The gamma-butyrolactone system from Streptomyces filipinensis reveals novel clues to understand secondary metabolism control.

Streptomyces γ-butyrolactones (GBLs) are quorum sensing communication signals triggering antibiotic production. The GBL system of S. filipinensis, the producer of the antifungal agent filipin, has been investigated. Inactivation of sfbR, a GBL receptor, resulted in a strong decreased production of filipin while deletion of sfbR2, a pseudo-receptor, boosted it, in agreement with lower or higher transcription of filipin biosynthetic genes respectively. Noteworthy, none of the mutations affected growth or morphological development. While no ARE (autoregulatory element)-like sequences were found in the promoters of filipin genes, suggesting an indirect control of production, five ARE sequences were found in five genes of the GBL cluster, whose transcription has been shown to be controlled by both SfbR and SfbR2. In vitro binding of recombinant SfbR and SfbR2 to such sequences indicated that such control is direct. Transcription start points were identified by 5'-rapid amplification of cDNA ends, and precise binding regions were investigated by DNase I protection studies. Binding of both regulators took place in the promoter of target genes and at the same sites. Information content analysis of protected sequences in target promoters yielded an 18-nucleotide consensus ARE sequence. Quantitative transcriptional analyses revealed that both regulators are self-regulated and repress each other transcription as well as that of the remaining target genes. Unlike other GBL receptor homologues, SfbR activates its own transcription, while SfbR2 has a canonical auto-repression profile. Additionally, SfbR2 was found here to bind antifungal antimycin A as a way to modulate its DNA-binding activity.IMPORTANCE: Streptomyces GBLs are important signalling molecules triggering antibiotic production in a quorum sensing dependent manner. We have characterized the GBL system from S. filipinensis, finding that two key players of this system, the GBL receptor and the pseudo-receptor, counteract each other transcription for the modulation of filipin production, and that such control over antifungal production involves an indirect effect on the transcription of filipin biosynthetic genes. Additionally, both regulators bind the same sites, are self-regulated and repress the transcription of three other genes of the GBL cluster, including that of the GBL synthase. Contrary to all the GBL receptors known, SfbR activates its own synthesis. Moreover, the pseudo-receptor was identified as the receptor of antimycin A, thus extending the number of examples supporting the signalling effects of antibiotics in Streptomyces The intricate regulatory network depicted here should provide important clues to understand the regulatory mechanism governing secondary metabolism.

while deletion of sfbR2, a pseudo-receptor, boosted it, in agreement with lower or higher 23 transcription of filipin biosynthetic genes respectively. Noteworthy, none of the 24 mutations affected growth or morphological development. While no ARE (autoregulatory 25 element)-like sequences were found in the promoters of filipin genes, suggesting an 26 indirect control of production, five ARE sequences were found in five genes of the GBL 27 cluster, whose transcription has been shown to be controlled by both SfbR and SfbR2. In 28 vitro binding of recombinant SfbR and SfbR2 to such sequences indicated that such 29 control is direct. Transcription start points were identified by 5´-rapid amplification of 30 cDNA ends, and precise binding regions were investigated by DNase I protection studies. 31 Binding of both regulators took place in the promoter of target genes and at the same 32 sites. Information content analysis of protected sequences in target promoters yielded an 33 18-nucleotide consensus ARE sequence. Quantitative transcriptional analyses revealed 34 that both regulators are self-regulated and repress each other transcription as well as that 35 of the remaining target genes. Unlike other GBL receptor homologues, SfbR activates its 36 own transcription, while SfbR2 has a canonical auto-repression profile. Additionally, 37 SfbR2 was found here to bind antifungal antimycin A as a way to modulate its DNA-38 binding activity. 39

INTRODUCTION 56 57
Streptomyces are soil-dwelling bacteria that undergo a rather complex 58 differentiation process which is usually accompanied by antibiotic and other secondary 59 metabolite-bioactive molecules production such as anticancer agents, 60 substantially higher than in other strains (26). Despite being a polyene and having a 102 potent antifungal activity derived from its interaction with the ergosterol of fungal 103 membranes, this pentaene also shows a rather high affinity for cholesterol, which makes 104 it useless in human therapy. Nonetheless, it is widely used for the detection and the 105 quantitation of cholesterol in biological membranes (27) and as a tool for the diagnosis of 106 Niemann-Pick type C disease (28). Its biosynthetic pathway in S. filipinensis has recently 107 been discovered (26), and besides a recent study on the mechanism of phosphate control 108 of filipin biosynthesis (29), there is an absolute lack of knowledge on regulatory 109 mechanisms of antibiotic production in this bacterium. It was therefore, of great interest 110 to study the GBL system of S. filipinensis and its role on filipin production. 111

RESULTS AND DISCUSSION 113
Cloning of a -butyrolactone gene cluster in S. filipinensis. 114 The GBL gene cluster was identified by hybridization using a cosmid library (26) and a 115 149 bp probe obtained by PCR amplification of S. filipinensis chromosomal DNA with 116 degenerate oligonucleotides derived from conserved stretches of the N-terminal region of 117 several GBL receptors (see "materials and methods"). Once a gene homologous to GBL 118 receptors was identified (sfbR), the remaining genes of the cluster were identified by 119 chromosome walking. The deduced gene organization within this region is shown in Fig  120   1A. 121 In silico analysis and arrangement of genes. 122 on November 3, 2020 by guest http://aem.asm.org/ Downloaded from molecule of the butenolide type (3). sfb4 also contains the rare leucine TTA codon, a 146 feature that shares with other cytochrome P450 monooxygenase-encoding genes from 147 GBL gene clusters such as orf16* from S. fradiae (34) or tsuB from S. tsukubaensis (36). 148 sfb5 is located downstream, it encodes a putative short-chain dehydrogenase/reductase 149 with similarity to Orf4 protein from Streptomyces sp. SBI034 (37) (51% identity) and 150 ScbC from S. coelicolor (38) (50% identity). These enzymes have been proposed to be 151 nucleoside-diphosphate-sugar epimerases involved in GBL biosynthesis, thus it is 152 conceivable that Sfb5 could have a role in the biosynthesis of S. filipinensis GBL. The 153 two genes that follow are oriented convergently (Fig. 1A), sfb6 that encodes a 154 hypothetical protein of unknown function, and sfb7 whose coding strand is in the 155 opposite strand of all the identified genes and encodes a small size SARP-like regulator. 156 This protein is highly similar to other SARPs belonging to GBL clusters such as BulY 157 from S. tsukubaensis (36), FarR4 from S. lavendulae (39) and SgvR2 from S. 158 griseoviridis (40) (66, 65 and 65 % identity respectively). 159 Downstream from sfb7 is sfbR2 (Fig. 1A), which encodes a putative GBL pseudo-160 receptor (see below). It showed highest similarity scores with PapR5 from 161 S. pristinaespiralis (33) (51 % identity) and TylQ from S. fradiae (34) (47 % identity). 162 in the same orientation lies sfbA, whose product shows high similarity with GBL 164 synthases such as SrrX from S. rochei (41) (57% identity) or Lct9 from S. rishiriensis 165 (42) (55% identity). Fig. S3 shows an alignment of SfbA with other GBL synthases. 166 When compared with other GBL gene clusters, no obvious gene synteny is 167 observed. Although the GBL receptor and pseudo-receptor tend to be clustered with the 168 on November 3, 2020 by guest http://aem.asm.org/ Downloaded from GBL biosynthesis gene/s (6) (Fig. 1B), many exceptions have been described. Thus, 169 S. avermitilis or S. fradiae GBL gene clusters lack GBL synthase-encoding genes and 170 harbour an acyl-CoA oxidase coding gene (aco) instead that possibly substitutes GBL 171 synthase (3,34). Another exception is the S. coelicolor GBL gene cluster where the scbR2 172 pseudo-receptor gene is separated from the GBL synthase and the receptor genes by the 173 coelimycin cpk cluster (Fig. 1B)

SfbR is a putative GBL receptor whereas SfbR2 is a pseudo-receptor. 176
Both GBL receptors and pseudo-receptors belong to the TetR family of regulators. These 177 comprise a conserved helix-turn-helix DNA-binding motif at the N-terminal region and a 178 variable C-terminal domain involved in ligand binding (44,45). The latter contains a 179 highly conserved tryptophan residue which is involved in ligand binding (46). 180 Interestingly, SfbR and SfbR2 share a 34% identity and both contain such conserved 181 tryptophan residue (W123 in SfbR and W128 in SfbR2 (Figs. S1 and S2)). 182 Despite their structural similarity, when compared with counterparts in the 183 databases, SfbR clusters with genuine GBL receptors whereas SfbR2 does it with pseudo-184 receptors. Fig. 2 shows a phylogenetic analysis with whole amino acid sequences. 185 Besides, SfbR has a calculated pI of 6.54, a slightly acidic-neutral value that is 186 characteristic of genuine receptors, whereas that of SfbR2 is 8.74, a basic value common 187 in pseudo-receptors (39). Most of the proteins analysed follow such paradigm but there 188 are a few exceptions (indicated in red in Fig. 2). Noteworthy, MmfR, an authentic 189 methylenomycin furan receptor from S. coelicolor (47) with a pI value of 5.99, as 190 expected, but which clusters with pseudo-receptors, and ScbR2 an S. coelicolor GBL 191 on November 3, 2020 by guest http://aem.asm.org/ Downloaded from pseudo-receptor (21) which clusters with pseudo-receptors, as expected, but which has a 192 slightly acidic pI value of 5.85. 193 Inactivation of sfbR reduces filipin production while sfbR2 deletion increases it. 194 To assess the function of sfbR or sfbR2, we deleted them by using the REDIRECT gene 195 replacement technology as indicated in "materials and methods". Double-crossover 196 mutants that had lost sfbR were screened by apramycin resistance whereas those lacking 197 sfbR2 were selected by their resistance to spectinomycin. All mutants were further 198 verified by PCR analysis (Fig. S4). 199 In order to study the effect that the inactivation of the sfbR and sfbR2 genes had 200 on the production of filipin, fermentation broths produced by the new mutant strains, 201 when grown in YEME medium, were extracted with methanol and analysed for the 202 presence of filipin III (the major component of the filipin complex). Results indicated that 203 S. filipinensis sfbR was impaired in filipin production, reaching ca. 50% of the 204 production observed in the parental strain (Fig. 3A), thus suggesting that SfbR is an 205 activator of filipin biosynthesis. On the contrary, S. filipinensis sfbR2 behaved as a To confirm that the gene deletions were directly responsible for the effects observed on 215 filipin III production, we complemented both mutants with the corresponding gene. For 216 that purpose, we introduced one copy of the gene including its promoter region into the 217 genome of the mutants using the integrative plasmids pSETneo::sfbR or pSETneo::sfbR2 218 respectively (see materials and methods). pSETneo (48) was also introduced into the 219 parental strain as control. No differences on growth were observed between the 220 complemented strains and the control. 221 Introduction of a copy of sfbR into S. filipinensis sfbR boosted its ability to 222 produce filipin III, almost restoring it to parental strain levels, whereas introduction of 223 sfbR2 into S. filipinensis sfbR2, reduced its ability to produce the antifungal to the same 224 values as the wild type strain (Fig. 3B). These results indicate that both regulators control 225 filipin biosynthesis in opposite ways. 226 Counteraction between GBL receptor and pseudo-receptor is not uncommon in 227 Streptomyces spp. In S. pristinaespiralis SpbR activates pristinamycin production 228 whereas both two pseudo-receptors PapR3 and PapR5 repress its production (14,49). 229 Similarly, in S. aureofaciens GBL receptor SagR and pseudo-receptor Aur1R behave as 230 auricin biosynthesis activator and repressor respectively (25,50). In S. venezuelae the 231 same type of competition has also been described (7,21). 232

SfbR and SfbR2 control fil genes expression indirectly. 233
In order to study if the effect on filipin production in the mutants was a direct 234 consequence of a higher or lower transcription of filipin biosynthetic genes when 235 on November 3, 2020 by guest http://aem.asm.org/ Downloaded from compared to the parental strain, we performed gene expression studies by  Total RNA was prepared from cultures after growth for 48 h in YEME medium without 237 sucrose, and used for analysis. The transcriptional levels of selected genes corresponding 238 to different operons governing filipin biosynthesis in the mutant strains were compared 239 with those of the wild strain, to which a relative expression value of 1 was assigned. The 240 genes selected included the polyketide synthases filA1 and filA2, the thioesterase 241 encoding filH, and two cluster situated regulators (filR and filF) (26). 242 In agreement with the overproduction of filipin in the sfbR2 mutant or the 243 decreased filipin production in the sfbR mutant, all the selected genes showed the same 244 pattern of expression, i.e. overexpression in the sfbR2 mutant and repression in the 245 sfbR mutant (Fig. 4), which indicates that SfbR2 is a repressor and SfbR an activator of 246 filipin production. 247 In the absence of their ligands, GBL receptors recognize and bind to palindromic 248 sequences rich in adenine and thymine called AREs (autoregulatory elements) present in 249 the promoter regions of target genes (in many occasions their own encoding genes), 250 repressing them (14,20,51,52). Recent studies have shown that pseudo-receptors are also 251 capable of recognizing and binding to the same operator sequences (24,25,36,53). SprA, a GBL receptor, stimulates transcription of several pimaricin biosynthetic genes 266 and antifungal production in an indirect manner (15). 267

Organization of transcriptional units within the GBL cluster 268
To obtain an overall picture of the transcriptional arrangement of the sfb genes in 269 S. filipinensis it was necessary to determine the operons governing their transcription. The sfb2 TSP is located at an adenine 35 nucleotides upstream the ATG start 290 codon. Analysis of the upstream sequence revealed TATCAT and AGTAGT as the -10 291 and -35 boxes. Both boxes are separated by 14 nucleotides, with the -10 hexamer centred 292 at 10 nucleotides from the TSP (Fig. 5). The TSP of sfbR is located at a thymine 59 293 nucleotides upstream from the GTG start codon. The sequence TAGCAT, centred at 294 position -9, constitutes the -10 consensus, and a -35 box CCGCCC was identified at 19 295 nucleotides distance. In the case of sfb7, the TSP was identified at a cytosine 84 296 nucleotides upstream from the ATG. The -10 and -35 boxes (TTTAAT and TCCACT, 297 respectively) were centred at positions -9 and -37 nucleotides from the TSP and are 298 separated by 22 nucleotides (Fig. 5). For its part, sfbR2 presented two TSPs, one at an 299 adenine and the second one at a guanine located 234 and 108 nucleotides, respectively, 300 upstream from the ATG start codon. The one at position -234 corresponds to a promoter 301 with -10 (CAGGGT) and -35 (CTGTCC) boxes separated by 19 nucleotides, while the 302 one at position -108 obeys to a promoter with -10 and -35 boxes (TTTGTT and 303 CGGAGC, respectively) separated by 21 nucleotides (Fig. 5). Finally, the TSP of sfbA 304 on November 3, 2020 by guest http://aem.asm.org/ Downloaded from was identified at a guanine 56 nucleotides upstream from the ATG codon. The analysis of 305 the upstream sequence revealed a clear promoter, with the -10 box TATATT located 10 306 nucleotides upstream from the observed TSP and the -35 box TCGCCC separated by 21 307 nucleotides (Fig. 5). This finding, together with former results indicates that sfbA can be 308 transcribed as a monocistronic transcript from its own promoter and as a bicistronic 309 transcript from the sfbR2 promoter. Taken together, these results suggest that S. filipinensis may present a rather more 314 complex control of GBL genes than other Streptomyces spp. On one side, while two 315 promoters could direct the transcription of the bicistronic sfbR2-sfbA mRNA, only one of 316 them (the one that is closer to the translation start) contains an ARE box. This suggests 317 that these genes could partially avoid self-regulation when being transcribed from the 318 distal promoter. On the other side, sfbA also has its own dedicated promoter which 319 contains an ARE box. This feature suggests that sfbA transcription could have various 320 points of control, either by SfbR or SfbR2, or both. 321

SfbR2. 323
In order to examine the role of SfbR and SfbR2 regulators on the transcription of the 324 genes whose promoters contained ARE sequences, we measured gene expression in the 325 mutants by RT-qPCR. Total RNAs obtained from 24-, 48-and 72-hour cultures were 326 used as a template, and the transcriptional levels of each gene in the different strains were 327 on November 3, 2020 by guest http://aem.asm.org/ Downloaded from compared with those of the parental strain, which was assigned a relative expression 328 value of 1. 329 In order to assess the transcription of the deleted genes primers were designed to 330 generate PCR products near the 5´-end of mRNA. Interestingly, transcription of GBL 331 genes was controlled by both SfbR and SfbR2. The absence of SfbR2 caused an increase 332 in the transcription of all the genes studied including that of its own gene, which indicates 333 that it behaves as a repressor of all these genes, in particular at 24 h. In contrast, the 334 absence of SfbR caused an increase in the transcription of all the genes except for its own 335 transcription, which was reduced (Fig. 6). This result indicates that SfbR is an activator of 336 its own synthesis and a repressor of the remaining genes studied. The self-activation of 337 sfbR transcription was completely unexpected since GBL receptors normally act as 338 repressors of their own synthesis (3,23,36,39). 339 In GBL regulatory systems, it is common for receptors and pseudo-receptors to 340 repress synthase expression (7,15,24,53,58) so that the GBL accumulates very slowly 341 until it reaches a critical concentration when it binds the receptor and releases it from 342 target promoters. According to our results, S. filipinensis follows such general model in 343 which the GBL receptors and pseudo-receptors act as repressors of the synthase gene 344 (20,39). Although normally these receptors regulate their own synthesis directly by 345 binding to ARE sequences located in their promoters, transcriptional analyses did not 346 allow to confirm such point. For this purpose, we decided to purify both regulators and 347 study their binding capacity to the ARE sequences identified. of dimers as has been proposed for the binding of BulR1, the GBL receptor of S. 366 tsukubaensis, to its targets (36). 367

DNase I protection studies reveal that both regulators bind the same sites. 368
To determine the precise binding sites of both regulators we carried out DNase I 369 footprinting assays. GST-SfbR or GST-SfbR2 protein (2M) was tested using 5´-end 370 fluorescein-labelled DNA fragments. All analyses were carried out by triplicate. These 371 analyses revealed that both fusion proteins protected a single site in the promoters 372 containing ARE sequences, and that both bound the same sites (Fig. 8). and both regions were displaced by 2 nucleotides (Fig. 8A). 377 Footprinting assays of the sfbRp region revealed a 29-nucleotide protection in the 378 coding strand (positions -136 to -164 with respect to the sfbR translation start site). In the 379 complementary strand the protected sequence was 30 bp long, spanning from position -380 128 to -157. In this case, both protected regions were displaced by 7-8 nucleotides (Fig.  381   8B). 382 In the case of the sfb7 promoter, a protected region of 28 nucleotides was 383 observed in the coding strand of sfb7 (positions -100 to -127 from sfb7 translation start 384 codon). In the bottom strand the protected sequence was 30 bp long, at positions -83 to -385 112 (Fig. 8 C). These protected regions were slightly displaced, i.e. they do overlap 13 386

nucleotides. 387
Results of the analysis of the sfbR2p promoter region showed a protected stretch 388 extending for 28 bp of the coding strand. This protected region is located at nucleotide 389 positions -117 to -144 with respect to the sfbR2 translational ATG start site. The 390 protection of the reverse strand was 29-nuclotide long (positions -108 to -136), both 391 regions being displaced by 8-9 nucleotides (Fig. 8D). 392 In the case of the sfbA promoter, a protected region of 28 nucleotides was 393 observed in the coding strand of sfbA (positions -63 to -90 from sfbA translation start 394 codon), and the same was also observed in the bottom strand (positions -63 and -90). 395 Typically, the protected region in the sense strand of the regulated gene is 397 accompanied by a protection in the complimentary strand, both protected regions being 398 slightly displaced (Fig. 8). 399

Information content analysis of the SfbR and SfbR2 operators. 400
An information-based model of the binding site was constructed, taking into account the 401 10 protected regions observed in the footprinting assays. A sequence logo (59) that 402 depicts the binding site is shown in Fig. 8F. This site spans 18 nucleotides and adjusts to 403 the consensus AAACVGNNBVNNCSGTTT (where V represents A, C, or G; S is C or 404 G; and B is C, G, or T). It is noteworthy that the binding site displays dyad symmetry and 405 is highly similar to consensus sequences recognized by other GBL receptors and pseudo-406 receptors from Streptomyces sp (23, 24). 407

SfbR2 is the receptor of antimycin A. 408
SfbR2 homologues ScbR2, JadR2 and AvaR2 have been described to bind antibiotics as 409 ligands (21-23). These antibiotics may be either endogenous (21) or exogenous (22). The 410 responses of SfbR2 to different antibiotics that rendered positive results with SfbR2 411 homologues (chloramphenicol (JadR2); kanamycin (AvaR2)) were analyzed by EMSAs 412 using two DNA targets of SfbR2, the promoter regions of sfbR and sfbA (Fig. 9). Other 413 exogenous antibiotics tested included tetracycline, other aminoglycosides such as 414 spectinomycin or apramycin, and betalactams such as ampicillin. The endogenous 415 macrolide filipin, and the polyketide-non ribosomal peptide antimycin A commonly 416 produced by Streptomyces spp. were also assayed. Dissociation of SfbR2-DNA 417 complexes was not induced by antibiotics chloramphenicol, kanamycin, spectinomycin, 418 apramycin, tetracycline or ampicillin, even at 20 mM concentration (Fig. 9), or filipin 419 which could only be tested up to 2 mM concentration given its low solubility (not 420 shown). In contrast, complexes were disrupted by antimycin A at 5 mM concentration 421 (Fig. 9). These results strongly suggest that antimycin A is recognized as ligand by 422 To further support the role of antimycin A as SfbR2 ligand, we performed in vivo 429 assays by testing its effect on the transcription of SfbR2 target genes. S. filipinensis was 430 grown for 24 h, added growing concentrations of antimycin A (or ethanol as solvent 431 control) and total RNA was isolated one hour later. Transcript quantity was assessed by 432 RT-qPCR using RNAs obtained as template. The transcriptional levels of each gene in 433 the presence of antimycin A were compared with those in its absence, which was 434 assigned a relative expression value of 1. Interestingly, the addition of 10 µM antimycin 435 A clearly increased transcription of every gene (Fig. 9). Moreover, transcription was 436 further incremented when we increased antimycin A concentration to 50 µM. Taken 437 together, these results clearly indicate that SfbR2 not only binds antimycin A but also 438 responds to it in a concentration dependent manner. 439 So far, and given that S. filipinensis genome is not sequenced, we do not know 440 whether it encodes antimycin A biosynthetic gene cluster. We tried to identify the producing organism, and the latter seems to be the case of S. filipinensis. Future 449 experimental studies will establish the molecular mechanism involved in this process. 450

GBL regulatory model in S. filipinensis. 451
Streptomyces GBLs are important signalling molecules to trigger antibiotic production in 452 a quorum sensing dependent manner. In this work, we have characterized the GBL 453 system from S. filipinensis, finding that two key players of this system, the GBL receptor 454 and the pseudo-receptor, counteract each other transcription for the modulation of filipin 455 production. Such control over antifungal production involves an indirect effect on the 456 transcription of filipin biosynthetic genes presumably through a yet unidentified regulator 457 (Fig. 10). In this scenario, the GBL receptor SfbR acts as an activator of filipin 458 biosynthesis, whereas the pseudo-receptor SfbR2 behaves as a repressor. Whether there is 459 a connection between this regulation of filipin biosynthesis and the recently described 460 Pho regulation for this strain (29) remains unknown, and further studies will be required 461 to check such possibility. 462 The structure of the GBL produced by S. filipinensis is currently unknown, but by 463 analogy of the genes found in the GBL gene cluster with other systems characterised, it is 464 likely synthesized by the concerted action of the GBL synthase SfbA, the cytochrome 465 on November 3, 2020 by guest http://aem.asm.org/ Downloaded from P450 monooxygenase Sfb4 and the nucleoside-diphosphate-sugar epimerase Sfb5 466 (3,16,17,38). 467 As occurs with other GBL systems, the GBL receptor and the pseudo-receptor 468 target the same sites at DNA, which are highly similar to previously identified binding 469 sites from other GBL receptors and pseudo-receptors. Both regulators are self-regulated 470 and repress the transcription of three other genes of the GBL cluster, specifically that of 471 the regulators sfb2 (StrR-family) and sfb7 (SARP-family), and the GBL synthase sfbA. 472 SfbR2 represses its own transcription as expected for a pseudo-receptor, but SfbR, 473 contrary to all the GBL receptors known up to date, activates its own synthesis rather 474 than repressing it (Fig. 10).   contaminating DNA in the RNA preparations. The identity of each amplified product was 513 corroborated by direct sequencing of the PCR product. 514

Reverse transcription-quantitative PCR. 515
Reverse transcription of total RNA was performed on selected samples with 5 µg of RNA 516 and 12.5 ng/µl of random hexamer primer (Invitrogen) using SuperScript™ III reverse 517 transcriptase (Invitrogen) as described (62). Reactions were carried out on three 518 biological replicates with three technical replicates each and appropriate controls were 519 included to verify the absence of gDNA contamination in RNA and primer-dimer 520 formation. Primers (Table 1)  Transcription start points were identified by using a 5′-RACE system for rapid 535 amplification of complementary DNA (cDNA) ends kit (Invitrogen) following the 536 manufacturer's instructions (version 2.0) and as described (61), using 5 μg of total RNA 537 for first-strand cDNA synthesis and the gene-specific primers listed in Table 1. 538

Assessment of filipin and antimycin A production. 539
Filipin and antimycin A production was assessed after growth at 30 ºC in YEME medium 540 without sucrose. To assay filipin in culture broths, 1 vol of culture was extracted with 1 541 vol of methanol, and further diluted with methanol to bring the absorbance at 338 nm in 542 the range of 0.1 to 0.4 units. Control solutions of pure filipin III (Sigma) were used as 543 controls. The identity of filipin was confirmed by its UV-visible absorption spectrum 544 (absorption peaks at 356, 338, 320 and 311 nm). Quantitative determination of filipin was 545 performed as previously described (26), using a Mediterranea Sea C18 column (4.6x150 546 mm, particle size, 3 mm) (Teknokroma). For antimycin A production assessment, 1 vol of 547 culture was extracted with 2 vol of ethyl acetate, and dried by rotary evaporation. The 548 pellet was the resuspended in methanol prior to HPLC analysis. The same 549 chromatographic method was used for estimation of antimycin A production at 318 nm. 550 Pure antimycin A (Sigma) was used as standard. 551

Construction of mutants. 552
Deletion of sfbR from S. filipinensis chromosome was made by replacing the wild-type 553 gene with a cassette containing an apramycin selective marker using a PCR based system 554 (66). The plasmid pIJ773 containing the apramycin resistance gene (aac(3)IV) and the 555 oriT replication origin was used as a template. The mutant was constructed using the 556 on November 3, 2020 by guest http://aem.asm.org/ Downloaded from oligonucleotides SfbR-Red-F and SfbR-Red-R (Table 1) as the forward and reverse 557 primers respectively (the sequence identical to the DNA segment upstream from the start 558 codon of sfbR is in lower case italics and the sequence identical to the segment 559 downstream from the stop codon of sfbR is underlined and in lower case). These two long 560 PCR primers were designed to produce a deletion of sfbR just after its start codon leaving 561 only its stop codon behind. The 3´ sequence of each primer matches the right or left end 562 of the disruption cassette (the sequence is shown uppercase in both primers). The 563 extended resistance cassette was amplified by PCR and E. coli BW25113 [pIJ790] 564 bearing cosmid 8H10 was electro-transformed with this cassette. The isolated mutant 565 cosmid was introduced into non-methylating E. coli ET12567 containing the RP4 566 derivative pUZ8002. The mutant cosmid was then transferred to S. filipinensis by 567 intergeneric conjugation. Double cross-over exconjugants were screened for their 568 apramycin resistance followed by confirmation by PCR. A similar strategy was used for 569 the deletion of sfbR2, but using the plasmid pIJ778 containing the 570 spectinomycin/streptomycin resistance gene (aadA) and the oriT as template and the 571 primers SfbR2-Red-F and SfbR2-Red-R (Table 1). In this case, double cross-over 572 exconjugants were screened for their spectinomycin resistance. 573

Construction of plasmids for gene complementation. 574
In order to complement sfbR replacement mutant, a 1261 bp DNA fragment containing 575 the entire sfbR gene plus its promoter region was amplified by PCR with primers 576 SfbR16F and SfbR12R (Table 1) using S. filipinensis chromosomal DNA as template. 577 The PCR product was cloned into an EcoRV-cut pSETneo (47) to yield pSETneo::sfbR. 578 on November 3, 2020 by guest http://aem.asm.org/

Downloaded from
Similarly, for S. filipinensis sfbR2 gene complementation, a 1489 bp DNA 579 fragment containing sfbR2 gene plus its promoter was amplified by PCR with primers 580 SfbR2F and SfbR2REcoRI (Table 1). The PCR product was cloned into an 581 EcoRI/EcoRV-cut pSETneo to yield pSETneo::sfbR2. 582

Construction of plasmids for protein expression. 583
The SfbR gene was amplified for insertion into the GST expression vector pGEX-2T 584 using PCR. The forward primer used SbRF-GST-F introduced a unique BamHI site at the 585 5´ end of the gene, while the reverse primer SbR-GST-R carries a EcoRI site 7 586 nucleotides downstream from the TGA translational stop codon (Table 1). The amplified 587 DNA fragment was digested with BamHI and EcoRI and cloned into the same sites of 588 pGEX-2T to generate pGEX-2T::sfbR. The amplified DNA fragment was sequenced 589 from the expression vector in order to discard any mistakes introduced by the DNA 590 polymerase. Similarly, SfbR2 was amplified using the forward primer SbR2-GST-F and 591 the reverse primer Sb2R-GST-R (Table 1) in 20% glycerol at -80ºC before use. Protein elution was quantified using Bradford 604 reagent and the presence of the fusion protein was assessed by SDS-PAGE (Fig. S6). 605

DNA-protein binding assays 606
DNA binding tests were performed by electrophoretic mobility shift assay (EMSA). The 607 DNA fragments used for EMSA were amplified by PCR using the primers listed in Table  608 1 and S. filipinensis genomic DNA as template, sequenced to confirm the absence of any 609 mutations, and then labelled at both ends with digoxigenin using the DIG Oligonucleotide