, J., Zhu, X., Kim, S. J. & Zhang, W. Antimycin-type depsipeptides: discovery, biosynthesis, chemical synthesis, and bioactivities. Nat. Prod. Rep. 33, 1146–1165 (2016).
ArticlePubMed Google Scholar20.
Wilson, M. C. & Moore, B. S. Beyond ethylmalonyl-CoA: the functional role of crotonyl-CoA carboxylase/reductase homologs in expanding polyketide diversity. Nat. Prod. Rep. 29, 72–86 (2012).
ArticlePubMed Google Scholar21.
Ray, L. & Moore, B. S. Recent advances in the biosynthesis of unusual polyketide synthase substrates. Nat. Prod. Rep. 33, 150–161 (2016).
ArticlePubMedPubMed Central Google Scholar22.
Peter, D. M. et al. Screening and engineering the synthetic potential of carboxylating reductases from central metabolism and polyketide biosynthesis. Angew. Chem. Int. Ed. 54, 13457–13461 (2015).
Article Google Scholar23.
Zhang, L. et al. Rational control of polyketide extender units by structure-based engineering of a crotonyl-CoA carboxylase/reductase in antimycin biosynthesis. Angew. Chem. Int. Ed. 127, 13664–13667 (2015).
Article Google Scholar24.
Sandy, M., Rui, Z., Gallagher, J. & Zhang, W. Enzymatic synthesis of dilactone scaffold of antimycins. ACS Chem. Biol. 7, 1956–1961 (2012).
ArticlePubMed Google Scholar25.
Yan, Y. et al. Biosynthetic pathway for high structural diversity of a common dilactone core in antimycin production. Org. Lett. 14, 4142–4145 (2012).
ArticlePubMed Google Scholar26.
Zhu, X., Liu, J. & Zhang, W. De novo biosynthesis of terminal alkyne-labeled natural products. Nat. Chem. Biol. 11, 115–120 (2015).
ArticlePubMed Google Scholar27.
Yan, Y. et al. Multiplexing of combinatorial chemistry in antimycin biosynthesis: expansion of molecular diversity and utility. Angew. Chem. Int. Ed. 52, 12308–12312 (2013).
ADSArticle Google Scholar28.
Ueda, J. Y. et al. A novel antimycin-like compound, JBIR-06, from Streptomyces sp. ML55. J. Antibiot. 61, 241–244 (2008).
Cassinelli, G., Grein, A., Orezzi, P., Pennella, P. & Sanfilippo, A. New antibiotics produced by Streptoverticillium orinoci, n.sp. Arch. Mikrobiol. 55, 358–368 (1967).
Lee, A. S. Glucose-regulated proteins in cancer: molecular mechanisms and therapeutic potential. Nat. Rev. Cancer 14, 263–276 (2014).
ArticlePubMedPubMed Central Google Scholar31.
Park, H. et al. Effect on tumor cells of blocking survival response to glucose deprivation. J. Natl Can. Inst. 96, 1300–1310 (2004).
Article Google Scholar32.
Vanner, S. A. et al. Chemical and biosynthetic evolution of the antimycin-type depsipeptides. Mol. Biosyst. 9, 2712 (2013).
ArticlePubMed Google Scholar33.
Skyrud, W. et al. Biosynthesis of the 15-membered ring depsipeptide neoantimycin. ACS Chem. Biol. 13, 1398 (2018).
ArticlePubMed Google Scholar34.
Komatsu, M., Uchiyama, T., Omura, S., Cane, D. E. & Ikeda, H. Genome-minimized Streptomyces host for the heterologous expression of secondary metabolism. Proc. Natl Acad. Sci. USA 107, 2646–2651 (2010).
ADSArticlePubMed Google Scholar35.
Komatsu, M. et al. Engineered Streptomyces avermitilis host for heterologous expression of biosynthetic gene cluster for secondary metabolites. ACS Synth. Biol. 2, 384–396 (2013).
ArticlePubMedPubMed Central Google Scholar36.
Takeda, Y. et al. Nuclear magnetic resonance and biosynthetic studies of neoantimycin and structure elucidation of isoneoantimycin, a minor metabolite related to neoantimycin. J. Nat. Prod. 61, 978–981 (1998).
ArticlePubMed Google Scholar37.
Li, X., Zvanych, R., Vanner, S. A., Wang, W. & Magarvey, N. A. Chemical variation from the neoantimycin depsipeptide assembly line. Bioorg. Med. Chem. Lett. 23, 5123–5127 (2013).
ArticlePubMed Google Scholar38.
Kapur, S., Chen, A. Y., Cane, D. E. & Khosla, C. Molecular recognition between ketosynthase and acyl carrier protein domains of the 6-deoxyerythronolide B synthase. Proc. Natl Acad. Sci. USA 107, 22066–22071 (2010).
ADSArticlePubMed Google Scholar39.
Dutta, S. et al. Structure of a modular polyketide synthase. Nature 510, 512–517 (2014).
ADSArticlePubMedPubMed Central Google Scholar40.
O’Connor, S. E., Walsh, C. T. & Liu, F. Biosynthesis of epothilone intermediates with alternate starter units: engineering polyketide-nonribosomal interfaces. Angew. Chem. Int. Ed. 42, 3917–3921 (2003).
Article Google Scholar41.
Rowe, C. J. et al. Engineering a polyketide with a longer chain by insertion of an extra module into the erythromycin-producing polyketide synthase. Chem. Biol. 8, 475–485 (2001).
ArticlePubMed Google Scholar42.
Tsai, S. et al. Crystal structure of the macrocycle-forming thioesterase domain of the erythromycin polyketide synthase: Versatility from a unique substrate channel. Proc. Natl Acad. Sci. USA 98, 14808–14813 (2001).
ADSArticlePubMed Google Scholar43.
Zhang, Y., Buchholz, F., Muyrers, J. P. P., & Stewart, A.F. New technology. A new logic for DNA engineering using recombination in Escherichia coli. Nat. Genet. 20, 123–128 (1998).
Bodur, C. & Basaga, H. Bcl-2 inhibitors: emerging drugs in cancer therapy. Curr. Med. Chem. 19, 1804–1820 (2012).
ArticlePubMed Google Scholar45.
Reeves, C. D. et al. Alteration of the substrate specificity of a modular polyketide synthase acyltransferase domain through site-specific mutations. Biochemistry 40, 15464–15470 (2001).
ArticlePubMed Google Scholar46.
Del Vecchio, F. et al. Active-site residue, domain and module swaps in modular polyketide synthases. J. Ind. Microbiol. Biotechnol. 30, 489–494 (2003).
ArticlePubMed Google Scholar47.
Sundermann, U. et al. Enzyme-directed mutasynthesis: a combined experimental and theoretical approach to substrate recognition of a polyketide synthase. ACS Chem. Biol. 8, 443–450 (2012).
ArticlePubMed Google Scholar48.
Koryakina, I. et al. Inversion of extender unit selectivity in the erythromycin polyketide synthase by acyltransferase domain engineering. ACS Chem. Biol. 12, 114–123 (2017).
ArticlePubMed Google Scholar49.
Sugimoto, Y., Ding, L., Ishida, K. & Hertweck, C. Rational design of modular polyketide synthases: morphing the aureothin pathway into a luteoreticulin assembly line. Angew. Chem. Int. Ed. 53, 1560–1564 (2014).
Article Google Scholar50.
Sugimoto, Y. et al. Freedom and constraint in engineered noncolinear polyketide assembly lines. Chem. Biol. 22, 229–240 (2015).
ArticlePubMed Google Scholar51.
Wlodek, A. et al. Diversity oriented biosynthesis via accelerated evolution of modular gene clusters. Nat. Commun. 8, 1206 (2017).
Li, Y. et al. Structural basis of a broadly selective acyltransferase from the polyketide synthase of splenocin. Angew. Chem. Int. Ed. https://doi.org/10.1002/anie.201802805 (2018).
Yuzawa, S. et al. Comprehensive in vitro analysis of acyltransferase domain exchanges in modular polyketide synthases and its application for short-chain ketone production. ACS Synth. Biol. 6, 139–147 (2017).
ArticlePubMed Google Scholar
This work was supported by a Grant-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology, Japan (JSPS KAKENHI Grant Number JP15H01836, JP16H06443, JP16K13084, and JP17H04763), JST/NSFC Strategic International Collaborative Research Program Japan-China, Kobayashi International Scholarship Foundation, and National Natural Science Foundation of China Grants (21520102004). We also thank Prof. Hiroyasu Onaka for providing genetic tools.
AffiliationsGraduate School of Pharmaceutical Sciences, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo, 113-0033, JapanTakayoshi Awakawa, Takuma Fujioka, Lihan Zhang, Shotaro Hoshino, Zhijuan Hu & Ikuro AbeCollaborative Research Institute for Innovative Microbiology, The University of Tokyo, Yayoi 1-1-1, Bunkyo-ku, Tokyo, 113-8657, JapanTakayoshi Awakawa, Kazuo Shin-Ya & Ikuro AbeJapan Biological Informatics Consortium, 2-4-7 Aomi, Koto-ku, Tokyo, 135-8073, JapanJunko Hashimoto & Ikuko KozoneLaboratory of Microbial Engineering, Kitasato Institute for Life Sciences, Kitasato University, Kanagawa, 252-0373, JapanHaruo IkedaNational Institute of Advanced Industrial Science and Technology (AIST), 2-4-7 Aomi, Koto-ku, Tokyo, 135 0064, JapanKazuo Shin-YaState Key Laboratory of Bio-Organic & Natural Products Chemistry, Shanghai Institute of Organic Chemistry, Chinese Academy of Sciences, Lingling Road 345, 200032, Shanghai, ChinaWen Liu
AuthorsSearch for Takayoshi Awakawa in:Nature Research journals • PubMed • Google Scholar Search for Takuma Fujioka in:Nature Research journals • PubMed • Google Scholar Search for Lihan Zhang in:Nature Research journals • PubMed • Google Scholar Search for Shotaro Hoshino in:Nature Research journals • PubMed • Google Scholar Search for Zhijuan Hu in:Nature Research journals • PubMed • Google Scholar Search for Junko Hashimoto in:Nature Research journals • PubMed • Google Scholar Search for Ikuko Kozone in:Nature Research journals • PubMed • Google Scholar Search for Haruo Ikeda in:Nature Research journals • PubMed • Google Scholar Search for Kazuo Shin-Ya in:Nature Research journals • PubMed • Google Scholar Search for Wen Liu in:Nature Research journals • PubMed • Google Scholar Search for Ikuro Abe in:Nature Research journals • PubMed • Google Scholar
T.A., L.Z., W.L., K.S. and I.A. designed the experiments. T.A., T.F., L.Z., S.H., Z.H., J.H., I.K. and H.I. performed the experiments. T.A., T.F., L.Z. and I.A. analyzed the data. T.A., K.S. and I.A. wrote the paper.
The authors declare no competing interests.
Correspondence to Takayoshi Awakawa or Kazuo Shin-Ya or Ikuro Abe.
Electronic supplementary material
Supplementary InformationPeer Review File
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
13 March 2018
01 August 2018
30 August 2018