Streptomyces venezuelae has 10 chromosomally encoded and one plasmid-encoded protein involved in c-di-GMP metabolism (Table 1). Four genes encode proteins with a GGDEF domain only, five genes encode composite proteins containing a GGDEF domain and an EAL domain, and two genes encode for proteins with a HD-GYP domain. For the genes encoding the GGDEF and EAL domain proteins we decided to keep the “cdg” designation, which stands for “c-di-GMP”, as is the case for the cdgA and cdgB homologues in S. coelicolor. We suggest hdgA and hdgB (HD-GYP), respectively, for vnz24090, coding for a conserved HD-GYP domain protein, and vnz24095 encoding a degenerate HD-GYP domain protein missing the key GYP residues.

The DGC function of CdgA and CdgB and the PDE activity of RmdA and RmdB of S. coelicolor have been demonstrated experimentally (den Hengst et al., 2010; Tran et al., 2011; Hull et al., 2012). CdgC has a highly degenerate EAL domain but a conserved GGDEF domain. CdgD, CdgE, and the plasmid-encoded pCdgG contain conserved GGDEF domains and can be assigned DGC function. The composite protein CdgF may represent a bifunctional protein since it has all the key residues in both the GGDEF and EAL domains to be enzymatically active. CdgC, CdgF, and RmdB are predicted to contain transmembrane helices. CdgA, CdgB, CdgC, CdgE, CdgF, and RmdA, all contain either PAS, PAC, GAF, or a combination of these N-terminal signaling domains but their specific functions and the signals they perceive remain unknown.

When grown in liquid, S. venezuelae begins to differentiate after ca. 14 h of growth when vegetative hyphae start to fragment and spores begin to form. To determine the stages of growth that individual GGDEF-, EAL-, HD-GYP proteins are present and thus may affect development, we have analyzed their expression patterns at both the transcript and protein levels (Figure 1). Time-course microarray experiments were performed in the study by Bibb et al. (2012), to determine the transcriptional profile of S. venezuelae during submerged sporulation. Seven samples were taken at 8, 10, 12, 14, 16, 18, and 20 h of growth, spanning the complete morphological differentiation states of the bacterium (Bibb et al., 2012). We have extracted the transcription data for the chromosomally encoded c-di-GMP signaling genes from this microarray experiment and carried out quantile normalization and median polishing using the RMA method. The expression profiles show that rmdA and rmdB, coding for active PDEs, represent the most highly expressed genes in the analyzed set. Their transcripts were abundant throughout the life cycle from vegetative growth to sporulation. Similarly, cdgB and cdgD, coding for an active and a predicted DGC, respectively, were also expressed in all developmental stages. Transcripts of cdgC accumulated along the life cycle and achieved highest levels in the sporulation phase, and this is also true for hdgB, coding for a degenerate HD-GYP protein. cdgA, cdgE, cdgF, and hdgA were the least expressed genes under the conditions tested.

To examine the expression of all the c-di-GMP-metabolizing proteins, we constructed alleles expressing C-terminally FLAG-tagged protein variants under the control of their native promoters. These alleles were cloned into the integrative vector pIJ10770 (Schlimpert et al., 2017), introduced into the corresponding deletion mutants at the ΦBT1 attachment site, and analyzed by Western blotting using a monoclonal anti-FLAG antibody. In each case, about 300 nucleotides upstream of the coding region were included in the constructs, in an attempt to incorporate the entire regulatory region. However, genomic organization suggests that cdgD is very likely arranged in one operon with vnz19765; therefore, the cdgD-FLAG construct contains the entire region upstream of cdgD up to vnz19770. Similarly, rmdA appears to be in an operon with vnz33670 and hdgB is very likely in an operon with hdgA and therefore these genes plus ca. 300 bp of their upstream region were included in the rmdA-FLAG and hdgB-FLAG constructs.

Due to pronounced differences in expression levels, the total protein amount used in this analysis was adjusted for individual proteins. For most of the GGDEF/EAL/HD-GYP domain proteins, the FLAG-tagged protein levels show that – considering the total protein loaded – the DGC CdgB and the PDE RmdB represent the most abundant c-di-GMP-signaling enzymes and CdgD, CdgF, and HdgA the least abundant c-di-GMP metabolizing proteins in S. venezuelae (Figure 1B). The protein levels of CdgA and RmdA appear to be constant in all developmental stages, whereas other proteins accumulate along the life cycle. Despite high transcription of cdgD (Figure 1A), we could hardly detect CdgD-FLAG in our Western blot analysis (Figure 1B). This suggests either that we have not included a relevant regulatory region in our construct or some post-transcriptional mechanism negatively affects CdgD levels.

CdgC, CdgF, RmdB, HdgA, and HdgB are predicted to contain transmembrane helices and thus to be bound to the cell membrane (Table 1). To experimentally validate this prediction, we separated the whole cell lysate from relevant strains expressing the FLAG-tagged alleles into soluble and membrane fraction using ultracentrifugation. Using the anti-FLAG antibody and Western blotting, we detected CdgC-FLAG, CdgF-FLAG, RmdB-FLAG, HdgA-FLAG, and HdgB-FLAG only in the membrane fraction (Supplementary Figure S2), confirming their membrane-bound localization.

To determine which of the DGC(s) and PDE(s) are involved in the regulation of differentiation in S. venezuelae, we created deletion mutants for all 10 chromosomal genes encoding GGDEF/EAL/HD-GYP proteins. Macrocolony morphology analysis of mutant strains grown for up to 5 days on MYM agar at 30°C revealed that deletions of cdgB, cdgC, rmdA, and rmdB result in morphological phenotypes which deviate from the wildtype phenotype (Figure 2 and Supplementary Figure S3). Complementation of these phenotypes with the FLAG-tagged alleles shows that the FLAG-tagged proteins are functional, and that the mutations are congenic with the wildtype strain (Supplementary Figure S4). S. venezuelae colonies turn green upon spore maturation due to the synthesis of a spore specific polyketide pigment. As shown in Figure 2, the cdgB mutant became green and produced mature spores reproducibly faster than the wildtype. Coverslip impression taken after 30 h of growth showed that ΔcdgB had already produced spores organized in characteristic chains whereas the wildtype strain still mainly consisted of non-divided filaments. Furthermore, the cdgB mutant formed colonies that are thicker compared to the wildtype colony and produced elaborated radial ridges on the surface.

Deletion of cdgC resulted in the most striking macrocolony phenotype. This strain formed very flat colonies that did not turn green even after prolonged incubation and formed many radial wrinkle-like structures on the colony surface. The cdgC mutant did not produce the classical hair-like aerial hyphae but sporulated precociously, as shown by coverslip images taken after 30 h of incubation (Figure 2). However, the spores were not in chains but appeared as an irregular mass, reminiscent of the phenotype of an S. venezuelae strain that overproduces the active Escherichia coli PDE PdeH (YhjH), leading to hypersporulation (Tschowri et al., 2014). Similarly, bldD and bldO deletion mutants bypass aerial mycelium formation and form spores precociously (Tschowri et al., 2014; Bush et al., 2017). The pleiotropic phenotype of the cdgC mutant suggests that CdgC acts as one of the major determinants of c-di-GMP levels in the cell.

Deletion of either rmdA or rmdB severely inhibited S. venezuelae development, resulting in a prolonged bald morphology. The rmdA and rmdB mutant macrocolonies matured after 4 and 5 days, respectively, whereas the wildtype strain formed mature spores after just 2.5 days (Figure 2A). As expected, coverslip impression images show that, after 50 h of incubation, the rmdA and rmdB colonies consisted of pure vegetative hyphae (Figure 2B). Interestingly, the rmdB mutant needed 1 day longer than the rmdA mutant to become green, i.e., develop mature spores. The RmdA and RmdB proteins share conserved GGDEF and EAL domains but their N-terminal domains differ, with RmdA having a PAS-PAC domain and RmdB having a transmembrane domain (Table 1). The phenotypic similarity of the two mutants suggests that the functions of these two genes are not redundant, which is also consistent with their spatial segregation since RmdA is a cytosolic and RmdB a membrane-associated protein (Supplementary Figure S2).

The architecture of CdgC strongly suggests that this protein is an active DGC since it contains a highly conserved GGDEF domain but a degenerate EAL domain missing all the key residues involved in c-di-GMP cleavage. However, the enzymatic activity of CdgC has not yet been addressed in any study. To validate the predicted DGC activity of CdgC experimentally, we have purified the N-terminally His6-tagged cytosolic fraction of CdgC (ΔTM-CdgC) and tested it for in vitro DGC activity. A mutationally activated form of the C. crescentus PleD protein (PleD^∗) served as our positive control (Paul et al., 2007). We used [³²P]-GTP as substrate and separated the reactions by thin layer chromatography (TLC) (Christen et al., 2005). As shown in Figure 3A, ΔTM-CdgC produced c-di-GMP from GTP and is thus an active DGC.

We confirmed DGC activity of CdgC also genetically. For that, we mutagenized the GGDEF active site to ALLEF in the chromosomal locus of cdgC and found that the resulting strain was phenotypically indistinguishable from the strain containing the complete cdgC deletion (Figure 3B). As in the null mutant, S. venezuelae carrying the mutagenized cdgC allele did not form aerial mycelium, remained gray on MYM plates even after prolonged (5 days) incubation (Figure 3B and data not shown) and showed precocious sporulation. For complementation, we cloned cdgC under the control of its native promoter into the integrative pMS82 vector and introduced this construct by conjugation into the ΔcdgC mutant and S. venezuelae expressing CdgC with the ALLEF motif instead of GGDEF in the active site. In contrast to the empty vector pMS82, expression of cdgC from the ΦBT1 phage integration site fully complemented both mutants.

To analyze whether the developmental phenotype caused by cdgC deletion can be suppressed by a heterologous DGC, we overexpressed CdgB from S. coelicolor using pIJ10350 described in Tran et al. (2011). pIJ10350 is a derivative of the integrative pIJ10257 vector containing cdgB under control of the strong constitutive ermEp^∗ promoter (Tran et al., 2011). Our data, presented in Figure 3B show that the cdgC phenotype could be partially suppressed by overexpressing the S. coelicolor CdgB, supporting the conclusion that changes in c-di-GMP levels is the main reason behind the hypersporulation phenotype of the cdgC mutant. However, the cdgC mutant overexpressing CdgB remains white even after 5 days of incubation. This could be due to differences in c-di-GMP levels resulting from cdgB overexpression, or to a CdgC-specific contribution to the developmental program that cannot be complemented by c-di-GMP synthesis derived from the overexpression of a heterologous DGC. Altogether, these results show that CdgC is an active DGC in vitro and fulfills its biological function via the GGDEF active site, but that it may also have a secondary role that is specific to this protein.

Having learned that CdgB, CdgC, RmdA, and RmdB have crucial roles in controlling the coordinated progress of the S. venezuelae life cycle, we wondered to what extent these four proteins are conserved across Streptomyces species. To identify the core set of c-di-GMP turnover proteins and their overall distribution in the genus Streptomyces, we analyzed the pan genome of 93 Streptomyces species with completely sequenced genomes available at the NCBI database¹. The genomes used in this study, and their accession numbers are listed in Supplementary Table S2. We functionally annotated the resulting gene clusters using the COG database and then searched for the relevant COG annotations for GGDEF, EAL, and HD-GYP domains (see Materials and Methods). We also included the developmental master regulator BldD in the analysis as it directly binds c-di-GMP and thus represents a key c-di-GMP signaling component in Streptomyces. We set chromosomally encoded c-di-GMP metabolism proteins of our model organism S. venezuelae as reference and used 45% identity cutoff to designate ortholog clusters (Chaudhari et al., 2016). By using these parameters, we identified 15 different GGDEF/EAL/HD-GYP domain proteins in the Streptomyces pan genome (Figure 4 and Supplementary Table S3). The number of c-di-GMP-turnover proteins per strain varies from 5 (e.g., in S. sp CLI2509) up to 12 (e.g., in S. lincolnensis NRRL2936) with most strains having nine GGDEF/EAL/HD-GYP domain proteins.

bldD, cdgA, and cdgB are present in all 93 Streptomyces species (Figure 4), and we defined these as “core” genes. The genes encoding the PDE RmdB, the DGC CdgC, and the putative DGC CdgE are present in at least 91 of the analyzed strains and belong to the “soft-core” category of genes (see Materials and Methods; Figure 4 and Supplementary Table S3). RmdB is only absent in S. violaceoruber S21 and CdgC cannot be found in Streptomyces sp CLI2509. S. vietnamensis GIM40001 and Streptomyces sp WAC00288 lack a CdgE ortholog. The PDE RmdA is present in 82 fully sequenced Streptomyces and coexists with RmdB in 81 out of 93 species. However, S. violaceoruber S21, lacking RmdB, contains the PDE RmdA but no other EAL domain proteins, suggesting that RmdA is the major EAL-type PDE in this strain. We found HdgA orthologs in 76 and HdgB orthologs in 79 strains, respectively. Thirteen strains do not have any of the HD-GYP proteins, supporting the hypothesis that HD-GYP proteins do not fulfill a conserved function in c-di-GMP metabolism in Streptomyces.

Since we have learned that cdgB, cdgC, rmdA, and rmdB are c-di-GMP-signaling genes that affect development in S. venezuelae and are widely distributed in the genus Streptomyces, we wondered whether these genes are encoded in a conserved genetic context. As such, we calculated the bidirectional blast hit of the three genes located both downstream and upstream of bldD, cdgB, cdgC, rmdA, and rmdB and plotted the blast identity percent as heatmaps. Our data show that, apart from rmdA, the analyzed genetic regions are highly conserved (Supplementary Figure S5). The overall conservation of the genetic context of these c-di-GMP metabolism genes further supports the accuracy of our orthologue groups.

Escherichia coli BW25113 (Datsenko and Wanner, 2000) containing a λ RED plasmid, pIJ790, was grown in LB under aeration at 30°C and used to create disruption cosmids for the generation of the DGC/PDE mutant strains and to replace the kan^R marker on the PI1_C11 cosmid with the apr-oriT region amplified from the pIJ773 template plasmid (Supplementary Table S1). HME68 (Thomason et al., 2014) was used for λ RED-mediated recombineering of the PI1_C11 apr-oriT cosmid with the single-strand oligonucleotide ssOligo ssRec5187 ALLEF (Supplementary Table S1) to create the G658A/G659L/D660L mutations in the cdgC locus on the cosmid. E. coli ET12567 containing pUZ8002 (Paget et al., 1999) was grown at 37°C and was used for conjugation experiments. S. venezuelae strains (Supplementary Table S1) were grown at 30°C in maltose-yeast extract-malt extract (MYM) medium (Stuttard, 1982) containing 50% tap water and 200 μl trace element solution (Kieser et al., 2000) per 100 ml and liquid cultures were inoculated with 10⁶ CFU/ml spores. Conjugations between E. coli and S. venezuelae were carried out as previously described (Bibb et al., 2012). For S. venezuelae macrocolonies or patches 12 μl spores (2 × 10⁵ CFU) in 20% glycerol were spotted on MYM-TAP agar and incubated at 30°C for up to 7 days.

To create the cdgC G658A/G659L/D660L mutation on the PI1_C11 kan::apr-oriT cosmid, recombineering using single-strand oligonucleotide ssOligo ssRec5187 ALLEF (Supplementary Table S1), and E. coli HME68 strain was performed as described in Costantino and Court (2003). The mutagenized cosmid was confirmed by PCR and Sanger sequencing and transformed into E. coli ET12567/pUZ8002 for conjugation into S. venezuelae. Conjugation plates were incubated at room temperature overnight before overlaying with apramycin. Ex-conjugants were re-streaked once on plates containing apramycin and nalidixic acid and then several times on non-selective medium. The desired mutants arising from a double crossing over were screened for apramycin sensitivity followed by PCR to confirm the cdgC G658A/G659L/D660L mutation. A PCR product comprising the cdgC PAS-PAC-GGDEF region was sequenced and the resulting strain was named SVSN1.

For complementation analysis, cdgC and its promoter region were amplified using primers 5187-HindIII-pMS82-f and 5187-KpnI-pMS82-rev listed in Supplementary Table S1 and cloned into HindIII-KpnI cut pMS82 to create pSVOJ1. The plasmid was introduced into S. venezuelae cdgC G658A/G659L/D660L and ΔcdgC by conjugation, and the strains were named SVSN55 and SVSN3, respectively. The strain SVABK-4 resulted from the conjugation of pIJ10350 containing cdgB from S. coelicolor on the pIJ10257 vector into S. venezuelae ΔcdgC. To engineer S. venezuelae strains expressing c-di-GMP-proteins with a C-terminal, triple FLAG (MDYKDHDGDYKDHDIDYKDDDDK) affinity tag, the FLAG tag sequence was cloned into the pMS82 derivative pIJ10770 cut with XhoI and KpnI by using the 3xFLAG_XhoI_f and 3xFLAG_KpnI_r primers, creating p3xFLAG. The genes of interest including approximate 300 bp upstream of the open reading frames were then amplified using primers listed in Supplementary Table S1 and cloned in frame with the FLAG tag into the p3xFLAG vector. In the case of cdgD, rmdA and hdgB the corresponding upstream genes and their 5′ UTR were included in the constructs, due to potential operon organization. The resulting plasmids listed in Supplementary Table S1 were conjugated into the adequate c-di-GMP mutants, and the strains were subsequently used for Western blot analysis.

For FLAG-tag-based Western blot analysis, samples were taken after 10 (or 12 in the case of CdgE-FLAG and CdgF-FLAG), 14, 16, 18, and 20 h of growth. S. venezuelae wildtype served as negative control and was harvested after 20 h of incubation. After washing in 1× PBS, the cell pellets were resuspended in 1× PBS containing cOmplete Protease Inhibitor Cocktail Tablets (Roche) and lysed using a BeadBeater (Biozym; four cycles at 6.00 m/s; 30 s pulse; 60 s interval). Lysates were then centrifuged at 4500 rpm for 10 min at 4°C to remove cell debris and total protein concentration was determined using Bradford assay (Roth). Following amounts of total protein were loaded and separated on a SDS polyacrylamide gel: 5 μg for CdgB-FLAG; 10 μg for CdgE-FLAG, RmdB-FLAG and HdgB-FLAG; 15 μg for CdgA-FLAG, CdgC-FLAG, and RmdA-FLAG; 20 μg for CdgD-FLAG, CdgF-FLAG, HdgA-FLAG, and wildtype. Loading controls are shown in Supplementary Figure S1. After electrophoresis, proteins were transferred to a polyvinylidene difluoride (PVDF, Roth) membrane by electroblotting. For detection, anti-FLAG antibody (Sigma) was used. Bound primary antibody was visualized by using anti-mouse IgG-HRP (Thermo Fisher Scientific) secondary antibody and ECL chemiluminescent detection reagent (Perkin Elmer).

For protein purification, the cytosolic part of cdgC was cloned into the pET15b (Novagen) vector using primers listed in Supplementary Table S1, which results in N-terminally His6-tagged gene product. E. coli Rosetta (Novagen) containing pET15b-ΔTM-cdgC (pSVSN2) was grown in LB with 100 μg/ml ampicillin at 37°C to an OD_578nm of 0.6. Protein expression was induced by adding IPTG (50 μM final concentration) and cultivation was continued overnight at 16°C. Cells were harvested and lysed by passage through a French press, and the soluble fraction was used for native protein purification using Ni-NTA affinity chromatography according to standard protocols (Qiagen). Buffer containing cOmplete Protease Inhibitor Cocktail Tablets (Roche) and 50 mM Tris, pH 8.0; 500 mM NaCl; 20 mM imidazole; 1 mM MgCl₂; 5 mM β-mercaptoethanol; 0.1% Triton X 100 as well as 5% glycerol was used for lysis and washing steps. Prior activity assays, the purified protein was dialysed against a modified cyclase reaction buffer (25 mM Tris-HCl pH 7.5, 100 mM NaCl, 10 mM MnCl₂, 5 mM β-mercaptoethanol, 5% glycerol).

Diguanylate cyclase reactions were performed according to Christen et al. (2005) with slight modifications. Two μM of purified His6-tagged ΔTM-CdgC in cyclase reaction buffer (see above) were incubated with 4 nM [α-³²P]-GTP (Hartmann Analytic GmbH) at 30°C for 60 min. The reaction was stopped by heating to 95°C for 5 min and by mixing of 5 μl reaction with equal volume of 0.5 M EDTA pH 8.0. A reaction with purified PleD^∗, a constitutively active DGC (Paul et al., 2007) was treated identically and served as positive control. Samples were separated on Polygram CEL 300 PEI cellulose TLC plates (Macherey–Nagel). Plates were developed in 1:1.5 (v/v) saturated NH₄SO₄ and 1.5 M KH₂PO₄, pH 3.6, and exposed on a Phosphor Imaging Screen (Fuji) after drying.

Coverslips were placed on the surface of S. venezuelae colony and put on a thin agarose pad on a microscope slide for microscopy. Cells were imaged using the Leica DM2000 LED microscope at 100× magnification. Digital images were organized using ADOBE photoshop software.

Streptomyces species that had a complete genome assembly were downloaded from NCBI. A total of 93 genomes (Supplementary Table S2) were analyzed after excluding redundant genomes. Plasmids were excluded from the analysis and plasmid genbank records were removed from genbank files prior to conservation analysis. We employed the BPGA analysis tool (Chaudhari et al., 2016) for conservation and functional annotation of genes encoding c-di-GMP turnover enzymes. We set the id cutoff at 0.45. Downstream analyses were carried out using in-house python and perl scripts. Clusters containing c-di-GMP related genes were determined according to their COG annotations. The COG references used included: COG2199, COG2200, COG3706, COG3887, COG5001, COG2206, COG3437, COG3434, and COG4943. In addition, BldD was also included as it is known to conduct regulatory activity, which is modulated by c-di-GMP (Tschowri et al., 2014). The c-di-GMP clusters were classified into four categories: “core” (conserved in all 93 species), “soft core” (conserved in 88–92 species), “accessory” (conserved in 2–87 species), and “unique” (exclusively present in one species). We found that BPGA had few inconsistencies when protein ids% were around 50 or below. For the c-di-GMP metabolisms genes, NCBI blastp was used to validate the BPGA pipeline output for proteins that had 50% id or below. The modified output is highlighted in red in Supplementary Table S3. We also found some errors in the genbank files, and we highlighted those in green in Supplementary Table S3. Protein domain organization analysis was performed using SMART².

Microarray data for the wildtype strain were acquired from the ArrayExpress database, accession number E-MEXP-3612 (Bibb et al., 2012). The raw data corresponding to each of the time course experiments were quantile normalized and median polished using the RMA method (Irizarry et al., 2003). The mean normalized expression values were calculated from the biological replicates and reported as log2. The expression figure was compiled using an in-house python script.