Bacterial/yeast strains, and plasmids used in the present study are listed in Supplementary Table S1. The composition of all culture media used in the present study is listed in Supplementary Table S2.

For standard experiments Streptomyces sp. BRA-346 was grown in 25 ml of A1 liquid medium with 1 ml of 20% (v/v) mycelial stock at 28°C for 48 h with shaking. Then, 1 ml of the seed culture was inoculated into 500 ml flasks (equipped with helical springs) with 100 ml of A1 liquid medium. Cultures were grown for 7 days at 28°C and 220 rpm. To test different media, BRA-346 was grown in A1, TSB, TSBY, or ISP2 media under the same conditions.

Testing of chemical elicitors was performed as described previously (de Felício et al., 2021). Briefly, BRA-346 was inoculated in 10 ml of A1 medium at 28°C, 200 rpm for 48 h. One-hundred microliter of culture was plated onto SFM agar media and incubated at 30°C for 72 h. The chemical elicitors were added to the solid media plates using an automated pipette to dispense 5 μl drops of each chemical elicitor onto the bacteria layer. The chemical elicitors used were: ampicillin 100 or 10 μg/ml, chloramphenicol 300 or 30 μg/ml (prepared in ethanol), kanamycin 100 or 10 μg/ml, streptomycin sulfate 50 or 5 μg/ml, sodium butyrate 50 or 5 μM, procaine 100 or 10 μM, DMSO 100 or 2% or EDTA 10 or 1 mM. Water and ethanol were used as controls. Plates were incubated at 30°C for 72 h. Alterations on colony morphology were visually inspected and recorded, following the described in de Felício et al. (2021), and used to select the chemical elicitors for liquid culture. The liquid cultures for the elicitation tests were performed as described above, with the addition of the chemical elicitors 72 h after BRA-346 inoculum.

The BRA-346 genomic DNA was extracted using the Wizard® Genomic DNA Purification Kit (Promega Corporation, Fitchburg, United States). Before sequencing, the quality of the extraction was verified by agarose gel electrophoresis. The genomic DNA was sequenced by the Macrogen laboratory (South Korea, Seoul), using the HiSeq System platform (Illumina Inc., San Diego, United States). Paired-end reads were quality checked using FastQC and processed with Trimmomatic (Bolger et al., 2014) to remove low quality reads and adapters sequences.¹ QC reads were de novo assembled using the A5-miseq pipeline (Coil et al., 2015). Genome quality and completeness were evaluated using both Quast and CheckM (Gurevich et al., 2013; Parks et al., 2015). Gene prediction and annotation were performed with Prokka (Seemann, 2014) and BGCs were further predicted using the antiSMASH web server version 6.0.1 (Blin et al., 2021). The draft genome of Streptomyces sp. BRA-346 is available at GenBank under the BioProject accession number PRJNA765318.

The epn/tmc BGC was cloned by PCR-based transformation-associated recombination (TAR) method, as described previously (Zhang et al., 2019; Sigrist et al., 2020), with minor modifications as detailed in the Supplementary Material.

Streptomyces coelicolor M1146 (host) and M1146-epn/tmc (heterologous organism) were cultivated in A1 medium for 72 h at 28°C and 220 rpm (100 ml in 500 ml flasks equipped with helical springs). M1146 and M1146-epn/tmc cultures were supplemented with 25 μg/μl nalidixic acid, and M1146-epn/tmc was also supplemented with 50 μg/ml kanamycin.

The cultures of Streptomyces sp. BRA-346 (100 ml), S. coelicolor M1146 (100 ml), and M1146-epn/tmc (100 ml), along with the culture medium A1 (control), were extracted by liquid–liquid partition with ethyl acetate. The biomass was filtered with glass filters (22 μm) after adding the solvent and before phase separation. The extraction was repeated twice by adding 40 ml of ethyl acetate to 100 ml of BRA-346 cultures. The organic phase was collected and dried under reduced pressure at 35°C. The crude extracts were submitted to solid-phase separation using C8 cartridge (Discovery® DSC-8 SPE Tube, Supelco) using H₂O:MeOH gradient, yielding three fractions: 5% (F5), 50% (F50), and 100% (F100) of methanol. Crude extract and fraction samples were resuspended in DMSO to a final concentration of 10 mg/ml. All samples were analyzed by UPLC–MS/MS.

LC-MS/MS data collection was performed as described previously (de Felício et al., 2021). For UPLC–MS/MS chemical profile, 2 μl aliquots of the samples (extracts and fractions) were injected on a BEH C18 reversed-phase column (1.7 μm, 2.1 × 100 mm) attached to a compatible pre-column, using an Acquity H-Class UPLC Waters (Waters, Milford, MA, United States) system coupled to a Bruker Impact II UHR-ESI-QqTOF mass spectrometer (Bruker Daltonics, Billerica, MA, United States). A solvent system of water (A), acetonitrile (B) 2% formic acid (C) methanol (D), was used to compose the following analytical method: 0.5 ml/min flow rate; column temperature 40°C; 0–1 min, 5% B; 1–6 min, 5% B to 35% B (curve 6); 6–10 min, 35% B to 95% B (curve 1); 10–12 min, 95% B; 12–15 min, 95% D (curve 1); 15–18 min for column equilibration on initial phase. C was kept constant, at 5% (final concentration of 0.1%). The mass spectrometer worked in positive ion mode scanning mass in the 30–2,000 Da range, with acquisition rate of 8 Hz. End plate offset = 500 Volts (V); V_cap 4,500 V; nebulizer 4.0 bar; drying gas (N₂) flow 10 L/min; drying gas temperature 200°C, followed by an MS/MS scan for the most intense ions in a cycle time of 1 s, absolute threshold (per 1,000 sum.) of 1,500 cts. As MS² rules, mass ratio (m/z) below 200 Da were excluded, and the “active exclusion” function was enabled. Each run was automatically calibrated using HCOONa (10 mM). Calibrated spectra were converted to mzXML files through Data Analysis 4.3 and included bio tools CompassXport (Bruker Daltonics, version 4.0.0.8).

Data processing and analysis were done with an in-house collection of scripts called NP³ Mass Spectrometry (MS) Workflow, as previously described (de Felício et al., 2021) with some parameter modifications to adjust sample properties. The beta version of the NP³ MS workflow is freely available in our repository (accessed on 13 Aug 2021) with a detailed manual and a command line interface.² The Ionization Variant Annotation Molecular Networking (IVAMN) was used as the base to select and exclude nodes present in chromatographic blanks (BLANKS) samples and their first immediate neighbors. Then, the remained nodes marked as “protonated representative” (column “protonated_representative” equals 1 at the nodes count table) were selected and filtered at the Spectra Similarity Molecular Network (SSMN). The raw SSMN was created with a minimum cosine similarity score for an edge connection of 0.6 and filtered with a maximum component size of 200 and a maximum number of connections for one single node of 15. IVAMN and SSMN construction at Cytoscape (Shannon et al., 2003) were done with an in-house python script that automates some tasks using the py2cytoscape library (accessed on 13 Aug 2021).³ After these automated steps, some other undesirable nodes like the ones marked as culture media (BED) and the single nodes were manually removed leaving a clean network to process and analyze. MS/MS chemical structure annotation was performed using spectra matches to the GNPS (Wang et al., 2016b) and the UNPD-ISDB (Allard et al., 2016) databases. Annotations were manually revised and the related MS/MS clusters were grouped and named by the chemical groups found in the SSMN.

Sequences were collected from antiSMASH ClusterBlast and KnownClusterBlast (Blin et al., 2021), from the antiSMASH-database (ClusterBlast Hit and KnownClusterBlast Hit for experimentally validated epoxyketone peptides BGCs; Blin et al., 2017), from the BiG-FAM database (querying the 29,955 Biosynthetic Gene Families, previously calculated using BiG-SLICE from a collection of more than 1.2 million Biosynthetic Gene Cluster; Kautsar et al., 2021a,b), and from MultiGeneBlast (Medema et al., 2013) run on the bacteria GenBank database from 2013 and a custom Streptomyces database from GenBank (2,526 NCBI GenBank Assemblies), using BRA-346 epn/tmc, Streptomyces hygroscopicus eponemycin and S. chromofuscus TMC-86A BGCs as probes. The best hit sequences, indicated by their closest relationship with BRA-346 epn/tmc BGC (low distance and high cumulative BLAST hit score), all carrying the acyl-CoA dehydrogenase gene, were aligned with BRA-346 epn/tmc BGC at nucleotide level with the multiple sequence alignment MAFFT v11 program (Katoh and Standley, 2013) using Geneious Prime (v.2021.2.2). Regions of poor alignment were inspected, and sequence refinement was done with Trimal software (Capella-Gutierrez et al., 2009; Method: Automated1). Construction of maximum likelihood trees from the alignment was conducted using the Molecular Evolutionary Genetics Analysis (MEGA) v11 software (Tamura et al., 2021) after statistical selection of the best-fit model of evolution with jModelTest2 tool (Darriba et al., 2012).

BRA-346 is a slow growing Actinobacteria, which achieves the stationary phase after at least 3 days of laboratory growth in highly aerated liquid cultures. Secondary metabolite production is expected by day 7, with epoxyketone peptides production by BRA-346 being culture and medium-dependent (Figure 2).

To exemplify this statement, four replicates of BRA-346 cultures in the A1 medium, the standard growth medium for BRA-346 (Furtado et al., 2021), were carried out (A1 C3a-c and A1 C5), in addition to the use of three different culture media (ISP2, TSBY, and TSB; Supplementary Table S2), in a total of seven culture conditions evaluated in this data set. The metabolome was assessed by LC–MS/MS analyses, being further evaluated by the construction and inspection of a spectra similarity molecular network (SSMN) of the representative [M + H]⁺ ions (de Felício et al., 2021; Figure 2A).

Notably, BRA-346 metabolome is shifted towards different classes of secondary metabolites, depending on the culture media used (Figures 2A,B). Specifically, the production of the target epoxyketone peptides was upregulated in the cultures using the A1 medium and downregulated when the other culture media were applied (Figures 2A,C). The downregulation of epoxyketone peptide production is concomitant with the overexpression of lipopeptides and macrolactams in the ISP2 and TSBY culture media (Figure 2A). In addition, the A1 C5 culture also presented a reduced diversity of epoxyketone peptides (Figures 2D,E), which was in turn aligned with an increase detection of lipopeptides, macrolides, and carotenoids (Figure 2A).

The epoxyketone peptide cluster containing 1 and 3 is highlighted in Figure 2D, spotting DHE (1, m/z 401.26) and DH-TMC (3, m/z 345.20). Two ions of m/z 401.26, which were clustered together in the SSMN, presented different retention times in the reverse-phase liquid chromatography (6.6 min and 7.4 min; Figure 2E). Their fragmentation spectra (Supplementary Figure S5) are similar (cosine = 0.88), suggesting that these ions represent stereoisomers. It is also important to comment that the bioactive epoxyketone peptide eponemycin (2, m/z 399.25), which was previously identified in BRA-346 cultures in low yields (Furtado et al., 2021), was not detected in the crude extracts of BRA-346 cultures presented here (Supplementary Figures S6, S7). In addition, TMC-86A (4, m/z 343.19), was identified in trace amounts (Supplementary Figure S8). Knowing that 1 and 3 are biosynthetic intermediates of 2 (Schorn et al., 2014) and 4 (Zabala et al., 2016), respectively, it is possible that the last step of 2 and 4 biosynthesis in wild type BRA-346 is not being completed under the growth conditions used.

In a further attempt to elicit secondary metabolite production by BRA-346, especially epoxyketone peptides, we used chemical elicitors in A1 cultures. This was done following precedents in the literature that some chemicals, as sodium butyrate (reviewed by Tomm et al., 2019), can act as epigenetic modulators, in addition to mimetics of the presence of competing organisms, as antibiotics produced by other micro-organisms (Abdelmohsen et al., 2015; de Felício et al., 2021). Following the approach described by de Felício et al. (2021) we tested several chemical elicitors in solid media, and selected ampicillin (100 μg/ml), sodium butyrate (50 μM), and procaine (100 μM) for eliciting BRA-346 secondary metabolites in A1 liquid medium. The metabolites were extracted, and the crude extracts were analysed by LC–MS/MS and SSMN (Supplementary Figure S9). The experiments (C5) were carried out in parallel, in the same incubator, and following the same procedures. In addition, the crude extracts were purified using C-8 cartridges and the water/methanol fractions (F50) were further analyzed (Figure 3).

It is worth noticing that 1 was produced in higher amounts in the ampicillin elicited cultures (Figure 3D), and TMC-86A (4), which had not been previously detected or sporadically detected in trace amounts (Figure 3C), was detected in the crude extracts and enriched in the F50 fraction of the ampicillin growth condition (Figure 3B). These results indicated that BRA-346 could produce 4, one of the expected final products of epoxyketone biosynthesis in Streptomyces bacteria. However, its production might be tightly regulated in BRA-346 and can be upregulated by an antibiotic chemical elicitor.

Analysing the groups of metabolites modulated by the chemical elicitors (Supplementary Figure S9), ampicillin increased epoxyketone and downregulated macrolides, terpenes, and diketopiperazines/alkaloids. Procaine, on the other hand, exerted an opposite effect, downregulating epoxyketone peptides and upregulating the three other major groups (macrolides, terpenes, and diketopiperazines/alkaloids). Macrolactams were not observed and just a small group of lipopeptides was observed.

Our efforts to understand epoxyketone biosynthesis and regulation by Streptomyces BRA-346 were further assisted by genome sequencing and mining, going beyond this class of secondary metabolites. BRA-346 draft genome contains 10 Mb and was presented on 180 contigs with 8,706 predicted coding genes.

BRA-346 genome mining and annotation using antiSMASH (Blin et al., 2021) show the presence of 66 putative biosynthetic gene clusters (BGCs). Although this is a draft genome, and some clusters are fragmented or duplicated, it was possible to find 35 complete BGCs encoding for ribosomally synthesized and post-translationally modified peptides (RiPPs), epoxyketone peptides, terpenes, ectoines, siderophores, non-ribosomal peptides (NRP) or polyketides (PK; Supplementary Table S6). In addition, 31 truncated BGCs encoding for PK or NRP were also annotated. This reflects a broad biosynthetic capacity of BRA-346, which is mostly hidden under laboratory conditions.

By applying the OSMAC (one strain many compounds) approach, using specific growth conditions (e.g., culture media—Figure 2A; Supplementary Table S4; and chemical elicitors—Figure 3; Supplementary Figure S9), we were able to reveal the different classes of secondary metabolites produced by this bacterium. Some of the BGCs annotated in BRA-346 draft genome could be linked to the metabolites presented in the SSMN (Figure 2A), providing additional insights into BRA-346 regulation and metabolic shifts.

In the regulatory perspective, the draft genome of Streptomyces BRA-346 shows a myriad of transcription regulators, which may be involved in the regulation of the specialized metabolism of BRA-346 under laboratory conditions. We highlight 70 LuxR-type transcription regulators, 118 TetR-type regulators, 2 gamma-butyrolactone (GBL) BGCs (region 35.1, candidate BGC = 11 kb; and region 2.1, candidate BGC = 11 kb, Supplementary Table S6; Supplementary Figure S10), which might be of particular importance for understanding BRA-346 regulation and, potentially quorum sensing.

For lipopeptides, which were evidenced in BRA-346 cultures using TSBY/ISP2 media (Figure 2A), a truncated NRPS BGC was found in contig 32 harboring the assembly line characteristic of lipopeptides, thus linking the metabolites detected by LC–MS/MS to BRA-346 genome. This BGC (region 32.1, candidate BGC 35 kb; Supplementary Table S6) contains a N-terminal acyl ligase domain, which is associated with lipopeptide biosynthesis (Kersten et al., 2011), in addition to NRPS modules. It further shares sequence similarity with Streptomyces RTd22 (Chagas et al., 2016) and ATCC53653 BGCs, the latter reported previously as a producer of stendomycin lipopeptides—Supplementary Figures S11, S12; Supplementary Table S5 (Kersten et al., 2011; Villadsen et al., 2018). A complex regulatory system is found within this BGC, which harbours TetR, LysR, sensor histidine kinase, SARP, and LuxR regulators.

Macrolactams, exemplified by heronamide B (with related ORFs present in two truncated regions: region 2.1, candidate BGC = 11 kb and region 63.1, candidate BGC = 27 kb; Supplementary Table S6) could also be linked to the BRA-346 metabolome (Figure 2A). This class of natural products was also only observed in the TSBY/ISP2 cultures. This BGC also presents a complex regulatory system with TetR, LuxR, LacI, AraC, SARP, and an AsfA butyrolactone producing gene.

Diketopiperazines, related to bicyclomycin (region 3.2, candidate BGC 43 kb), and carotenoids, related to isorenieratene (region 34.1, candidate BGC 21 kb), annotated at the SSMN (Figure 2A), were additionally located in BRA-346 draft genome (Supplementary Table S6). Bicyclomycin could be linked to an alkaloid/diketopiperazine SSMN cluster expressed in the ISP2 medium; whereas carotenoids were mostly observed in the TSBY and A1C5 cultures (Figure 2A) and in A1 cultures induced with procaine (Supplementary Figure S9). Curiously, these differentially expressed BGCs are also under the control of TetR (Table 1), a family of transcription factors, generally repressors, regulated by gamma-butyrolactones (GBL; Cuthbertson and Nodwell, 2013).

For the epoxyketone peptides, mostly observed when BRA-346 was grown in the A1 medium (cultures C3 a–c, Figure 2) or upregulated in the presence of ampicillin (Figure 3; Supplementary Figure S9), a single complete BGC was found (BGC number 42.1, Supplementary Table S6). The BGC, herein named epn/tmc, is predicted to contain 25 open reading frames (ORFs). From these, 12 ORFs are related to previously reported eponemycin, epn (Schorn et al., 2014) and TMC-86A, tmc (Zabala et al., 2016) epoxyketone gene clusters (Figure 4; Supplementary Table S7). This BGC has a simpler regulatory system, compared to the BGCs presented above, with only one LuxR-type transcription factor (epn/tmcA) and TTA codons.

Although eponemycin (2) and TMC-86A (4) BGCs in Streptomyces bacteria have been already described in the literature, these BGCs were individually reported as producers of eponemycin or TMC-86A natural products. BRA-346 was shown to produce both 1 and 3, the last step biosynthetic intermediates of 2 and 4, suggesting multiple epoxyketone BGCs or a multifunctional BGC. In addition, we found the production of these specialized metabolites is tightly regulated in wild type BRA-346 and influenced by its metabolic shifts, limiting their full obtention for additional studies.

We then cloned the epn/tmc BGC of BRA-346 using the transformation-associated recombination (TAR) approach (Zhang et al., 2019; Sigrist et al., 2020), inserting the target BGC under the control of a strong promoter into a S. coelicolor M1146 “antibiotic null” host organism (Gomez-Escribano and Bibb, 2011). The heterologous organism S. coelicolor M1446-epn/tmc was cultured in A1 medium and its metabolites were analyzed in parallel with the wild-type BRA-346 and the host organism Streptomyces coelicolor M1446 (Figure 5A; Supplementary Figure S12). M1146-epn/tmc reaches the stationary phase in less than 3 days, with full production of the epoxyketone peptides on day 3, accelerating laboratory cultivation and chemical sample preparation.

The high yields obtained for both final products (2 and 4) in the heterologous organism containing the BRA-346 epn/tmc BGC (M1146-epn/tmc, Figure 5; Supplementary Figure S13) are worth noticing, confirming that the epn/tmc BGC of BRA-346 is a multifunctional BGC capable of producing both TMC-86A and EPN. It was further possible to detect several [M + H]⁺ MS/MS clusters related to epoxyketone peptides (grouped in Figure 5A), suggesting a collection of epoxyketone peptides that can be formed from BRA-346 epn/tmc BGC expression.

In comparison to the wild-type BRA-346, the heterologous organism M1146-epn/tmc showed higher amounts of the biosynthetic final products 2 and 4 and lower amounts of the dihydro leucine intermediates 1 and 3 (Figure 5; Supplementary Figure S13). Although wild-type BRA-346 was reported to produce 2 (Furtado et al., 2021) it did not produce this compound in our experiments (Figure 3A; Supplementary Figures S6, S7), and produced 4 in low yields (Supplementary Figure S8), even when induced by ampicillin (Figure 3C). Together, these results may point to regulation of the last step of 2 and 4 biosynthesis in wild type BRA-346, which was overcome by heterologous expression of the epn/tmc BGC in S. coelicolor M1146.

Wild type BRA-346 also presented the production of other antibiotics, what in turn can limit the production of epoxyketone peptides by this highly diverse antibiotic producer (Figure 5A in pink). On the other hand, heterologous expression of the epn/tmc BGC in an “antibiotic null” strain resulted in epoxyketone peptide biosynthetic final products overproduction. This suggests that the availability of biosynthetic precursors and/or crosstalk of regulatory mechanisms are controlling epoxyketone peptide production by wild type BRA-346.

Aiming to gain further insights on the occurrence and phylogenetic relationship of the BRA-346 epn/tmc BGC we additionally searched for related BGCs in public databases and constructed a phylogenetic tree for this BGC (Figure 6). Aquimarina sp. putative BGC was assigned as the outgroup of the tree, since it was the most distant BGC to BRA346 epn/tmc, while sharing sequence similarities. This analysis showed that the BRA-346 epn/tmc BGC is closely related to other Streptomyces BGCs linked to EPN and TMC specialized metabolites, as expected.

It is worth noticing that from all the genomes and BGC sequences evaluated (see “Materials and Methods” section for details) only 16 share sequence similarity and minimal BGC characteristics related to BRA-346 epn/tmc BGC. Even within 50 Streptomyces spp. closely related to BRA-346 (vide phylogenetic analysis of Streptomyces sp. BRA-346; Supplementary Figure S15) only two related BGCs could be found. Most of the organisms containing BGCs related to BRA-346 epn/tmc are of the phylum Actinobacteria, covering two orders: Streptomycetales and Pseudonocardiales. Interesting, Coralloccocus sp., and Chondromyces sp., both Proteobacteria in the order Myxococcales, and Aquimarina sp., a Bacteroidetes in the order Flavobacteriales, were also pointed as presenting a BGC related to the BRA-346 epn/tmc (Figure 6). This indicates the occurrence of putative epoxyketone peptide BGCs also in Gram negative bacteria. Together, these results indicated that this BGC is rare in nature, not frequent observed in closely related Streptomyces strains, and not very widespread across Bacteria, being so far restricted to three Phyla (Actinobacteria, Bacteriodetes, and Proteobacteria). The scarce presence of epn/tmc related BGCs in the BRA-346 species tree (Supplementary Figure S15), in addition to the finding that this BGC was found in unrelated Actinobacteria and Proteobacteria members (Figure 6), point to horizontal gene transfer of this BGC, which is a common feature of Actinobacteria BGCs linked to specialized metabolites (Ziemert et al., 2014).

Regarding the environment from which these bacteria or DNA samples were isolated, Streptomyces BRA-346 (marine endemic tunicate Euherdmania sp., Brazil), Streptomyces luteovercillatus CGMCC 15060 (ocean sediment, China) and Aquimarina sp. AU119 (Marine sponge Tedania sp., Australia) were isolated from the marine environment, the others being obtained from soil (most organisms, including eponemycin, TMC-86A, epoxomicin, clarepoxcin, and landepoxcin producers), river sediment (one organism), and decayed wood (one organism) samples.

Biosynthetically, the NRPS responsible for the biosynthesis of the peptide core of epoxyketone peptides present up to four condensation domains (C-domains; Supplementary Figure S15). Clades with BGCs sharing C-domain pattern similarity are indicated, as the NRPs encoding the tetrapeptides epoxomicin and clarepoxcin. One clade harbours BGCs with shorter NRPSs, with two C-domains, as Epn/TmcG and the NRPS from BRA-346 epn/tmc BGC, which are responsible for the production of the dipeptides eponemycin and TMC-86A. From these analyses it is possible to suggest that this BGC could suffer module duplication or loss in its evolutionary history.

The Epn/TmcG multi-modular enzymes, which add leucine and serine amino acids to a fatty acid precursor in EPN and TMC biosynthesis (Liu et al., 2015; Zabala et al., 2016), share higher sequence identity between Streptomyces BRA-346 Epn/TmcG and S. hygroscopicus EpnG (91% ID) than between BRA-346 Epn/TmcG and S. chromofuscus TmcG (76% ID). In BRA-346, Epn/TmcG might promiscuously accept both longer (EPN) and shorter (TMC) fatty acid precursors, therefore being able to synthesize both specialized metabolites. This potentially indicates an evolutionary path for this enzyme on the flexibility for TMC or EPN biosynthesis, a feature that has been involved in the “evolutionary dynamics of specialized metabolites biosynthesis” (Chevrette et al., 2020).

A common characteristic of the retrieved BGCs is the presence of a ACAD enzyme (ORF coloured yellow in Figure 6). The ACAD is related to Epn/TmcF in BRA-346 epn-tmc BGC and is linked to the biosynthesis of the epoxyketone pharmacophore in epoxyketone peptide proteasome inhibitors as EPN (Zettler et al., 2016) and TMC (Zabala et al., 2016). Except for the ACAD from Chondromyces apiculatus (NZ_ASRX01000038), the other ACAD enzymes showed at least 65 sequence similarity (Supplementary Table S8) reflecting a similar biosynthetic role.

From the 17 BGCs analyzed, nine also present an ORF encoding for a beta-subunit of the proteasome (related to Epn/TmcC; Figure 6B in cyan). This suggests the presence of a self-resistant gene in these BGCs and, possibly, the biological target for the resulting biosynthetic product (D’Costa et al., 2011; Kale et al., 2011; Ziemert et al., 2016; Alanjary et al., 2017). Indeed, six of the retrieved BGCs have a specialized metabolite assigned, which is proteasome inhibitors: eponemycin, TMC-86A, landepoxcin A, epoxomicin, and clarepoxcin C (Figure 6C). From these, the epoxomicin BGC was the only one that did not show a proteasome beta-subunit within the BGC.

Interesting, orphan BGCs from three other Streptomyces species (Streptomyces sp. NBS 14/10, S. bingchenggensis BCW-1, and S. sparsogenes ATCC 25498) and the BGC from Actinocrispum wychmicini DSM 45934 also carry a beta-subunit of the proteasome, showing high level of synteny to the epn/tmc BGCs. The first three orphan BGCs are further clustered with the epoxomicin and clarepoxcin C BGCs in the phylogenetic tree (Figure 6). This indicates that these Actinobacteria BGCs also produce epoxyketone peptides, potentially acting as proteasome inhibitors, thus deserving future research.

Regarding BGC regulation, all BGCs retrieved show regulatory genes flanking the BGC, which belong to a diverse set of transcriptional regulators (Figure 6). The three epn/tmc BGCs show regulatory genes upstream the BGC, belonging to the LuxR-type of transcription factors with HTH DNA-binding domains (Epn/TmcA). These DNA binding domains are conserved among the three Epn/TmcA sequences analysed (Supplementary Figure S17). BRA-346 Epn/TmcA is the longer sequence with 190 amino acids, displaying an additional PAS-fold regulatory domain at the N-terminal region of the coded protein. A deep inspection of the epn BGCs also reveals a PAS domain 275–20 bases upstream the epnA ORF. The same analyses could not be performed to the tmc BGC as its sequence is truncated just upstream the tmcA ORF in the public databases consulted. It can be suggested, at least for the BRA-346 epn/tmc BGC, that the PAS domain is an additional sensor for tight regulation of this BGC transcription. The BRA-346 Epn/TmcA PAS domain is predicted to bind at least one ligand, being a heme or flavin ligands (Supplementary Figure S18), which are related to light or redox regulation, respectively. PAS domains can also bind quorum sensing molecules (Pappas et al., 2004), a property that cannot be ruled out to BRA-346 Epn/TmcA at this point.

Although little is known about LuxR regulation in Gram-positive bacteria (Santos et al., 2012; Polkade et al., 2016; Rajput and Kumar, 2017), there are precedents in the literature that salinity increases the production of LuxR transcription factor activator molecule, acyl homoserine lactone (AHL), which in turn is produced by LuxI in Gram-negative bacteria (Bjelland et al., 2012; Sivakumar et al., 2019). Phylogenetic analysis of Gram-positive bacteria, however, point that the LuxI/LuxR system is absent in this group, although an analogous signalling molecule, gamma-butyrolactone (GBL), is found in Gram-positive bacteria (Santos et al., 2012; Rajput and Kumar, 2017). Indeed, we could not find a luxI related gene in BRA-346 genome, but we found two butyrolactone BGCs in BRA-346 genome (region 35.1, candidate BGC = 11 kb; and region 2.1, candidate BGC = 11 kb; Supplementary Figure S10). As mentioned above, these GBL BGCs are most certainly related to lipopeptides and macrolactam BGC activation in BRA-346, which in turn shuts-off the production of epoxyketone peptides by this bacterium. It is not clear however if these GBLs, or associated regulatory proteins, have a direct role on the downregulation of the epn/tmc BGC or if its downregulation is simply a consequence of biosynthetic precursor availability, once it competes with the GBL-regulated BRA-346 BGCs for amino and fatty acids.

In addition to LuxR transcriptional regulation, the epn/tmcA gene displays three TTA codons, further indicating regulation of this BGC at the translational level. The TTA codon is a rare codon in GC rich genomes, as those of Actinobacteria, which is linked to secondary metabolism regulation (Takano et al., 2003; Sen et al., 2012). Apart from Aquamarina sp. AU119 (33%), the other BGCs analysed here show high GC content (>67%) and TTA regulatory codons were frequently found within the epoxyketone peptide related BGCs, some of them also within the ORF encoding a transcription factor. Considering that Epn/TmcA is a positive regulator of the epn/tmc BGC, the TTA codon on its regulator is also a point of competition during translation. In summary, the analyses of the 16 BGCs related to BRA epn/tmc BGC suggest this BGC is tightly regulated, at both the transcription and translational levels, what can further explain the low reproducibility of wild type BRA-346 cultures in producing epoxyketone peptides.