A collection of seventy one Streptomyces strains was used in this work. For sporulation, strains were grown for 7 days at 30°C on MA plates (Sánchez and Braña, 1996). For metabolite production studies, spores were added to flasks containing 40 ml TSB medium and grown at 30°C and 250 rpm. After 24 h, this culture was used to inoculate 50 mL of five different media to a final OD_600nm = 0.2: R5A medium (Fernández et al., 1998); SM medium (Zhang et al., 1992); SM10 medium (per liter): 20.9 g MOPS, 11.5 g L-Proline, 23 g glycerol, 0.5 g NaCl, 2.1 g K₂HPO₄, 0.28 g Na₂SO₄, 1 mL MgSO₄ 0.2 M, 1 mL CaCl₂ 0.2 M and 5 mL trace elements [composition per liter: 0.04 g ZnCl₂, 0.2 g FeCl₃.6H₂O, 0.01 g CuCl₂.2H₂O, 0.01 g MnCl₂.4H₂O, 0.01 g Na₂B₄O₇.10H₂O, 0.01 g (NH₄)₆Mo₇O₂₄.4H₂O], pH 6.5; SM17 medium (per liter): 2 g glucose, 40 g glycerol, 2 g soluble starch, 5 g soybean flour, 5 g peptone, 5 g yeast extract, 5 g NaCl and 2 g CaCO₃ in tap water; and SM19 medium (per liter): 40 g tomato juice, 15 g oat flour and 2 g glucose. For intergeneric conjugation, MS medium (Hobbs et al., 1989) was used. Escherichia coli DH5α (Hanahan, 1983) was used as host for cloning and E. coli ET12567/pUB307 for intergeneric conjugation (ET12567/pUB307) (Kieser et al., 2000) and both were grown in 2×TY medium supplemented, when required, with the appropriate antibiotic.

Culture media were supplemented with antibiotics when plasmid-bearing strains were used: apramycin (100 μg/mL for E. coli, 25 μg/mL for Streptomyces), kanamycin (25 μg/mL), tetracycline (10 μg/mL), chloramphenicol (25 μg/mL), and/or nalidixic acid (50 μg/mL).

DNA manipulations were performed according to standard procedures for E. coli (Sambrook et al., 1989) and Streptomyces (Hanahan, 1983). Plasmids used in this work were pOJ260 (Bierman et al., 1992) for gene interruption and pEM4ATc, pSETec (Cano-Prieto et al., 2015) and pOJ260P (Olano et al., 2004) for gene expression under the control of ermE^∗ promoter.

pEM4ATc was generated by cloning a BamHI-HindIII fragment containing the apramycin resistance gene aac(3)IV from pEFBA (Lozano et al., 2000) (ends were filled by Klenow activity following manufacturer indications) into pEM4T (Menendez et al., 2006) digested with EcoRV.

Probes 2,3DH and GS-DH were obtained by PCR amplification of total DNA with degenerate oligonucleotide pairs PCR-D1a/PCR-D1b (Trefzer et al., 2002) and AG5F/4,6DH2 (Hyun et al., 2000; Ogasawara et al., 2004), respectively (Supplementary Table 12). PCR conditions were: initial denaturation at 94°C, 5 min; 35 cycles of 94°C, 1 min, 60°C, 1 min and 72°C, 1 min and a final extension cycle at 72°C, 10 min. DreamTaq DNA polymerase (Thermo Scientific) and 2.5% dimethyl sulfoxide (DMSO) were used. Amplicons were verified by sequencing, digested with EcoRI/HindIII and subcloned into pOJ260. The resulting plasmids were introduced into Streptomyces strains by intergeneric conjugation.

The Streptomyces sp. CS057, CS113, CS159 and CS227 chromosomes were sequenced at Department of Biochemistry, University of Cambridge (Cambridge, United Kingdom) using Illumina MiSeq Sequencing technology. De novo assemblies were achieved using default parameters in Newbler assembler software version 2.9. Annotation was performed using the PGAAP pipeline¹ (Angiuoli et al., 2008). Database searching and sequence analysis were carried out with the bioinformatic tool antiSMASH (Blin et al., 2017).

These Whole Genome Shotgun projects have been deposited at DDBJ/ENA/GenBank under the accessions NEVF00000000 (Streptomyces sp. CS057), NEVC00000000 (Streptomyces sp. CS113), NEVD00000000 (Streptomyces sp. CS159) and NEVE01000000 (Streptomyces sp. CS227). The versions described in this paper are version NEVF01000000, NEVC01000000, NEVD01000000, and NEVE01000000, respectively.

Gene clusters for warkmycin and cervimycin have been deposited at Minimum Information about a Biosynthetic Gene Cluster (MIBiG) repository (Medema et al., 2015) under the accessions BGC0001438 and BGC0001439, respectively.

For cluster activation in CS113, cmv22 was expressed under the control of ermE^∗p. A 0.96 kb fragment was amplified with oligonucleotides 113-eSARP78.5b and 113-eSARP78.3b, gel purified, digested with BamHI/EcoRI and cloned into pSETec, pEM4ATc, and pOJ260P, leading to final plasmids pSET113eSARP, pEM113eSARP, and pOJ113eSARP, respectively.

orf15 of CS159 was amplified with oligonucleotides pErmE6RT15.5b and pErmE6RT15.3b. The resulting 0.79 kb fragment was gel purified, digested with BamHI/EcoRI and cloned into pSETec, pEM4ATc, and pOJ260P, leading to final plasmids pSET159eRT15, pEM159eRT15, and pOJ159eRT15, respectively.

Using genomic DNA of CS227, orf2 (0.92 kb) and orf37 (2.8 kb) were amplified with oligonucleotide pairs 227-eSARP2.5/227-eSARP2.3 and 227-eLuxR37.5/227-eLuxR37.3, respectively. Both amplicons were gel purified, digested with BamHI/EcoRI and cloned into pSETec, pEM4ATc, and pOJ260P, leading to final plasmids pSET227eSARP, pEM227eSARP, pOJ227eSARP, pSET227eLuxR, pEM227eLuxR, and pOJ227eLuxR.

wmc32 of CS057 was amplified with oligonucleotide 57-eReg.5b and 57-eReg.3b. A 0.84 kb fragment was gel purified, digested with XbaI and cloned into pSETec, pEM4ATc, and pOJ260P, leading to final plasmids pSET57eReg, pEM57eReg, and pOJ57eReg, respectively. The correct orientation of wmc32 was confirmed by restriction enzyme digestion and sequencing.

All amplifications were carried out with the high fidelity polymerase Herculase II Fusion (Agilent Technologies) and PCR conditions were: initial denaturation at 98°C, 2 min.; 30 cycles of 98°C, 30 s, 58°C, 30 s and 72°C, 30 s (3 min for orf37) and a final extension at 72°C, 3 min. The final plasmids were introduced into their respective strains by intergeneric conjugation. All oligonucleotides used are listed in Supplementary Table 12.

Whole cultures (1 mL) of selected strains were extracted at three different times (3, 5, and 7 days of culture) and with three different organic solvents [ethyl acetate, ethyl acetate containing formic acid (1%) or butanol] and analyzed by reversed phase chromatography in an Acquity UPLC instrument fitted with a BEHC18 column (1.7 μm, 2.1 mm × 100 mm, Waters), with acetonitrile and MQ water + 0.1% TFA as solvents. Samples were eluted with acetonitrile (10%) for 1 min, followed by a linear gradient of acetonitrile (10–100%) over 7 min (flow rate of 0.5 mL/min; column temperature 35°C). For HPLC-MS analysis, an Alliance chromatographic system coupled to a ZQ4000 mass spectrometer and a SunFire C18 column (3.5 μm, 2.1 mm × 150 mm, Waters) was used with the above solvents. Elution was performed with an initial isocratic hold with acetonitrile (10%) for 4 min followed by a linear gradient of acetonitrile (10–88%) over 30 min (0.25 mL/min). MS analyses were done by ESI in the positive mode (capillary voltage 3 kV, cone voltage 20 V). Detection and spectral characterization of peaks were performed by photodiode array detection with Empower software (Waters).

Products 11 and 12 were isolated from two-liter culture of CS149 in R5A medium (40 mL × 50 mL, in 250 mL flasks). After the mycelium was removed by centrifugation, supernatants were filtered and then a pass-through cleanup strategy was used on a C18 cartridge (10 g, Waters) to eliminate non-desirable compounds. The sample was extracted with twice volume of ethyl acetate, concentrated under vacuo, resuspended in 1.5 mL 50:50 DMSO:MeOH and subsequently subjected to semipreparative HPLC using an Atlantis dC18 column (10 μm, 10 mm × 150 mm, Waters) and MQ water + 0.05% TFA/ACN (55:45) as a mobile phase at 5 mL/min.

Products 22 and 24 were isolated from 5 mL × 400 mL (in 2 L flasks) cultures of CS057 in R5A medium whose supernatants were first filtered, then concentrated on a C18 cartridge (10 g, Waters), and subsequently fractioned on a 0.1% TFA–MeOH gradient. Fractions containing desired compounds were dried in vacuo, resuspended in 1mL of 50:50 DMSO:MeOH, and processed on a preparative HPLC using a SunFire C18 column (10 μm, 10 mm × 250 mm, Waters) and an isocratic mixture 60:40 MQ water:ACN at 5 mL/min.

The purity of the isolated peaks was determined by HPLC–MS before structural elucidation. The isolated compounds were then dried in vacuo, resuspended in 50:50 tert-butanol:water and lyophilised. Yield and productivity of each compound was as follows: 11, 1.1 mg (0.55 μg/mL); 12, 0.5 mg (0.25 μg/mL); 22, 2.1 mg (1.05 μg/mL); and 24, 5.8 mg (2.9 μg/mL).

The new compounds sipanmycin A (11), sipanmycin B (12), warkmycin CS1 (22) and warkmycin CS2 (24) were analyzed by LC-DAD-ESI-TOF to obtain their UV-vis (DAD) spectra and determine their molecular formula based on the experimental accurate masses and the corresponding isotopic distribution. The structural elucidation of these compounds was carried out by detailed analysis of 1D (¹H) and 2D (COSY, TOCSY, NOESY, HSQC and HMBC) NMR spectra further assisted by comparison with the spectral data reported for incednine (Futamura et al., 2008), silvalactam (Schultz et al., 2012), warkmycin and 4-O-deacetyl-warkmycin (Helaly et al., 2013). Solvents used in the NMR analyses were deuterated methanol (CD₃OD) for compounds 11 and 12, or deuterated chloroform (CDCl₃) for compounds 22 and 24. Relative configurations were determined by coupling constants and NOEs analyses. The molecular model of 11 was generated in Chem3D starting from the reported model for incednine (Futamura et al., 2008).

LC-DAD-ESI-TOF analyses were performed using an Agilent 1200RR HPLC equipped with a SB-C8 column (2.1 mm × 30 mm, Zorbax) coupled to a Bruker maXis Spectrometer. Solvent A consisted of 10% acetonitrile and 90% water with 1.3 mM trifluoroacetic acid (TFA) and ammonium formate, and solvent B was 90% acetonitrile and 10% water with 1.3 mM TFA and ammonium formate. The gradient started at 10% B and went to 100% B in 6 min, kept at 100% B for 2 min and returned to 10% B for 2 min to initialize the system. The mass spectrometer was operated in positive ESI mode. The instrumental parameters were: 4 kV capillary voltage, drying gas flow of 11 L/min at 200°C, nebulizer pressure at 2.8 bars. TFA-Na cluster ions were used for mass calibration of the instrument prior to sample injection. Pre-run calibration was by infusion with the same TFA-Na calibrant.

NMR spectra were recorded at 24°C on a Bruker AVANCE III-500 MHz (500 and 125 MHz for ¹H and ¹³C NMR, respectively) equipped with a 1.7 mm TCI MicroCryoProbe^TM, using the residual solvent signal as internal reference. Structural features were compared using the Dictionary of Natural Products (DPN) NMR Features database created by Prof. John Blunt, as described in Pérez-Victoria et al. (2016).

In the context of our study, a collection of seventy one Streptomyces strains were screened for the presence of genes potentially involved in the biosynthesis of 6DOH. Total DNA from these strains was extracted and purified, and the presence of 6DOH biosynthesis genes determined by PCR analysis using different primers. On one hand, two primers located one within conserved regions in GSs (forward primer) and a second one (reverse primer) located within conserved regions for NDP-D-glycose-4,6-dehydratases (primers “GS-DH”). These two genes are usually (but not exclusively) adjacent in the chromosome and they are common to the biosynthesis of all 6DOH. The second pair of primers corresponded to internal and conserved regions for NDP-D-glucose-2,3-dehydratases, enzymes catalyzing the 2,3-dehydration step in the biosynthesis of 2,6-DOH (primers “2,3DH”). PCR analysis of the strains showed that forty nine strains gave positive amplicons with the “GS-DH” primers and fifty-two strains with the “2,3DH” primers, with a total of fifty-five strains being positive in at least one of the assays. The presence of the expected DNA sequence was verified by sequencing each amplicon.

The amplicons of the fifty five strains giving positive amplification with any of the two pairs of oligo primers were subcloned as EcoRI-HindIII fragments in pOJ260, an E. coli vector unable of replication in Streptomyces. The resultant constructs were introduced in the corresponding strains from which they were originated by intergeneric conjugation E. coli-Streptomyces and transconjugants selected for resistance to apramycin (resistance marker in the vector). Because pOJ260 is a suicide vector in Streptomyces spp., these transconjugants must be the consequence of an integration of the construct in the chromosome by homologous recombination through the insert in the vector. After several conjugation experiments, transconjugants were obtained in ten of the fifty-five strains tested. The lack of transconjugants in the remaining strains could be due to low efficiency of transconjugation or recombination, presence of DNA restriction systems, etc., but this was not further studied. Four mutants were obtained using the corresponding “GS-DH inserts” in the host strains CS65a, CS90a, CS149, and CS227, and seven mutants using the “2,3DH inserts” in the host strains CS57, CS92, CS113, CS123, CS147, CS149, and CS159. In one case, strain CS149, mutants were obtained with both “GS-DH” and “2,3DH” inserts.

In order to determine if inactivation of the selected 6DOH biosynthesis genes gave rise to non-producing mutants and thus to the identification of the produced compounds, the ten strains and their corresponding mutants were grown on R5A liquid medium. Cultures were extracted with three different solvent systems (ethyl acetate, ethyl acetate plus 1% formic acid or butanol) at 3, 5, and 7 days and the metabolic profiles analyzed by UPLC. The different strains fall down in two different groups. The first group included strains showing some differential peaks, present in the wild type and absent in the mutant; this was the case for six strains: CS065a, CS090a, CS92, CS123, CS147, and CS149 (Figures 1A–F). The second group included strains not showing differential peaks between the wild type and the mutant strains in samples extracted with any of the solvents described above or at any time tested; this was the case for four strains: CS057, CS113, CS159, and CS227.

Extracts of the six strains of the first group and their corresponding mutants were dereplicated by LC-UV-LRMS and LC-HRMS against our in-house database (Pérez-Victoria et al., 2016) and the Chapman & Hall Dictionary of Natural Products (Buckingham, 2013) trying to determine if the differential compounds were already known. In five out of the six strains, compounds were identified in the differential peaks as: chromomycins A3, Ap and A2 (peaks 1, 2 and 4 in strain CS065a), benzanthrins A and B (both under the same peak 5 in strain CS090a), lobophorin A and B (peaks 6 and 8 in strains CS092 and CS123) and vicenistatin (peak 10 in strain CS147) (Figures 1A–E). In some cases, small minor peaks with characteristic spectrum of the family were also detected and were not further characterized (peak 3 from CS065a, 7 from CS092, and 7 and 9 from CS123). All these identified compounds contain 6DOH attached to their respective aglyca and therefore it can be assumed that the generated mutants were non-producer of the corresponding compounds.

The sixth strain (CS149) showing differential peaks that were not identified by dereplication was characterized in more detail. Two different mutants were obtained from this strain in which the glucose synthase and 4,6-dehydratase genes (mutant int149-1) or the 2,3-dehydratase gene were inactivated (mutant int149-2). Comparative analysis of the UPLC profiles of wild type strain CS149 and both mutants (Figure 1F) showed two differential peaks, 11 and 12, absent in the mutant strain. From these peaks, two compounds were isolated, purified and their structures elucidated. These compounds, designated sipanmycin A (11) and B (12), belong to a family of 24-membered macrolactams that include incednine (13) (Futamura et al., 2008) and silvalactam (14) (Schultz et al., 2012) (Figure 2).

The molecular formula of 11 was established as C₄₆H₇₁N₃O₈ based on the observed ion [M+Na]⁺ at 816.5143 (calcd. for C₄₆H₇₁N₃O₈Na+ = 816.5133, Δm = 1.2 ppm) (Supplementary Figure 1). Its UV (DAD) spectrum (Supplementary Figure 2) shows a main maximum at 320 nm with a secondary maximum (shoulder) at 360 nm, suggesting the presence of a polyene moiety. Complete NMR spectral data including ¹H, COSY, TOCSY, NOESY, HSQC, and HMBC spectra were acquired to determine the structure (Supplementary Figures 3–9). Although the ¹H spectrum showed broad signals, in the HSQC spectrum fourteen sp² methines could be identified which, according to the COSY and TOCSY spectra, would belong to two independent spin systems. This structural feature agrees with the conjugated polyene motif compatible with the UV (DAD) spectrum. On the other hand, two anomeric methine groups from two different spin systems were likewise deduced from the ¹H, COSY, TOCSY and HSQC spectra, suggesting the presence of two monosaccharide (aldose type) units. After searching the DPN NMR Features database (Pérez-Victoria et al., 2016), these structural features are found in incednine, a glycosylated 24-membered macrolactam natural product isolated from the culture broth of Streptomyces sp. ML694-90F3 (Schultz et al., 2012). Direct comparison of the NMR spectra of 11 with the corresponding data reported for incednine confirmed the structural similarity of both compounds. Thorough analysis of the 2D NMR spectra assisted by comparisons with the NMR data of incednine allowed stablishing the connectivity of 1 (Supplementary Figure 10). The 24-membered macrolactam ring of 11 is identical to that of incednine differing just in the side chain substituent at C-2, which for 11 is an isobutyl chain instead of a methoxy group. Interestingly, the same macrolactam ring aglycon is found in silvalactam (Schultz et al., 2012), another closely related antibiotic which bears just one monosaccharide. The pentopyranose (2-deoxy-2-amino-β-D-xylopyranose) directly attached to the aglycon of 11 is likewise almost identical to that found in incednine just lacking the N-methylation. However, the second monosaccharide unit (at the non-reducing end of the disaccharide moiety) shows important differences. Compared to the corresponding one in incednine it displays N-dimethylation rather than monomethylation and it is not a deoxy sugar at C-3″ carrying at this position both a hydroxyl and a methyl group. The relative configuration of the monosaccharides in 11 was determined based on coupling constants and NOEs analysis (Supplementary Figure 11). The absolute stereochemistry of the chiral centers in 11 was assumed to be identical to that reported for incednine (Futamura et al., 2008; Ohtani et al., 2010) based on the key NOEs observed for 11 which turned out to be identical to those reported for incednine. Molecular modeling of 11 based on the reported models for incednine and incednam (its aglycon) (Pérez-Victoria et al., 2016), perfectly match these key NOEs and confirms that the aglycon of 11 and incednam share the same absolute stereochemistry (Supplementary Figure 12). In a similar fashion, modeling also shows a perfect match with the key NOEs involving the monosaccharides confirming the D-configuration of both sugar units (Supplementary Figures 13, 14). The monosaccharide glycosylating the macrolactam must share the same biosynthetic pathway in incednine and 11 since they just differ in an N-methylation. On the other hand, the non-reducing monosaccharide in incednine is identical to D-forosamine carrying just N-monomethylation instead of N-dimethylation. It is reasonable to assume that the non-reducing monosaccharide in 11 is being biosynthesized from D-glucose via a route related to that of forosamine (Zhao et al., 2005; Hong et al., 2008).

The molecular formula of 12 was established as C₄₇H₇₃N₃O₈ based on the observed ion [M+H]⁺ at 808.5475 (calcd. for C₄₇H₇₄N₃O⁸⁺ = 808.5470, Δm = 0.6 ppm) (Supplementary Figure 15). This molecular formula contains one extra C atom and two extra H atoms compared to 11. The UV (DAD) spectrum (Supplementary Figure 16) was identical to that of 11. Likewise the complete NMR spectral set (Supplementary Figures 17–22) showed just minor differences compared to 11. The structural relationship of 11 and 12 was obvious and the difference was easily located, after analysis of the 2D NMR spectra, in the side chain at C-2 being in this case a 2-methyl-1-butyl group (Supplementary Figures 23, 24). The coincidence in chemical shifts of 11 and 12 (Supplementary Tables 1, 2) indicates they share the same absolute configuration. The absolute configuration of the extra chiral center in 2 (C-30) could not be determined though.

The four strains that did not show differential peaks by comparison of the UPLC metabolic profiles (strains CS057, CS113, CS159, and CS227) were further characterized in an attempt to identify the BGC (if any) where the mutant was generated and subsequently to identify the corresponding compound. With this aim, we used two different strategies trying to activate the expression of the corresponding silent clusters: a nutritional and a genetic approach.

In the nutritional approach four different media (SM, SM10, SM17, and SM19) differing in pH and in the carbon and nitrogen sources were used. The strains were grown in these media for several days and, after extraction with three different solvent systems, the UPLC profiles were compared between the wild type and the mutants. This kind of strategy has proved to be successful in a wide range of microorganisms (Bode et al., 2002). In the case of three strains (CS057, CS159, and CS227) there were no significant differences suggesting that none of the possible BGCs for secondary metabolites was activated by changing the nutritional conditions. Consequently, it was not possible to identify the compound coded by the mutated cluster. However, UPLC profiles comparison between strain CS113 and int113 mutant showed several differential peaks when they were grown on SM19 medium (Figure 3A). UPLC analysis and dereplication of the compounds present in five peaks against our in-house database (Pérez-Victoria et al., 2016), and the Chapman & Hall Dictionary of Natural Products Compounds (Buckingham, 2013) revealed that, based on the absorption spectrum and mass analysis, they contain compounds compatible with members of the cervimycins family (Figure 3B), a tetracycline-type compound with six 6DOHs attached. They were identified as antibiotic HKI 10311129 (15, m/z = 1113.4765 [M+H]⁺), antibiotic A2121-1 (16, m/z = 1199.4794 [M+H]⁺), cervimycin D (17, m/z = 1213.4939 [M+H]⁺), cervimycin C (18, m/z = 1227.5115 [M+H]⁺) and cervimycin A (19, m/z = 1226.5152 [M+H]⁺) (Herold et al., 2004, 2005).

In the second approach we applied two different genetic strategies that require the previous knowledge of the DNA sequence and genetic organization of the targeted cluster. For this approach, the genomes of the four strains were sequenced and searched for the presence of putative BGCs using the bioinformatic tool “antibiotics and Secondary Metabolite Analysis Shell” (antiSMASH v.4) (Blin et al., 2017). All strains showed the presence of a large number of BCGs (between 20 and 33) with only one glycosylated BGC in strains CS113 (cluster 23, Supplementary Table 3), CS159 (cluster 26, Supplementary Table 4) and CS227 (cluster 21, Supplementary Table 5), and two glycosylated BCGs in strain CS057 (clusters 14 and 19, Supplementary Table 6). Sequence analysis allowed us to confirm that the mutants in strains CS113, CS159, and CS227 were effectively generated in the corresponding glycosylated BGC and, in the case of CS057, at cluster 14. The genetic organization of these four clusters is shown in Figure 4. All four clusters seemed to contain at least one putative pathway-specific positive regulatory gene for the putative control of secondary metabolite biosynthesis. In order to attempt the activation of the expression of these clusters we used two approaches: (i) inserting a strong and constitutive promotor (ermE^∗p) in the chromosome to control the expression of the putative regulatory gene or (ii) ectopic expression of the regulatory gene under the control of ermE^∗p in the wild type strain either using a replicative (pEM4AT) or an integrative (pSET152e) vector.

This 38 kb BGC could code for a type II polyketide (Figure 4D and Supplementary Table 10). AntiSMASH analysis showed 68% of genes in this cluster presenting similarity with genes involved in the biosynthesis of the angucycline landomycin. There are four glycosyltransferase genes at this cluster that suggested the presence of a minimum of four 6DOH in the final compound. Cluster 14 also contains a possible transcriptional activator gene of the OmpR-family (wmc32). By using both genetic strategies mentioned above, we found five differential peaks (20–24) between the experimental strains and the corresponding controls, all of them sharing the same absorption spectrum (Figure 5A). Dereplication of the compounds in these peaks against our in-house database (Pérez-Victoria et al., 2016) and the Chapman & Hall Dictionary of Natural Products (Buckingham, 2013) showed that they could be related to the glycosylated angucycline warkmycin (Helaly et al., 2013). To confirm the novelty of these compounds, two of them (22, m/z = 1025.4341 [M+NH₄]⁺; and 24, m/z = 1049.4104 [M+H]⁺) were purified and their structures elucidated and named warkmycin CS1 and CS2, respectively (Figure 5B).

Compound 22 showed a UV (DAD) spectrum (Supplementary Figure 25) with a maximum at 220 nm identical to that reported for warkmycin. Its molecular formula was established as C₄₈H₆₅NO₂₂ based on the observed ion [M+NH4]⁺ at 1025.4341 (calcd. for C₄₈H₆₉N₂O₂₂⁺ = 1025.4336, Δm = 0.5 ppm) (Supplementary Figures 26, 27). The molecular formula matches that of 4-O-deacetyl-warkmycin. However, this known warkmycin could not explain the fragments observed in-source in the HRMS experiment. Thus, a complete set of NMR spectra including ¹H, COSY, TOCSY, NOESY, HSQC and HMBC spectra (Supplementary Figures 29–34) was acquired to ensure the novelty of 22 and determine its structure. The analysis of the NMR data confirmed 22 as a novel warkmycin displaying for most signals almost identical chemical shift compared with 4-O-deacetyl-warkmycin. The main differences were observed for the resonances of sugar rings A and D. Such difference turned out to be a consequence of the different position of the carbamoyl functional group which is located at position 4 of the sugar ring D in 22 while in 4-O-deacetyl-warkmycin is placed at position 4 of the sugar ring A. The key HMBC observed between the proton at position C-4D and the carbonyl carbon of the carbamoyl functionality unequivocally distinguishes 22 from its regioisomer (4-O-deacetyl-warkmycin) (Supplementary Figure 35). The mentioned in-source fragments observed in its HRMS spectrum likewise unequivocally located the carbamoyl functionality at sugar ring D (Supplementary Figure 28).

The UV spectrum of compound 24 (Supplementary Figure 36) was identical to 22. Its molecular formula was established as C₅₀H₆₇NO₂₃ based on the observed ion [M+NH4]⁺ at 1067.4450 (calcd. for C₅₀H₇₁N₂O₂₃⁺ = 1067.4442, Δm = 0.9 ppm) (Supplementary Figure 37). The difference in its molecular formula compared with 22 immediately indicated that 24 bears an extra acetyl group. Once again 24 shares the same molecular formula as warkmycin but the analysis of the NMR spectral data (Supplementary Figures 38–45) and the in-source fragmentation observed in its HRMS spectrum unequivocally located the position of the carbamoyl in the same place found in 22. Thus 24 is a regioisomer of warkmycin and is identical to 22 carrying an extra acyl group at C-4. The effect of the extra acylation is reflected in the corresponding shifting in resonance frequency for the nearby positions to C-4 (Supplementary Table 11).

The absolute stereochemistry of 22 and 24 is identical to that reported for the known warkmycins based on chemical shift comparisons and the observation of the same coupling constants pattern and key NOEs as those reported for warkmycin (Supplementary Figures 46, 47).