BldD is a transcriptional regulator essential for morphological development and antibiotic production in Streptomyces coelicolor. Here we identify the BldD regulon by means of chromatin immunoprecipitation-microarray analysis (ChIP-chip). The BldD regulon encompasses ∼167 transcriptional units, of which more than 20 are known to play important roles in development (e.g. bldA, bldC, bldH/adpA, bldM, bldN, ssgA, ssgB, ftsZ, whiB, whiG, smeA-ssfA) and/or secondary metabolism (e.g. nsdA, cvn9, bldA, bldC, leuA). Strikingly, 42 BldD target genes (∼25% of the regulon) encode regulatory proteins, stressing the central, pleiotropic role of BldD. Almost all BldD binding sites identified by ChIP-chip are present in the promoters of the target genes. An exception is the tRNA gene bldA, where BldD binds within the region encoding the primary transcript, immediately downstream of the position corresponding to the processed, mature 3′ end of the tRNA. Through gene overexpression, we identified a novel BldD target gene (cdgA) that influences differentiation and antibiotic production. cdgA encodes a GGDEF domain protein, implicating c-di-GMP in the regulation of Streptomyces development. Sequence analysis of the upstream regions of the complete regulon identified a 15 bp inverted repeat that functions as a high-affinity binding site for BldD, as was shown by electrophoretic mobility shift assays and DNase I footprinting analysis. High-scoring copies of the BldD binding site were found at relevant positions in the genomes of other bacteria containing a BldD homologue, suggesting the role of BldD is conserved in sporulating actinomycetes.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Streptomyces spp. are Gram-positive soil bacteria that grow as a vegetative mycelium of branching hyphae. For dispersion, spores are formed on specialized reproductive structures called aerial hyphae, which emerge from the colony surface into the air. Genetic analysis of differentiation in Streptomyces coelicolor has identified two classes of regulatory mutants, blocked in distinct stages of differentiation. White (whi) mutants form aerial hyphae in the normal way but these hyphae are unable to complete the developmental process to form mature chains of spores (Flärdh and Buttner, 2009). They appear white when grown on solid media because they fail to produce the grey polyketide pigment associated with mature, wild-type spores. bld mutants are blocked at an earlier stage of development; they are unable to erect aerial hyphae and therefore appear ‘bald’, lacking the characteristic fuzzy morphology of the wild type (Flärdh and Buttner, 2009). Streptomycetes are renowned for their ability to produce clinically important antibiotics and other bioactive compounds, and the production of these molecules is temporally and genetically co-ordinated with the developmental programme. Thus, in addition to causing loss of aerial mycelium formation, mutations in many bld loci, including bldD, the focus of this report, pleiotropically block antibiotic production.
One of the bld genes relevant to the work described here is bldA. Unlike almost all the other developmental regulatory genes in Streptomyces, bldA does not encode a transcription factor. Instead, bldA encodes the only tRNA that can efficiently translate the leucine codon UUA, a codon rarely used in streptomycete genomes, which typically have ∼73% G+C content (Lawlor et al., 1987). In S. coelicolor, only ∼2% of the genes contain a TTA codon (Chandra and Chater, 2008) and these depend on bldA for translation (Leskiw et al., 1991). No essential genes contain a TTA codon and so bldA null mutants are viable, but they are blocked in morphological differentiation and the production of the four antibiotics made by the wild-type strain: actinorhodin, the prodigionines, methylenomycin and calcium-dependent antibiotic (CDA) (Chater and Chandra, 2008). Recently, a fifth antibiotic produced by S. coelicolor was described, originating from the cpk gene cluster (Gottelt et al., 2010). The requirement for the BldA tRNA in the production of actinorhodin and the prodigionines can be attributed to the presence of TTA codons in the genes encoding pathway-specific regulators of the corresponding antibiotic biosynthetic gene clusters, actII-ORF4 and redZ respectively. The role of BldA in differentiation is mediated, at least in large part, by the presence of a TTA codon in the global regulator BldH (also called AdpA), as changing the codon to CTC or TTG substantially suppresses the morphological defect of bldA mutants (Nguyen et al., 2003; Takano et al., 2003).
The principal focus of this work is BldD. BldD is a small (18 kDa) DNA-binding protein that exists in solution predominantly as a homodimer (Elliot et al., 2003b). BldD has two distinct domains, a C-terminal domain of unknown function and an N-terminal domain that mediates DNA binding and dimerization (Lee et al., 2007). The crystal structure of the N-terminal domain has been determined and shows similarity with lambda repressor, a major-groove binding transcriptional regulator containing a classical helix–turn–helix DNA-binding motif (Kim et al., 2006). Interestingly, BldD shows homology to Bacillus subtilis SinR, which functions as a transition-state regulator that represses sporulation and biofilm formation during vegetative growth, and promotes competence and motility functions (Bai et al., 1993; Chu et al., 2006). Like bldA mutants, S. coelicolor bldD mutants are blocked at the earliest stage of morphological differentiation and also fail to produce actinorhodin, the prodigionines, methylenomycin and CDA (Elliot et al., 1998). Expression of bldD is temporally regulated and BldD has been shown to directly repress four developmental genes, as well at its own expression, during vegetative growth by binding to sequences overlapping the promoter and/or the transcriptional start site (Elliot and Leskiw, 1999; Elliot et al., 2001; 2003b). Among the known BldD-regulated targets are two genes encoding sigma factors (σBldN and σWhiG) that, in turn, play crucial roles in Streptomyces development (Chater et al., 1989; Bibb et al., 2000; Elliot et al., 2001). It has also been suggested that BldD links differentiation to stress, as it regulates expression of a third sigma factor gene, sigH, encoding σH, involved in the regulation of stress responses (Kelemen et al., 2001).
Through the work of many laboratories, significant progress is being made in understanding the cell biological processes underlying morphogenesis in Streptomyces (Flärdh and Buttner, 2009), and all of the bld and whi developmental master regulators defined by classical mutant hunts have been cloned and characterized. However, in total, only a handful of direct targets of these regulators has been identified. As a consequence, what is strikingly needed is an understanding of the regulatory networks that connect the developmental cell biological processes to the master regulators. Here we apply global approaches to identify the complete regulon of genes targeted by BldD, which reveal unexpected connections between BldD and a multitude of well-characterized genes involved in developmentally co-ordinated processes.
Results and discussion
Characterizing the regulatory networks that govern morphological development in S. coelicolor poses certain logistical problems. S. coelicolor sporulates only on solid medium and the differentiating part of the colony (the aerial mycelium) constitutes just ∼10% of the total biomass (Flärdh and Buttner, 2009), making the application of global techniques like chromatin immunoprecipitation-microarray analysis (ChIP-chip) to development-specific transcription factors problematic. Characterizing the regulon of BldD was feasible, however, because previous work suggested that BldD functions to repress developmental genes during vegetative growth, and BldD is therefore active in vegetative cells grown in liquid culture where ChIP-chip can readily be applied. However, transcriptional profiling under these conditions has drawbacks because genes that are repressed by BldD but also activated by a developmental regulator, absent from liquid cultures, are unlikely to be differentially expressed between wild type and a bldD mutant. Finally, the highly pleiotropic role of BldD leads to a significantly reduced growth rate for bldD mutants on solid and in liquid media, with the potential to cause many indirect effects on the transcriptome. With these issues in mind, we attempted to characterize the full extent of the BldD regulon.
Effects of bldD inactivation on the S. coelicolor transcriptome
To examine the effects of BldD on genome-wide transcription, DNA microarray experiments were performed, comparing the transcriptional profile of S. coelicolor M600 (a plasmid-free derivative of the wild type) with that of a congenic ΔbldD null mutant in rapidly growing cultures in liquid-rich medium. RNA samples were prepared from both strains and, following cDNA synthesis and labelling, were hybridized to DNA microarrays. Analysis of the DNA microarray data from three independent experiments revealed that the expression of 359 genes (located in 261 putative transcription units) was changed over twofold as a result of the bldD mutation (Table S1). Of these transcription units, 118 were upregulated in the mutant, whereas 143 of them were downregulated in the mutant.
Among the genes that were most severely affected by bldD (Table 1), several are involved in primary metabolism. Expression of leuB and leuC, genes required for branched-chain amino acid biosynthesis, was increased 52-fold (see also below). Genes belonging to central pathways of primary carbon metabolism (eno2, glycolysis/gluconeogenesis; tktB, pentose phosphate pathway; sacA, citric acid cycle) were downregulated in the bldD null mutant. This is likely to be an indirect consequence of the slow growth rate of the ΔbldD strain, which is apparent from its reduced colony size, compared with the wild-type, when grown on plates (Elliot et al., 2003b). Reduced expression of several ABC transport systems and amino acid permeases could also reflect the slow growth rate of the mutant. The genes encoding putative 50S and 30S ribosomal proteins, rpmEGB and rpsN, were strongly upregulated in the bldD mutant, which could again be an indirect effect, indicative of a global change in physiology associated with bldD disruption.
Table 1. Transcriptome comparison of S. coelicolor M600 ΔbldD and S. coelicolor M600.
One of the established roles of BldD in morphological differentiation is to repress bldN expression during vegetative growth (Elliot et al., 2001). bldN encodes a developmentally regulated σ factor, σBldN, required for aerial mycelium formation (Bibb et al., 2000). Consistent with this, the expression of bldN was 25-fold derepressed in the bldD mutant (Table 1). As part of the developmental cascade, σBldN in turn directly activates expression of bldM, which encodes a response regulator also required for aerial mycelium formation (Bibb et al., 2000; Molle and Buttner, 2000). bldM expression was 10-fold derepressed in the bldD mutant (Table 1), probably reflecting a combination of increased σBldN expression and the fact that bldM, like bldN, is directly repressed by BldD (see below). Expression of the prsH-sigH operon, encoding an anti-sigma factor/sigma factor pair directly regulated by BldD, was increased sixfold in the mutant. Differential expression of two other previously identified targets of BldD (whiG and bdtA) could not be detected, however.
The chaplins are a family of eight hydrophobic cell wall-associated proteins that confer hydrophobicity on aerial hyphae and spores, and constructed strains lacking most or all of the chaplin (chp) genes fail to form aerial hyphae on most media (Claessen et al., 2003; Elliot et al., 2003a). The chpE and chpH genes are strongly derepressed in the bldD mutant (Table 1 and Table S1). Of the chaplin genes, expression of chpA, chpB, chpC, chpD, chpF and chpG is strictly correlated with formation of aerial hyphae, whereas chpE and chpH are expressed in submerged hyphae as well (Claessen et al., 2003), perhaps explaining why only the latter two were upregulated in the bldD mutant under the conditions of our transcriptome analysis.
The BldD regulon contains ∼167 genes, including 42 genes that encode DNA-binding proteins
To distinguish between indirect and direct effects, BldD ChIP-chip experiments were performed. A culture of S. coelicolor M600 grown under the same conditions as for the transcriptional profiling experiments was subjected to formaldehyde cross-linking and sonication as described in Experimental procedures. After immunoprecipitation using a BldD-specific polyclonal antibody, the DNA was labelled with Cy3 and hybridized together with a total-DNA control, labelled with Cy5, to a high-density microarray representing the S. coelicolor genome. Figure 1 presents representative results from three independent biological experiments. Peaks in the fluorescence ratio indicate regions of the chromosome that are bound by BldD. As a control, a ChIP-chip experiment was performed using the congenic bldD null mutant to eliminate any signals arising from cross-reaction of the antibody with other transcription factors. A total of ∼172 peaks were detected throughout the chromosome in the wild-type strain. Five of these peaks were also found in the bldD null mutant control experiment and these were removed from the data set, leaving ∼167 putative BldD binding sites, of which 29 of the downstream genes were identified in the transcriptome analysis described above. BldD binding sites were found across the genome but the density of sites was greater in the core essential genome and less in the two arms of the linear chromosome (Fig. 1A), regions that can be lost without affecting viability under laboratory conditions and that contain a preponderance of conditionally adaptive genes (Bentley et al., 2002).
Strikingly, 42 of the BldD target genes (∼25% of the regulon) encode regulatory proteins themselves, stressing the central, highly pleiotropic role of BldD. The BldD targets include genes involved in morphological differentiation, energy storage, cell division, cell wall modification, signalling pathways, antibiotic production, and proteolytic and regulatory functions. Of the BldD targets, ∼19% encode proteins of unknown function.
BldD targets involved in morphological differentiation
All of the five previously identified BldD-regulated genes (bldN, sigH, bdtA, whiG and bldD) were among the enriched targets. In addition, strong signals for BldD binding were found in the upstream regions of 16 other genes or operons that are known to be involved in morphological differentiation and/or antibiotic production (Table 2 and Fig. 1B). Among these are the genes ftsZ, ssgA, ssgB and smeA-ssfA. Each of these genes is known to be under developmental control, but none of the direct regulators involved has previously been identified, except that ssgA is activated by SsgR (Traag et al., 2004). Sporogenic aerial hyphae undergo a synchronous round of cell division, initiated by the polymerization of a ladder of 50 or more FtsZ rings, and FtsZ ring placement is strongly influenced by SsgA and SsgB (Schwedock et al., 1997; Flärdh et al., 2000; van Wezel et al., 2000; Keijser et al., 2003; Grantcharova et al., 2005; Noens et al., 2005). The smeA-sffA operon encodes a DNA translocase (SffA), involved in chromosome segregation into spores, which is specifically targeted to sporulation septa by the small membrane protein SmeA (Ausmees et al., 2007). BldD also binds near SCO2525, a recently identified gene that encodes a putative methyltransferase that has effects on several aspects of colony growth (Gehring et al., 2004).
Table 2. BldD binding sites upstream of genes involved in sporulation and/or antibiotic production.
Our analysis also identified bldC as a member of the BldD regulon. bldC, which is required for normal morphological differentiation and antibiotic production in S. coelicolor (Hunt et al., 2005), encodes a member of a family of small DNA-binding proteins that are related to the DNA-binding domains of the MerR family of transcriptional activators. The S. coelicolor genome carries one homologue of bldC, bdtA (Hunt et al., 2005), which is also a target for BldD (Elliot et al., 2001). We constructed a bdtA deletion mutant, but it did not display any apparent phenotype (data not shown).
BldD binds the region lying upstream of bldM (Table 2), which encodes a response regulator required for aerial mycelium formation (Molle and Buttner, 2000). bldM transcription was 10-fold derepressed in the bldD mutant, suggesting that BldD functions to repress bldM expression during vegetative growth. Consistent with this, a bioinformatically predicted BldD binding site (see below) was found overlapping the −10 region of the bldMp2 promoter (Table 2). BldD also binds the upstream region of whiB (Table 2), which is required for the early stages of sporulation in aerial hyphae (Davis and Chater, 1992). Although the role of the actinomycete-specific WhiB-like (Wbl) family of proteins has been controversial (den Hengst and Buttner, 2008), recent biochemical experiments suggest they function as transcription factors (Guo et al., 2009; Singh et al., 2009).
σF is required for spore maturation in Streptomyces (Potuckova et al., 1995; Sun et al., 1999). One of the strongest BldD targets, both in our ChIP-chip and in our transcriptome analysis, is SCO4677, which encodes a protein that was recently found to interact with σF as a potential antagonistic regulator (anti-sigma) (Kim et al., 2008b). In turn, the SCO4677 anti-σF interacts with SCO0869 and SCO0781, which have similarity with anti-anti-sigma factors such as SpoIIAA of B. subtilis (Kim et al., 2008b). Thus, although BldD does not directly regulate expression of sigF, their regulons are connected through SCO4677.
BldD targets involved in antibiotic production
BldD is essential for the production of secondary metabolites in different species of actinomycetes (Elliot et al., 1998; Chng et al., 2008). Our ChIP-chip analysis pinpointed several genes and operons, in addition to bldC discussed above, that are known to influence antibiotic production in S. coelicolor (Table 2 and Fig. 1B). A strong ChIP-chip signal was detected upstream of the five-gene ‘conservon’, cvn9, disruption of which causes precocious aerial mycelium formation and conditional overproduction of actinorhodin (Komatsu et al., 2006). The cvn9 operon is likely to encode a membrane-associated heterocomplex, resembling the eukaryotic G-protein-coupled receptor system, and is therefore likely to function in signal transduction (Komatsu et al., 2006). BldD binds to the upstream region of nsdA, a gene that negatively influences antibiotic production in multiple species of actinomycetes. Disruption of nsdA in S. coelicolor causes hypersporulation and overproduction of the antibiotics actinorhodin, methylenomycin and CDA (Li et al., 2006; Wang et al., 2009). The genes leuB and leuC, required for branched-chain amino acid biosynthesis, are strongly derepressed in the bldD mutant and BldD binds upstream of leuA. Recently it was shown that 50% of the acetate for synthesis of the type II polyketide actinorhodin is derived from the catabolism of branched amino acids (Sprusansky et al., 2005; Stirrett et al., 2009). Additionally, antibiotic production might be influenced by BldD through regulation of the mobilization of triacylglycerol (TAG) during the stationary phase of growth in submerged liquid cultures. TAG has been proposed to serve as a carbon source for antibiotic biosynthesis in S. coelicolor (Olukoshi and Packter, 1994; Packter and Olukoshi, 1995). The product of SCO0958, which carries out the esterification of diacylglycerol with a fatty acid molecule, is largely responsible for TAG production (Arabolaza et al., 2008) and SCO0958 showed a strong signal in the BldD ChIP-chip experiment.
BldD targets involved in polysaccharide metabolism
Besides neutral lipid storage compounds such as TAG, S. coelicolor synthesizes polysaccharides, the interconversion of which may also be influenced by BldD. BldD targets the region between pepA and pepA2 (SCO7336 and SCO7337), the most upstream genes of two divergent operons (glgBII locus) that are involved in the deposition of glycogen as carbon reserves in the apical compartments of the aerial mycelium (Schneider et al., 2000; Yeo and Chater, 2005). BldD could also influence utilization of cellulose as it might bind to the intergenic region of the divergently transcribed genes SCO6546 and SCO6548, encoding secreted cellulases, and to the upstream region of msiK, a gene required for growth on minimal medium containing cellulose or other polysaccharides (Schlosser et al., 1999; Saito et al., 2008). Recently, an important additional role for cellulose in S. coelicolor has been established. It was shown that extracellular cellulose fibrils mediate the attachment of hyphae to hydrophobic surfaces, and that the enzyme responsible for its production, cellulose synthase, localizes to the tip of growing hyphae and interacts with DivIVA, the essential cytoskeletal protein that directs apical growth in Streptomyces (Hempel et al., 2008; Xu et al., 2008; de Jong et al., 2009b).
Overexpression of a novel BldD target gene implicates c-di-GMP in morphological differentiation and antibiotic production
Cyclic-di-GMP (c-di-GMP) has been shown to be a second messenger primarily involved in the developmental switch to biofilm formation in a wide variety of bacteria (Jenal and Malone, 2006). c-di-GMP levels are controlled through enzymes containing GGDEF domains, needed for synthesis, and EAL domains, needed for hydrolysis of the molecule. In Streptomyces, a role for c-di-GMP has not been established, but the S. coelicolor genome encodes several proteins with GGDEF and/or EAL domains. Interestingly, the upstream regions of three of these genes (SCO2817, SCO5511 and SCO4281) are bound by BldD. Although expression of SCO2817, here designated cdgA (c-di-GMP), is unaffected by BldD in liquid cultures (this work), transcriptome analysis of wild-type S. coelicolor cells, grown on solid medium, showed that cdgA transcription is developmentally regulated (Hesketh et al., 2007). We overexpressed cdgA in S. coelicolor M600, and found that the resulting strain was unable to form aerial hyphae (the bald phenotype) (Fig. 2A). Moreover, this strain displayed reduced production of the blue-pigmented antibiotic actinorhodin.
To address whether overexpression of cdgA is associated with an increase in c-di-GMP levels, site-directed mutagenesis was carried out. First, we repeated the cdgA overexpression experiment with a C-terminally FLAG-tagged version of the wild-type allele and obtained exactly the same results, showing that the FLAG tag does not interfere with CdgA activity (Fig. 2A). Then we introduced a GG to AA mutation into the GGDEF motif of the FLAG-tagged cdgA overexpression construct and introduced it into S. coelicolor. This mutation has been shown to abolish the c-di-GMP biosynthetic activity of several diguanylate cyclases in other organisms (e.g. Paul et al., 2004). The substitutions abolished the bald phenotype caused by overexpression of the wild-type cdgA allele and restored normal levels of production of the blue-pigmented antibiotic actinorhodin (Fig. 2A). Immunoblot analysis, using anti-FLAG antibody, showed that the levels of CdgA protein produced in the strains expressing the wild-type allele and the GG to AA mutant allele were the same, showing that the GG to AA mutation did not cause CdgA instability (Fig. 2B). From this we conclude that the block to aerial mycelium formation caused by CdgA overexpression results from c-di-GMP synthesis.
cdgA is a predicted 67.2 kDa cytoplasmic protein that contains a PAS domain in addition to its GGDEF diguanylate cyclase and EAL domains. PAS domains are often associated with GGDEF and EAL domain proteins and can be involved in signal transduction, because of its ability to bind a wide variety of cofactors that can modulate the catalytic activity of the protein in response to environmental stimuli. The data presented here suggest for the first time a role for c-di-GMP in Streptomyces development, and the involvement of BldD in the regulation of c-di-GMP metabolism.
Identification of a consensus BldD binding site
The oligonucleotides on the DNA microarrays used in this study have a dense coverage of the genome, but do not precisely define the location of each of the BldD binding sites. To validate the ChIP-chip data, and to establish a consensus BldD operator site, a computational analysis of the targeted DNA regions was performed. Defining the exact position of the BldD binding site with respect to the transcriptional start site could also help to explain the role of BldD as a transcriptional repressor or activator. Previously, BldD binding to the promoters of bldN, sigH, bdtA, whiG and bldD itself was examined by DNase I footprinting analysis (Elliot and Leskiw, 1999; Elliot et al., 2001; Kelemen et al., 2001). These experiments clearly showed that the BldD binding sites overlap the promoters or the transcriptional start sites, explaining how BldD functions as a repressor of these genes. Alignment of the BldD-bound sequences identified an imperfect inverted repeat sequence, AGTgA N(m) TCACc, that might function as the consensus BldD binding site (Elliot et al., 2001). This sequence has similarity with a motif, [TA]GTGAN(18,20)TN(2)C, that was found using the SIGffRid algorithm, employed to compare sequences from related genomes (Touzain et al., 2008). However, both of these proposed motifs implied the length of the spacer separating the inverted repeat sequences was variable. The ChIP-chip experiments conducted here identified 162 potential new direct targets of BldD, creating a much larger data set that allowed us to identify a consensus BldD binding site.
Because operator sites of regulatory proteins in bacteria are usually located close to their target promoters, fragments of up to 200 bp were chosen such that they encompassed the known or predicted promoter sequences of their cognate genes. The data set, comprised of all genes directly targeted by BldD, was examined using the MEME algorithm (Bailey and Elkan, 1994). Over 80% of the DNA sequences contains a well-conserved copy of a 15 bp palindromic sequence, 5′-nTnACnC(A/T)GnGTnAn-3′, for which the sequence logo is shown in Fig. 3. To validate this sequence (designated the BldD box), and to confirm and extend the ChIP-chip analysis, DNase I footprinting analysis using purified histidine-tagged BldD was carried out on the intergenic region between pepA and pepA2, involved in glycogen branching (Schneider et al., 2000), which was 8.5-fold enriched in the ChIP-chip experiment (Fig. 1 and Table 2). Addition of BldD protein resulted in protection of a region containing a well-conserved copy of the BldD box (Fig. 4 and Table 2), consistent with this sequence serving as a high-affinity binding site for BldD. Similarly, DNase I footprinting showed that BldD protected a region containing a predicted BldD box upstream of cdgA and of ftsZ (Fig. 4).
Using the new BldD-binding consensus sequence, it can be explained why mutagenesis of the bldD promoter in a previous study (Elliot and Leskiw, 1999) did not have a clear effect on BldD binding, since the introduced mutations were located just outside the BldD box identified here, or at position 3 (Fig. 3), which is not conserved. In some cases (e.g. SCO5887, whiG, bldN, and bldD itself), two copies of the BldD box could be identified bioinformatically within a single promoter region (Table 2). For whiG, bldN, and bldD, these predicted binding sites correlate well with DNase I footprinting analyses (Elliot and Leskiw, 1999; Elliot et al., 2001), where, in each case, two separate BldD binding sites were found, and with our in vivo data, where very strong ChIP-chip signals for these targets were found, suggesting particularly tight binding by BldD (Fig. 1). Stringent regulation of the regulons of σBldN and σWhiG by BldD also becomes apparent from the fact that their downstream targets, bldM and the glgBII locus (Bibb et al., 2000; Yeo and Chater, 2005), respectively, are also targets of BldD (Fig. 1), although direct regulation of glgBII by σWhiG has not been verified experimentally.
To test the predictive power of the derived consensus sequence, the genome of S. coelicolor was searched for the occurrence of BldD boxes using Virtual Footprint (Munch et al., 2005). When only the intergenic regions were scanned, hundreds of well-conserved BldD binding sites were detected (Table S3). These were ranked according to their similarity with the consensus. Obviously, high-scoring copies were found in the promoter sequences that were used to build the weight matrix. However, the BldD box was also found upstream of genes that were not identified as BldD targets in the ChIP-chip experiment, and these might constitute additional targets for BldD. This could happen if they were subject to complex regulation involving other transcription factors; such promoters might appear as BldD targets if the ChIP-chip experiments were performed under different growth conditions.
Strikingly, when the entire genome was scanned, hundreds of BldD boxes were found within coding regions, although these intragenic sites were not bound by BldD in vivo, as determined by our ChIP-chip experiment. Topological constraints of the DNA, or the presence of nucleoid proteins (McArthur and Bibb, 2006), might prevent BldD from interacting with predicted binding sites present within ORFs.
Conversely, a well-conserved BldD box could not be distinguished, by the bioinformatic tool used in this study, in the upstream regions of approximately 20% of the target genes identified in the ChIP-chip experiment. We cannot therefore rule out the possibility that other sequences play a role in the recognition and regulation of these genes by BldD. Alternatively, it could be that interaction with other DNA-binding proteins is important for recruitment of BldD to these promoters.
As judged by ChIP-chip analysis, BldD binding is restricted to promoter regions in all but three cases. Two of these exceptions correspond to the BldD binding sites in the intergenic regions downstream of the convergently transcribed genes SCO2445/SCO2446, and the convergently transcribed genes SCO5809/SCO5810. The third exception is BldD binding to the bldA gene.
The bldA tRNA gene is a BldD target
A very strong ChIP-chip signal was detected for bldA (Fig. 1) and bldA contains one of the bioinformatically highest scoring copies of a BldD box (Table 2), suggesting a good correlation between the level of sequence conservation in a BldD box and the affinity of BldD for its target. Electrophoretic mobility shift assays (EMSAs) demonstrated binding of BldD to a DNA fragment spanning the bioinformatically identified BldD box in bldA (Fig. 4B), and deletion of the BldD box from the fragment abolished BldD binding (Fig. 4B). This result was extended by DNase I footprinting, showing that this region in bldA indeed serves as a high-affinity binding site for BldD (Fig. 5B). Intriguingly, the BldD box in bldA is not present in the promoter but within a region encoding the primary transcript of bldA, immediately downstream of the position corresponding to the site where the primary transcript would be processed to give rise to the mature 3′ end of the tRNA (Fig. 4B).
The regulatory significance of BldD binding to bldA will be the subject of future investigation. It is not excluded that BldD binding could influence transcription initiation at the bldA promoter. However, because of the location of the BldD binding site, another speculative possibility is that BldD affects transcriptional termination, and thus influences the post-transcriptional processing of the bldA primary transcript. Regardless of mechanism, the fact that BldD targets bldA has important implications for our understanding of the regulatory network that controls Streptomyces development. Ever since bldA was first shown to encode a tRNA, there has been extensive debate as to whether BldA constitutes a true regulatory device. In other words, although it is apparent that bldA null mutants cannot differentiate, it has never been clear if BldA availability regulates (limits) translation of TTA-containing genes during development of wild-type S. coelicolor (Leskiw et al., 1991; 1993; Gramajo et al., 1993; Rebets et al., 2006). The fact that bldA is a BldD target suggests that BldA is integrated into the regulatory network that controls Streptomyces development, consistent with BldA tRNA functioning as a true regulatory device. This suggestion is further supported by the evident integration of the BldD, BldH and BldA regulons.
The regulons of BldD, BldH and the BldA tRNA overlap
Conserved BldD binding sites were detected upstream of 11 genes containing a TTA codon (Fig. 5A). As only 146 out of the 7825 (< 2%) predicted genes in the S. coelicolor genome contain a TTA codon, TTA genes are overrepresented in the BldD regulon (11 out of ∼167 targets). In both Streptomyces griseus and S. coelicolor the bldA‘regulon’ is extended through AdpA (also called BldH in S. coelicolor), a global regulator containing a TTA codon (Nguyen et al., 2003; Takano et al., 2003; Kim et al., 2005; Ohnishi et al., 2005). In S. griseus, AdpA directly regulates at least 37 transcriptional units, several of which are involved in morphological development and secondary metabolism (Yamazaki et al., 2003; 2004; Ohnishi et al., 2005; Akanuma et al., 2009). Among the S. coelicolor BldD-regulated genes (Table 2) is the S. coelicolor adpA orthologue (bldH) itself, as was confirmed by DNase I footprinting (Fig. 5B). These results suggest that BldD is intertwined with the BldH regulon, both directly, by regulating transcription of bldH, and indirectly, by regulating bldA, on which BldH depends for translation. In addition to these genes, BldD binds upstream of the BldH targets SCO5821, a SprU-like serine protease, and SCO0762, encoding STI (Table 2), as was also confirmed by DNase I footprinting (Fig. 5B). STI is a protease inhibitor that forms part of a complex cascade (Kim et al., 2008a; Chater et al., 2010) involving multiple extracellular proteases (Fig. 5A). STI activity, and its subsequent degradation by an unidentified protease, are important for differentiation, at least under certain conditions (Kato et al., 2005; Kim et al., 2008a).
Expression of sti is significantly downregulated in a chp null mutant that forms few aerial structures (de Jong et al., 2009a) as is SCO1978, encoding a predicted secreted hydrolase, which was enriched in the BldD ChIP-chip experiment. In addition, ChIP-chip peaks and conserved BldD binding sites were detected upstream of several other genes encoding putatively secreted proteins (SCO2818, SCO3268 and SCO4132) and putative intra- and extracellular proteases (SCO3373, SCO3540, SCO0752 and SCO7226). The intracellular signals and regulatory pathways leading to the initiation of extracellular events involved in developmental and catabolic processes are largely unknown (Chater et al., 2010), but our results strongly implicate BldD in regulating the composition of the extracellular proteome.
The S. coelicolor BldD binding site is conserved among sporulating actinomycetes
Although the BldH/AdpA regulons in S. coelicolor and S. griseus overlap significantly, there are strain-specific differences. For example, in contrast to S. griseus, S. coelicolor BldH does not depend on the bacterial hormone A-factor for its expression to become derepressed (Takano et al., 2003). S. griseus contains an orthologue of BldD (Ohnishi et al., 2008), but, thus far, only one target gene has been identified (Ueda et al., 2005). In this organism, BldD targets ramCSAB, a gene cluster required for the formation of AmfS, which acts as an extracellular surfactant peptide that stimulates aerial growth. In contrast, the orthologous system in S. coelicolor, ramCSAB, encoding the SapB surfactant peptide, is probably not targeted by BldD (Ueda et al., 2005). We did not identify ramCSAB by ChIP-chip with BldD, and sequence inspection of its regulatory region revealed no apparent BldD box.
Using blast, we found that BldD is present throughout the sporulating actinomycetes, but also in the actinomycete Acidothermus, which is not known to form spores, and species of Kribella, Nakamurella and Stackebrandtia (Table S2). S. coelicolor BldD is 97–98% identical to BldD from other sequenced Streptomyces species, and more distantly related to orthologues from other sporulating actinomycete genera (e.g. 77% identical to Saccharopolyspora erythraea BldD). To assess the extent to which the BldD regulatory network is conserved in sporulating actinomycetes, several available genomes were searched for potential BldD binding sites, using the weight matrix obtained from the S. coelicolor data set (Fig. 3). The results of these searches are available through Table S3. Among the highest-scoring hits in S. griseus were sites located upstream of genes orthologous to the S. coelicolor developmental loci whiB, bldH, bldM, ftsZ, ssgA, rarA, bldA and bldD itself, all of which were identified as BldD targets in the current study. Potential BldD binding sites in S. griseus were also found upstream of the cytoskeletal protein gene mreB, and sprV, encoding a putative trypsin-like protease. In Streptomyces avermitilis, conserved binding sites were found upstream of, among others, minD2, leuA, ftsZ, whiG, sprB, ssgA, wblB, cvnA5, bldK and whiP. whiP is the orthologue of S. coelicolor crgA, important for co-ordinating growth and cell division in sporogenic hyphae (Del Sol et al., 2006). Similarly, high-scoring copies of the BldD binding site were found at relevant positions in Salinispora tropica and Frankia. These data suggest the function of BldD is conserved in sporulating actinomycetes.
We also searched the genome of S. erythraea. A constructed bldD null mutant of S. erythraea has a bald phenotype and fails to make erythromycin (Chng et al., 2008). The erythromycin (ery) biosynthetic cluster lacks a pathway-specific regulatory gene and in vitro DNA-binding assays suggest that BldD binds to five promoter regions in the ery cluster (Chng et al., 2008). Given that a S. erythraea bldD mutant fails to make erythromycin, these results were interpreted to imply that BldD functions as an activator of the ery gene cluster (Chng et al., 2008). However, the only DNase I footprinting analysis performed, on eryBVI, showed that BldD protects a region covering the −35, the −10 and the transcription start site (Chng et al., 2008), which is perhaps more consistent with BldD acting as a repressor of eryBVI. This is strengthened by the presence of a BldD box (TTTCCCCGCCGTGAC) located between the −35 and −10 sequences of the eryBVI promoter. We would like to raise the speculative possibility that BldD functions to repress the ery cluster during vegetative growth and the inability of S. erythraea bldD mutants to make erythromycin arises indirectly.
For those BldD targets whose transcriptional start sites have been determined experimentally in S. coelicolor, we examined the positions of the predicted BldD binding sites with respect to the promoters. In most cases, BldD binding sites overlapped with the promoter elements or transcriptional start sites, consistent with BldD acting as a repressor by preventing RNA polymerase from binding DNA and/or initiating transcription. In contrast, in the case of sti, we found that BldD binds to a BldD box centred 63 bp upstream of the transcriptional start site (Fig. 5), suggesting that BldD might function as activator of sti. Consistent with this suggestion, sti expression was 41-fold downregulated in the bldD mutant (Table 1). This gene is co-regulated by BldH (AdpA), for which predicted binding sites are located 84 bp and 159 bp upstream of the transcriptional start site (Kato et al., 2005). Therefore, activation of sti might result from interplay between BldH, BldD and the sti promoter, but such a model is further complicated by the fact that the bldH is also a BldD target.
BldD acts as a repressor of key developmental genes during vegetative growth
Our data suggest that the main function of BldD is to repress, during vegetative growth, genes that are needed for morphological differentiation and antibiotic production. BldD transcript levels are highest during this stage of the S. coelicolor life cycle (Elliot et al., 1998), and immunoblot analysis of BldD protein during development of S. coelicolor on solid medium shows that BldD protein levels correlate with the bldD transcription pattern (Fig. 6). BldD protein is present predominantly during vegetative growth, although some BldD could still be detected even after onset of aerial mycelium formation and initiation of antibiotic production (Fig. 6).
The proposed role of BldD in Streptomyces is reminiscent of the role the B. subtilis transition state regulator AbrB, which functions to prevent premature expression of genes required for developmental processes during vegetative growth (Strauch, 1993). Like BldD, expression of abrB decreases during entry into stationary phase, but this is not a complete explanation of how AbrB targets become derepressed. In addition, AbrB is inactivated by AbrA, which acts as an anti-repressor, interacting with AbrB to disable DNA binding (Banse et al., 2008). Transcription factor inactivation has also been shown for another B. subtilis transition state regulator, SinR, which controls biofilm formation. SinR is antagonized through hetero-dimerization with an anti-repressor, SinI (Bai et al., 1993). Interestingly, the N-terminal domain of BldD has similarity with the Xre family of transcriptional repressors, which includes SinR (Kelemen et al., 2001). However, our attempts to identify a BldD partner protein have, thus far, been unsuccessful.
In summary, we show that BldD is a key regulator of morphological differentiation and antibiotic production and that it connects the regulons of several other regulators that play pivotal roles in these two central aspects of Streptomyces biology (Fig. 7). Our ChIP-chip approach allowed the identification of many new candidate developmental genes. Overexpression of one of these genes (cdgA) influences both differentiation and antibiotic production and implies a role for c-di-GMP in Streptomyces development. Given the conservation of BldD and the presence of predicted BldD binding sites at relevant positions in other genomes, it is likely that BldD plays a similarly important role across sporulating actinomycetes.
Bacterial strains, plasmids, oligonucleotides and growth conditions
Strains, plasmids and oligonucleotides, used in this study, are described in Table 3. Escherichia coli strains were grown in Luria–Bertani medium for routine purposes. Streptomyces were grown on R2YE or R5 solid media (Kieser et al., 2000). Strains used for transcriptional profiling, and ChIP-chip analysis, were grown in a 1:1 mixture of YEME and TSB liquid media (Kieser et al., 2000) at 30°C and 250 r.p.m. For immunoblot detection, suspensions of 6 × 108 spores from wild-type or bldD null mutant strains were grown on R2YE solid medium, overlain with sterile cellophane, and incubated at 30°C. Samples of mycelium were harvested after 14, 17, 20, 24 and 38 h, by using a spatula, and immediately flash-frozen in liquid nitrogen.
For overproduction in S. coelicolor, SCO2817 was amplified by PCR using oligonucleotides SCO2817OVERFW and SCO2817OVERRV, carrying NdeI and HindIII restriction sites respectively. The resulting fragments were first cloned into SmaI-cut PUC19, and transferred to NdeI–HindIII-cut pIJ10257 to generate pIJ10350. The plasmid was introduced into E. coli ET12567 (pUZ8002) to allow transfer into S. coelicolor M600 by conjugation (Kieser et al., 2000). Similarly, a strain containing a C-terminally FLAG-tagged version of the wild-type allele (pIJ10354) was obtained using oligonucleotides SCO2817OVERFW and SCO2817OVERFLAGRV. The QuikChange® Site-Directed Mutagenesis Kit (Stratagene) was used to replace the GGDEF motif in cdgA by a AADEF sequence to generate pIJ10366.
Frozen mycelium was ground in liquid nitrogen with a pre-cooled mortar and pestle and resuspended in lysis buffer (0.005% KH2PO4, 0.37% CaCl2, 0.57% TES) containing 2 mg ml−1 lysozyme (Sigma). Samples were incubated for 1 h on ice and sonicated at 10 microns for four cycles of 15 s with 1 min intervals. Cell debris was removed by centrifugation at 14 000 r.p.m. for 15 min at 4°C, after which the protein concentration of the cleared extract was determined with Bradford reagent (Bio-Rad). Equal amounts (25 µg) of protein from each sample were loaded onto a 15% polyacrylamide gel and, after electrophoresis, transferred to a Hybond-C Extra nylon membrane (Amersham Pharmacia Biotech), and probed with a 1:10 000 dilution of anti-BldD antibody that was raised against His6-tagged BldD in rabbits and affinity purified. BldD was detected by chemiluminescence with ECL Western blotting detection reagents following the manufacturer's instruction (Amersham Pharmacia Biotech) using horseradish peroxidase-coupled secondary antibody. Similarly, FLAG–CdgA fusion proteins were detected in liquid-grown cells using ANTI-FLAG® antibodies (Sigma) that were raised in rabbits.
RNA isolation and DNA microarray analysis
RNA isolation and DNA microarray analysis were essentially carried out as described previously (Hesketh et al., 2009). In short, S. coelicolor M600 and a S. coelicolor M600 bldD null mutant were grown in triplicate for 15 h in YEME/TSB liquid medium. Total RNA was isolated using the RNeasy midi kit (Qiagen) according to the manufacturer's instructions, with modifications as described (Hesketh et al., 2009). Subsequently, single-strand reverse transcription (amplification) and indirect labelling of 10 µg of total RNA were performed for hybridization to Streptomyces diS_div712a GeneChip arrays according to the manufacturer's published protocol (Affymetrix). The GeneChips were washed and stained using a GeneChip fluidics workstation model 450, and then scanned with a Gene Array Scanner. After preprocessing, the data were imported into GeneSpring 9.0 (Agilent Technologies), converted to log2 values and normalized per gene to the median. Error models based on replicate values were implemented and statistical calculations, on the filtered data, were performed in GeneSpring by two-way anova as described previously (Hesketh et al., 2009). Transcriptome data have been deposited at the MIAME-compliant ArrayExpress database under accession number E-MEXP-2853.
ChIP-chip culture conditions were exactly the same as those for transcriptional profiling and were done in triplicate for the wild-type strain. Formaldehyde was added to cultures at a final concentration of 1% (v/v) and incubation was continued for 30 min. Glycine was then added to a concentration of 125 mM to stop the cross-linking. The sample was left at room temperature for 5 min and washed twice in PBS buffer (pH 7.4). The pellet was resuspended in 0.5 ml of lysis buffer (10 mM Tris-HCl pH 8, 50 mM NaCl), containing 15 mg ml−1 lysozyme and protease inhibitor (Roche Applied Science), and incubated at 25°C for 1 h. Subsequently, 0.5 ml of IP buffer (100 mM Tris-HCl pH 8, 250 mM NaCl, 0.5% Triton X-100, 0.1% SDS), containing protease inhibitor, was added and chilled on ice. The sample was sonicated for seven cycles of 15 s each at 10 microns to shear chromosomal DNA into fragments ranging from 300 to 1000 bp on average. The sample was then centrifuged twice at 13 000 r.p.m. at 4°C for 15 min to clear the cell extract, after which 10 µl of cell extract was set aside for total-DNA extraction. The remainder (900 µl) was incubated with 45 µl of protein A-sepharose (Sigma) for 1 h on a rotating wheel to clear from non-specifically binding proteins. Samples were then centrifuged for 15 min at 4°C and 13 000 r.p.m. to remove the beads. The supernatant was incubated with 50 µl of anti-BldD antibody overnight at 4°C with rotation. Subsequently, 90 µl of protein A-sepharose was added to precipitate BldD and incubation was continued for 4 h. The sample was centrifuged at 3500 r.p.m. for 5 min and the pellet was washed twice with 0.5×× IP buffer, then twice with 1× IP buffer, and transferred to a new tube after the first washing step. The pellet, and 10 µl of total cell extract, were eluted overnight at 65°C in 150 µl of IP elution buffer (50 mM Tris-HCl pH 7.6, 10 mM EDTA, 1% SDS) to reverse cross-links. The sample was centrifuged at 13 000 r.p.m. for 5 min to remove the beads. The pellet was re-extracted with 50 µl of TE buffer (10 mM Tris-HCl pH 7.4, 1 mM EDTA) and incubated with 0.2 mg ml−1 Proteinase K (Roche) for 2 h at 55°C. The samples were extracted with phenol-chloroform and further purified using QiaQuick columns (Qiagen). DNA was eluted in 50 µl of EB buffer and quantified using a NanoDrop spectrophotometer (Thermo Scientific).
DNA labelling and hybridization to DNA microarrays were performed by Oxford Gene Technology (OGT) and were essentially carried out as described previously (Bucca et al., 2009). Briefly, using the Bio-Prime kit (Invitrogen), 800 ng of the total and immunoprecipitated DNA was labelled with Cy5-dCTP and Cy3-dCTP respectively. Labelled DNA was hybridized, in an Agilent Technologies hybridization oven, to high-density DNA microarrays, representing the genome of S. coelicolor A3(2) (Bentley et al., 2002), that were manufactured by OGT. Following washing, the arrays were read out using an Agilent Technologies scanner and Cy5 and Cy3 signals were quantified using Agilent's Feature Extraction software. Data were analysed using a combination of bespoke Perl and r programs. Essentially, the following steps were carried out. Columns containing the positions of probes in the genome sequence, the green median signals and the red median signals were extracted from the data files received from OGT. Ratios of green median signals to the red median signals were calculated. For the bldD mutant control and for each of the three replicates of the wild type, the signal ratios were read into a data frame in r and scaled to make the mean as well as the standard deviation for each fall within the range of 0 to 1. For each probe, the mean of the three (scaled) bldD ratios was calculated and from it the control ratio was subtracted to arrive at a vector of the differences between the bldD means and the control. For each of the difference so calculated, a P-value was calculated assuming a population mean equal to the mean of all differences and a population standard deviation equal to the standard deviation of all the differences. These P-values were adjusted for multiple testing by the method of Benjamini and Hochberg to arrive at adjusted P-values. Data were ordered by increasing (adjusted) P-values and all probes having an adjusted P-value < 5e−3 were manually inspected to determine whether they were a part of a signal peak or not. ChIP-chip data have been deposited at the MIAME-compliant Gene Expression Omnibus database under accession number GSE23401.
Identification of DNA motifs
To identify conserved DNA motifs, DNA sequences encompassing up to 200 bp of the upstream regions of the genes, found in the ChIP-chip experiment, were collected from the genome sequence of S. coelicolor (Bentley et al., 2002). This data set was used as input for the MEME software tool (Bailey and Elkan, 1994) to search for over-represented sequences. A search for the occurrence of identified motifs was performed using Virtual Footprint software suite (Munch et al., 2005) in the genome of S. coelicolor. Alternatively, the entire genomes of S. griseus NBRC 13350, S. avermitilis NRRL 8165, S. erythraea NRRL 23338, S. tropica CNB-440 and Frankia sp. EAN1 were used as templates. Results of all motif searches are available in Table S3.
In vitro DNA-binding assays
Electrophoretic mobility shift assays (EMSAs) were carried out as described previously (Elliot and Leskiw, 1999). In short, DNA fragments containing a predicted BldD binding site downstream of bldA, and a truncated derivate, were generated by PCR using Expand DNA polymerase (Roche) with combinations of oligonucleotides bldA_F1 and bldA_R1, and bldA_F1 and bldA_R2 respectively. Radioactive probes were generated from these fragments by end-labelling and incubated with various amounts of histidine-tagged BldD, purified as described previously (Elliot and Leskiw, 1999), at 30°C in a 20 µl volume containing 10 mM Tris-HCl (pH 7.8), 150 mM NaCl, 2 mM dithiothreitol, 1 µg of poly[dI-dC] (Roche) and 10% glycerol. Protein–DNA complexes were separated in 4% polyacrylamide gels, run in TBE buffer at 100 V for 1 h, which were dried after electrophoresis and used for autoradiography using a FLA-7000 phosphorimager (Fujifilm).
DNase I footprinting experiments were carried out essentially as described previously (den Hengst et al., 2005) and according to the description supplied with the Sure Track footprinting kit (Amersham Pharmacia Biotech). DNA fragments were prepared by PCR with combinations of oligonucleotides 7773_F1 and 7773_R1, 0792_F1 and 0792_R1, 2792_F1 and 2792_R1, 2082_F1 and 2082_R1, and bldA_F3 and bldA_R3 to generate DNA probes overlapping potential BldD binding sites near the pepA, sti, bldH, ftsZ and bldA genes respectively. Oligonucleotides were first end-labelled with T4 polynucleotide kinase (Amersham Pharmacia Biotech) and [γ-32P]-ATP as described by the manufacturer. Binding reactions were identical to those used in EMSAs, but in a total volume of 40 µl, and in the presence of approximately 110 000 cpm of the DNA probe.
We thank Paul Dyson for cosmid StE60, Marie Elliot for helpful suggestions and critical reading of the manuscript, and OGT for expertly handling ChIP samples. This work was funded by BBSRC Grant BB/H006125/1 and by a grant-in-aid to the John Innes Centre from the BBSRC.