The novel ECF56 SigG1-RsfG system modulates morphological differentiation and metal-ion homeostasis in Streptomyces tsukubaensis

Extracytoplasmic function (ECF) sigma factors are key transcriptional regulators that prokaryotes have evolved to respond to environmental challenges. Streptomyces tsukubaensis harbours 42 ECFs to reprogram stress-responsive gene expression. Among them, SigG1 features a minimal conserved ECF σ2–σ4 architecture and an additional C-terminal extension that encodes a SnoaL_2 domain, which is characteristic for ECF σ factors of group ECF56. Although proteins with such domain organisation are widely found among Actinobacteria, the functional role of ECFs with a fused SnoaL_2 domain remains unknown. Our results show that in addition to predicted self-regulatory intramolecular amino acid interactions between the SnoaL_2 domain and the ECF core, SigG1 activity is controlled by the cognate anti-sigma protein RsfG, encoded by a co-transcribed sigG1-neighbouring gene. Characterisation of ∆sigG1 and ∆rsfG strains combined with RNA-seq and ChIP-seq experiments, suggests the involvement of SigG1 in the morphological differentiation programme of S. tsukubaensis. SigG1 regulates the expression of alanine dehydrogenase, ald and the WhiB-like regulator, wblC required for differentiation, in addition to iron and copper trafficking systems. Overall, our work establishes a model in which the activity of a σ factor of group ECF56, regulates morphogenesis and metal-ions homeostasis during development to ensure the timely progression of multicellular differentiation.

Streptomycetes are unique soil-dwelling bacteria with a prolific biosynthetic capacity responsible for hundreds of widely used bioactive compounds. The onset of streptomycetes specialized metabolism and the production of secondary metabolites is intimately related with the morphological differentiation process.
Streptomyces species follow a complex multicellular filamentous lifecycle that starts with the germination of a spore when the environmental conditions are favourable. Hyphae emerge that grow by tip-extension and branching, to form a dense filamentous vegetative mycelium that facilitates nutrient scavenging. When nutrients are depleted, a highly regulated process of differentiation is initiated. Vegetative growth is halted before the erection of aerial hyphae that extend upwards, away from the vegetative mycelium. Previously, we demonstrated that differentiation in S. natalensis is tightly controlled by oxidative stress 1 . We illustrated that the transition from first mycelium stage (MI) to secondary metabolite producing differentiated mycelium (MII) is preceded by phenomena of cell death in the vegetative mycelia, which reroutes nutrients to differentiate MII hyphae into nascent aerial mycelium. Also, pivotal to this substrate to aerial mycelium transition is the production of a hydrophobic coat σ 2 ) and σ 70 _r 4 (PF04545, σ 4 ) conserved domains, and a C-terminal (CT) extension, encoding a SnoaL_2 domain (PF12680), which makes it distinctive from other typical ECF σ factors (Fig. 1a). Domains σ 2 and σ 4 are separated by an unusually long (53 aa) spacer. The presence of the CT SnoaL_2 domain in ECF sigma factors is distributed among different phyla 17,18 . They are predominantly found in the Actinobacteria and Firmicutes and, in particular, they are included within groups ECF41, ECF56, ECF205, ECF294 and ECF295 of the general ECF group classification ( Supplementary Fig. S1) 17 . Besides STSU_11560, we identified three additional ECFs with a SnoaL_2 extension in the S. tsukubaensis genome: STSU_14518 and STSU_17474, that belong to the ECF41 group; and STSU_12530, a closer homologue to STSU_11560. These last two ECFs are enclosed within group ECF56, which is exclusively composed of ECFs with SnoaL_2 domains at the CT extension. STSU_11560 belongs to subgroup 3 of the ECF56 family ( Supplementary Fig. S1).
The pool of SnoaL_2-containing ECFs in S. tsukubaensis resembles the organisation of orthologous ECFs experimentally characterised in the Mycobacterium genus. Sequence homology revealed that the full-length proteins encoded by STSU_11560 and STSU_12530 are similar to SigG from Mycobacterium tuberculosis H37Rv (40% and 47% identity, respectively), while STSU_14518 and STSU_17474 showed identity to SigJ from M. tuberculosis H37Rv (43% and 38% identity, respectively). The closest homologous protein to STSU_11560 found in the Streptomyces genus was a sigma-70 family RNA polymerase sigma factor from Streptomyces katsurahamanus (75% identity, E value: 2e -168 ). Following these analyses, we named the gene STSU_11560 as sigG1 and STSU_12530 as sigG2. This work focuses on the physiological characterisation of sigG1.
Direct coupling analyses (DCA) predict intramolecular contacts between SigG1 ECF core domains and SnoaL_2. ECFs with additional domains have been proposed to promote intramolecular contacts that regulate ECF activity 24,28,29 . Making use of the statistical method called direct-coupling analysis (DCA) we assessed whether the SnoaL_2 domain is folded towards the ECF domains σ 2 and σ 4 . DCA is based on the covariation observed between pairs of residues that interact in large families of homologous proteins. When two residues interact, mutations in one of them need to be compensated by mutations in the second in order to preserve the contact. DCA is able to find pairs of amino acids that covariate due to their direct interaction 34 . DCA has been successfully used for the prediction of contacts between the core ECF and the SnoaL-like CT of ECF41 and the TPR-containing CT of ECF42, which later were proven to have an important regulatory role on ECF activity 23 . Here we found that DCA predicted a SigG1 tertiary structure that suggests contacts between the first half of SnoaL_2 CT and the linker that separates σ 2 and σ 4 (Fig. 1b, areas B and C). Additionally, DCA revealed a high probability of contacts between the N-terminal part of SnoaL_2 and the σ 4 region (Fig. 1b, area  A). These results suggest a potential self-regulatory function for SnoaL_2 (Fig. 1c).
SigG1 interacts with a putative anti-sigma factor, RsfG. Downstream from sigG1 we identified an additional uncharacterised gene, STSU_11555, that encodes an hypothetical protein with no conserved Pfam domains. Many known ECFs are bound by their respective cognate anti-sigma factor (ASF), an interaction that Contact map for predicted contacts between ECF core regions and the SnoaL_2 extension, using Direct Coupling Analyses (DCA). Each axis corresponds to the amino acids of SigG1 and the heatmap represents the DCA output, where darker spots correspond to higher scores. Scores > 1 are set to 1 in order to allow for the observation of smaller scores. Dashed lines split the core ECF regions from the C-terminal extension. Significant contacts between SnoaL_2 and core ECF regions are indicated by circles. (c) Model for compact conformation of SigG1 as predicted by DCA. DCA rank and scores of the significant contacts between ECF core regions and the SnoaL_2 domain are shown.
Scientific Reports | (2020) 10:21728 | https://doi.org/10.1038/s41598-020-78520-x www.nature.com/scientificreports/ renders the ECF inactive until it is released to direct transcription from stress-responsive promoters. The antisigma is usually encoded by a gene adjacent to the ecf and can be transcribed independently or co-transcribed in a single mRNA like for example, the bldN-rsbN pair in S. venezuelae 35 or the sigI-rsgI pair in Bacillus subtillis 36 .
To investigate whether sigG1 and STSU_11555 are co-transcribed, we assessed the transcription of both genes by RT-PCR using primers pairs that amplified at different locations in the mRNA. The RT-PCR revealed a longer transcript that included both sequences, indicating that sigG1 and STSU_11555 are co-transcribed (Fig. 2a). The transcriptional start site (TSS) of sigG1 was mapped by 5′ RACE 118 nucleotides upstream of the annotated start codon. This result was confirmed by analysis of existing RNA sequencing (RNA-seq) data 37 . In addition, we identified a TSS for STSU_11555, 74 nucleotides upstream of its coding region, indicating that it can also be transcribed from its own dedicated promoter, independently of sigG1.
Scientific Reports | (2020) 10:21728 | https://doi.org/10.1038/s41598-020-78520-x www.nature.com/scientificreports/ between the two proteins in the BACTH experiments, possibly because specific residues required for complex assembly were missing, or due to the loss of stable conformations of the proteins. The formation of a SigG1-RsfG complex was confirmed via co-expression of the two coding sequences in E. coli using the pRSFDuet™-1 system. The coding sequences for SigG1 and RsfG were cloned into the IPTGinducible vector to co-produce RsfG with an amino terminal (NT) histidine tag (6His-RsfG) and an untagged version of SigG1 in E. coli. By immobilizing the 6His-RsfG on a nickel agarose affinity resin (Ni-NTA) it was possible to pull-down the untagged sigma factor, indicating that RsfG and SigG1 form a complex (Fig. 2c). Elution of the purified proteins revealed two isoforms of 6His-RsfG, as identified by Peptide Mass Fingerprinting (PMF) and Western blot: one protein with the predicted 17 kDa and a second isoform with approximately 19 kDa. Size exclusion chromatography displayed a single peak corresponding to the SigG1/RsfG complex ( Supplementary  Fig. S2)).

SigG1 binds to target promoters in vivo.
To identify the set of genes under the direct control of SigG1 we generated a polyclonal antibody against SigG1 (α-SigG1) to be used in Chromatin Immunoprecipitation and Sequencing (ChIP-seq). Prior to ChIP-seq, we first replaced the sigG1 and rsfG coding regions with an apramycin resistance cassette and generated single deletion mutants for these genes (ΔsigG1 and ΔrsfG, respectively). The ChIP-seq experiments were performed in the rsfG null mutant to ensure SigG1 was available (i.e. not bound to the anti-sigma) to allow detection when bound to its target promoters. Analysis of sigG1 transcript levels by RT-qPCR confirmed expression of sigG1 in the rsfG mutant, and SigG1 protein was similarly detected by automated western blot (WES Simple) ( Supplementary Fig. S3) confirming that the polyclonal antibody was suitable for use in ChIP-seq experiments. The sigG1 null mutant was used as a negative control. DNA purified following immunoprecipitation using α-SigG1 was sequenced. The genomic regions with a fold-change enrichment of at least 1.5 (ΔrsfG vs ΔsigG1) were considered significant for peak calling. Using this approach we identified six peaks dispersed throughout the genome ( Fig. 3a; Table 1).
Consistent with its function as a sigma factor, with the exception of peak 3, the remaining peaks are located in promoter regions (Fig. 3b). Strikingly, no clear peak was identified near the sigG1 promoter suggesting that this sigma factor might not be subject to autoregulation via a positive feedback loop, a common feature of ECFs. SigG1 binds at the promoters of two genes related with morphological differentiation in streptomycetes-alanine dehydrogenase, ald 38 and the WhiB-like regulator C, wblC 39 . We also identified three additional SigG1 target genes involved in metal-ion dependent homeostasis, STSU_23024, STSU_32197 and aconitate hydratase, acn. Furthermore, peak 3 is located in the intergenic region between STSU_22045 and STSU_22050 with no apparent promoter region (Fig. 3b).
To investigate which genes were dependent on sigG1 regulation, we assessed genome-wide transcription by total RNA sequencing (RNA-seq). We compared transcript levels in the wild-type and in ΔsigG1 by analysing gene expression in MGm-2.5 liquid cultures at mid-exponential phase of growth-72 h ( Supplementary Fig. S5). Gene expression results were validated by RT-qPCR on a selection of candidate genes ( Supplementary Fig. S6) and the pattern of expression correlated with the one observed in the RNA-seq data. The level of transcripts for hrdB and rspP genes, encoding the primary essential sigma factor (HrdB) and the 30S ribosomal protein S16 (RspP) respectively, was monitored as an internal normalization control. A combination of the data obtained from the RNA-seq and the ChIP-seq experiments indicates that at least four target genes are members of the SigG1 regulon (Table 1). In addition, a careful inspection of RNA-seq data allowed to identify a putative promoter region for peak 3 identified in the ChIP-seq experiment. The data suggests that SigG1 binds to the upstream region of a putative antisense RNA of unknown function at the STSU_22050 locus ( Supplementary Fig. S4A). Surprisingly, inspection of both ChIP-seq and RNA-seq NGS data unveiled an unexpected genomic rearrangement in the ΔsigG1 mutant. Genomic instability resulting in excision of large fragments of DNA has previously been described in the Streptomyces genus and other bacteria [40][41][42][43][44] . Through NGS re-sequencing of the full genome of this strain, we confirmed a 228 kb deletion in the terminal region of the chromosome from position 349,601 to 577,663 bp (Fig. 3a). No deletion was observed in the ∆rsfG strain.
Considering the central core of the ChIP-seq peaks, and the identification of the TSSs for each gene 37 , we were able to isolate and align promoter sequences harbouring the six identified peaks. By inputting the sequences for the sigG1-dependent genes into the MEME Suite tool 45 , we identified a putative consensus sequence for SigG1 DNA-binding, defined by the conserved residues CCG in the -35 element and the residues CCC TCC in the -10 element (Fig. 3c), separated by 17 bp non-conserved residues. The spacer size and the -10 box are in accordance with what was previously proposed for ECFs in the ECF56 group families 18 . sigG1 is important for progression of morphological differentiation. To validate the sigG1-dependent recruitment of RNAP to the target promoters we have functionally characterised the deletion mutants for the newly identified ECF/ASF pair. One of the most significant peaks obtained with ChIP-seq was found in the promoter region of ald that codes for an alanine dehydrogenase enzyme, Ald1. Ald1 converts L-alanine residues, generated by protein turnover, into extra pools of pyruvate that help fuel metabolism for growth progression in B. subtilis 38 . Moreover, Ald1 is required for maturation of spores in S. coelicolor 46 . On ISP4 media, the sigG1 null mutant exhibited impaired morphological development of sporogenic hyphae. After 6 days of growth on ISP4 agar, the wild-type strain had sporulated, whereas ΔsigG1 lacked the spore-associated pigmented appearance of the wild-type (Fig. 4a). This phenotype was fully restored on agar plates by the introduction of a single copy of the sigG1 wild-type allele under the control of its native promoter expressed from the ɸC31 integration site, located within STSU_17107 in S. tsukubaensis (integration was confirmed by PCR). The delay in differentiation was further studied by scanning electron microscopy (SEM). SEM revealed that despite their initial white appearance, colonies of the sigG1 mutant eventually fully differentiate leading to long chains of spores (Fig. 4b).
Scientific Reports | (2020) 10:21728 | https://doi.org/10.1038/s41598-020-78520-x www.nature.com/scientificreports/  We conclude that SigG1 is required for the timely differentiation of S. tsukubaensis. Strikingly, ΔrsfG grown on solid media exhibited a darker pigmentation that is usually associated with spores, whereas a double mutant for both sigG1 and rsfG exhibited a partial reversal of the phenotype, as it resembled the wild-type morphology. Although some influence of the genomic gap in ΔsigG1 strain would be expected, the phenotypic rescue observed in the complemented strain (that also contains the genomic deletion) shows that the morphological phenotype was mostly due to the absence of SigG1. www.nature.com/scientificreports/ Using the RNA-seq data, we evaluated transcription in the ΔrsfG mutant and compared it to transcript levels in the wild-type. The identification of 18 significantly deregulated genes ( Table 2; Supplementary Fig. S5; FDR < 0.05) revealed that rsfG plays a targeted role in the cell that is confined to the regulation of small operons. These encode for instance different types of proteases or, more interestingly, an operon encoding a putative ECF/ ASF system of unknown function. Moreover, Pre-ranked Gene Set Enrichment Analysis (GSEA 47 ) of ΔrsfG mRNA-seq profile showed an activation of genes involved in the TCA cycle ( Supplementary Fig. S7). sigG1 is required for the maintenance of metal-ion homeostasis. The SigG1 regulon includes two genes whose products are involved in the maintenace of metal-ion availability. STSU_23024 shares homology to genes that have been linked to metal-ion dependent drug efflux systems 48 . It encodes a Major Facilitator Superfamily (MFS) transporter with 76% identity to the multidrug resistance protein Bmr3_2 from Streptomyces sp. AVP053U2. STSU_32197, is a newly identified gene that had not been annotated before ( Supplementary  Fig. S4B). It codes for an hypothetical protein and is part of an operon encoding a copper transport system. Hence, and given the high identity shared by the product of STSU_23024 with metal-ion dependent transporter systems, we further asked if metal-ion homeostasis was compromised in the sigG1 mutants. Using antiSMASH 4.0 49 we identified at least three siderophore biosynthetic clusters in S. tsukubaensis predicted to encode Fe-enterobactin (STSU_33190-STSU_33135 FecCD), Fe-desferrioxamine B (STSU_23636-STSU_23676) and a second Fe-hydroxamate (STSU_16507-STSU_16442) siderophores. We assessed the ability of S. tsukubaensis to release siderophores, high affinity chelators that sequester ferric iron at very low concentrations to assist in metal-ion internalisation 50 . These small molecules are therefore secreted in iron-limiting conditions. Culture supernatants from cells grown in iron-limited conditions exhibited substantially higher levels of siderophores when sigG1 was absent (as shown by the CAS assay in Fig. 5a) and, more strikingly, ΔsigG1 produced siderophores in ironreplete cultures, revealing that these bacteria were under iron starvation. These results were corroborated by a 30% decrease in the intracellular iron content, as measured by flame atomic spectrometry (Fig. 5b). Noticeably, the intracellular iron content was restored upon complementation of the mutant strain (Fig. 5b), corroborating the importance of SigG1 in the maintenace of iron levels.
The emergence of a copper transporter system as a putative SigG1 target, prompted us to assess sigG1dependent copper tolerance. Serial dilutions of spores of the wild-type, ΔsigG1, ΔrsfG, ΔsigG1-rsfG and the ΔsigG1 strain complemented with sigG1, were inoculated on ISP4 copper-replete media (ISP4 supplemented with 100 µM CuSO 4 ). After 6 days of growth, all sigG1/rsfG-related mutant strains exhibited an impaired growth when compared to the wild-type (Fig. 5c). Moreover, ΔsigG1 spores were more sensitive towards copper than any of the other strains, in which spores fail to grow only in the highest dilutions tested, due to the toxicity of detrimetal concentrations of CuSO 4 . The dramatic arrest of development of aerial mycelium observed in ΔsigG1 was reverted by expressing the sigG1 wild-type allele in this mutant.
The S. tsukubaensis oxidative stress response is dependent on sigG1. The disruption of metal ions homeostasis is frequently associated with concurrent impairment of the cell redox state. A major cause of toxicity is the cross-reaction of intracellular iron with H 2 O 2 -through the Fenton reaction 51 -that results in harmful levels of oxidative stress. To examine the impact of the fluctuations in iron availability upon sigG1 dele- In order to investigate the role of SigG1 in the oxidative stress response, we exposed the cells to exogenous H 2 O 2 . After 15 min of H 2 O 2 exposure, we examined sigG1 transcript levels by RT-qPCR. Upon disruption of the major regulator of H 2 O 2 , oxyR, we observed an upregulation of sigG1 transcripts in standard conditions, indicating that sigG1 is expressed during the response to the oxidative stress damage (Fig. 6). Although transcription in the wild-type was not affected by the treatment, sigG1 levels were strongly induced by exogenous H 2 O 2 in the absence of OxyR, indicating that sigG1 is activated in response to H 2 O 2 -mediated oxidative stress. The transcript levels of rsfG were unchanged upon the ROS insult in these strains. To evaluate if the morphological impairment observed on solid medium and the defective oxidative stress response displayed by ∆sigG1 had an impact on secondary metabolism we assessed the production of FK506. In FK506-producing conditions (MGm-2.5 liquid cultures 54 ), ∆sigG1 showed no significant differences in growth when compared to the wild-type ( Supplementary Fig. S8A) however we detected a delayed onset of FK506 biosynthesis, which resulted in a significant reduction (> 30%) in the total production when compared to the wildtype ( Supplementary Fig. S8B). Nevertheless, we cannot exlude that the effect observed upon the production of FK506 could be influenced by the genomic deletion displayed by the ∆sigG1 strain.

Discussion
ECFs with additional domains have been predicted to rely on the autoregulatory activity of the interaction of the ECF conserved core with domains at the C-termini. SigG1 lacks the conserved NPDKL motif at the CT extension predicted to interact with the YVGPWLPEP motif in the linker between σ 2 and σ 4 of the σ ECF41 ( Supplementary  Fig. S9) 22,23 . In fact, DCA analysis predicted that SigG1 conformation is established by pairs of amino acid residues that are distinct from the ones that are in contact in σ ECF41 . This suggests an alternative regulation of SigG1 in respect to ECF41. In addition, nearby sigG1, we found an adjacent gene encoding the RsfG protein, which was co-transcribed with sigG1. We observed that RsfG interacts with SigG1 acting as the cognate sigma antagonist to control SigG1 activity. Hence, we propose that intramolecular regulation in SigG1 might not be sufficient to fully inhibit transcription, as it requires an anti-sigma factor to prevent its runaway activity.
In previous studies in S. natalensis, we showed that the cell death of the vegetative mycelia that precedes the emergence of aerial mycelium results from an increase of the H 2 O 2 intracellular pool 1 . Like in S. coelicolor 55 , in S. tsukubaensis the LysR regulator OxyR perceives the H 2 O 2 signal and drives transcription of alkylhydroperoxidases to detoxify the excess of these reactive oxygen species (ROS) (Pires et al. unpublished data). In parallel, the OxyR regulon includes other players that maintain metal-ion homeostasis. In this study, we provide evidence that the antioxidant response in S. tsukubaensis harbours novel players that compose an ECF sigma/anti-sigma system. In the 147 Streptomyces genomes within ECF56 family in our database, we found a gene encoding an OxyR-like transcriptional regulator in the proximity of a SigG1 homologue-encoding gene. We hypothesize that SigG1 is likely to be an ECF that responds to H 2 O 2 intracellular signals, generated by developmental transitions, to activate the cascades of genes that regulate the metabolic shift during growth. M. tuberculosis SigJ has also been implicated in the response to oxidative stress. Despite the susceptibility to H 2 O 2 exhibited by a ΔsigJ mutant, sigJ mRNA levels are not responsive to H 2 O 2 -induced stress 56 .
The control of intracellular H 2 O 2 homeostasis is linked with the homeostasis of metal-ion content as the cells face emergent toxicity due to the Fenton reaction 51 . Characterisation of ∆sigG1 strain suggests that SigG1 participates in the control of fluctuations in iron and copper availability in S. tsukubaensis. Note that despite the genome gap displayed by the ∆sigG1 strain, complementation with a copy of sigG1 restored the wild-type phenotype, not only regarding intracellular iron and copper toxicity (Fig. 5) but also the morphological phenotype (Fig. 4). The SigG1 regulon includes genes encoding proteins with iron-sulphur centres that are known regulators Expression of the target genes was calculated relatively to the transcription of the reference genes (rpsP and hrdB) and internal normalisation was performed using the wild-type as the control. Results are representative of three biological replicates. Statistically significant differences to the wild-type were determined by one-way ANOVA followed by a Dunnet's test (GraphPad Prism).
Scientific Reports | (2020) 10:21728 | https://doi.org/10.1038/s41598-020-78520-x www.nature.com/scientificreports/ of metal-ion homeostasis, including AcnA and WblC. In streptomycetes, WblC is involved in developmental cascades and, moreover, it regulates the expression of the gene encoding SigR, a well-studied sigma factor responsible for coordinating the response to oxidative stress induced by thiol agents 57 . Furthermore, SigG1 regulates the expression of two genes encoding putative metal-ion dependent efflux systems described in other bacteria. Together with a disrupted iron balance, we show that strains defective in sigG1 were significantly more sensitive to exogenous copper than the wild-type. The mechanisms controlling copper uptake or secretion in Streptomyces species were recently identified [58][59][60][61] . The function of the product of STSU_32197 remains unknown since no clear homologues were identified for this protein. However, one possibility is that SigG1 mediates copper tolerance by activating an operon for the coding sequences for the copper chaperone/P-type ATPase system CopA/CopZ (STSU_32197-STSU_32205) responsible for the secretion of these metal-ion in S. lividans. Similarly to ferrous iron, copper ions can also react with intracellular thiols to produce hydroxyl radicals (the Haber Weiss reaction). Moreover, by occupying a high position in the Irving-Williams series, copper can substitute iron as a cofactor for transcriptional regulators or enzyme activity 62 , which damages iron-sulphur centres due to displacement of the iron ions 63 . Our hypothesis is that to accompany the transcriptional activation of iron-sulphur clusters containing proteins, SigG1 activates the copA/copZ copper efflux locus to limit copper concentration in the cytoplasm and avoid toxicity. In the absence of sigG1, cells fail to export copper and to avoid an increase in oxidative stress, cells can trigger the export of iron to balance redox active metals in the cytoplasm. A low intracellular iron content prompts bacteria to recover environmental iron bound to siderophores, which in turn suppress copper-induced ROS toxicity 64 . Previous works have demonstrated that cuproproteins are recruited for the developmental switch from vegetative to aerial mycelium 60,65 , which is consistent with the timeframe of SigG1 activity. Overall, our results suggest that SigG1 might be involved in regulating iron and copper homeostasis, probably in an effort to counteract the metal-ion imbalance caused by oxidative damage. Curiosly, no SigG1 binding site was identified upstream of sigG1. Instead, sigG1 promoter harbours a conserved motif described for the sigma-factor BldN, a developmental determinant regulated by iron availability at the interface of vegetative to aerial growth 35,66 . The potential BldN-dependent regulation of sigG1 could explain the iron limitation and the impaired aerial growth observed for S. tsukubaensis ΔsigG1.
Altogether, these findings provide an indication as to how Streptomyces could protect itself from oxidative stress generated endogenously at the colony core. We hypothesize that as the rate of primary metabolism slows down, nutrient depletion provides the oxidative trigger required to activate SigG1. SigG1 responds by activating the expression of Ald1, which catalyses the alanine from protein turnover, leading to increased pyruvate production during sporulation. AcnA will then promote pyruvate conversion into energy through the TCA cycle. This extra energy, together with a balance in intracellular metal-ion promoted by other SigG1 targets, will support the metabolic processes responsible for the developmental shift to aerial differentiation. The presence of a SigG1dependent promoter upstream of the aconitate hydratase encoding gene and the upregulation of the genes that composed the TCA cycle molecular signature, observed when deleted rsfG, support this idea ( Supplementary  Fig. S7). Moreover, we observed the enhanced transcription of genes encoding other proteins that are directly involved in metal-ion homeostasis (sodA, bfr) in the ΔrsfG mutant (p > 0.05).
Transitions between mycelia differentiation stages are associated with the onset of secondary metabolism. Another important finding in our study using the industrial antibiotic producing conditions was that disrupting the sigma factor decreased the production of FK506 (Supplementary Fig. S8). This further reinforces SigG1 as an important regulator for the timely progression of growth in S. tsukubaensis and, by association, for the production of FK506. Although in the latter case we cannot completely exclude an indirect effect of the genomic deletion displayed by ∆sigG1.
To the best of our knowledge, this is the first work describing the physiological role of an ECF with a SnoaL_2 extension in bacteria. Overall, the insights gained in this work strengthens the link between morphological development, metal-ion homeostasis and oxidative stress in Streptomyces tsukubaensis, in which, via the activation of a unique SnoaL_2-containing ECF, S. tsukubaensis can respond to stress and thrive in the complex environment of the soil.

Materials and methods
Bacterial strains and growth conditions. Bacterial strains used in this study are described in Table 3.
Escherichia coli strains were grown under aeration in LB (Lysogeny Broth) liquid medium or on LB agar at 30 °C or 37 °C. Streptomyces tsukubaensis strains were grown at 30 °C on Difco ISP4 solid medium (BD, NJ, USA) for spore production. Liquid cultures were grown in MGm-2.5 medium at 28 °C, under aeration at 220 rpm 54 . Ampicillin at 100 μg/mL, kanamycin 50 μg/mL, hygromycin 25 μg/mL, apramycin 50 μg/mL, chloramphenicol at 25 μg/mL or thiostreptone at 40 μg/mL where added to the media when required. Phenotypic analysis of S. tsukubaensis and derived strains was conducted on ISP4 solid media or in MGm-2.5 liquid cultures.

Construction of deletion mutant strains.
In-frame deletions ΔsigG1::apr, ΔrsfG::apr and ΔsigG1-rsfG::apr were generated using REDIRECT PCR targeting technology 67 . Briefly, native sigG1 was replaced with the apr-oriT cassette on the 15C1 cosmid using primers red_sigG1_F/Rand red_rsfG_F/R (Supplementary  Table S1). Modified cosmid was transfer to S. tsukubaensis WT by intergeneric conjugation 68 and apramycin resistant recombinants selected. In-frame deletions were confirmed by PCR (primers red_conf_sigG1_F/R and red_conf_rsfG_F/R; Supplementary Table S1) and by Southern hybridization. For complementation of the sigG1 deletion mutant strain, sigG1 and its promoter region were amplified using oligonucleotides SigG1_SpeI_F and SigG1_SpeI_R, cut with SpeI and cloned into the XbaI site of pIJ12333. The construct was introduced into S. tsukubaensis ΔsigG1 by conjugation.

Polymerase chain reaction (PCR) and oligonucleotides used. DNA fragments used in this study
were obtained by PCR using the G2 GoTaq (Promega, WI, USA) or Q5 High-Fidelity DNA Polymerase (New England Biolabs, MA, USA). All oligonucleotides used are listed in Supplementary Table S1. www.nature.com/scientificreports/ ture aliquots were mixed with two volumes of RNA Protect Bacteria Reagent (Qiagen, Hilden, Germany) and maintained for 5 min at room temperature. Cells were harvested by centrifugation and immediately frozen by immersion in liquid nitrogen. The total RNA was isolated using the RNeasy Mini kit (Qiagen) according to manufacturer instructions with modifications described in Beites, et al. 52 . RNA quality and integrity were evaluated in an Experion Automated Electrophoresis System (Bio-Rad, CA, USA). For cDNA synthesis, 1 μg of DNase I-treated total RNA was transcribed with the iScript Select cDNA Synthesis Super Mix Kit (Bio-Rad). RT-qPCR amplifications were performed in an iCycler iQ5 Real-Time PCR detection system (Bio-Rad) using 0.2 μM of each primer (Supplementary Table S1) and using, 10 μL of KAPA SYBR FAST RT-qPCR Master Mix (KAPA Biosystems, MA, USA) and 2 μL of template cDNA. Standard serial dilutions of the cDNA were used to check the relative efficiency and quality of each primer pair. Non-template controls were included. A melting curve analysis was performed at the end of each RT-qPCR to exclude the formation of nonspecific products. Analysis included three biological replicates and technical triplicates for each cDNA. The data obtained was analysed using the method described by Pfaffl 70 . For each analysis rpsP and hrdB mRNAs were used for normalization.

RNA isolation and RT-qPCR analyses.
Mapping of the 5′ terminus of mRNA by Rapid Amplification of cDNA Ends (5′ RACE). Transcriptional start site (TSS) identification was performed using the 5′ RACE System for Rapid Amplification of cDNA Ends, Version 2.0 kit (Thermo Fisher Scientific), following the manufacturer's instructions. First strand cDNA synthesis was carried out using the gene-specific primer GSP1 (Supplementary Table S1). PCR amplification of tailed cDNA was carried out using the 5'-RACE abridged anchor primer (AAP) with the GSP2 nested primer (Supplementary Table S1). Specificity of the PCR products was confirmed by re-amplification using the AUAP primer and a GSP3 nested primer, and by sequencing. Preparation of cell-free protein extracts. Cells were suspended in lysis buffer containing 50 mM potassium phosphate buffer (pH 6.8) supplemented with Complete EDTA-free protease inhibitor cocktail (Roche, Mannheim, Germany) and lysed on ice using a sonifier. Protein concentration was determined using the BCA Protein Assay Reagent (Thermo Fisher Scientific). Bovine serum albumin was used to determine standard curves.

RNA sequencing.
Secreted siderophores detection. Siderophore production was assessed using the Chrome Azurol S (CAS) assay 71 . Cultures were grown in MGm-2.5 media with or without ferrous iron supplementation. Samples were harvested at 96 h and supernatants placed on CAS agar plates at 30 °C.
Determination of total iron content. Extracellular iron levels were determined in samples harvested throughout growth in MGm-2.5. Cells were centrifuged and the supernatant was recovered to determine total extracellular iron content, using the QuantiChrom Iron Assay Kit (BioAssay Systems, CA, USA) according to the manufacturer's instructions. Intracellular iron levels were measured by flame atomic absorption spectrometry (F-AAS) following a procedure adapted from Yang et al. 72  Production of SigG1-specific polyclonal IgY antibodies. The 6His-SigG1 purified protein and the V16-A39 epitope (SigG1 V16-A39 ) were used to induce immunization of quail for the production of the anti-SigG1 polyclonal antibodies in egg yolks at HenBiotech (Coimbra, Portugal).
Immunoblot detection of SigG1 and 6His-RsfG. Total protein extracts were run on 10% SDS-PAGE and transferred to a nitrocellulose membrane. Membranes were incubated with blocking buffer (5% (w/v) dried milk in TPBS/TBST), rinsed twice in TPBS/TBST and further incubated with blocking buffer supplemented with anti-SigG1 polyclonal quail antibody (final dilution 1:1000) or tetra-his mouse monoclonal antibody (Qiagen), final dilution 1:5000. After incubation, the membranes were rinsed twice in TPBS/TBST, and incubated with secondary antibody-rabbit anti-chicken IgY (IgG) coupled to peroxidase (A9046, Sigma-Aldrich, MI, USA) or anti-mouse IgG conjugated to HRP (Santa Cruz Biotechnology, CA, USA). Signals were revealed with Prime Plus ECL detection kit (Bio-Rad). ChIP-sequencing. For the Chromatin Immunoprecipitation (ChIP) assays S. tsukubaensis strains were grown in MGm-2.5 and samples were harvested and prepared as described in Gallagher, et al. 74 with the following modifications. For immunoprecipitation of SigG1 cross-linked DNA, the total extract was incubated overnight with polyclonal anti-SigG1 V16-A39 antibody at 4 °C. Immunoprecipitation was carried out using a goat anti-chicken IgY agarose (ab76444, Abcam) for 4 h. Genomic DNA libraries enriched for SigG1 binding were produced from these samples, size selected to ~ 100-500 bp, and sequenced on an Illumina NovaSeq 6000 platform (Novogene, Hong Kong), using 150 bp paired-end reads. These experiments were performed in two biological replicates.

Recombinant expression and purification of
Bacterial two-hybrid assays. Bacterial adenylate cyclase two-hybrid assays (BACTH, Euromedex, Souffelweyersheim, France) were performed as described previously 75 . Briefly, the DNA fragments that encode the full-length RsfG and SigG1 proteins were cloned into BACTH T18 and T25 containing vectors. The empty plasmids and the zip plasmids were used as negative control and positive control, respectively. Plasmids were co-transformed into E. coli BTH101 and incubated at 30 °C for 2-3 days. Transformants were grown on M63/ MacConkey agar supplemented with 0.3% lactose, 0.5 mM Isopropyl-b-D-1-thiogalactopyranoside (IPTG), the β-galactosidase chromogenic indicator 5-bromo-4-chloro-3-indolyl-b-D-galacto-pyranoside (X-Gal, 40 µg/ml) and appropriate antibiotics, at 30 °C for 3-7 days. The positive clones were assessed using the β-galactosidase assay 76 . Single colonies of the co-transformed bacteria were expanded overnight in LB broth containing X-Gal at 100 mg/ml, 0.5 mM IPTG and antibiotics for selection. Cells were harvested and resuspended in Z buffer (60 mM Na 2 HPO 4 , 40 mM NaH 2 PO 4 , 10 mM KCl, 1 mM MgSO 4 , 50 mM b-mercaptoethanol pH 7.0). Cells were permeabilized by mixing with toluene. Permeabilized cells were mixed with 10 mg/ml ortho-nitrophenylβ-galactoside (ONPG) and OD 405nm was measured over time. β-galactosidase activity units were calculated as previously described by Miller 76 .
Scientific Reports | (2020) 10:21728 | https://doi.org/10.1038/s41598-020-78520-x www.nature.com/scientificreports/ Bioinformatics procedures. ECF phylogenetic trees were built as described in Casas-Pastor et al. 17 . For Direct Coupling Analysis (DCA) protein sequences of members of groups ECF41 and ECF56 were retrieved from the most recent ECF classification 17 and aligned using Clustal Omega 1.2.3 77 . DCA was performed using Gaussian DCA 78 as described in Wu, et al. 23 . Calculation of differentially expressed genes: after trimming (ambiguous limit = 2 nt, quality limit = 0.01), RNA-seq generated reads were mapped onto the concatenated version of the reference genome (see ChIP-sequencing part in this section). Gene expression levels were determined based on the trancripts per Million (TPM). Expression with log2 fold change ≥ log2 (1.5) and q-value ≤ 0.05 or log2 fold change ≤ -log2 (1.5) and q-value ≤ 0.05 was considered as differentially expressed. Additional analyses of the generated sequence raw data were carried out using CLC Genomics Workbench 20 (Qiagen). Functional enrichment analyses: regulons were categorized according to their associated gene ontology IDs (GO) retrieved from the information available at Uniprot database. GSEA analysis was performed using the Broad Institute GSEA software 47 using Gene set collections obtained by searching the S.tsukubaensis genome for homologues of proteins involved in sporulation and TCA through NCBI BLASTp analyses (https ://blast .ncbi.nlm.nih.gov). The ΔsigG1 vs WT full transcript list was ranked according to the log2 fold-change expression values and probed against the indicated gene signatures using the GSEA Pre-ranked mode with the following parameters: 10,000 permutations, classic scoring schem and meandiv normalization. ChIP-seq data analysis: reads resulting from paired-end sequencing were aligned to the genome of S. tsukubaensis using the bowtie2 79 and further treated as described in Gallagher et al. 74 . Enrichment for the ΔsigG1 control samples was subtracted from the enrichment in ΔrsfG samples. Significance of enrichment values were calculated assuming normal distribution of the enrichment values. Results were visualized in the Integrated Genome Browser 80 , or in the CLC Genomics Workbench 20 (Qiagen). The reference genome was the concatenation of all contigs available on NCBI at the time (genome assembly ASM29715v2; WGS project AJSZ01). However, after this article was submitted for review, genome assembly ASM29715v2 was superceded by a new version (ASM29715v3) with an updated annotation (GenBank accession numbers: CP029157, CP029158 and CP029159). Whilst This does not affect the results or claims of this article, the locus tags that identify genes have changed. For instance, the locus tag for STSU_11560 has changed to STSU_011570 in the latest genome assembly version.

Statistical analyses.
For each experiment, we assayed at least three independent biological replicates. Statistical significance was addressed through the GraphPad Prism 8 software, according to the requirements of each data set.

Data availability
Sequencing data was submitted in GEO under the following accession codes: GSE144815 for RNA-seq data and GSE144907 for ChIP-seq data.