The region located upstream of the DX gene (Supplementary Figure S1C) and the DX gene (codon optimized for expression in Streptomyces, codon adaptation value = 1.0) (Wright and Bibb, 1992) (Supplementary Figure S1B) were synthetized as a 395 nt long fragment and cloned as a PstI-HindIII fragment into the Escherichia coli-Streptomyces shuttle vector pOSV10 (Supplementary Figure S1A) (Juguet et al., 2009). The resulting plasmid pOSV10/DX and pOSV10 were transformed into the E. coli strain S17-1 carrying an integrated trans-acting RP4 transfer function (Simon et al., 1983), then the plasmids were transferred into Streptomyces albogriseolus/viridodiastaticus as described in Mazodier et al. (1989), yielding A37 and A36, respectively. All constructs were sequence verified.

Genomic DNA of the strains was prepared as described in Kieser et al. (2000) and PCR was carried out as described in Shikura et al. (2021) using the primers mentionned pOSV10 F1 and R1 located from either site of the PstI and HindIII cloning sites of pOSV10 (Table 2).

Three independent biological RNA samples were prepared as described in Kieser et al. (2000) from the strains A36 and A37 grown for 40 h at 28°C on solid R2YE medium limited in Pi (1 mM final, no K₂HPO₄ added) and RT-qPCR experiments were carried out as described in Lejeune et al. (Lejeune et al., 2021a) using the forward and direct primers internal to the reference genes as well as to dx, phoR and phoP (Table 1). phoR (sco4229) and phoP (sco4230) of S. coelicolor were used to identify by BLAST (Johnson et al., 2008) phoR and phoP of Streptomyces albogriseolus/viridodiastaticus whose genome sequence was deposited in NCBI Gen bank (SUB12396509 waiting for validation).

10³ (Figures 1–8A) or 10⁵ (Figures 8B and 9) spores of each strains were plated on the surface of a cellophane disk laid onto plates of solid R2YE medium (Kieser et al., 2000) limited (1 mM free phosphate coming from the constituents of the growth medium, no K₂HPO₄ added) (Figures 1–8A) or proficient in phosphate (4 mM K₂HPO₄ added) (Figure 9) and incubated for various times to establish a growth curve (Figure 2), for 40 h (Figures 1 and 3) or for 72 h (Figures 4–9) at 28°C. The mycelium of 20 agar plates scraped of the cellophane disks with a spatula and the agar of the plates cut into small pieces were transferred into two individual Erlenmeyer flasks containing 2 L of ethyl acetate. The flasks were placed on a shaking table for 24 h at 28°C then the solvent was filtered. The organic phase was washed three times with Milli-Q water using a separating funnel and then concentrated in vacuo to obtain the four following crude extracts (Biomass + Gelose): m_A36 (- Pi) = 17.0 + 10.1 = 27.1 mg, m_A36 (+ Pi) = 5.8 + 4.8 = 10.6 mg and m_A37 (- Pi) = 9.3 + 14.1 = 23.4 mg, m_A37 (+ Pi) = 3.2 + 4.8 = 8.0 mg.

Mycelium of each strain originating from 4 different cultures grown on plates of solid R2YE with no K₂HPO₄ added (situation of phosphate limitation, 1 mM free phosphate coming from the components of the growth medium) for 72 h at 28°C were collected and immediately frozen in liquid nitrogen and lyophilized.

Extraction of the metabolites was carried out from 5 mg (DW) of the ground lyophilized samples in 2 mL Safelock Eppendorf tubes. The samples were re-suspended in 1 mL of frozen (−20°C) Water:Acetonitrile:Isopropanol (2:3:3) containing Ribitol at 4 µg/mL and extracted for 10 min at 4°C with shaking at 1500 rpm in an Eppendorf Thermomixer. Insoluble material was removed by centrifugation at 13,500 rpm for 10 min 700 µL were collected and 70 μL of myristic acid d27 at 0.3 mg/mL was added as an internal standard for retention time locking. After centrifugation, aliquots of each extract (100 µL) were dried for 4 h at 30°C in a Speed-Vac and stored at −80°C. Two Blank tubes underwent the same steps as the samples.

Subsequently all steps of GC-MS analyses were carried out as described in Fiehn et al. (Fiehn, 2006; Fiehn et al., 2008). Samples were taken out of −80°C, warmed 15 min before opening and speed-vac dried again for 1 h at 35°C before adding 10 µL of 20 mg/mL methoxyamine in pyridine to the samples and the reaction was performed for 90 min at 30°C under continuous shaking in an Eppendorf thermomixer. 90 µL of N-methyl-N-trimethylsilyl-trifluoroacetamide (MSTFA) (REGIS TECHNOLOGIES 1–270590-200 -10 × 1 g) were then added and the reaction continued for 30 min at 37°C. After cooling, 90 µL were transferred to an Agilent vial for injection. 4 h after derivatization 1 µL of sample were injected in splitless mode on an Agilent 7890B gas chromatograph coupled to an Agilent 5977A mass spectrometer. The column was a ZORBAX DB5-ms from Agilent (30 m × 0.25 mm × 0.25 µm with a 10 m guard). An injection in split mode with a ratio of 1:30 was systematically performed for saturated compounds quantification. Oven temperature ramp was 60°C for 1 min then 10 °C/min to 325°C for 10 min. Helium constant flow was 1.1 mL/min. Temperatures were the following: injector: 250°C, transfer line: 290°C, source: 230°C and quadripole 150°C. The quadrupole mass spectrometer was switched on after a 5.90 min solvent delay time, scanning from 50–600 u. Absolute retention times were locked to the internal standard d27-myristic acid using the RTL system provided in Agilent’s Masshunter software. Retention time locking reduces run-to-run retention time variation. Samples were randomized. A fatty acid methyl esters mix (C₈, C₉, C₁₀, C₁₂, C₁₄, C₁₆, C₁₈, C₂₀, C₂₂, C₂₄, C₂₆, C₂₈, C₃₀) was injected in the middle of the queue for external RI calibration.

Raw Agilent data files were analysed with AMDIS http://chemdata.nist.gov/mass-spc/amdis/. The Agilent Fiehn GC/MS Metabolomics RTL Library (version June 2008) was employed for metabolite identifications. This library is one of the most comprehensive library of metabolite GC/MS spectra that is commercially available. It contains searchable GC/MS EI spectra and retention time indexes from approximately 700 common metabolites. Peak areas were determined with the Masshunter Quantitative Analysis (Agilent) in splitless and split 30 modes. Because automated peak integration was occasionally erroneous, integration was verified manually for each compound in all analyses. Resulting areas were compiled in one single Excel File for comparison. Peak areas were normalized to Ribitol and Dry Weight. Metabolite contents are expressed in arbitrary units (semi-quantitative determination).

All experiments were done in triplicates. For metabolomics, normalized data (mean-center) were drawn as a clustered metabolomic array (heat map) using MeV 4.1 open source software (Saeed et al., 2003). The clustering was based on the cosine correlation method. Student’s t-test was performed using this software to compare strains A36 and A37 responses. Metabolites considered to vary significantly between strains A36 and A37 were those with p < 0.05 using crossed Student-Welsh tests.

These experiments were carried out as described in Lejeune et al. (2021a).

LC-MS analysis of A36 and A37 metabolome was carried out as described in (Guerard et al., 2011). Approximately 50 mg (DW) of the lyophilized mycelial samples of A36 and A37 were ground in a mortar using liquid nitrogen, with 1 mL perchloric acid 17%. After centrifugation (14,000 g, 20 min, 4°C) the supernatant was neutralized with potassium hydroxide, adjusted to pH 7 with potassium bicarbonate and centrifuged (14,000 g, 20 min, 4°C). The aliquot was cleaned up with two solid-phase extractions, using the columns Strata-X-C and Strata-X-AW (Phenomenal, Le Pecq, France). 10 mL H₃PO₄ 50% were added to the extract that was then loaded on the Strata-X-C column (flow rate 1 mL/min) after conditioning with methanol and water. The loading fraction was collected in a tube. The column was washed with 1 mL H₃PO₄ 0.1% and the washing fraction was collected. Elution was carried out with NH₃/MeOH/H₂O (0.5 mL 5/25/70 v/v/v, 0.5 mL 10/25/65, 1 mL 20/25/55; flow rate 1 mL/min). The eluate was collected in an Eppendorf tube. Both loading and washing fractions were gathered and 500 mL phosphate buffer (0.01 mol/L) were added. The resulting buffered fraction was loaded (1 mL/min) on the Strata-X-AW column after conditioning with pure methanol, methanol/formic acid/water (2/25/73 v/v/v) and then phosphate buffer. The column was then washed with 1 mL phosphate buffer. Elution was carried out with NH₃/MeOH/H₂O (0.5 mL 5/25/70 v/v/v, 0.5 mL 10/25/65, 1 mL 20/25/55; flow rate 1 mL/min). The eluate was collected in an Eppendorf tube. The two eluates (C and AW) were spin-dried (speed-vac) and used immediately for LC-TOF analysis.

In order to carry out LC-TOF analyses, spin-dried extracts were suspended in 100 µL milli-Q water, centrifuged (8000 g, 4°C, 20 min) and the supernatant was transferred in a glass vial. 10 µL Ara-C (cytosine b-D-arabinofuranoside, 0.1 mmol/L) were added as internal standard. Samples were injected in the UHPLC 1290 infinity II (Agilent Technologies, Les Ulis, France) equipped with the column UPLC-HSS T3 (2.1 × 100 mm, 1.8 mm) (Waters, Guyancourt, France) under an ammonium acetate/methanol gradient (99/1–0/100%, 6 min). The UPLC was connected to the injection orifice of the mass spectrometer via a capillary tube. The flow rate was 0.4 mL/min. Electrospray (Jet Stream Technology Ion Source) mass spectrometry was carried out with the 6540 Q-TOF (Agilent Technologies, Les Ulis, France) with N₂ as a spray gas, at 325°C (Gas Temp), 10 L/min (Drying Gas), 33 psi (Nebulizer), 400°C (Sheat Gas Temp) and 12 L/min (Sheath Gas Flow), upon negative source ionization (3500 V capillary voltage). Resolution parameters were adjusted to achieve optimal mass resolution in the mass range 50–1100 amu. Mass spectra were recorded with an acquisition rate of 2 spectra/s and acquisition time 500 m/spectrum. Data were acquired and processed using the softwares MassHunter Data Acquisition, MassHunter Qualitative Analysis and MassHunter Quantitative Analysis. All compounds analyzed here were measured and calibrated with pure standards purchased at Sigma-Aldrich/Fluka (Saint-Quentin Fallavier, France).

An overnight culture of Micrococcus luteus in LB carried out at 37°C was diluted 10 fold in fresh LB medium and grown at 37°C up to OD600 nm 0.4.100ul of this culture was added to 3 mL of SNA resulting into 104 cfu/mL and was poured onto the surface of LB agar plates. Four agar plugs of A36 and A37 cultures grown for 72 h on R2YE limited in phosphate were deposited on the surface of the solidified and dryed SNA top. The plates were incubated at 37°C for 24 h. Halos of growth inhition around the plugs indicated antibiotics production (Figures 8A and B). However, this test was judged poorly sensitive since the increase of the diameter of inhibition zone around agar plugs of the strains was shown to be linearly proportional to the log of the concentration of the antibiotic produced (Salgado et al., 2006).

LC-ESI-HRMS² analysis of A36 and A37 extracts were performed using two different equipments.

For the extracts of the strains spread at 10³ spores per plate (Figure 8A), LC analyses were performed with an Agilent Infinity II 1260 system (Agilent Technologies, Waldbronn, Germany) equipped with an Accucore C₁₈ RP-MS column (2.1 × 100 mm; 2.6 μm, ThermoScientific, Les Ulis, France). The mobile phase consisted of a gradient water-acetonitrile (H₂O-CH₃CN) acidified with 0.1% formic acid (95:5) to 0:100 in 20 min, then held at 0:100 for 3 min at a flow rate of 350 µL min^-1. The temperature of the column oven was set at 45°C and the injection volume at 5 µL. LC-ESI-HRMS² analyses were achieved by coupling the LC system to an Agilent 6540 hybrid quadrupole time-of-flight mass spectrometer (Agilent Technologies, Waldbronn, Germany) equipped with an electrospray ionization (ESI) dual source, operating in the positive-ion mode. Source parameters were set as follows: gas temperature 325°C, sheath gas flow rate 10 L min⁻¹, nebulizer pressure 30 psi, sheath gas temperature 350°C, capillary voltage 3500 V, skimmer voltage 45 V, fragmentor voltage 130 V, nozzle voltage 500 V and octopole 1 RF voltage 750 V. The data-dependent MS/MS events were acquired for the five most intense compounds detected by full-scan MS, from 200–1000 m/z range, above an absolute threshold of 1 000 counts. Selected precursor compounds were fragmented at a fixed collision energy of 30 eV and with an isolation window of 1.3 amu. Use of a calibration solution, containing two internal reference masses (purine, C₅H₄N₄, m/z 121.0509, and HP-921 [hexakis-(1H,1H,3H-tetrafluoropentoxy) phosphazene], C₁₈H₁₈O₆N₃P₃F₂₄, m/z 922.0098), routinely led to mass accuracy below 3 ppm. LC-UV and MS data acquisition were performed using MassHunter Workstation Data Acquisition and Qualitative Navigator B.08.00 software (Agilent Technologies).

For the extracts of the strains spread at 10⁵ spores per plate (Figure 8B, Figure 9), LC analyses were performed with a Thermo Ultimate 3000 system equipped with a Cortecs C₁₈ column (2.1 × 100 mm; 2.7 μm, Waters). The mobile phase consisted of water-acetonitrile (H₂O-CH₃CN) acidified with 0.1% formic acid (90:10) held for 2 min, then a gradient from 90:10 to 0:100 in 20 min held at 0:100 for 8 min, at a flow rate of 600 µL min⁻¹. The temperature of the column oven was set at 40°C and the injection volume at 5 µL. LC-ESI-HRMS² analyses were achieved by coupling the LC system to an Impact II Bruker quadrupole time-of-flight mass spectrometer (Bruker Daltonics, Bremen, Germany) equipped with an ESI dual source, operating in the positive-ion mode. Source parameters were set as follows: end plate offset 350 V, capillary voltage 4500 V, nebulizer pressure 60 psi, drying-gas flow rate 10 L min⁻¹, drying-gas temperature 240°C. MS scans were operated in the full-scan mode from m/z 100 to 1700 (at 6 Hz). MS¹ scan was followed by MS² scans of the five most intense ions above an absolute threshold of 2000 counts. Selected parent compounds were fragmented with a collision energy fixed at 30 eV and using an m/z dependent isolation window of 2–4 amu. The mass accuracy is guaranteed by injection of a calibration solution, from sodium formate clusters with external (at the beginning of each run) and internal (segment 0.1 at 0.4 min of each sample) calibration by a High Precision Calibration (HPC) equation with a maximum mass delta of 1 ppm and 7 as the minimal number of calibration points. LC-UV and MS data acquisition and processing were performed using DataAnalysis 4.4 software (Bruker Daltonics, Bremen, Germany).

MZmine 2 Parameters: The MS² data files were converted from the. d Agilent standard data format to. mzXML format using MSConvert software, part of the ProteoWizard package (Chambers et al., 2012). The. d Bruker data files were converted to. mzXML format using DataAnalysis 4.4 software. All. mzXML were then processed using MZmine 2 v53 (Pluskal et al., 2010). The mass detection was conducted using a noise level of 400 counts for MS and 0 count for MSMS dimension. The ADAP chromatogram builder was used with a minimum group size of scans of 4, a group intensity threshold of 3000, a minimum highest intensity of 4000 and an m/z tolerance of 15 ppm (Myers et al., 2017). The ADAP wavelets deconvolution algorithm was used with the following standard settings: S/N threshold = 8, minimum feature height = 3000, coefficient/area threshold = 10, peak duration range 0.02–1.0 min, RT wavelet range 0.01–0.07. Isotopologues were grouped using the isotopic peaks grouper algorithm, with an m/z tolerance of 15 ppm and an RT tolerance of 0.1 min. MS² scans were paired using an m/z tolerance range of 0.025 Da and RT tolerance range of 0.1 min. Peak alignment was performed using the join aligner module (m/z tolerance = 15 ppm, weight for m/z = 1, weight for RT = 1, absolute RT tolerance = 0.1 min). The peak list was gap-filled with the peak finder module (m/z tolerance = 5 ppm and RT tolerance = 0.05 min). The ions corresponding to the culture media were removed from the feature list. Eventually, the. mgf spectral data file and its corresponding. csv metadata file (containing RT and peak areas) were exported using the dedicated “Export to GNPS-FBMN” built-in module (Nothias et al., 2020; Wang et al., 2016).

Molecular Network Analysis: The two files mentioned above were imported into MetGem 1.3.6. (Olivon et al., 2018). MS² spectra were window-filtered by choosing only the top ten peaks within the ±50 Da window throughout the spectrum. The data were filtered by removing all peaks in the ± 5 Da range around the precursor m/z. The m/z tolerance window used to find the matching peaks was set to 0.02 Da, and cosine scores were kept under consideration for spectra sharing 4 matching peaks at least. The t-SNE networks were created with nodes sharing at least 4 cosine scores above 0.65 with others (“at least 6 cosine score above 0.5” parameter). The number of iterations, perplexity, learning rate, and early exaggeration parameters were set to 10,000, 6, 100, and 12, respectively. The Barnes-Hut approximation was disabled.

Details of construction of the plasmid pOSV10/DX are provided in Material and Methods and map of pOSV10/DX is provided in Supplementary Figure S1. Attempts were made to conjugate the plasmids pOSV10 (control) and pOSV10/DX into 19 Streptomyces strains listed in Table 2. These strains were specifically chosen for their high total lipid content which is correlated with their low ability to produce antibiotics (David et al., 2020). Of these 19 strains, 14 have an insertion site known to be identical to that of pOSV10 (Raynal et al., 1998) (Table 2) that was successfully conjugated into 12 of these 14 strains.

In contrast, the conjugation of pOSV10/DX was only successful in 2 species: Streptomyces albogriseolus/viridodiastaticus and Streptomyces pristinaespiralis. Our inability to obtain a greater number of Streptomyces pOSV10/DX conjugants suggested a toxicity of DX that might be due to leaky dx expression leading to ATP depletion. In the 2 strains successfully transformed by pOSV10/DX, we speculate that the ErmE promoter is poorly recognized by the RNA polymerase of these strains. This resulted into a weak dx expression, limiting DX toxicity. Furthermore, the apparent toxicity of DX in the majority of the Streptomyces strains suggested that the riboswitch located downstream of the ErmE promoter did not function to repress dx expression.

In this study, we focused on the consequence(s) of the presence of the DX protein on the physiology as well as on the primary and specialized metabolisms of S. albogriseolus/viridodiastaticus. The presence of the pOSV10 (strain A36) and pOSV10/DX (strain A37) in S. albogriseolus/viridodiastaticus was confirmed by PCR (Figure 1A) and the expression of dx was verified by RT-qPCR (Figure 1B). This confirmed that the expression of dx occured in A37 in the absence of theophylline, the inducer of the riboswitch, and was thus constitutive.

In order to assess the consequence(s) of DX over-expression on the physiology of S. albogriseolus/viridodiastaticus, a growth curve was first established for the two strains grown on the classical solid R2YE medium limited in phosphate (Kieser et al., 2000) in absence of theophyllin and subsequently various comparative lipidomic and metabolomic analysis were carried out.

As seen in Figure 2, the growth rates of A36 and A37 were similar up to 60 h of incubation. Beyond this point, the growth rate of A37 was significantly higher than that of A36. This suggested that the DX-mediated artificial ATP spilling led to a stimulation of the metabolism of A37, as reported in other micro-organisms (McKinlay et al., 2020).

Low ATP levels (energetic stress) are usually correlated with low phosphate availability (Esnault et al., 2017) that is known to trigger the expression of phoR (sco4229)/phoP (sco4230), genes encoding the two components system that governs positively the expression of genes of the Pho regulon involved in Pi scavenging and up-take in S. coelicolor (Martin and Liras, 2021). The level of expression of the orthologues of phoR/phoP of S. coelicolor in Streptomyces albogriseolus/viridodiastaticus was assessed in A36 and A37 by RT-qPCR. To do so, RNA was prepared in the same growth conditions as those of Figure 1 and specific primers, internal to each gene (Table 1) were designed to assess their transcriptional activity.

The expression level of the usual reference genes (SCO5820, SCO3795, SCO3873, SCO3874 and SCO2126) was first tested using the oligos listed in Table 1. This analysis revealed that the expression of most of them was higher in A37 than in A36 (Supplementary Figure S3A). This might be due to the general stimulation of the metabolism occuring in A37. The only gene with a similar expression pattern in both strains was SCO3795. SCO3795 was thus used as reference gene to assess phoR and phoP expression (Figure 3). As expected, data of Figure 3 clearly indicated the upregulation of phoP and phoR expression in A37 compared to A36. A similar calculation was carried out with the three reference genes showing an enhanced expression pattern in A37 (SCO3873, SCO3874 and SCO2126) compared to A36 and data shown in Supplementary Figure S3B also indicated a more active transcription of the regulator phoP in A37 than in A36 whereas the transcription level of the kinase PhoR remained similar in the two strains. These data clearly indicated that a DX-mediated ATP shortage triggers phoR/phoP expression.

Since we previously established, in the model strain S. lividans, that a reduced phosphate availability, that resulted into a low ATP content (energetic stress) (Esnault et al., 2017), was correlated with a low phospholipid content (Lejeune et al., 2021a), a comparative lipidomic analysis in LC/MS of A36 and A37 grown for 72 h à 28°C was carried out. This analysis revealed that A37 had a 2 to 3 fold lower content in the phospholipids phosphatidyl ethanolamine (PE), phosphatidyl inositol (PI) and cardiolipin (CL), compared to A36 (Figure 4). This suggested that DX induces a state of energetic stress in S. albogriseolus/viridodiastaticus. In contrast, triacylglycerol (TAG), a neutral storage lipid that does not contain a phosphate moiety was 1.3 fold high in A37 than in A36 (Figure 4).

The comparative analysis of the metabolome of A36 and A37 revealed that the abundance of numerous metabolites was highly contrasted between the two strains (Figures 5, 6, 8). We choose to classify these metabolites in 5 different groups according to what they tell us about the state of the metabolism.

The first class encompasses many intermediates of glycolysis and of the tricarboxylic acid cycle (TCA) as well as amino acids. These metabolites were less abundant in A37 than in A36 (Figure 6) since they were likely consumed for biomass generation in A37. In contrast, molecules in classes 2 to 5 were more abundant in A37 than in A36 (Figures 5, 6, 8).

Class 2 includes molecules such as ADP, AMP, UMP, ornithine (Figure 5) and thymine (Figure 6) The high abundance of the weakly phosphorylated molecules, ADP, AMP and UMP is consistent with the existence of an energy deficit in A37 as is the lower abundance of phospholipids in A37 than in A36 (Figure 4). The amino acid ornithine, also more abundant in A37 than in A36 (Figure 5), is known to be synthesized under phosphate and thus ATP limitation in various microorganisms (Lejeune et al., 2021a), including Streptomyces (Sandoval-Calderon et al., 2015; Barbosa et al., 2018). In such condition, the replacement of membrane phospholipids by ornithine lipids that do not contain phosphate is usually promoted. Thymine is an intermediate of the biosynthesis or degradation of thymidine and thymidine triphosphate (TTP) or uracil triphosphate (UTP), constituents of DNA or RNA. The high abundance of thymine in A37 suggests a dephosphorylating of TTP/UTP that is likely to be promoted by the DX-mediated energetic depletion. Indeed, a limitation in phosphate, and thus in ATP, leads to the induction of the expression of genes of the Pho regulon including various phosphatases (Apel et al., 2007) in order to support ATP replenishment. Unfortunately, our LC/MS analysis could not reveal the presence of ATP in either of the two strains. The cause of this failure is not known but since the presence of DX has a positive impact on growth, this suggests that the DX-mediated ATP spilling most likely triggers an activation of the oxidative metabolism of the strain resulting into ATP generation as in other micro-organisms. Indeed Stomel et al reported that the ATP content of and E. coli strain expressing the DX protein could be 10 fold higher than that of the strain containing the empty plasmid (Stomel et al., 2009).

Class 3 includes NAD and threonic acid. The greater abundance of NAD and its biosynthetic intermediates in A37 then in A36 (Figure 5) is consistent with the activation of the oxidative metabolism in A37. Oxidative phosphorylation requires the biosynthesis of NAD, which is reduced in NADH by reactions of the TCA cycle and subsequently re-oxidized by the respiratory chain to generate ATP. Threonic acid (Figure 6) is a major product of the degradation of ascorbic acid (vitamin C), which generally takes place under conditions of OxS (Deutsch, 1998). Thus, the high abundance of threonic acid in A37 indicates a situation of OxS, likely to be associated with the activation of its oxidative metabolism. Indeed, a limitation in phosphate, and thus in ATP, leads to an activation of oxidative metabolism to restore the energetic balance of the cell via ATP generation but also generates ROS/NOS (OxS) (Esnault et al., 2017; Millan-Oropeza et al., 2017; Millan-Oropeza et al., 2020; Lejeune et al., 2021b).

Class 4 molecules include fatty acids (lauric and myristic acids, Figure 6) and TAG (Figure 4) as well as citramalic acid and L-alanine (Figure 6). Lauric acid (C12) and myristic acid (C14) result from the condensation of acetyl-CoA units. These fatty acids may contribute to the biosynthesis of TAGs that were approximately 1.3 fold more abundant in A37 than in A36 (Figure 4). Citramalic acid is produced from the condensation of pyruvate and acetyl-CoA catalyzed by the citramalate synthase (Webb et al., 2018). Its biosynthesis therefore also consumes acetyl-CoA. Finally, L-alanine was reported to inhibit the activity of pyruvate kinase in yeast (Taber et al., 1998). This enzyme converts PEP into pyruvate that itself is converted into acetyl-CoA by the pyruvate dehydrogenase (PDH). If such inhibition of the activity of pyruvate kinase by L-alanine also occurs in Streptomyces, it might also indirectly contribute to limit the generation of acetyl-CoA. In A37, the synthesis of class 4 metabolites is predicted to limit the availability of acetyl-CoA to feed and thus activate the TCA cycle as summarized in Figure 7. These strategies might be used to temper the activation of the oxidative metabolism of A37 and thus the generation of ATP and OxS.

The level of production of bioactive specialized metabolites (group 5) was first assessed by a basic Micrococcus luteus growth-inhibitory test (pictures of plates in Figures 8A, B) described in Material and Methods. This study revealed that when plated at 10³ spores per plate, A37 had a stronger inhibitory effect on Micrococcus luteus growth than A36, indicating a higher antibiotic production in A37 than in A36 (Figure 8A). In contrast, when the spores were plated at a spreading density of 10⁵ per plate, the inhibition zones of Micrococcus luteus growth around agar plugs of both strains were of similar size (Figure 8B).

The molecules produced in these two conditions were extracted with EtOAc from the mycelium and the solid growth medium of the strains. The extracts were analyzed by LC/HRMS² in data-dependent acquisition mode. Spectral data were preprocessed with MZmine 2 (Pluskal et al., 2010) and organized as molecular networks using MetGem (Olivon et al., 2018). The metabolite annotation was first performed by searching for standards and analogs in GNPS spectral libraries (https://gnps.ucsd.edu/ProteoSAFe/libraries.jsp) using the similarity cosine scoring of MS² spectra (Supplementary Tables S1, S2). Subsequently, molecular formulas of compounds of interest were predicted with the Sirius software tool combining isotope and fragmentation pattern analysis (Duhrkop et al., 2019). Class annotation was confirmed with Canopus (Duhrkop et al., 2021; Djoumbou Feunang et al., 2016). The non-exhaustive list of compounds matching predicted molecular formulas were obtained from Scifinderⁿ (https://scifinder-n.cas.org/). Note that potential adduct ions were maintained in Supplementary Tables S1, S2.

Molecular networks (Wang et al., 2016; Olivon et al., 2018) shown in Figures 8A, B (t-SNE projections) indicate the diversity of molecules present in the analyzed samples. Media components and single nodes were removed from the resulting networks. Each node represents an ion detected in LC-HRMS² analysis. Compounds sharing a similar fragmentation pattern, and therefore structural similarities, gather in the same area of the t-SNE graph. Molecular families can thus be organised and easily visualised when analysing a complex mixture. In Figure 8, a different color was assigned to the samples originating from each strain (A36/pink, A37/green) allowing the visualisation of the origin of the ion. A unicolor node represents an ion only detected in one of the two strains whereas a node with two colors indicates that the ion is present in both strains with a specific ratio.

In both conditions and both strains, most of the characterized molecules were polycyclic aromatic polyketides belonging to the angucycline class (Rohr and Thiericke, 1992; Kharel et al., 2012) usually synthesized by type II polyketide synthases (PKS) (Zhang et al., 2017). Interestingly, 2 PKS of type II were detected by anti-Smash in the genome of S. albogriseolus/viridodiastaticus (Supplementary Figure S2). The molecular network analysis of the angucyclines and related metabolites, at a plating density of 10³ spores per plate, revealed that over half of the detected ions originated exclusively from A37 (Supplementary Table S1). Among those originating from both strains, over 2/3 were 1.2 to 8.75 fold more abundant in A37 than in A36.

At a plating density of 10⁵, nearly 10 fold more features (350 versus 38) were detected than at a plating density of 10³ (Supplementary Table S2). This far greater quantity and diversity of molecules cannot be exclusively attributed to the higher sensitivity of the apparatus used in this second experiment (see Materials and Methods) but most likely results from a more severe nutritional limitation, linked to higher cell density, that led to a more severe ATP limitation. Most of the detected angucyclines were present in both A36 and A37 but the majority of them were more abundant in A37 than in A36, and as much as 10 to 33 fold for some of them (Supplementary Table S2). A large number of these compounds were dereplicated as high-molecular weight angucyclines, showing a higher degree of glycosylation than the compounds produced at a plating density of 10³. At 10⁵ spores per plate, only a few features (7%) were exclusively detected in A37 (Supplementary Table S2).

This study indicated that high spreading density or the over-expression of DX had similar effects on the production of specialized metabolites. A reduced generation of ATP due to nutritional limitation in condition of high cell density or DX-mediated ATP degradation were correlated with the triggering of the production of specific antibiotics. This is consistent with previous work that demonstrated that the over-expression of an ATPase triggers antibiotic production in the weak producer Streptomyces lividans (Seghezzi et al., 2022). However, even at high plating density, A37 produces higher amounts of the same specialized metabolites than A36 (Supplementary Table S2). The DX-mediated earlier and more severe ATP depletion occurring in A37 leads to earlier and higher biosynthesis of specialized metabolites in this strain than in A36% and 6.8% of the detected ions remain specific to A37 (Supplementary Table S2). Interestingly and consistently with previous work that established that the phospholipid content was lower in antibiotic producing strains than in strains that do not produce antibiotics (David et al., 2020; Lejeune et al., 2021a), the phospholipid content of A37 was shown to be lower than that of A36 (Figure 4).

To determine the impact of phosphate availability on the production of specialized metabolites in A36 and A37, spores of these strains were plated at a spreading density of 10⁵ spores per plate, on solid R2YE medium either limited (1 mM) or proficient (5 mM) in phosphate (Pi). Results are shown in Figure 9. The same color code as in Figure 8 was applied in Figure 9A, while the color code in Figure 9B indicates the molecules preferentially produced under conditions of phosphate proficiency (red) or limitation (blue). The comparative LC/HRMS² analysis clearly indicated that most angucyclines detected were produced in condition of phosphate limitation and thus that high Pi availability strongly repressed the biosynthesis of these metabolites in both strains. Approximately half of the angucycline derivatives detected were exclusively produced in condition of Pi limitation whereas the remaining half produced in both Pi conditions were approximately 10.5 fold more abundantly produced in Pi limitation than in Pi proficiency (data not shown). This indicated that in condition of Pi proficiency, the intracellular ATP levels are high and that DX chelation and subsequent reduction in bioavailable ATP is not sufficient to reduce intracellular ATP levels below a certain threshold required to induce the activation of the oxidative metabolism. Such activation generates ATP but also OxS. The latter was proposed to be an important trigger of the biosynthesis of a specific class of quinone antibiotics with anti-oxidant properties (Esnault et al., 2017; Millan-Oropeza et al., 2020; Pires et al., 2020; Beites et al., 2015; Prajapati et al., 2018; Beites et al., 2011; Virolle, 2020).