Abstract
Reprogramming biosynthetic assembly-lines is a topic of intense interest. This is unsurprising as the scaffolds of most antibiotics in current clinical use are produced by such pathways. The modular nature of assembly-lines provides a direct relationship between the sequence of enzymatic domains and the chemical structure of the product, but rational reprogramming efforts have been met with limited success. To gain greater insight into the design process, we wanted to examine how Nature creates assembly-lines and searched for biosynthetic pathways that might represent evolutionary transitions. By examining the biosynthesis of the anti-tubercular wollamides, we show how whole gene duplication and neofunctionalization can result in pathway bifurcation. Importantly, we show that neofunctionalization occurs primarily through intragenomic recombination. This pathway bifurcation leads to redundancy, providing the genetic robustness required to enable large structural changes during the evolution of antibiotic structures. Should the new product be none-functional, gene loss can restore the original genotype. However, if the new product confers an advantage, depreciation and eventual loss of the original gene creates a new linear pathway. This provides the blind watchmaker equivalent to the ‘design, build, test’ cycle of synthetic biology.
Introduction
Microbial natural products produced by modular biosynthetic assembly-lines, i.e. (type I) polyketide synthases (PKSs)1 and non-ribosomal peptide synthetases (NRPSs)2, are an important source of pharmaceutical and agrochemical agents. Examples include well known antibacterial molecules such as the polyketide insecticide spinosyn3 and NRPS derived penicillins4. Importantly, biosynthetic assembly-lines provide thousands of natural product scaffolds, including many of our essential clinical agents.
Essentially, NRPS and PKS modular megasynth(et)ases give rise to highly functionalised biopolymers from a broad variety of monomers, referred to as extender units. Hundreds of extender units have been reported, typically derived from malonate in the case of PKSs5 or amino acids in the case of NRPSs6,7. They are likened to assembly-line processes due to their hierarchical and modular structures in which multiple, repeating modules of enzymatic domains catalyse the incorporation of an extender unit into the growing chain, along with any programmed additional chemical modifications, before transferring the elongated chain to the next module. The archetypical minimal assembly-line module consists of three ‘core’ domains. Firstly, a domain for the selection and activation of an extender unit, the acyltransferase (AT) domains for PKSs or adenylation (A) domains for NRPSs. The activated substrate is then covalently attached to a prosthetic phosphopantetheine group of a small acyl carrier protein (ACP; PKSs) or peptidyl carrier protein (PCP; NRPSs) domain. Finally, the ketosynthase (KS; PKSs) or condensation (C; NRPSs) domains then link the covalently bound substrates to the growing polyketide or peptide chain. Although the exact mechanisms and ancillary domains of PKSs and NRPSs differ, the fundamental principal is that modules condense covalently bound substrates in a linear fashion. The inherent logic of this mechanism means that there is a direct relationship between the sequence of domains in an assembly line and the chemical structure of the resulting molecule1,8,9. In principle this relationship enables the prediction of natural product chemical structures directly from DNA sequences. In turn, this logic has inspired efforts to rationally reprogramme assembly-lines to produce tailor-made molecules.
Numerous examples of assembly line engineering have been reported, however many display productivities well below that of the parental (wild type) system. Insights into structural flexibility, inter-domain communication, and the role of proof-reading by catalytic domains preempted novel strategies to engineer assembly-line proteins10–14. There is also an increasing body of evidence suggesting that we might further improve our ability to engineer these systems if we had a better understanding of their evolution1,15–18.
With this latter point in mind, we have been searching for biosynthetic pathways that may represent transitional evolutionary states and provide exemplar systems to inform future work. Natural selection acts upon phenotypes, yet even a small change to the structure of a natural product can have profound effects on bioactivity and, thus, the fitness of the producing organism. Therefore, we hypothesised that strains encoding BGCs evolving new functionalities might be expected to produce multiple related products (congeners). Here, we describe the BGC encoding the wollamide-desotamide family of antibiotics, which represents an evolutionary snapshot of a modular NRPS assembly line.
The wollamides are cyclic hexapeptides (Fig. 1a) first reported in 201419. The only known producer of the wollamides is Streptomyces sp. MST-11508819 and they exhibit potent antimycobacterial activity that has attracted the attention of synthetic medicinal chemists20–22. Along with the wollamides, Streptomyces sp. MST-115088 produces a related group of hexapeptides, the desotamides23. The desotamides and wollamides share a common peptide scaffold, except for a single residue change from glycine in the desotamides to D-ornithine in the wollamides (Fig. 1a). It is important to note that the previously identified desotamide producer, Streptomyces scopuliridis SCSIO ZJ46, is not reported to produce wollamides and the desotamide (dsa) BGC follows canonical NRPS logic24. The NRPS is encoded by 3 genes (dsaI, dsaH and dsaG) encoding two modules each. Importantly, module six, encoded by DsaG, incorporates glycine as the final amino acid. The ability of Streptomyces sp. MST-115088 to produce congeners with D-ornithine and glycine in the same position is therefore difficult to rationalise as, under a canonical NRPS mechanism, A-domains are responsible for the selection and activation of specific amino acid substrates. While A-domains can activate structurally related amino acids giving rise to families of structurally similar congeners (for example the combinations of valine, leucine, or allo-leucine at positions 3 and 4 of the wollamides-desotamides), the ability of an A-domain to activate substrates with such varying physico-chemical properties as glycine and ornithine is without precedent. Therefore, to explain the production of wollamides and desotamides by a single strain we hypothesised three scenarios (Fig. 1b): dual specificity of module six for glycine and ornithine; duplicated genes encoding module six homologues, each specific to glycine or ornithine respectively; or duplicated BGCs where, as above, the final modules of each BGC are specific to glycine or ornithine.
Herein, we describe our combined bioinformatics, in vivo engineering, and biochemical analysis demonstrating how dsa evolved from an ancestral wollamide producing BGC via the process of gene duplication and intragenomic recombination with a second NRPS encoding locus. This allows us to propose an update to the current model for the evolution of assembly line BGCs through which duplication and the resulting bifurcation reduces selective pressure and drives the evolution of new functions.
Results
Desotamides and wollamides are produced by a bifurcated NRPS assembly-line
The genome of Streptomyces sp. MST-115088 was sequenced using Pacific Biosciences RS225 single molecule technology and assembled using the HGAP326 pipeline to generate a single 7.9 Mb chromosomal assembly (Supplementary Table 1)(GenBank:CP074380). Analysis of the assembly using antiSMASH v4.08 allowed us to rapidly identify the wollamide (wol) BGC (Supplementary Table 2), which was then compared to the desotamide (dsa) BGC previously reported from Streptomyces scopuliridis SCSIO ZJ46 (GenBank: KP769807) (Fig. 2ab). Five additional desotamide producers (MST-70754, MST-71458, MST-71321, MST-94754 and MST-127221) were identified through metabolomic analysis of Microbial Screening Technologies unique collection of more than fifty thousand Australian actinomycete strains (Supplementary Information). The desotamide BGC (Supplementary Tables 3-7) of these strains display functionally identical architectures (GenBank: MZ093610-MZ093614) to that reported from Streptomyces scopuliridis SCSIO ZJ4624 (>95% nucleotide identity) (Supplementary Fig. 1). The wol BGC displays a similar architecture to the dsa BGCs but contains two genes wolG1 and wolG2 that are duplicates of the dsaG gene which encodes the final two modules of the NRPS. Similarly, there are duplicated homologues of the dsaF gene (wolF1 and wolF2), which encodes an MbtH-like protein involved in A-domain functionality27,28. The wol BGC also contains six additional genes wolRSTUVW predicted to be involved in the biosynthesis of L-ornithine consistent with the presence of ornithine in position 6 of the wollamides.
During desotamide biosynthesis, DsaG is responsible for the final two rounds of peptide chain elongation. We hypothesised, therefore, that WolG1 and WolG2 may encode two forks of a bifurcated biosynthetic pathway where the first four rounds of peptide elongation proceed via the colinear activity of WolI and WoH, but the final two elongation steps are catalysed either by WolG1 or WolG2, yielding desotamide or wollamide products respectively (Fig. 2c). Consistent with this hypothesis, in silico analysis of A-domain specificities29,30 predicted the substrates for the final A-domains of WolG1 and WolG2 (henceforth WolG1A2 and WolG2A2) to be glycine and L-ornithine (Supplementary Table 8).
Engineering wollamide production in a desotamide-only producing strain
To confirm our biosynthetic hypothesis and to explore the evolutionary relationship between the wollamide and desotamide pathways, we sought to engineer the co-production of wollamides into a desotamide-only producing strain through the heterologous expression of wolG2.
As NRPS genes are typically large (for reference, wolG2 is 7.9 kb) and difficult to clone through conventional methods we generated pBO1 (Supplementary Fig. 2), a Streptomyces expression vector capable of propagating in both E. coli and S. cerevisiae. This allows larger genes to be assembled efficiently in yeast via transformation associated homologous recombination31 downstream of a constitutive promotor and subsequently transferred from E. coli to Streptomyces spp. through conjugal transfer (Supplementary Information). To assemble a wolG2 expression plasmid, pBO1 along with target fragments amplified from genomic DNA were transformed into S. cerevisae CEN.PK 2-1C32. The resulting plasmid pBO1_wolG2 was transformed into the desotamide producer Streptomyces sp. MST-7075433 via conjugation.
Streptomyces sp. MST-70754 and progeny carrying pBO1_wolG2 were grown under desotamide producing conditions and methanolic culture extracts were then analysed by LCMS in comparison to the wollamide producer Streptomyces sp. MST-115088. The presence of both wollamide and desotamide congeners was confirmed for the engineered strains by comparison of retention time, isotopic masses and MS/MS fragmentation of the compounds produced by the native wollamide producer (desotamide: [M+H]+ = 697.4047, [M+Na]+ = 719.382; wollamide A [M+H]+ 754.4655, [M+Na]+ = 776.4419) (Fig. 3) (Supplementary Fig. 3).
This engineering strategy relied on the assumption that the docking domains (DDs) of DsaH and WolG2 that mediate NRPS interactions would still function as a pair34,35. While the production of wollamides shows that the interaction is still possible, the low relative yields of wollamides versus desotamides suggested that some depreciation has occurred.
To investigate this, we compared the N-terminal DD primary sequences of WolG1 and WolG2 (Supplementary Fig. 4), as well as calculated homology models of the DDs in complex with the WolH DD (Supplementary Fig. 5). Alignment of the DDs revealed a glutamate to alanine mutation at position (E16A). Homology modelling with the previously characterised PaxB docking domain (6TRP_1)36 revealed that the E16A amino acid change in WolG2 leads to the loss of a salt bridge formed with R3504 from WolH, most likely causing decreased DD pair affinities, consistent with the observed differences in titers.
Evolution of adenylation domain specificity via intragenomic recombination
The unusual architecture of the wol BGC led us to consider potential mechanisms for its evolution. The high similarity of wolG1 to wolG2 (80.9%) and wolG1/wolG2 to dsaG (74.3%71.2% respectively) is indicative of an ancestral gene duplication event37. However, despite this high similarity, there is a notable drop in nucleotide identity within the region coding for the final adenylation domains which is also manifest in the gene products (54.6% nucleotide and 29.2% protein sequence similarity) (Supplementary Fig. 6). Given the overall similarity of these genes we deemed it unlikely that such high sequence variation could emerge through speciation (point mutation) alone. Furthermore, phylogenetic reconstructions of A-domains hinted at independent evolutionary histories when compared to the rest of the assembly line (Supplementary Figures 7-9). Similar patterns have been observed in other bacterial NRPS clusters18,38–40
Many studies have highlighted the role recombination plays in the evolution of assembly-lines1,41. More specifically, given the reduced rate of horizontal gene transfer between distantly related taxa and the high rate of heterogeneity between recombinant sequences has, it has been speculated that intragenomic recombination within ancestral strains can provide opportunities for assembly-line diversification 18,42. To assess the possible role of intragenomic recombination in the evolution of the wol BGC we generated a nucleotide sequence alignment of all thirty-four NRPS-associated adenylation domains present in the Streptomyces sp. MST110588 genome for analysis using the Recombination Detection Programme 4 (RDP4)43.
Using our dataset, RDP443 predicted 29 potential intragenomic recombination events (Supplementary Fig. 10 and Supplementary Table 9), 11 of which were supported by two or more methods, allowing recombination breakpoints to be predicted. Only two recombination events were supported unanimously, including one event, in which wolG1A2 was identified as a recombinant sequence with wolG2A2 as the major parent and an adenylation domain encoding sequence from elsewhere in the genome, encoded by orf6595A, as the minor parent (Fig 4a-c and Supplementary Table 9). The predicted recombinant region lies between the conserved motif A2 (N-terminal) and motifs A5 and A6 (C-terminal) of the A-domain, thus comprising the flavodoxin subdomain18,38,40 and a large portion of the N-terminal subdomain (Fig. 4d). Crucially, this would allow for the substitution of the amino acid binding pocket and catalytic P-loop while maintaining the C-A -domain interface (Supplementary Fig. 11). This pattern is seen frequently in our dataset (Supplementary Fig. 10) and may suggest an advantage for maintaining the structural relationships between the P-loop and substrate binding pocket.
Importantly, the adenylation domain encoded by orf6595 was predicted in silico to select for glycine (Supplementary Table 8) meaning that the predicted recombination event could theoretically convert a wollamide producing assembly line into a desotamide producer.
Biochemical validation of A-domain substrate specificities
To gain insight into the function of the minor parent BGC, the sequences of Orf6595 and proteins encoded by the surrounding genes were searched against the MIBiG database44 using cblaster45. The search identified the BGC as homologous to the rimosamide (rmo) BGC from Streptomyces rimosus ATCC 1097046 (Supplementary Table 10). More specifically, Orf6595 is a homologue of RmoG which encodes a single NRPS module (C-A-T) that is known to incorporate glycine into the rimosamide peptide chain. This was consistent with bioinformatic predictions of the Orf6595 adenylation domain active site providing strong evidence that recombination between the adenylation domains of orf6595 and wolG2 could confer specificity to glycine (Supplementary Fig. 12).
To verify this prediction and examine the substrate specificity of all of the A-domains of interest, pET28a hexa-histidine tagged WolG1A2, WolG2A2 and Orf6595A constructs were cloned for expression based upon the A-domain boundaries as described in Crüsemann et al38. These were expressed in E. coli Rosetta 2(DE3)pLysS and purified using Ni-Affinity chromatography. Initially, the resulting protein was insoluble, however co-expression with the MbtH-like protein WolF2 (expressed from pCDFDuet-1) enabled the purification of soluble protein in each case. The ability of these isolated A-domains to activate each of the twenty proteinogenic amino acids and L-ornithine was then measured using an hydroxylamine trapping assay47 (Supplementary Table 11). WolG2A2 adenylates L-ornithine, in line with our hypothesis (Fig. 5a); however, it was also capable of activating other substrates albeit with lower efficiency. Most noticeably, WolG2A2 accepted L-aspartate (58% activity relative to L-ornithine) and L-asparagine (44% activity relative to L-ornithine) as substrates, but wollamide analogues in which aspartate or asparagine were substituted for ornithine were not identified in culture extracts of Streptomyces sp. MST-110588 despite targeted LCMS analysis. In contrast, both WolG1A2 and Orf6595A activate glycine in a highly specific manner (Fig. 5a). These data corroborate our hypothesis that an historic recombination with orf6595 could alter the substrate specificity of the module six adenylation domain of WolG2 from L-ornithine to glycine. To test this hypothesis further, we produced a hybrid A-domain encoding gene sequence representing the hypothetical ancestral recombinant, based on wolG2A2 and orf6595 sequences (Fig. 5b). The resulting gene product was purified and assayed as above and found to be highly selective for glycine (Fig. 5a).
Taken together with combined genomic, in silico and in vivo data above, these biochemical data show how the contemporary desotamide BGCs have evolved from an ancestral wollamide-like BGC through the process of gene duplication, intragenomic recombination and gene loss.
Discussion
How NRPS and PKS assembly line biosynthetic pathways evolve is a question of perennial interest in natural products research. An understanding of this process offers new avenues for developing rational approaches for bioengineering and the targeted production of new molecules. Evolutionary analyses of publicly available BGC sequences are common in the literature18,48,49, but it is unprecedented to find an example of a BGC that represents a snapshot of assembly line evolution. In this work, we studied Streptomyces sp. MST110588, a co-producer of desotamide and wollamide hexapeptide antibiotics. Despite their close structural relationship, it was not obvious how such a mixture of congeners might be assembled based on canonical NRPS function, pointing to an unusual biosynthetic pathway.
The architecture of the wollamide BGC is highly similar to the previously described desotamide BGC, however a striking difference is that it contains duplicates (wolG1 and wolG2) of dsaG which encodes the final NRPS protein of the desotamide assembly line. This suggested a bifurcated assembly-line was responsible for the observed mixture of desotamide and wollamide congeners (Fig. 2c). In this scenario, WolIH is responsible for condensation of the first four amino acids prior to extension with either WolG1, producing desotamides, or WolG2, producing wollamides. This hypothesis was supported by bioinformatic (Fig. 4) and biochemical analysis (Fig. 5a) of the A-domains WolG1A2 and WolG2A2, which encode for selection and activation of glycine and L-ornithine respectively. Further in vivo evidence came from strain engineering in which expression of WolG2 in a desotamide-only producing strain led to biosynthesis of additional wollamide congeners (Fig. 3).
Comparison of the protein sequences of the NRPSs encoded by wolG1/G2 indicated a high degree of sequence conservation except for the region coding for the final adenylation domains. These data were indicative of a recombination event rather than of divergence through speciation alone, and we subsequently analysed all 34 A-domain sequences present in the Streptomyces sp. MST110588 genome using the recombination detection program RDP443. This identified wolG2 as the major parent, and an NRPS gene orf6595 (encoded approximately 3 Mb away on the chromosome) as the minor parent, of a recombination event that formed wolG1. Using the cblaster tool we identified the BGC containing orf6595 as a homologue of the rimosamide BGC46. Orf6595 is a homologue of RmoG that selects for glycine. This was verified by subsequent bioinformatics and biochemical analysis. To validate the recombination event predicted by RDP4 we used the contemporary sequences of wolG2A2 and orf6595 to recapitulate the predicted ancestor and generated a synthetic A-domain. Subsequent biochemical analysis of the isolated protein showed it was selective for glycine activation as predicted.
Based on these combined data we can confidently trace the evolutionary history of the wollamide and desotamide BGCs (Fig. 6a). First, a gene duplication event in an ancestral wollamide (or wollamide-like) BGC resulted in a redundant copy of the bimodular NRPS encoding modules 5 and 6 of the assembly line. Subsequently, an intragenomic recombination event between the DNA encoding the adenylation domains of the duplicated module six in the wolG homologue and orf6595 resulted in an intermediate NRPS selective for glycine. This progenitor is the ancestor of all wollamide and desotamide producing strains which, through additional selection and mutations, became the lineage of the wol BGC observed in the contemporary genome of Streptomyces sp. MST110588 capable of bifurcated biosynthesis. In a divergent lineage, the ancestral gene encoding the L-ornithine-specific adenylation domain along with the duplicated MbtH-like protein encoding gene and associated genes encoding L-ornithine biosynthesis were lost through gene deletion, ultimately resulting in the contemporary dsa BGCs producing only desotamides. These observations are consistent with the presence of a redundant epimerase domain located in the second modules of WolG1 and all DsaG homologues50,51. Moreover, in Streptomyces MST110588 production of wollamide congeners was more than an order of magnitude lower than that of desotamides. Our engineered strain of the desotamide producer, Streptomyces sp. MST-70754, expressing wolG2 also showed this large difference in titer. This observation suggested that the deprecation of interactions between the C-and N-terminal docking domains of WolH/DsaH and WolG2 respectively. To assess this, we calculated homology models which revealed that a key E16A amino acid change in WolG2 leads to the loss of a salt bridge formed with R3504 from WolH, most likely causing decreased DD pair affinities, explaining the observed differences in peptide titers. The depreciation of this interaction is consistent with a loss of selective pressure for wollamide production and may indicate drift towards a desotamide-only BGC.
Gene duplication and divergence is a central mechanism behind the evolution of new gene functions37,52,53. Gene duplications may occur as whole gene duplications (characterised by the duplication of an entire gene), or as partial or intragenic duplications (characterised by incomplete duplication of the ancestral gene and can result in attenuated or lengthened duplicates respectively37). There is convincing evidence for the role of intragenic duplications during assembly line evolution (Fig. 6b). Such processes are often cited as the origin of multi-modularity1,54 and there is significant evidence for this process, especially in PKS assembly lines. In contrast, whole gene duplication has the potential to create bifurcated biosynthetic pathways, as evidenced here by the wol BGC, or parallel pathways, as evidenced by the recently published BE-18257A-C and pentaminomycins A-E BGC55. Through either mechanism, whole gene duplication creates redundancy and has the potential to reduce selective pressure for maintenance on both alleles37,53. In this fashion, a gene may evolve new functionality without losing the original product. Thus, neofunctionalization of one copy becomes possible without strong selective pressure against the loss of a phenotype and may follow a more gradual route for the emergence of a new activity. Subsequently, the original allele can be lost if there is no advantage and selection pressure for its maintenance leading to a new linear pathway (Fig. 6b). Moreover, this model provides a scenario where rounds of duplication and neofunctionalization can occur but if no advantage is gained then the duplicated allele can be lost, restoring the original genotype. As genes and pathways are subject to multiple selective forces and it is unclear how common this mechanism may be in Nature. This is compounded by the homogenising effects of gene conversion/concerted evolution that can make it difficult to reliably distinguish duplication events1,54,56,57. However, the wol BGC provides compelling evidence that this process has occurred at least once, during the evolution of desotamide biosynthesis.
Finally, the question arises as to whether our knowledge of the NRPS evolution can also be used to develop new engineering approaches. Recently, several efficient and/or highly productive engineering strategies have been published58–61. Yet efficient engineering of these often-huge biosynthetic machineries to produce novel bioactive NRPs is an ongoing challenge. Understanding the evolution of these multifunctional enzymes might provide new insight for engineering and discovering new peptide based therapeutic agents62. A key feature of desotamide evolution that could be exploited for future NRPS engineering efforts is the intragenomic recombination between distinct NRPS encoding BGCs. The recombination event we predicted takes place within the boundaries of A domains, consistent with previous38,39,63 and more recent work18,40. More specifically, the predicted recombination allows for the substitution of the substrate specificity conferring active site and the ATP/phosphate binding catalytic P-loop64 while functionally maintaining most of the A-T65 and all of the C-A interdomain contacts66–68. This pattern is frequently seen in our dataset, suggesting an evolutionary advantage for maintaining the structural relationships between the P-loop and substrate binding pocket. Our results are supported by two other independent studies, further highlighting its significance18,40.
Calcott and co-workers analysis of recombination hotspots within C-A-T tri-domains (modules) from Pseudomonas, Bacillus and Streptomyces species identified the same recombination sites as predicted for the formation of wolG1A240. Subsequently, Baunach and co-workers systematic in silico dissection of many individual recombination events unveiled the striking commonality of A domain recombination events in nature18. Specifically, these recombination events target variable portions of the Acore domains to modulate A domain substrates while domain-domain interactions and the flexible Asub domain largely remained unaffected. These studies, taken together with the evolutionary and biochemical evidence presented here, must inevitably lead to a change in paradigm of future NRPS engineering experiments. Established engineering principles are being overhauled by the increasing evidence base. In particular, the idea that C and A domains have coevolved69,70, resulting in strong acceptor site specificity of C domains71,72 is in question18,40,60. This hypothesis is not congruent with the observation that A domain recombination are responsible for the diversification of many NRPs18,38,73, not least the L-ornithine to glycine change demonstrated here. Such diversification would be impossible if C domain ‘gatekeeping’ was universal.
The ability to exchange A domains with greater accuracy should aid engineering efforts in the future, however many issues remain. As we have observed in nature, there is unlikely to be a single recombination site that will work in every case. Nevertheless, the growing body of genomic data will continue to reveal evolutionary snapshots akin to the wollamide BGC presented here. Over a decade of sequencing has provided us with only a threadbare sample of microbial genomic diversity and, as a result, we are still largely ignorant of the evolutionary processes that govern small molecule biosynthesis. Fortunately, this means there remains wealth of information yet to be gleaned from Nature.
Methods
For methods and experimental details, please see the supplementary information.
Contributions
B.W. and E.L. conceived the study. B.W., T.J.B., K.A.J.B. and J.L. designed the experiments, and T.J.B., K.A.J.B. and J.L. performed the experiments and analysed the data. B.W. and T.J.B. wrote the paper and all authors edited the paper.
Competing interests
E.L. is a co-owner of Microbial Screening Technologies. The other authors have no competing interests.
Acknowledgements
This work was supported by the Biotechnology and Biological Sciences Research Council (BBSRC) via Strategic Program Project BBS/E/J/000PR9790 to the John Innes Centre (JIC), and by Norwich Research Park Doctoral Training Program Studentship BB/M011216/1. We also acknowledge the support of the Cooperative Research Centres Projects Scheme (CRCPFIVE000119). We thank the Earlham Institute (Norwich UK) for sequencing and assembly of the Streptomyces sp. MST110588 genome. We also thank the JIC metabolomics platform at for excellent mass spectrometry support.
References
- (1).↵
- (2).↵
- (3).↵
- (4).↵
- (5).↵
- (6).↵
- (7).↵
- (8).↵
- (9).↵
- (10).↵
- (11).
- (12).
- (13).
- (14).↵
- (15).↵
- (16).
- (17).
- (18).↵
- (19).↵
- (20).↵
- (21).
- (22).↵
- (23).↵
- (24).↵
- (25).↵
- (26).↵
- (27).↵
- (28).↵
- (29).↵
- (30).↵
- (31).↵
- (32).↵
- (33).↵
- (34).↵
- (35).↵
- (36).↵
- (37).↵
- (38).↵
- (39).↵
- (40).↵
- (41).↵
- (42).↵
- (43).↵
- (44).↵
- (45).↵
- (46).↵
- (47).↵
- (48).↵
- (49).↵
- (50).↵
- (51).↵
- (52).↵
- (53).↵
- (54).↵
- (55).↵
- (56).↵
- (57).↵
- (58).↵
- (59).
- (60).↵
- (61).↵
- (62).↵
- (63).↵
- (64).↵
- (65).↵
- (66).↵
- (67).
- (68).↵
- (69).↵
- (70).↵
- (71).↵
- (72).↵
- (73).↵
- (74).