Triggering the expression of a silent gene cluster from genetically intractable bacteria results in scleric acid discovery† †Electronic supplementary information (ESI) available: Supplementary methods and results; Tables S1–S6; Fig. S1–S22. See DOI: 10.1039/c8sc03814g

The characterisation of scleric acid, a new natural product from a silent and cryptic gene cluster from genetically intractable bacteria, and its biosynthesis are reported.


Introduction
Actinomycete bacteria have been the foremost producers of antibiotics since the mid-1940s. In the last decade, highthroughput DNA sequencing technologies and novel bioinformatics tools have highlighted an immense number of uncharacterised biosynthetic gene clusters (BGCs) predicted to direct the assembly of bioactive natural products. 1 The presence of BGCs has not only been revealed in actinomycete genomes but also in those of human commensal and pathogenic bacteria (i.e. Staphylococcus lugdunensis, Burkholderia cepacia complex), as well as in the genomes of unculturable bacteria and in metagenomic libraries. [2][3][4][5][6][7] Despite the conspicuous number of specialised metabolites isolated from actinomycetes, only a small fraction of the natural products 'encrypted' at the DNA level has been exploited to date. Experimental characterisation of the biosynthetic product of a BGC is oen laborious and time-consuming particularly due to the uniqueness of every microorganism. Protocols for introducing DNA into bacterial cells are species-dependent and oen ineffective. Their optimisation can take years but many culturable micro-organisms remain genetically intractable. This prevents the exploitation of BGCs using many of the previously reported strategies. 1 In addition, the biosynthesis of specialised metabolites is oen tightly controlled at the transcriptional level. Cluster-associated transcriptional regulators that belong to the TetR-family of transcriptional repressors are particularly numerous. 8 Deletions of cluster-specic TetR-like transcriptional repressors have been shown to trigger overproduction of the corresponding specialised metabolites, as previously reported for the antibiotics methylenomycin and coelimycin in Streptomyces coelicolor A3(2) and for the urea-containing gaburedins in Streptomyces venezuelae. [9][10][11][12] Genetic manipulation of Streptomyces genomes has classically been accomplished using established but oen laborious protocols optimised for specic bacterial strains. 13 In recent years however, targeted genome editing has been revolutionised by the advent of clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated (Cas) systems, which allow generation of clean genomic deletions/insertions. 14 Toolkits for editing streptomycete genomes have been developed, [15][16][17] enabling researchers to overcome the issues associated with classical methods of gene disruption, in particular when multiple mutation events are desirable. 18 The number of available selectable markers and issues with potential restoration of the wild-type conguration due to occurrence of single-crossover events were notable limitations.
Here, we report a genome mining strategy based on the identication of a conserved regulatory cassette for selecting and characterising BGCs. A specic BGC was rst captured and transferred into a validated Streptomyces heterologous host where CRISPR/Cas9-mediated genome editing was employed to rationally derepress the expression of silent biosynthetic genes (Fig. 1). This approach was applied for the identication, isolation and structural elucidation of a novel structural class of natural products from a silent and cryptic gene cluster found in the soil-dwelling species Streptomyces sclerotialus NRRL ISP-5269, a species of lamentous bacteria rst isolated in Poona (India). 19

Results and discussion
Identication of the scl gene cluster in S. sclerotialus NRRL ISP-5269 The prioritisation of the gene cluster under study was guided by the presence of a specic set of ve regulatory genes that we had previously characterised in S. coelicolor A3(2), 9 and subsequently exploited to trigger the expression of a silent and cryptic biosynthetic gene cluster in Streptomyces venezuelae. 12 In S. coelicolor A3(2), these ve genes are responsible for regulation of methylenomycin biosynthesis: mmyR and mmfR both code for TetR-like transcriptional repressors; mmfLHP are responsible for the biosynthesis of signalling molecules, known as methylenomycin furans (MMFs) that trigger production of the methylenomycin antibiotics. 9 A mathematical model that explains in detail the methylenomycin regulatory system involving these ve genes has been developed and matched with experimental data. 20 In previous studies we have exploited this regulatory cassette and shown that inactivation of the mmyR-like transcriptional repressors is an effective approach for derepressing silent gene clusters in actinomycetes. 10,12 In order to nd gene clusters that contained regulatory cassettes homologous to the one found in the methylenomycin cluster and to assess how widespread these are, we performed searches with Multigene BLAST 21 using as a query the DNA sequences of mmyR, mmfR and mmfLHP, as well as with Clus-terTools, 22 using as a query the protein sequences of MmyR, MmfR and MmfLHP. Fourteen actinomycete genomes were found that contained orthologues of all ve genes coding for MmyR, MmfR and MmfLHP within a 50-kb region (ESI Table  S4 †). Remarkably, a total of 98 actinomycete genomes were found that contained orthologues of at least mmyR, mmfR and mmfL within a 50-kb region. We have previously shown that the butenolide synthase MmfL alone is sufficient to give production of functional MMF signalling molecules in S. coelicolor A3(2). 9 Additionally, a functional regulatory system that controls biosynthesis of coelimycin antibiotics in S. coelicolor A3(2) has been characterised that includes the butenolide synthase ScbA and the two TetR-like transcriptional regulators ScbR and ScbR2, 11 which are orthologues of MmfL, MmfR and MmyR respectively. Hence the regulatory systems that include orthologues of mmyR, mmfR and mmfL are also putatively functional, and the biosynthetic products they regulate expression of could also be studied via manipulation of these regulatory cassettes.
Among the hits generated, the gene cluster from S. sclerotialus NRRL ISP-5269, named hereaer scl cluster, was chosen for further study; the nucleotide sequence containing the scl cluster was available from the GenBank accession number JOBC01000043.1. In addition to homologues of the ve genes used as query, the genetic organisation of the scl gene cluster included two adjacent and divergent operons of biosynthetic genes (Fig. 2a). A combination of AntiSMASH 23 and manual BLASTp 24 analyses indicated the putative borders of the scl cluster ( Fig. 2a and Table 1). The predictive power of these modern bioinformatics tools oen permits to deduce the chemical structure(s) of cryptic gene cluster products, particularly when modular systems such as type I modular polyketide synthases (PKS) or non-ribosomal peptide synthetases (NRPS) direct the biosynthesis. 25 However, the originality of the scl cluster prevented such predictions and we worked on the assumption that a lack of bioinformatics prediction was more likely to result in a more structurally diverse and therefore truly novel natural product. The cluster spanned a region of 19 782 bp, and comprised 18 putative genes: 11 biosynthetic genes, 6 genes for regulation and 1 gene coding for a membrane transporter ( Fig. 2a and Table 1).

Capture of the scl gene cluster and introduction into heterologous hosts
In order to characterise the product of the scl cluster, we rst set out to examine the genetic tractability of S. sclerotialus NRRL ISP-5269. To do so, we attempted introduction of plasmid pCRISPomyces-2 (ref. 15) for CRISPR/Cas9-based genome engineering via intergeneric-conjugation of mycelia with E. coli ET12567/pUZ8002 or protoplast transformation but these processes were found ineffective. 13 The preparation of S. sclerotialus NRRL ISP-5269 spores was also attempted but despite screening various culture media in the aim to induce sporulation, this strain showed little aerial growth. We therefore set out to capture and heterologously express the scl gene cluster, with the aim of characterising its biosynthetic product(s) in a well-characterised and genetically amenable host. A 33-kb region of genomic DNA that included the 19.8-kb scl BGC was captured from the puried genomic DNA of  Fig. S1 †). 26 A pCAP03-derived plasmid was rst assembled to specically recombine in yeast with both extremities of the 33-kb target DNA fragment. 27 The pCAP03-scl construct was introduced and stably integrated into the genome of Streptomyces albus J1074 and S. coelicolor M1152 via intergeneric tri-parental conjugation with E. coli ET12567/pCAP03-scl and E. coli ET12567/pUB307. These two strains were chosen as they are well characterised chassis for heterologous expression of actinomycete gene clusters. 28,29 In parallel, the negative control strains S. albus/pCAP03 and S. coelicolor M1152/pCAP03 were generated using the "empty" pCAP03 plasmid. The four Streptomyces strains were grown on supplemented minimal agar medium for 5 days, and the acidied agar medium was extracted with ethyl acetate. Their metabolic proles were analysed by ultra-high-pressure liquid chromatography-high resolution mass spectrometry (UHPLC-HRMS). Comparison of the MS chromatograms failed to reveal any new compounds in the heterologous hosts where the scl BGC had been integrated.
Derepression of the scl gene cluster and characterisation of scleric acid The lack of accumulation of novel metabolites in the scl-containing strains was likely to be due to the transcriptional repression activity of the TetR-like regulators encoded in the scl cluster. We had previously shown that genetic inactivation of mmyR-like genes in S. coelicolor A3(2) and in S. venezuelae would specically derepress the expression of adjacent BGCs. 10, 12 We therefore decided to genetically inactivate the mmyR homologous gene, sclM4 using the CRISPR/Cas9-based plasmid pCRISPomyces-2 (pCm2). 15 For this purpose, we assembled the plasmid pCm2-sclM4 and attempted intergeneric-conjugation of S. albus/scl and S. coelicolor M1152/scl with E. coli ET12567/ pUZ8002/pCm2-sclM4. Ex-conjugants for strain S. albus/scl DsclM4 (see Fig. 2b for a representation of the genotype) were readily obtained, however no ex-conjugants could be obtained when attempting insertion of plasmid pCm2-sclM4 into S. coelicolor M1152/scl. The desired 20-bp out-of-frame deletion of sclM4 was conrmed by sequencing of a PCR product using S. albus/scl DsclM4 genome as a template (ESI Fig. S2 †). Metabolites produced by this strain were compared by UHPLC-HRMS to those of S. albus/scl and S. albus/pCAP03 grown under the same conditions as those described earlier. S. albus/scl DsclM4 showed accumulation of a major metabolite (Fig. 3a- . This compound was puried using a combination of ash chromatography on C 18 -silica column and HPLC. Its structure was then elucidated by a combination of 1D-and 2D-NMR spectroscopy experiments (ESI Fig. S5-S9 †). The novel compound was characterised as being (2-(benzoyloxy)acetyl)-L-proline and named scleric acid (Fig. 3d).
In order to establish the stereochemistry of the proline residue, scleric acid was hydrolysed and derivatised with Marfey's reagent. 30 L-and D-proline were also derivatised using the same procedure and used as standards for HPLC comparison. Approximately 95% of the proline residue of scleric acid puried from S. albus/scl DsclM4 was found to correspond to Lproline (ESI Fig. S10 †). To conrm the proposed structure of scleric acid, an authentic standard was synthesised (see ESI Fig. S20 † for a schematic representation of the synthetic route). A structural analogue, 2-((benzoyl-L-prolyl)oxy)acetic acid, possibly consistent with the initial NMR data obtained, was also synthesised (see ESI Fig. S21 † for a schematic representation of the synthetic route and ESI Fig. S11-S15 † for NMR spectra). LC-MS analyses and NMR data of both of these compounds unequivocally conrmed the proposed structure for scleric acid (Fig. 3); the analogue revealed different NMR spectra and its physico-chemical properties resulted in a different retention time on LC-MS. Moreover, two sets of NMR signals were observed for the puried natural product as well as for the synthetic standard and revealed that scleric acid existed as two different rotamers, trans-and cis-scleric acid ( Fig. 3e and ESI  Fig. S5 †). This is consistent with literature data for synthetic N-benzoyl-L-proline methyl ester where a 4 : 1 mixture of the two rotamers was observed. 31 Dening key biosynthetic genes in the scl cluster BLASTp analyses 24 combined with the elucidated structure for scleric acid allowed us to propose plausible functional assignments for all the biosynthetic enzymes coded in the scl BGC (Table 1 and Fig. 4). [32][33][34] Three main building blocks were identied as being part of scleric acid: a glycolic acid unit (highlighted in red in Fig. 4), a benzoic acid unit (in magenta in Fig. 4) and an L-proline residue (in blue in Fig. 4). The origin of these three building blocks and the overall biosynthesis of scleric acid are discussed in detail here. The four proteins encoded by genes sclQ1-4 showed high homology to a set of three enzymes -QncN, QncL and QncMfrom Streptomyces melanovinaceus. This set of enzymes has been shown to direct the biosynthesis and attachment of a C 2 -glycolicacyl unit to a non-ribosomal peptide. 34 More specically, SclQ2 was homologous to QncN, a thiamin diphosphate (ThDP) binding domain. SclQ3 was homologous to the rst two Nterminal domains of QncL, a pyruvate dehydrogenase/ transketolase pyrimidine binding domain and a transketolase C-terminal domain while SclQ4 was homologous to the last two domains of QncL, a lipoyl attachment domain and an acyltransferase catalytic domain. Lastly, SclQ1 was homologous to the acyl carrier protein (ACP) QncM. Based on the homology of SclQ1-4 with QncN, QncL and QncM, we propose that SclQ1-4 are overall responsible for converting of a ketose phosphate from the primary metabolism (such as xylulose-5-phosphate) into the activated glycolic acid unit found in scleric acid.
The three-gene cassette made of sclA, sclD and sclI showed high homology to genes involved in biosynthesis of a benzoic acid unit in Streptomyces sp. YN86. 33 Specically, SclA showed high similarity to the anthranilate synthase enzyme PauY18, SclD to the DAHP synthase PauY21 and SclI to the isochorismatase PauY19. Overall these three enzymes were hypothesised to be responsible for biosynthesis of the benzoyl group found in scleric acid via chorismate as an intermediate.
SclN was predicted to be a NRPS enzyme consisting of a single minimal elongation module: a putative, atypical condensation domain (C), an adenylation domain (A) and a peptidyl carrier protein (PCP) domain. 35 The SclN A-domain was predicted to specically activate L-proline, which was in accordance with the presence of a L-proline residue in scleric acid. 25 The SclN C-domain was proposed to catalyse the amide bond formation between L-proline and glycolic acid.
Other genes putatively involved in biosynthesis and export of scleric acid and present in the scl cluster are: sclT, sclG and sclE. The thioesterase SclT is predicted to release the L-proline-oxyacetic acid intermediate from the PCP-domain of SclN. We propose that the ATP-grasp family enzyme SclG would bind and activate the benzoic acid unit produced by SclADI. That same enzyme would also promote condensation of the benzoyl unit with L-proline-oxyacetic acid, giving scleric acid. This would be exported out of the cell by the putative MFS transporter SclE.
In order to conrm the proposed involvement of the enzymes SclN, SclA and SclQ1-4 in the biosynthesis of the building blocks that make up scleric acid, we constructed gene deletion mutants in strains where the transcriptional repressor sclM4 has also been inactivated (S. albus/scl DsclM4 background). Plasmids pCm2-sclN, pCm2-sclA and pCm2-sclQ1-4 were assembled and used to generate double mutant strains S. albus/scl DsclM4 DsclN, S. albus/scl DsclM4 DsclA and S. albus/scl DsclM4 DsclQ1-4. Deletions were conrmed by PCR screening (ESI Fig. S3 †). UHPLC-HRMS analysis revealed that production of scleric acid was abolished in S. albus/scl DsclM4 DsclN and S. albus/scl DsclM4 DsclA (Fig. 3a), conrming the essential role of SclN and SclA. Residual scleric acid production was detected from S. albus/scl DsclM4 DsclQ1-4; this could be explained by the fact that glycolic acid is known to be produced by Streptomyces species in particular for the biosynthesis of N-glycolylmuramic acid. 36 Addition of 5 mM glycolic acid to the culture medium of S. albus/scl DsclM4 DsclQ1-4 also resulted in scleric acid being produced in similar level to that observed with S. albus/scl DsclM4 (ESI Fig. S16 †).
The identication of key precursors in scleric acid biosynthesis was also exploited to further increase the titres of scleric acid produced by S. albus/scl DsclM4. Enriching the culture medium with 5 mM L-proline, 5 mM benzoic acid or 5 mM glycolic acid signicantly increased levels of scleric acid observed upon UHPLC-HRMS analysis of the ethyl acetate extracts compared to those observed with S. albus/scl DsclM4 grown on the standard supplemented minimal medium (ESI Fig. S17 †). The strategy of manipulating the pathway-specic transcriptional regulatory system also makes scleric acid production not reliant on a complex culture medium. Importantly the utilisation of supplemented minimal media does signicantly facilitate the isolation of the natural product of interest.

Evidence for proposed biosynthetic intermediates
Based on the predicted function of the scl biosynthetic genes, we therefore proposed a putative biosynthetic route to scleric acid as shown in Fig. 4. A glycolic acid unit is believed to be produced by SclQ2-4 and loaded onto the SclQ1 ACP. SclP would convert apo-SclQ1 into holo-SclQ1 by phosphopantetheinylation. Meanwhile the adenylation domain of the NRPS SclN would activate L-proline, which could then be condensed to the glycolic acid unit through the activity of the C-domain of SclN. The L-proline-oxyacetic acid intermediate would then be released from the PCP-domain of SclN by the activity of the thioesterase SclT. The three enzymes SclADI would direct the biosynthesis of the benzoic acid moiety, which would be activated by the ATP-grasp family enzyme SclG. That same enzyme would also catalyse the condensation of the benzoyl unit with L-proline-oxyacetic acid, giving scleric acid. Lastly, scleric acid would be exported out of the cell by the putative MFS transporter SclE.
In support to this proposed pathway, we investigated by UHPLC-HRMS accumulation of the L-proline-oxyacetic acid intermediate from the ethyl acetate extracts of the scleric acid producing strain, as well as of the double mutant strains. A compound with a retention time of 3.0 minutes on C 18 reverse phase HPLC column and an m/z value of 174.0763 [M(C 7 H 12 NO 4 ) + H] + (calculated m/z of 174.0761) was detected in the scleric acid producing strain S. albus/scl DsclM4, as well as in the double mutants S. albus/scl DsclM4 DsclQ1-4 and S. albus/scl DsclM4 DsclA (ESI Fig. S18 †). Consistent with the predicted function of the L-proline-activating NRPS SclN, strain S. albus/scl DsclM4 DsclN did not show any accumulation of L-proline-oxyacetic acid. In order to further conrm the identity of this intermediate, which was detected from the crude extracts in amounts not sufficient for HPLC purication and subsequent NMR characterisation, a synthetic standard was prepared (see ESI Fig. S22 † for a schematic representation of the synthetic route) and run alongside the crude extracts on UHPLC-HRMS. This showed the same retention time and mass spectrum as the natural product L-proline-oxyacetic acid intermediate (ESI Fig. S18 †). Moreover, we grew S. albus/scl DsclM4 DsclN, unable to produce scleric acid, in the presence of 5 mM L-prolineoxyacetic acid. Feeding the intermediate to the mutant strain restored production of scleric acid, as visible from UHPLC-HRMS analysis of its acidied ethyl acetate extracts (ESI Fig. S19 †). This provides additional evidence that L-prolineoxyacetic acid is a true intermediate in scleric acid biosynthesis. It also suggests that the L-proline-oxyacetic acid is released from the C-domain of SclN prior to SclG catalysing its condensation with the benzoyl group, in accordance with the order of reactions proposed in Fig. 4.

Biological activity of scleric acid
The antimicrobial activity of scleric acid was assessed. Inhibitory activity against representatives of the ESKAPE panel of pathogens: Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, and Enterobacter cloacae was rst screened but all strains appeared to be resistant to scleric acid, giving no observable MIC (ESI Table S6 †).
Scleric acid was then tested for a broader range of pharmaceutically relevant bioactivities through the Eli Lilly Open Innovation Drug Discovery (OIDD) Program. In a single point (20 mM) primary assay, scleric acid showed moderate antibacterial activity against Mycobacterium tuberculosis (H37Rv), exhibiting a 32% inhibition on the growth of this strain.

Conclusions
Specialised metabolites from actinomycete bacteria are one of the most valuable sources of novel antibiotics, as well as of other useful bioactive compounds employed in various elds, from human medicine to crop protection. High-throughput sequencing of bacterial genomes/metagenomes has become quick and inexpensive, and is unearthing a myriad of putative gene clusters that are awaiting to be characterised and exploited. In this study, we have demonstrated that a cryptic gene clusters from a genetically intractable actinomycete could be successfully exploited for characterising the biosynthetic pathway it encodes as well as the resulting natural product. In principle, the same approach could be used with metagenomic DNA. Our strategy rst relied on the bioinformatics identication and selection of cryptic gene clusters containing a characterised regulatory system (Fig. 1). A DNA fragment containing the entire gene cluster was then captured via TAR cloning and introduced in the genome of a validated heterologous host. Expression of the biosynthetic genes was then triggered through CRISPR/Cas9-mediated editing of the key mmyR-like transcriptional repressor. Analytical chemistry procedures were then undertaken to identify, isolate and characterise the new natural product, overproduced in the heterologous host. Finally, subsequent rounds of CRISPR/Cas9-mediated gene deletions targeted putative biosynthetic genes and afforded mutant strains. In turn, a biosynthetic route to the novel natural product could be proposed. This approach was validated by the discovery of (2-(benzoyloxy)acetyl)-L-proline, named scleric acid, from the genome of the soil-dwelling lamentous bacterium Streptomyces sclerotialus NRRL ISP-5269. Based on the predicted function of the Scl enzymes and the detection of a key intermediate from culture extracts of selected strains a plausible biosynthetic route to scleric acid was proposed (Fig. 4). In addition to the novel biochemistry this biosynthetic pathway offers, scleric acid has been shown to exhibit moderate antibacterial activity against M. tuberculosis, as well inhibition on the cancer-associated enzyme nicotinamide N-methyltransferase (NNMT). We are currently investigating scleric acid bioactivity further through the Eli Lilly Open Innovation Drug Discovery (OIDD) Program.
The widespread presence of orthologues of the methylenomycin regulatory genes among actinomycete genomes (ESI Table S4 †) revealed that the approach described herein might be very promising for the discovery and characterisation of novel natural products, and therefore, of novel biocatalysts. Comparative genomics analyses also indicated that there is no apparent correlation between the presence of the regulatory cassette we targeted and the type of natural products that they regulate production ofboth in relation to biosynthesis and bioactivitymethylenomycin, 9 gaburedins 12 and scleric acid being examples of natural products characterised so far.
In conclusion, beyond the discovery of this specialised metabolite, we strongly believe that targeting conserved pathway-specic regulatory elements, as opposed to mining BGCs encoding dened enzymatic machineries (i.e. PKS, NRPS), will lead to the identication and characterisation of microbial natural products assembled by truly novel types of biocatalysts.

Bioinformatics analysis
The whole genome sequence of Streptomyces sclerotialus NRRL ISP-5269 was downloaded from the Genomes Online Database of the JGI Portal (U.S. Department of Energy, https:// gold.jgi.doe.gov/, GOLD Project ID Gp0187859). Contig43 of the genome contained the scl gene cluster object of this study. A combination of AntiSMASH 23 and manual BLASTp 24 analyses allowed to dene the putative borders of the scl cluster and predict the function of the encoded enzymes. The website for PKS/NRPS analysis from the University of Maryland was used for domain prediction of SclN. 25 SerialCloner 2.6.1 (SerialBasics) was used for DNA sequence analysis and plasmid design. Multigene BLAST 21 was used to search for gene clusters that included homologues of the methylenomycin regulatory genes from S. coelicolor A3(2) and allowed to pinpoint the scl BGC from S. sclerotialus NRRL ISP-5269. ClusterTools was used to search orthologues of the proteins involved in regulation of methylenomycin production in other actinomycetes (ESI Table S4 †), as well as orthologues of the scl cluster proteins in other actinomycetes (ESI Table S5 †). 22

Reagents
All chemicals were purchased from Sigma-Aldrich, unless otherwise stated. Phusion DNA polymerase, as well as all restriction endonucleases, T4 DNA ligase, shrimp alkaline phosphatase (rSAP) and Gibson Assembly cloning kit, were purchased from New England Biolabs. Zymolyase 20T was purchased from MP Biomedicals, 5-uoroorotic acid (5-FOA) was purchased from Thermo Fisher Scientic. Primers for PCR amplication were purchased from Sigma-Aldrich (see ESI  Table S3 † for a list of oligonucleotides used in this study).

Culturing and engineering of microorganisms
Streptomyces sclerotialus NRRL ISP-5269 was obtained from JCM (Japan Collection of Microorganisms, culture collection number 4828 T ) (see ESI Table S1 † for a list of strains used in this study). Streptomyces albus J1074 and Streptomyces coelicolor M1152 were used for heterologous expression. All Streptomyces strains were grown on soya our mannitol (SFM) agar medium (20 g L À1 soya our, 20 g L À1 mannitol, 20 g L À1 agar), with appropriate antibiotic selection upon insertion of plasmid DNA (50 mg mL À1 apramycin when transformed with pCRISPomyces-2-derived plasmids; 25 mg mL À1 kanamycin when transformed with pCAP03-derived plasmids; 25 mg mL À1 nalidixic acid on the rst round of subculture aer intergeneric conjugation). E. coli ET12567 and ET12567/pUB307 were used for the purpose of intergeneric tri-parental conjugation. One Shot TOP10 chemically competent E. coli cells (Thermo Fisher Scientic) were used for cloning and storage of plasmid DNA. All E. coli strains were grown on lysogeny broth (LB) medium (10 g L À1 tryptone, 5 g L À1 yeast extract, 10 g L À1 NaCl) or LB agar medium (same as LB medium, with 15 g L À1 agar), with appropriate antibiotic selection (50 mg mL À1 apramycin when transformed with pCRISPomyces2-derived plasmids, 25 mg mL À1 chloramphenicol to maintain the dam mutation in E. coli ET12567, 25 mg mL À1 kanamycin either to maintain helper plasmid pUB307 or aer insertion of pCAP03-derived plasmids).
S. cerevisiae VL6-48N was used for TAR cloning and grown on yeast extract peptone (YPD) broth (5 g L À1 yeast extract, 10 g L À1 peptone, 2% w/v glucose) or YPD agar (same as YPD, with 15 g L À1 agar). Purication of genomic DNA from S. sclerotialus was performed from a 100 mL liquid culture by phenolchloroform extraction. 13 The scl gene cluster was captured using TAR cloning. 26 Assembly of plasmid pCAP03-scl was performed following the procedure described by Moore and colleagues; pCAP03 was a gi from Bradley Moore (Addgene plasmid # 69862) (see ESI Table S2 † for a list of plasmids used in this study). 27 For this purpose String DNA fragments (Thermo Fisher Scientic) were ordered to include 60-bp hooks homologous to either side of the scl cluster (ESI Table S2 †) and introduced into pCAP03 via Gibson Assembly (New England Biolabs). The identity of the captured cluster was conrmed by PCR amplication and restriction digestion (ESI Fig. S1 †). Insertion of the scl gene cluster in the genome of the heterologous hosts S. albus and S. coelicolor was accomplished via intergeneric tri-parental conjugation following the protocol described by Moore and colleagues. 27 CRISPR/Cas9-based engineering of S. albus strains was performed using plasmids pCm2-sclM4, pCm2-sclN, pCm2-sclA and pCm2-sclQ1-4. Golden Gate Assembly was rst performed to insert the specic sgRNAs into the backbone pCm2 plasmid, then Gibson Assembly was used to include 800-bp homologous recombination arms, all following the procedure described by Zhao and colleagues; pCRISPomyces-2 was a gi from Huimin Zhao (Addgene plasmid # 61737). 15 Clearance of temperature sensitive plasmids based on pCm2 was achieved by culturing the mutant strains on SFM agar medium non-selectively at 39 C.
Identication, isolation and structure elucidation of scleric acid S. albus strains were cultured for 5 days at 30 C on supplemented solid minimal (SM) medium (2 g L À1 casaminoacids, 8.68 g L À1 TES buffer, 15 g L À1 agar; aer autoclaving, and just before use, 10 mL of 50 mM NaH 2 PO 4 + K 2 HPO 4 , 5 mL of 1 M MgSO 4 , 18 mL of 50% w/v glucose and 1 mL of trace element solution [0.1 g L À1 each of ZnSO 4 $7H 2 O, FeSO 4 $7H 2 O, MnCl 2 -$4H 2 O, CaCl 2 $6H 2 O and NaCl] were added) as described by Hopwood and colleagues. 13 Ethyl acetate was added in equal volume to the volume of SM medium used, and acidied to pH 3 by the addition of 37% HCl. The ethyl acetate layer was removed and evaporated under reduced pressure. The remaining residue was dissolved in 500 mL of 50 : 50 (v/v) HPLC grade methanol/ water for UHPLC-HRMS analysis. For purication of scleric acid, 2 L of SM culture medium was used for organic extractions, and the remaining residue aer evaporation of ethyl acetate was dissolved in 10 mL 50 : 50 HPLC grade methanol/ water for silica column pre-purication.
UHPLC-HRMS analyses were carried out with 20 mL of prepared extracts injected through a reverse phase column (Zorbax Eclipse Plus C18, size 2.1 Â 100 mm, particle size 1.8 mm) connected to a Dionex 3000RS UHPLC coupled to Bruker Ultra High Resolution (UHR) Q-TOF MS MaXis II mass spectrometer with an electrospray source. Sodium formate (10 mM) was used for internal calibration and a m/z scan range of 50-1500 was used with a gradient elution from 95 : 5 solvent A/solvent B to 0 : 100 solvent A/solvent B over 10 minutes.
Pre-purication of crude extract containing scleric acid was performed using ash chromatography. A column was loaded with C18-reversed phase silica gel, preconditioned with one volume of methanol, activated with one volume of solvent B (0.045% v/v triuoroacetic acid in acetonitrile), and equilibrated with two volumes of solvent A (0.045% v/v triuoroacetic acid in water). Crude extract was loaded onto the column. Compounds were then eluted with ve different consecutive solvent systems: two volumes of 20 : 80 solvent B/solvent A, two volumes of 40 : 60 solvent B/solvent A, two volumes of 50 : 50 solvent B/solvent A, two volumes of 60 : 40 solvent B/solvent A and two volumes of 80 : 20 solvent B/solvent A. Fractions were collected throughout the elution steps, evaporated under reduced pressure and dissolved in 500 mL of 50 : 50 (v/v) HPLC grade methanol/water for UHPLC-HRMS analysis. Fractions containing scleric acid were combined and used for HPLC purication.
Reverse-phase HPLC was performed using a Zorbax XBD-C18 column (212 Â 150 mm, particle size 5 mm) connected to an Agilent 1200 HPLC equipped with a binary pump and DAD detector. Solvent A: 0.1% TFA water, solvent B: 0.1% TFA in acetonitrile, 5% B to 95% B in 45 min. Retention time compound 1: 29.7 min, retention time compound 2 (scleric acid): 34.4 min. Gradient elution was used (solvent A: water with 0.1% HCOOH, solvent B: methanol) with a ow rate of 10 mL min À1 . Fractions were collected by time or absorbance at 210 nm using an automated fraction collector. The fractions collected containing scleric acid were pooled, methanol removed under reduce pressure and scleric acid was re-extracted from the remaining water (2 Â 50 mL ethyl acetate). The ethyl acetate was removed under reduced pressure and the sample re-dissolved in deuterated methanol for NMR analysis.

MIC testing
The susceptibility of bacterial strains was investigated in collaboration with the Warwick Antimicrobial Screening Facility in a 96-well plate experiment, according to the Clinical & Laboratory Standards Institute (CLSI) guidelines (M7-A9 2012). Scleric acid was diluted to a concentration of 7.5 mg mL À1 in 25% DMSO in distilled water. To further prevent toxicity effects from DMSO in MIC testing, 27 mL of the natural product stock was combined with 173 mL of cation adjusted Muller-Hinton broth to a nal concentration of 1024 mg mL À1 of compound in 200 mL with 3% DMSO. This was then further doubling diluted throughout the MIC. Meropenem and cefoxitin were used as positive controls during MIC testing.

Conflicts of interest
There are no conicts to declare.