Specialized and shared functions of diguanylate cyclases and phosphodiesterases in Streptomyces development

Levels of the second messenger bis-3’-5’-cyclic di-guanosinemonophosphate (c-di-GMP) determine when Streptomyces initiate sporulation to survive under adverse conditions. c-di-GMP signals are integrated into the genetic differentiation network by the regulator BldD and the sigma factor σWhiG. However, functions of the development-specific c-di-GMP diguanylate cyclases (DGCs) CdgB and CdgC, and the phosphodiesterases (PDEs) RmdA and RmdB, are poorly understood. Here, we provide biochemical evidence that the GGDEF-EAL domain protein RmdB from S. venezuelae is a monofunctional PDE that hydrolyzes c-di-GMP to 5’pGpG. Despite having an equivalent GGDEF-EAL domain arrangement, RmdA cleaves c-di-GMP to GMP and exhibits residual DGC activity. We show that an intact EAL motif is crucial for the in vivo function of both enzymes since strains expressing protein variants with an AAA motif instead of EAL are delayed in development, similar to null mutants. Global transcriptome analysis of ΔcdgB, ΔcdgC, ΔrmdA and ΔrmdB strains revealed that the c-di-GMP specified by these enzymes has a global regulatory role, with about 20 % of all S. venezuelae genes being differentially expressed in the cdgC mutant. Our data suggest that the major c-di-GMP-controlled targets determining the timing and mode of sporulation are genes involved cell division and the production of the hydrophobic sheath that covers Streptomyces aerial hyphae and spores. Altogether, this study provides a global view of the c-di-GMP-dependent genes that contribute to the hyphae-to-spores transition and sheds light on the shared and specific functions of the key enzymes involved in c-di-GMP metabolism in S. venezuelae. Importance Streptomyces are important producers of clinical antibiotics. The ability to synthesize these natural products is connected to their developmental biology, which includes a transition from filamentous cells to spores. The widespread bacterial second messenger c-di-GMP controls this complex switch and is a promising tool to improve antibiotic production. Here, we analyzed the enzymes that make and break c-di-GMP in S. venezuelae by studying the genome-wide transcriptional effects of the DGCs CdgB and CdgC and the PDEs RmdA and RmdB. We found that the c-di-GMP specified by these enzymes has a global regulatory role. However, despite shared enzymatic activities, the four c-di-GMP enzymes have specialized inputs into differentiation. Altogether, we demonstrate that altering c-di-GMP levels through the action of selected enzymes yields characteristically distinct transcriptional profiles; this can be an important consideration when modulating c-di-GMP for the purposes of natural product synthesis in Streptomyces.


50
Streptomyces are important producers of clinical antibiotics. The ability to synthesize these 51 natural products is connected to their developmental biology, which includes a transition from 52 filamentous cells to spores. The widespread bacterial second messenger c-di-GMP controls 53 this complex switch and is a promising tool to improve antibiotic production. Here, we 54 analyzed the enzymes that make and break c-di-GMP in S. venezuelae by studying the 55 genome-wide transcriptional effects of the DGCs CdgB and CdgC and the PDEs RmdA and 56 RmdB. We found that the c-di-GMP specified by these enzymes has a global regulatory role. 57 Introduction 67 Cellular levels of the bacterial second messenger bis-3´-5´-cyclic di-guanosinemonophosphate 68 are controlled by the competing activities of GGDEF-domain containing diguanylate cyclases 69 (DGCs) that produce the molecule out of GTP, and by phosphodiesterases (PDEs) that carry 70 an EAL or HD-GYP domain to degrade the second messenger (1). Multiplicity of c-di-GMP-71 turnover genes within a genome is widespread in bacteria, making it challenging to 72 understand how individual DGCs and PDEs control specific cellular responses while sharing 73 common enzymatic activities (2). For example, Escherichia coli K-12 has 12 DGCs and 13 74 PDEs. While deleting distinct DGCs and PDEs has no effect on cellular c-di-GMP levels, it 75 has drastic consequences for E. coli biofilm formation (3). In Vibrio cholerae, which 76 possesses 53 proteins with c-di-GMP-metabolizing domains, only a subset of these proteins 77 affects motility, biofilm formation, or both (4). 78 c-di-GMP is renowned for its function in guiding the transition between motility and 79 sessility in most bacteria (5). High levels of the second messenger favor the switch to 80 sessility, a process that often involves formation of self-organized, structured biofilms as a 81 stationary-phase induced survival strategy. In the non-motile streptomycetes, c-di-GMP is a 82 key factor controlling the transition between their filamentous lifestyle and spore formation. 83 However, in these bacteria, low levels of the molecule favor initiation of their sporulation 84 survival strategy. For example, overexpression of the E. coli PDE PdeH in S. venezuelae 85 induces premature, massive sporulation (6). A classical Streptomyces life cycle includes the 86 erection of hyphae into the air when the bacteria switch to their stationary growth phase, 87 followed by the morphogenesis of these aerial filaments into chains of spores. In S. 88 venezuelae, aerial mycelium formation is completely bypassed when c-di-GMP levels are too 89 low, due to PDE overexpression (6). A phenotypically identical response can be caused 90 through deleting the DGC-encoding gene cdgC -one of the 10 chromosomally-encoded 91 GGDEF/EAL/HD-GYP genes in S. venezuelae (7,8). Deletion of yet another DGC-encoding 92 gene, cdgB, also leads to precocious sporulation; however, the cdgB mutant still undergoes 93 the classical Streptomyces life cycle and forms spores on reproductive aerial hyphae like the 94 wild type. Therefore, deleting cdgB shifts sporulation timing but does not affect the principle 95 mode of spore formation, i.e. transition of aerial hyphae into chains of spores. Conversely, 96 overexpressing the S. coelicolor DGC, CdgB, causes an opposing phenotype in S. venezuelae, 97 in that it prolongs filamentous, vegetative growth (8); this process can be mimicked by 98 deleting either rmdA or rmdB, which encode functional PDEs (9) (Fig. 1). 99

Biochemical and physiological activities of the GGDEF and EAL domains of RmdA and 135
RmdB 136   137 The cytosolic RmdA and the membrane-bound RmdB are composite  proteins that are functional PDEs in S. coelicolor (9)  that arose due to rapid degradation of [ 32 P]c-di-GMP by its EAL domain (Fig. 1B). To 164 confirm that c-di-GMP production by RmdA required an intact GGDEF site, we mutagenized 165 the GGDEF to GGAAF motif and used purified MBP-RmdA GGAAF in the DGC assays. As 166 expected, neither c-di-GMP nor pGpG were detectable in the reaction containing the 167 mutagenized RmdA GGAAF protein (Fig. 1B). Altogether, these data show that RmdB from S. 168 venezuelae is a monofunctional PDE that cleaves c-di-GMP to the linear pGpG. Conversely, 169 RmdA hydrolyzes c-di-GMP to GMP via pGpG and has weak DGC activity that likely 170 remains cryptic, since the c-di-GMP generated by the GGDEF domain appears to be 171 immediately hydrolyzed by the PDE activity of the EAL domain. Such residual DGC activity 172 in tandem proteins is not uncommon and has also been reported for the GGDEF-EAL protein 173 PdeR from E. coli (15). However, we cannot exclude the possibility that, under certain 174 conditions, the DGC activity of RmdA becomes dominant over the PDE function. 175 To assess the impact of the individual GGDEF and EAL domains of RmdA and RmdB 176 on developmental control in vivo, we generated strains carrying chromosomal mutations in 177 either GGDEF or EAL active sites. The strain expressing rmdA with an AAA motif instead of 178 the EAL motif (rmdA AAA ) showed a delay in development, similar to that of the rmdA mutant 179 strain (Fig. 1C). In contrast, mutagenizing the GGDEF motif to ALLEF in the chromosomal 180 locus of rmdA (rmdA ALLEF ) had no effect on differentiation compared to wild type (Fig. 1C). 181 Similarly, a strain carrying the mutant AAA motif (in place of the EAL motif) in the EAL 182 domain of rmdB (rmdB AAA ) was delayed in development, like the rmdB null mutant. We were 183 unable to generate a strain expressing the rmdB ALLEF allele from the chromosome, so instead 184 we applied complementation analysis. We found that an rmdB allele carrying the mutagenized 185 GGAAF motif in the GGDEF site could complement the differentiation defect of the rmdB 186 mutant (Fig. 1C). These data suggest that a functional EAL domain is crucial for the in vivo agar. Hence, for the RNA-seq analyses, we harvested macrocolonies from plates that were 209 inoculated with identical numbers of spores (12 µl of 2x10 5 CFU/µl) and were grown for 30 210 hours at 30 °C. For each strain, three independent macrocolonies were pooled for RNA-211 isolation and two samples were sequenced per strain. Thus, the resulting transcriptional 212 profiles would be representative of six (combined) biological replicates. At the time of 213 harvest, wild type, ∆rmdA and ∆rmdB were at a vegetative stage of growth, while ∆cdgB and 214 ∆cdgC had already sporulated (Fig. S2). 215 We were specifically interested in genes that are known components of cascades 216 controlling differentiation (10); however, a complete table of differentially expressed genes is 217 presented in Dataset S1. Genes that exhibited a more than 2-fold (log2 >1/<-1; p<0.05) 218 increase or decrease in expression in the mutants relative to the wild type were considered as 219 significant. Impressively, in the cdgC mutant, 1458 genes exhibited significant changes in 220 transcription, with 844 genes being up-and 616 downregulated in comparison to the wild type 221 PdeR, antagonistically controlling the biosynthesis of adhesive curli-fibers, act in a precise 227 and non-global manner on few specific targets (19). 228 By comparing the transcriptomes of ∆cdgB and ∆cdgC we found only 92 upregulated 229 and 41 downregulated genes that were shared in the two DGC mutants ( Fig. 2B and C). Thus, 230 out of the 1770 genes that are in sum differentially expressed in the two mutants, only ~8 % 231 of genes overlapped. When examining the transcription profiles of the ∆rmdA and ∆rmdB 232 mutants, we found 52 upregulated genes and 51 downregulated genes that were common to 233 both strains ( Fig. 2B and C). In total, this corresponds to ~23 % of all differentially expressed 234 genes being similarly impacted by both RmdA and RmdB. This shows that despite a shared 235 enzymatic activity, the DGCs and the PDEs, respectively, control characteristic sets of genes. 236 The N-termini of CdgC and RmdB are anchored in the cell membrane, CdgB has GAF-PAS-237 PAC N-terminal sensory domains and RmdA contains PAS-PAC domains at the N-terminus 238 (7). Likely, the signals perceived by the characteristic sensory domains specify the distinct 239 functions of CdgB, CdgC, RmdA and RmdB.  (18). Unexpectedly, we found few bld and whi genes to be differentially expressed 248 in the studied mutants. In agreement with a delay in development, bldN, bldM, whiD, whiH 249 and whiI were downregulated in ∆rmdA; however, of these, only whiH was also 250 downregulated in ∆rmdB. In ∆cdgB, only whiI was upregulated at the tested time-point, while 251 in ∆cdgC, both whiI and whiD were upregulated, while bldH and bldN were downregulated 252 ( Fig. 2D and S3A). 253 The expression of whiI and whiH is directly activated by the sigma factor σ WhiG , whose 254 activity is controlled by the RsiG-(c-di-GMP) anti-sigma factor. Expression of whiI 255 completely depends on whiG, whereas whiH expression is only partially dependent on the 256 sigma factor (11). Thus, the fact that whiI was upregulated in both ∆cdgB and ∆cdgC, 257 reflected the activation of σ WhiG in the two DGC mutants. whiH and whiI were, however, both 258 downregulated in ∆rmdA; whiH was also less expressed in ∆rmdB. This collectively suggests 259 reduced activity of σ WhiG in the two PDE mutant strains. The inversely correlated transcription 260 profiles of these σ WhiG -dependent genes imply that the two DGCs and two PDEs all contribute 261 to modulating σ WhiG -activity. 262 The BldN ECF sigma factor activates the expression of the chaplin and rodlin genes, 263 which encode the hydrophobic sheath proteins that encase aerial hyphae and spores (22). 264 BldD-(c-di-GMP) directly represses bldN expression (23) (18, 21). Thus, we expected 265 increased transcription of bldN in the DGC mutants, due to loss of BldD repressive activities, 266 and reduced expression of bldN in the PDE mutants. It was, therefore, a surprise that bldN 267 expression was downregulated in the ∆cdgC strain. Because of that we set out to examine the 268 expression patterns of all known BldD-(c-di-GMP) target genes in our different mutants. Of 269 the 282 direct BldD-(c-di-GMP) targets in S. venezuelae, we found 19, 57, 27 and 8 genes to 270 be differentially expressed in ∆cdgB, ∆cdgC, ∆rmdA and ∆rmdB, respectively ( Fig. 2A). 271 These analyses revealed that, at least under the conditions tested, only a relatively 272 minor fraction of all BldD-(c-di-GMP)-targets responded to c-di-GMP changes in the studied 273 mutants. Notably, the direct BldD-regulon was determined in S. venezuelae grown in liquid 274 culture, and some of the observed differences may be explained by the fact that here, colonies 275 grown on solid medium were analyzed. However, many direct BldD-targets are co-regulated 276 by multiple transcription factors in a hierarchical manner (10) and ChpC) and five short (ChpD-H) chaplins, and these proteins are expected to self-assemble 290 into amyloid-like filaments on the cell surface, where they then permit the aerial hyphae to 291 escape the surface tension. As further components of the hydrophobic layer, S. venezuelae 292 produces three rodlin proteins (RdlA-C), which are proposed to organize the chaplin filaments 293 into so-called rodlets. Unlike the chaplins, however, the rodlins are dispensable for aerial 294 development and surface hydrophobicity (28). Moreover, when grown on rich medium, 295 Streptomyces secrete an additional surfactant peptide, SapB (product of the ramCSAB operon) 296 (29). 297 Expression of genes encoding the different hydrophobic sheath components was 298 significantly affected in the four tested mutants. As shown using RNA-seq, deleting cdgB 299 resulted in upregulation of chpD, chpF, chpG, rdlA, rdlC, ramS and ramC (Fig. 2D). In 300 addition, quantitative RT-PCR (qRT-PCR) data revealed that chpH was also upregulated in a 301 cdgB mutant (Fig. 3A). Surprisingly, our data showed that in contrast to ∆cdgB, all chaplin 302 genes, except chpB, chpD, and the three rodlin genes, rdlA-C, were downregulated in the 303 cdgC mutant (Fig. 2D), despite this strain having the same rapid sporulation phenotype as the 304 cdgB mutant. qRT-PCR data confirmed that expression of chpC, chpE and chpH was 11-fold, 305 21-fold and 11-fold, respectively, lower in ∆cdgC than in wild type (Fig. 3A). We also 306 detected a strong downregulation of the chaplin and rodlin genes in both ∆rmdA and ∆rmdB 307 strains ( Fig. 2D and 3A). 308 We tested the water repellent properties of the colony surface of the different wild type 309 and mutant strains, and found that wild type and ∆cdgB both repelled aqueous solutions (seen 310 as pearl droplet formation on the colony surface), suggesting that they possessed a 311 hydrophobic layer atop their colonies. In contrast, ∆rmdA, ∆rmdB and ∆cdgC colonies were 312 highly hydrophilic, with water droplets immediately dispersing (Fig. 3B). The observed 313 properties associated with these colony surfaces are consistent with expression of chp genes 314 in wild type and ∆cdgB, and reduced expression of the chaplin genes in ∆cdgC, ∆rmdA and 315 ∆rmdB. 316 We wondered whether chaplin overexpression could restore the inability of ∆cdgC, 317 ∆rmdA and ∆rmdB to form aerial mycelium. To test this, we introduced chpB-F and chpH, 318 under the control of the constitutive ermE* promoter, on the integrative pMS82 vector into 319 each mutant strain. Colony morphology analysis revealed that none of the overexpressed 320 chaplin genes could fully restore aerial mycelium formation to the studied mutants, when 321 overexpressed individually (Fig. S4). Presumably, fine-tuned expression of multiple chp 322 genes is needed to overcome this developmental defect (30). 323 In conclusion, our data revealed that production of the amyloid-forming chaplin and 324 rodlin proteins is controlled by c-di-GMP in S. venezuelae. This is reminiscent of many 325 bacteria, in which the synthesis of equivalent extracellular matrix components is activated by 326 c-di-GMP. For example, in E. coli, expression of csgA and csgB, encoding the main 327 components of the amyloid curli fibers, is activated by c-di-GMP (13). However, strikingly, 328 chp and rdl genes are downregulated upon deletion of the DGC cdgC, while deletion of the 329 DGC cdgB has the opposite effects, leading to upregulation of these genes. The contrasting 330 expression profile of these genes in the two DGC mutants explains the morphological 331 difference between them. Obviously, lack of a hydrophobic layer means ∆cdgC is unable to 332 break the surface tension at the air-agar interface and raise aerial hyphae, so that instead the 333 spores are formed on the upper layer of the substrate mycelium. The downregulation of chp 334 and rdl genes in ∆cdgC is likely a result of bldN downregulation in this strain ( Fig. 2D and  335 S3A), where bldN encodes an ECF sigma factor needed for expression of these genes. bldN 336 expression is governed by BldD-(c-di-GMP), while BldN activity is controlled by the 337 membrane-bound anti-sigma factor, RsbN (31). Since CdgC is associated with the membrane 338 via its transmembrane helices, it will be interesting to test whether this enzyme affects chp 339 and rdl expression through its modulation of RsbN activity. Our RNA-seq data showed that multiple genes encoding components of the cell division, cell 345 wall synthesis and chromosome segregation machineries, were upregulated upon deletion of 346 cdgC (Fig. 2D). Among these targets were ssgB, whose product is important for the assembly 347 of FtsZ rings at cell division sites (32); ssgD, encoding a protein that appears to be involved in 348 lateral cell wall synthesis; and ssgE, whose product was proposed to control the correct timing 349 of spore dissociation (33). In addition, the three Streptomyces mreB-like genes (mreB, 350 vnz35885 and mbl) and mreC were upregulated in ∆cdgC (Fig. 2D) GTPase FtsZ, which polymerizes into filaments, called Z-rings, close to the membrane and 359 recruits additional cell division proteins (37, 38). Ladder-like array of multiple FtsZ rings 360 define the future sporulation septa. In S. coelicolor, ftsZ expression is controlled by three 361 promoters (39); the same organization was observed for the ftsZ promoter region in S. 362 venezuelae (Fig. 4A). The onset of sporulation coincides with a strong upregulation of ftsZ 363 transcription, and this increased expression is crucial for sporulation septation (39). We 364 expected to detect increased ftsZ transcript levels in the cdgB and cdgC mutants that sporulate 365 precociously, but RNA-seq did not reveal significant changes in ftsZ expression in any of the 366 mutants. Since the two DGC mutant strains have already formed spores when harvested for 367 RNA-isolation from plates after 30 h of growth, we suspected that harvesting at an earlier 368 time point may have revealed changes in ftsZ transcript levels. 369 Given this, we sought to address ftsZ expression in our mutant strains using an 370 alternative approach. We introduced an ftsZ-ypet translational fusion, under the control of the 371 native ftsZ promoter on the pSS5 plasmid (40), into the F BT1 phage integration site in the wild 372 type strain, alongside the cdgB, cdgC, rmdA and rmdB mutants. After 12 h of growth in liquid 373 MYM medium, wild type and the two PDE mutant strains grew vegetatively and only weak 374 FtsZ-YPet signals were detected. In contrast, in the two DGC mutants, the ftsZ::ypet fusion 375 was highly upregulated, with abundant Z-ring ladders observed, signaling the initiation of 376 sporulation septation. In ∆cdgC, single spores were already detectable at this early stage of 377 growth (Fig. 4B). Immunoblot analysis using an anti-GFP antibody confirmed that FtsZ::YPet 378 was most abundant in ∆cdgC, and was elevated in ∆cdgB relative to the wild type. In contrast, 379 in ∆rmdA and ∆rmdB, FtsZ::YPet levels were strongly reduced when compared with wild 380 type levels (Fig. 4C). In vivo ChIP-seq analysis identified cdgA, cdgB, cdgC and cdgE as direct BldD-(c-di-GMP) 399 targets in S. venezuelae (6). For cdgB, this finding was confirmed biochemically using 400 EMSAs (23), but such confirmation had not been performed for cdgA, cdgC and cdgE. We 401 systematically tested binding of BldD to promoters of all genes coding for c-di-GMP-402 metabolizing enzymes in S. venezuelae using EMSAs. Our in vitro data confirmed that BldD 403 bound in a c-di-GMP-responsive manner to the promoter regions of cdgA, cdgC and cdgE 404 (Fig. 5A), but we did not detect any protein binding to the promoters of cdgD, cdgF, rmdA, 405 rmdB and hdgAB (data not shown). BldD binds to a pseudo-palindromic sequence, designated 406 the BldD box; such boxes were located 215, 224 and 59 bp upstream of the translational start 407 codons of cdgA, cdgC and cdgE, respectively (Fig. 5B). CdgA, CdgB and CdgC are active 408 DGCs (8,16,20). We sought to test the DGC activity for CdgE (possessing GAF-GGDEF 409 domains), and found that indeed it too had DGC activity (Fig. 5C). Intriguingly, CdgE activity 410 was subject to product inhibition, since added non-labelled c-di-GMP inhibited conversion of 411 [ 32 P]GTP into [ 32 P]c-di-GMP (Fig. 5C). 412 This regulatory feedback loop comprising BldD as c-di-GMP sensor that controls 413 expression of four active DGCs let us hypothesize that expression of cdgA, cdgB, cdgC and 414 cdgE may be altered in the analyzed DGC / PDE mutant strains. However, according to RNA-415 seq, neither transcript abundance of cdgA, nor that of cdgE, was affected at the tested time 416 point in any of the mutants (Fig. 2D). cdgC expression was reduced upon rmdA deletion, 417 while cdgB transcript levels were lower in ∆cdgC than in wild type ( Fig. 2D and S3B). 418 Deleting cdgC also resulted in downregulation of rmdA and upregulation of cdgF (Fig. 2D  419 and S3B), which codes for a PAS-PAC-GGDEF-EAL protein that contains 10 predicted 420 transmembrane helices (7). 421 Transcriptional regulation of c-di-GMP-metabolizing enzymes in S. venezuelae is 422 complex and involves the action of multiple global regulators, likely explaining why BldD 423 activity modulation due to changes in c-di-GMP levels in the tested DGC / PDE mutants was 424 not associated with significant transcriptional changes in these genes, at least under the 425 studied conditions. The four direct BldD-(c-di-GMP) targets (cdgA, cdgB, cdgC and cdgE) 426 are also directly controlled by the response regulator MtrA, which further binds to the 427 promoters of cdgF and rmdB (24). Moreover, cdgB is directly repressed by the transcription 428 factor WhiA, while cdgE is directly activated by the MerR-like regulator BldC (41, 42). Such 429 multi-layered transcriptional control of c-di-GMP synthesis and degradation suggests that 430 levels of this molecule are fine-tuned in response to disparate signal transduction cascades. 431 In addition to genes coding for c-di-GMP-turnover enzymes, we found that rshA, 432 encoding a RelA / SpoT homologue containing a conserved HD-domain for hydrolysis of the 433 alarmone (p)ppGpp (7) was downregulated in ∆cdgC and ∆rmdA (Fig. 2D). In addition, cya, 434 encoding a cAMP synthetase was upregulated in ∆cdgC, suggesting that CdgC links c-di-435 GMP-signaling to (p)ppGpp and cAMP metabolism. 436 437 Natural product genes differentially expressed in ∆cdgB and ∆cdgC 438 439 Streptomyces spore pigments are frequently aromatic polyketides that are produced by 440 enzymes encoded in the highly conserved whiE cluster. In S. coelicolor, this cluster comprises 441 an operon of seven genes (whiE-ORFI to whiE-ORFVII; sco_5320 -sco_5314) and the 442 divergently transcribed gene whiE-ORFVIII (sco_5321) (43). In S. venezuelae, the 443 homologous cluster is similarly organized and encompasses the genes vnz_33525 to 444 vnz_33490. In the cdgB and cdgC mutants, whiE-ORFI to whiE-ORFVII genes were up to 12-445 fold upregulated (Fig. 2D). 446 Since the whiE genes are developmentally regulated and expressed only in spores (43) All strains, plasmids and oligonucleotides used in this study are listed in Tables S1 and S2 in 484 the supplemental material. E. coli strains were grown in LB medium under aerobic conditions. 485 When required, LB was supplemented with 100 µg/ml ampicillin (Amp), 50 µg/ml 486 kanamycin (Kan), 50 µg/ml apramycin (Apr) or 15 µg/ml chloramphenicol (Cam). For 487 hygromycin B (Hyg) -based selection, nutrient agar (NA; Roth) or LB without NaCl (LBon) 488 were used, to which 16 µg/ml or 22 µg/ml, respectively, Hyg were added. S. venezuelae 489 strains (Table S2)  To generate rmdA ALLEF , rmdA AAA and rmdB AAA mutations on the SV3-B05 and SV2-B03 502 cosmid, respectively, recombineering using single-strand oligonucleotides (Table S1) in E. 503 coli HME68 was performed as described in (48). Prior to this, the kan-resistance cassette of 504 both cosmids was replaced by apr-oriT in E. coli BW25113/pIJ790. For that, the apr-oriT 505 sequence with neo-specific extensions was amplified by PCR from pIJ773 (Table S1 and S2). 506 Successful mutagenesis was confirmed by PCR and Sanger sequencing and the confirmed 507 mutant cosmids were transformed into E. coli ET12567/pUZ8002 for conjugation into S. 508 venezuelae, as described in (22). Conjugation plates were incubated at room temperature 509 overnight, and then overlayed with Apr. Ex-conjugants were re-streaked once on plates 510 containing Apr and nalidixic acid, and then several times on non-selective medium. The 511 desired mutants arising from a double crossing over were screened for Apr-sensitivity 512 followed by PCR to confirm the desired mutations. PCR products comprising the 513 mutagenized regions were sequenced and the resulting strains were named SVJH29 514 (rmdA ALLEF ), SVJH30 (rmdA AAA ) and SVJH31 (rmdB AAA ). 515 516

Complementation of ∆rmdB 517
For complementation analysis of ∆rmdB with rmdB GGAAF -FLAG, pIJ10170-rmdB GGAAF -518 FLAG was constructed using PCR with pSVJH02 containing rmdB-FLAG under the control 519 of the native promoter (8) as a template and the PRJH36 / PRJH37 primer pair (Table S1). 520 The resulting pSVJH03 plasmid was introduced into the phage integration site F BT1 in the 521 ∆rmdB mutant by conjugation and the strain was named SVJH4. 522 523

Immunoblot analysis 524
For detection of FtsZ-YPet, S. venezuelae strains expressing ftsZ-ypet controlled by the native 525 ftsZ promoter on the pSS5 vector (40) integrated at the F BT1 phage site, were grown in liquid 526 MYM for 12 h. Two ml were harvested, washed and homogenized in lysis buffer (20 mM 527 Tris, pH 8, 0.5 mM EDTA and cOmplete protease inhibitor cocktail tablets, EDTA-free 528 (Roche)) using a BeadBeater (Biozym; six cycles at 6,00 m/s; 30 s pulse; 60 s interval). Total 529 protein concentration was determined using the Bradford Assay (Roth) and each sample was 530 adjusted to 1 mg/ml. Fourteen µg total protein were loaded per lane and separated on a 12% 531 SDS polyacrylamide gel via electrophoresis and transferred to a polyvinylidene difluoride 532 membrane (PVDF, Roth). For immunodetection, anti-GFP antibody was used and bound 533 primary antibody was detected using anti-rabbit IgG-HRP secondary antibody following 534 visualization with the Clarity TM Western ECL Substrate (BioRad) and subsequent detection in 535 a ECL Chemocam Imager (Intas Pharmaceuticals Limited). For semi-quantitative 536 densitometric evaluation of detected FtsZ-YPet, ImageQuantTL software (GE Healthcare Life 537 Sciences) was used to calculate the amount of pixel per band in equal sized areas indicated as 538 arbitrary intensity units (AIU). Signals were normalized to FtsZ-YPet in wild type that were 539 set to 100%. 540 541

Data analysis 591
Reads were aligned to the Streptomyces venezuelae strain NRRL B-65442 genome (accession 592 no. CP018074) using Bowtie 2, with one mismatch allowed. Samtools (version 1.4.1) was 593 used for downstream coverage calculation. The number of reads per gene was obtained using 594 featureCounts (version 1.5.0-p1). The aligned reads were normalized per kilobase per million 595 (RPKM). Differentially transcribed genes were identified using DESeq2 package in R using 596 P-values <0.05 and log2 fold-change <-1 for (downregulated genes) or >1 (for upregulated 597 genes) as significance thresholds. To generate a heat map of differentially expressed 598 genes, we first grouped the targets into selected functional groups. Then we plotted the 599 RPKM normalized values of those genes if they were differentially transcribed in at least one 600 of the cdgB, cdgC, rmdA or rmdB mutants, using seaborn (version 0.9.0) in Python.   region from S. coelicolor (ftsZ P Sco) and S. venezuelae (ftsZ P Sve). The TGA stop codon from 823 ftsQ and the GTG start codon from ftsZ are shown in blue. Three ftsZ mRNA 5´ends were 824 mapped in the study by Flärdh et al., 2000 (39) and are highlighted in green (ftsZ 1P , ftsZ 2P , 825 ftsZ 3P ). Putative -10 and -35 promoter regions are underlined and marked in red. BldD-826 binding site was determined by den Hengst et al., 2010 (20)  Fourteen µg of total protein was loaded per lane (see Fig. S5 for loading control). WT free of 833 the FtsZ-YPet fusion was used as negative control. For quantification, arbitrary units (AIU) 834 were determined using ImageQuantTL. CPM: color prestained protein marker (NEB). 835