Genome-Wide Analysis of In Vivo Binding of the Master Regulator DasR in Streptomyces coelicolor Identifies Novel Non-Canonical Targets

Streptomycetes produce a wealth of natural products, including over half of all known antibiotics. It was previously demonstrated that N-acetylglucosamine and secondary metabolism are closely entwined in streptomycetes. Here we show that DNA recognition by the N-acetylglucosamine-responsive regulator DasR is growth-phase dependent, and that DasR can bind to sites in the S. coelicolor genome that have no obvious resemblance to previously identified DasR-responsive elements. Thus, the regulon of DasR extends well beyond what was previously predicted and includes a large number of genes with functions far removed from N-acetylglucosamine metabolism, such as genes for small RNAs and DNA transposases. Conversely, the DasR regulon during vegetative growth largely correlates to the presence of canonical DasR-responsive elements. The changes in DasR binding in vivo following N-acetylglucosamine induction were studied in detail and a possible molecular mechanism by which the influence of DasR is extended is discussed. Discussion of DasR binding was further informed by a parallel transcriptome analysis of the respective cultures. Evidence is provided that DasR binds directly to the promoters of all genes encoding pathway-specific regulators of antibiotic production in S. coelicolor, thereby providing an exquisitely simple link between nutritional control and secondary metabolism.


Introduction
The ability of bacteria to adapt to a wide range of nutritional environments requires intricate regulatory systems to facilitate the key routes of primary metabolism. The flux of carbon and nitrogen is typically controlled by a number of global, pleiotropic regulators, each targetting a number of specific, non-pleiotropic regulators in their response regulon. The global regulators ensure that the flux of essential metabolites is closely monitored and tuned [1,2]. Streptomycetes are mycelial bacteria with a complex multicellular life style [3]. Adaptation to changing conditions in the environment is particularly challenging and requires a network of sensory systems, well represented in the Streptomyces coelicolor genome by 45 ECF (extracytoplasmic function) sigma factors genes and some 70 two-component regulatory systems [4]. As producers of over half of known antibiotics, streptomycetes are a paradigm of secondary metaboliteproducing microorganisms [5,6]. Typically, antibiotic biosynthetic gene clusters are, at the very least, activated by a pathway-specific regulator within or adjacent to the cluster, and in a growth-phase dependent manner [7]. A pivotal question to address is how primary environmental signals are relayed to global carbon and nitrogen control and further to the control of development and antibiotic production. The elucidation of the S. coelicolor genome sequencethe first from an antibiotic-producing actinomycete [8]-was a landmark event, and showed that the antibiotic-producing potential of actinomycetes had been underestimated. Indeed, actinomycete genomes typically contain some 20 sets of putative biosynthetic genes for secondary metabolites [8,9,10,11], and it is yet unclear how many of these are antibiotics [12]. Many of these gene clusters are not expressed under typical laboratory conditions of rapid growth on nutrient-rich media and are often referred to as sleeping or cryptic antibiotic clusters. This may explain why they have been missed in the massive screening efforts performed by 'Big Pharma' [13,14]. Many antibiotic biosynthetic gene clusters are regulated in subtle ways so as to allow the organisms to adapt to life in the soil, with its varying physical, chemical and biological stresses, and the challenge is now to exploit this untapped source of potentially valuable natural products [6,15,16].
To allow activation and screening of cryptic antibiotic biosynthetic clusters, detailed insight is required into the linkages between environmental (nutritional) signals and secondary metabolite production. Carbon source utilization is a major determining factor in the metabolic control of antibiotic production [17,18]. The major control system for carbon utilization in bacteria is carbon catabolite repression (CCR). In the model organisms Escherichia coli and Bacillus subtilis, CCR is mediated via the phosphoenolpyruvate-dependent phosphotransferase system (PTS; reviewed in [19,20,21]). In streptomycetes, the PTS does not seem to play a major role in carbon control and CCR is mediated through glucose kinase, although the mechanism is still unknown [22,23]. Additionally, glucose is not internalised via the PTS but rather via the major facilitator GlcP [24]. In streptomycetes the PTS fulfills a dual role, and is involved in both N-acetylglucosamine (GlcNAc) uptake [25,26] as well as in the switch from normal growth to development [27]. This switch is somehow mediated via the accumulation of extracellular GlcNAc, which causes developmental arrest and blockage of antibiotic production in S. coelicolor [27,28]. GlcNAc is the monomer of the abundant natural polymer chitin-after cellulose the second most abundant carbohydrate on earth-and also a major constituent of cellwall peptidoglycan. GlcNAc is a primary source of carbon and nitrogen for streptomycetes. The GlcNAc regulon is controlled by the GntR-family regulator DasR [27,29,30], and DNAbinding activity is inhibited in vitro by glucosamine-6-phosphate [27], a metabolic product derived from GlcNAc. in silico and in vitro evidence revealed that direct repression of the actII-ORF4 and redZ, the pathway-specific regulators of actinorhodin and prodiginine production, respectively [28], was likely.
Previous studies showed the involvement of DasR in the control of genes involved in GlcNAc/chitin catabolism [27,31], antibiotic production [28], siderophore biosynthesis [32] and stress response [33]. Additionally, bioinformatic predictions using the PREDetector algorithm [34] revealed some 200 sequences that conformed to the consensus binding site for DasR (dre, for DasR responsive element), namely the palindromic 16-bp consensus sequence A(G/C) TGGTCTAGACCA(G/C)T. These target sequences are primarily associated with genes related to primary metabolism, sugar transport and extracellular polysaccharide hydrolysis, as well as antibiotic production [32]. Moreover, evidence for the functional importance of several of these sequences was demonstrated by showing DasR binding in vitro and changes in the level of associated transcript(s) in a DasR knockout.
In this work we use a genome-wide analysis to determine directly the sites of DasR binding in vivo and correlate them with changes in gene expression. This has highlighted major and highly dynamic changes in its binding pattern both in a growth phase-dependent manner (binding during vegetative growth or sporulation) and after the exogenous addition of the ligand GlcNAc.

Results
We previously showed that the phenotype of dasR null mutants and the extracellular accumulation of N-acetylglucosamine resulted in similar phenotypes, with enhanced antibiotic production and accelerated development under poor growth conditions, and blocked development and antibiotic production on rich media [28]. Linkage between the effect of GlcNAc and the phenotype of dasR null mutants is most likely provided by GlcN-6P, which acts as a ligand for DasR [27]. However, the differential effect of GlcNAc under different growth conditions and at different time points remained unexplained and detailed biological insights into the genome-wide distribution of DasR binding in vivo is therefore a prerequisite to better understand the biological role of DasR. We therefore conducted ChIP-on-chip analysis on wild-type S. coelicolor M145 carrying the integrative vector pGAM29, which expresses C-terminally 3xFLAG-tagged DasR, using the dasR null mutant GAM29 as the control. Two different types of genome-wide DNA-binding studies were performed; first we analysed DasR binding in cultures grown on minimal media (MM) agar plates with mannitol as the sole carbon source for the comparative analysis of DasR binding during vegetative and developmental growth, and subsequently we performed induction experiments, adding GlcNAc to liquidgrown cultures to monitor the changes on DasR binding in both rich and poor media.
Direct control by DasR of its core regulon and of developmental genes depends on the developmental status of the cell Firstly, we analysed DasR binding during growth on MM agar plates with mannitol as the sole carbon source, whereby two biological replicates were analysed. Samples were collected at 24 h (vegetative growth) and 54 h (sporulation), with similar growth rates and timing of development for both strains. Using a combined approach of statistical (two or more significantly enriched probes) and manual inspection (clear peaks were located by visual inspection of profile data on a genome browser), DasR binding events were identified at 21 and 51 promoter-proximal regions bound in the 24 h and 54 h old samples, respectively (Table 1 and S1 Table; rRNA operons were not taken into account). In total 10 binding events (nagE1, nagE2, nagKA, scr3092, dasR-dasA intergenic region, scr5239, tRNAGln-Glu, ptsH, SCO6032-6033 intergenic region and SCO7056) were shared between the 24 and 54 h samples. It should be noted at this point that we also performed eight additional ChIP-on-chip experiments on liquid-grown NMMP and R5 cultures, to analyse the effect of GlcNAc on in vivo binding (see below). Taken together, this provided very good validation for the in vivo binding of DasR, with up to eight independent verifications of the target genes ( Table 1).
As expected, clear enrichments for DasR-binding were found in the upstream regions of many genes that belong to the nag regulon, the pts genes and the chitinolytic system, which can be described as the core regulon for DasR [27,28], and coincide with bioinformatically predicted dre elements (using the PREDetector algorithm [34]) with generally high scores (Table 1). These Development and secondary metabolism Transposases and DNA topology include the upstream regions for the nag metabolic genes nagKA and nagB, the pts genes nagE1, nagE2, ptsH, and members of the chitinolytic system, namely chiA, chiC, chiD, chiH, chiJ, SCO6032 (for β-N-acetylhexosaminidase), SCO6300 (for a secreted β-N-acetyl-glucosaminidase), SCO6345, the SCO6005-6007 operon (the SCO6004-SCO6005 intergenic region has only one probe significantly enriched), dasA and dasR itself (Fig 1 and S1 Fig show targets with two or more significantly enriched probes). The previously experimentally verified DasR targets crr-ptsI, chiI, chiB and chb [27,30,31] also came out as hits, but only had one probe significantly enriched, and therefore failed to meet our rigorous statistics (see Experimental Procedures). Surprisingly, the genomic distribution of the observed DasR binding showed major divergence between the cultures harvested at 24 h and 54 h, which highlights growth-phase dependent differences in the DasR-binding pattern. In particular, all of the genes of the core regulon  T0 T1 T2 T3 T0 T1 T2 T3   SCO1471 - * one probe; *54h one probe at 54h a, +, binding;-, no binding (nag, ptsH, chi) that were bound by DasR at 24 h were either not bound after 54 h of growth, or (in the case of the chitinolytic genes chiC, chiH, SCO6300, SCO6345 and SCO7225) showed reduced binding. From the perspective of developmental control, only after 54 h did we see binding of DasR to any of the genes involved in development and secondary metabolism, as well as many other targets outside the core regulon. Binding was observed to the intragenic regions of three biosynthetic genes (SCO6273-6275) of the cpk gene cluster for the cryptic type I polyketide synthase (Cpk) and of cdaPSI and cdaPSII (SCO3230 and SCO3231) encoding the peptide synthetases I and II for the synthesis of calcium-dependent antibiotic (CDA). Additionally, in liquid-grown cultures binding of DasR was detected in the promoter proximal region of actII-ORF4 and redD, redZ, encoding the pathway-specific activator genes for Act and Red biosynthetic clusters, respectively (see Table 1, Fig 2 and below). dre elements were identified in silico for actII-ORF4, redZ, and cpkC, and actII-ORF4 and redZ were also corroborated in vitro by EMSAs. DasR binding was observed for the promoter regions of many genes for transfer-RNA (tRNA) and noncoding RNA (ncRNA). As for ncRNAs, binding was observed to dre elements upstream of scr3092 [35] and scr5239 [36] (Table 1). scr5239 decreases the production of the antibiotic actinorhodin, and represses expression of the extracellular agarase gene dagA (SCO3471) at the post-transcriptional level [36]. Additionally, DasR likely controls the  expression of several tRNA genes, as suggested by binding to multiple sequences associated with highly conserved dre elements-and in several independent ChIP-on-chip experimentswithin the five-membered tRNA Gln/Glu gene cluster, to a gene for tRNA Gly upstream of SCO3679, to tRNA Met and to the tRNA Glu/Asp/Phe gene cluster upstream of SCO4092. Additionally, after 54 h of growth on MM agar plates, DasR binding was observed upstream of SCO2504 (glyS) encoding glycyl-tRNA synthetase (Table 1). Finally, binding was also observed to rRNA operons. S. coelicolor has six rRNA operons [37] and binding to them was observed previously for HspR [38]. To validate binding of DasR to scr5239, we tested if indeed DasR could bind specifically to a probe encompassing the upstream region of scr5239. Phylogenic analysis of upstream sequences revealed a strongly conserved dre element upstream of scr5239 in streptomycetes, with core sequence 5'-TGGTCTAGTCCA, spaced 49 nt away from the start of scr5239 (S2 Fig). EMSAs using purified His 6 -tagged DasR showed that indeed the protein bound in vitro to a double-stranded oligonucleotide encompassing the dre element, with equal efficiency as to the positive control (a fragment of the same length encompassing the dre element of dasA). The negative control (a fragment of the bla gene for a beta-lactamase) was not bound by DasR.

GlcNAc-mediated alteration of DasR binding in vivo
Having established DasR binding during normal growth, we then wondered how in vivo DasR binding would respond to the exogenous addition of GlcNAc. In particular, we sought to investigate the molecular basis for the differential effect of GlcNAc addition under feast (rich media) and famine (poor media) conditions, with block or stimulation of development and antibiotic production, respectively. To address this important question, we performed ChIP-on-chip experiments for DasR following the extracellular addition of GlcNAc. For this purpose, mycelia of S. coelicolor M145 harbouring pGAM29 and the dasR mutant were each grown in liquid NMMP mannitol or R5 cultures, and cells were induced by the addition of 50 mM GlcNAc. Samples were collected prior to (T 0 ) and 30 (T 1 ), 60 (T 2 ) or 120 min (T 3 ) after GlcNAc addition, and DasR-enriched chromatin was analysed as described in the Experimental Procedures section. Prior to GlcNAc induction (T 0 ) high ChIP-enrichment ratios demonstrated DasR binding at 72 and 66 sites in NMMP-and R5-grown mycelia, respectively (S2 Table). In line with the experiments performed on mycelia from surface-grown cultures, these sites included its core regulon pts, nag and chi (S3 Fig).
Interestingly, after GlcNAc addition to NMMP-grown mycelia, the DasR protein dissociated from these target sites, providing direct evidence that indeed DasR responds to the addition of GlcNAc in vivo, thereby reducing its affinity for its target DNA. This also provided the first in vivo evidence for the previously proposed signaling cascade [28] from extracellular accumulation of GlcNAc to the onset of antibiotic production via the GlcNAc-mediated release of DasR from promoters of: (i) genes for GlcNAc transport and phosphorylation by the PTS (crr-ptsI and nagE2); (ii) the transcription unit that includes nagA which is necessary to provide the DasR effector GlcN-6P; and (iii) actII-ORF4 and of redD and redZ, encoding the pathway-specific activator genes for Act and Red biosynthetic clusters, respectively, as well as from the coding sequences of cdaPSI and cdaPSII (SCO3230-3231) for CDA peptide synthetases I and II, and for cpkBC (for Cpk polyketide synthases; [39]). Furthermore, this also presents the first evidence that in fact DasR directly binds to genes belonging to all of the known antibiotic biosynthetic clusters of S. coelicolor.
In the above mentioned DasR-mediated GlcNAc signaling cascade from nutrient availability to antibiotic production, internalization via the nagE2 product is the first step, while control of pathway-specific regulators is the last step. We previously showed that DasR and AtrA have antagonistic actions towards nagE2 and actII-ORF4. AtrA is required for actinorhodin production via the direct activation of actII-ORF4 [40], and also activates the GlcNAc transporter gene nagE2 [26], while DasR does the opposite. Interestingly, our data show that in fact DasR binds to the upstream region of atrA, and that this binding is reduced by the addition of GlcNAc to NMMP-grown mycelia and relieved when GlcNAc was added to R5 cultures. Thus, AtrA and DasR antagonise each other in terms of their influence on this important signaling cascade.
Finally, over 30 cases were found in nutrient rich R5-grown cultures where DasR binding was not relieved by the addition of GlcNAc, of which nearly 20 in fact showed enhanced binding. Examples of constant binding include pyrR-bldD, the upstream regions of the tRNA genes/ operons tRNA Glu/Asp/Phe , tRNA Gln/Glu and tRNA Met

Transcriptomic analysis of the dasR null mutant
The ChIP-on-chip experiments provided insight into the global DasR binding. To obtain complementary insight into the global changes at the transcriptional level following the deletion of dasR, microarray analysis was performed on wild-type and dasR mutant strains grown on MM mannitol agar plates. Under these standard growth conditions both the wild-type and mutant displayed similar development [27]. RNA was isolated from mycelia harvested after 24 h (vegetative growth), 30 h (early aerial growth), 36 h (early aerial growth), 42 h (late aerial growth) and 54 h of growth (sporulation). Microarray analysis revealed some 1200 genes that were significantly differentially expressed in the dasR null mutant of S. coelicolor relative to the parent M145, and the transcription of over 400 genes was more than two-fold up-or down-regulated at one or more time points in the dasR mutant (across two biological replicates). Some 100 genes showed more than three-fold and 26 genes showed more than five-fold change in transcription relative to the parental strain (S3 Table). Classes of genes that were significantly differentially expressed in the dasR strain compared to wt related to: primary metabolism; development and secondary metabolism; SapB, chaplins, rodlins (development-related peptides) and small peptides with unknown functions; transposases and DNA recombination. The most notable changes are discussed below; for details see S3 Table. Genes of the DasR core regulon were up-regulated in the dasR mutant and mostly at all time points, including nagE2, nagF, ptsH (encoding PTS components); nagB, nagK (encoding GlcNAc metabolic enzymes); glmS1 and glmS2 (opposite catalytic activity as nagB); and dasA-dasBC (chitobiose transporter [41,42]) (Fig 3A). Additionally, several genes involved in central metabolism, such as glycolysis, the TCA cycle, gluconeogenesis and pyruvate metabolism, as well as for biosynthetic pathways for amino acids, nucleotides and fatty acids were up-regulated in the dasR mutant (S3 Table). The ABC transporter operons, in particular the branched-chain amino acid transporter operon SCO2008-SCO2012, the SCO6005-SCO6007 operon (Fig 3A) that is orthologous to the ngcEFG chitobiose/GlcNAc transport operon in S. olivaceoviridis [43], and some of the genes in the oligopeptide transporter operon SCO5476-SCO5480, were strongly up-regulated in the dasR mutant (S3 Table).
As expected, on the basis of the direct binding of DasR as observed by ChIP-on-chip analysis, transcription of some of the genes for the biosynthesis of actinorhodin (SCO5070-SCO5091) and prodiginines (SCO5877-SCO5898) was enhanced in the dasR mutant (S3 Table). However, while changes for many act and red genes were statistically relevant, the majority did not meet the criterion of a minimum of two-fold change in expression. Transcription of scbA (SCO6266) for γ-butyrolactone synthase implicated in the control of the antibiotic clusters act, red and cpk Important categories of genes whose transcription was altered significantly in microarray experiments when comparing the dasR mutant GAM29 to its parent S. coelicolor M145 are shown; these are (A) genes for amino acid and (amino-)sugar transport and metabolism; (B) genes related to development and antibiotic production; (C) genes for chaplins, rodlins and SapB. RNA was isolated from S. coelicolor M145 or its dasR mutant GAM29 cultivated on MM mannitol agar plates after 24 h (vegetative growth), 30 h (vegetative/aerial growth), 36 h (aerial growth), 42 h (aerial growth/sporulation) and 54 h (sporulation). A heat map is provided as reference for the fold changes in log scale (log 2 (mutant/wild-type)).Note that many genes shown a larger than four-fold change in transcription. Some key genes that were experimentally validated DasR targets but with <2-fold change are also presented. For exact values see S3 Table. doi:10.1371/journal.pone.0122479.g003 Nutrient Sensing in Streptomyces in S. coelicolor [44], and cpkO (SCO6280), for the transcriptional activator of the cpk cluster, was up-regulated at 24 h, followed by enhanced transcription of the cpk cluster in the following time point (30 h) (Fig 3B and S3 Table). Other antibiotic-related genes whose transcription was upregulated were nsdA (SCO5582), a negative regulator of antibiotic production in S. coelicolor [45] and its paralogue nsdB (SCO7252) [46]. Several genes of the deoxysugar gene cluster SCO0381-SCO0401, which shows similarity to the ste gene cluster of Streptomyces sp. 139 required for Ebosin biosynthesis, were also up-regulated in the dasR mutant (S3 Table). Ebosin is an exopolysaccharide (EPS) which shows anti-rheumatic arthritis activity in vivo [47]. It is interesting to note that in Streptomyces sp. 139 the dasR gene lies immediately adjacent to the orthologous cluster [47].
The DasR regulon includes a large number of putative transcriptional regulators, with around 40 regulatory genes at least two-fold up-or down-regulated in the mutant at one or more time points, which forms a likely explanation for the large number of differentially expressed genes in the mutant. Several early developmental genes were up-regulated in the dasR mutant at later time points, namely bldK, bldM, bldN, crp and wblA (Fig 3B), while expression of sporulation genes (e.g. whi, ssg) was left unchanged. The pentacistronic bldK operon encodes an oligopeptide transporter required for early stages of development [48,49], bldN encodes a developmentally regulated σ factor, σ BldN , required for aerial mycelium formation, which in turn is required for the transcription of the response regulator gene bldM [50], crp is involved in the control of spore germination [51,52], and the WhiB-like protein WblA represses antibiotic production and may also be involved in the transition from early aerial hyphal cells to the subapical stem and apical compartments that precede sporulation [53]. Some of the most highly up-regulated genes were wblC (SCO5190) and its flanking genes SCO5189 and SCO5191.
Strikingly, chpABCDEFGH, rdlAB and ramS, encoding chaplins, rodlins and SapB, respectively, were very strongly down-regulated during early time points. The chaplins and rodlins are hydrophobic 'coat' proteins that assemble into a rodlet layer that provides surface hydrophobicity to aerial hyphae and spores, thus allowing them to break through the moist soil surface [54,55]; the spore-associated protein SapB is a lantibiotic-type signaling peptide required for the onset of development [56]. Several of these genes featured among the 10 most strongly down-regulated genes during early time points (Fig 3C).
Finally, transcription of two other gene clusters encoding small (60-100 aa) secreted proteins and a GntR regulator [57], was also strongly repressed, with transcription of the SCO3982-SCO3988 cluster so strongly enhanced that it suggests that the cluster is completely silenced in wild-type cells by DasR (S3 Table). This gene cluster is most likely directly controlled by DasR, because the gntR regulatory gene SCO3986, encoding the likely regulator of this cluster, was identified as a DasR target in the ChIP-on-chip experiments on MM agar plates with mannitol at 54h (S1 Table). Having said that, binding of DasR upstream of SCO3986 was not detected at 24 h, despite that also during vegetative growth the SCO3982-SCO3988 cluster is strongly up-regulated. The regulation of the cluster therefore requires further investigation. Transcription of the paralogous gene cluster SCO3263-3268 (with SCO3982 similar to SCO3268 and so on) was also repressed by DasR S3 Table. The function of the gene clusters is yet unknown, but their strong up-regulation in the dasR mutant is intriguing.

DasR has two classes of direct targets and binds in a developmentdependent manner
Streptomycetes usually live in an environmental context of poor nitrogen and rich carbon availability [58]. In this respect, GlcNAc is a highly appropriate energy source for these bacteria, as it can be used as both a nitrogen and a carbon source. The core of the DasR regulon is formed by genes related to amino sugar metabolism, namely the nag and pts genes for GlcNAc metabolism and transport, respectively, and the chi regulon involved in the catabolism of polymers of GlcNAc. Since Fructose-6P, acetate and ammonia are end products of GlcNAc catabolism, the DasR core regulon is immediately connected to major metabolic pathways such as glycolysis, nucleic acids, nitrogen, and fatty acid metabolism, the tricarboxylic acid (TCA) cycle and cellwall biosynthesis. There is an excellent correlation between the presence of a dre motif and DasR binding in vitro and in vivo, and indeed we have never observed a predicted dre with a high statistical score that is not bound by DasR in vitro. There were however a few examples of known core target genes that did not display ChIP-enrichment peaks across multiple samples (e.g. binding to crr-ptsI was only seen once); false negatives are not uncommon in ChIP-based studies and may reflect cases where the DasR is physically hidden within the cross-linked chromatin (e.g. encompassed by looped DNA and/or other transcription factors). A notable example in this context is the failure to detect a ChIP-enrichment peak for CRP bound to the E. coli lac operon promoter [59].
Novel functional categories of Class I targets (i.e. preceded by dre elements) identified here include genes for ncRNAs and tRNAs. Additionally, we here provide evidence that in S. coelicolor all of the known antibiotic biosynthetic gene clusters are directly controlled by DasR, namely act, red, cpk and cda for actinorhodin, undecylprodigiosin, a cryptic polyketide and calciumdependent antibiotic, respectively. Most of these are associated with dre elements. However, many binding events, referred to as 'Class II' targets, were identified in vivo that are not associated with a canonical dre element. Some of these were identified as highly-enriched probes in as many as eight biologically independent ChIP-on-chip experiments, providing a strong case that these are indeed bona fide binding sites. Genome-wide analysis of the regulons of several other globally acting bacterial transcription factors show similar duality, with binding to both their cognate (consensus) target sequences and to non-canonical sites in vivo, even though the latter binding sites show no similarity to the consensus sequence and they are not bound in vitro.
Other examples include LexA [60], Crp [61] and FNR [62] in E. coli, Spo0A in Bacillus subtilis [63], CtrA in Caulobacter crescentus [64] and notably also Crp [65], GlnR [66] and PhoP [67] in Streptomyces. For B. subtilis Spo0A, some 15% of the total number of binding sites was not bound in vitro [63]. AdpA of Streptomyces griseus and PhoP of S. coelicolor represent examples of master regulators in streptomycetes with a significant proportion of putative target genes without recognisable consensus binding sites [67,68]. It is likely that binding by these regulators to non-canonical sites requires additional factors, such as cooperative interaction with other proteins and/or changes in DNA conformation. An example of cooperative binding that is required for target recognition is that for CRP in E. coli, which binds to a noncanonical site in the melAB promoter in concert with MelR associated with adjacent target sites [69].
The conventional Class I targets are generally bound more avidly or, in a stochastic binding model, more frequently than the unconventional Class II targets, as the most enriched probes in the ChIP-on-chip experiments were typically those for Class I targets. This is similar to the situation observed for LexA in E. coli, where unconventional targets were also not bound by LexA in vitro, and bound less tightly in vivo [60]. A proteomics-based approach is currently being employed to elucidate how DasR binds to Class II targets in vivo, with particular focus on identifying possible protein partners that bind cooperatively with DasR.
The in vivo DNA-binding experiments revealed major differences in the binding pattern between cultures harvested at 24 h and 54 h, which highlights the rather expected growth-phase dependence of DasR binding. The notion that the affinity of a certain regulator for its binding site changes during growth is a well-accepted principle, but a growth-phase dependent change in binding, with other sites bound at different developmental stages, is to the best of our knowledge a novel concept. All of the genes of the core regulon (nag, pts, chi) were bound by DasR at 24 h, but after 54 h, using the same detection methods, no evident binding was found for chiC, chiH, SCO6300, SCO6345 and SCO7225. Conversely, DasR binding to genes involved development and antibiotic production (and other processes) was only found after 54 h. The pleiotropic role of DasR in the control of antibiotic production at least in S. coelicolor was further underlined by the binding of DasR to the promoter-proximal regions of actII-ORF4, redD and redZ, encoding the activator genes for Act and Red biosynthetic clusters, respectively [28].

The DasR response regulon and multi-level control of its target genes
The core regulon of the DasR homologue NagR in Bacillus also revolves around GlcNAc metabolism, but in contrast to dre elements, the NagR-responsive elements are found scarcely (up to three sites per genome [70]). Comparative genomics analysis suggests that similar small DasR regulons are found in other low G+C Gram-positive bacteria such as Streptococcus and Listeria (not shown). Our work shows that streptomycetes have adopted the GlcNAc regulatory system to build a much more intricate system that controls a wide range of genes. More than 1,200 genes were significantly differentially expressed in the dasR null mutant of S. coelicolor (i. e. 15% of the genome) and transcription of over 400 genes altered more than two-fold, which indicates the large changes in gene expression resulting from the absence of the metabolic master regulator DasR. In a recent environmental study, microarray data on the dasR null mutant in soil-grown cultures in the presence of chitin revealed some 700 genes that were differentially expressed [33]. In general, deletion of developmental regulators has a major impact on gene expression in Streptomyces, exemplified by the more than 1000 significantly differentially expressed genes in mutants lacking the morphogene ssgA, which controls processes relating to cell-wall remodeling such as germination, tip growth, branching and cell division [71,72], or the very large response regulon of BldD, which directly controls a large number of developmental genes including some 40 regulatory genes [73].
The DasR regulon incorporates around 40 transcriptional regulatory genes, suggesting an extensive secondary response. Global transcriptional regulators provide an additional layer of control on top of the control mediated by specific activators and/or repressors [1]. Therefore, an 'all or nothing' change in transcription following deletion of dasR was not anticipated. The specific environmental conditions are an important parameter for the response of the regulon, and additional control systems operate, such as substrate induction for induction of sugar utilization operons, or growth phase-dependent control of antibiotic production. The additional requirement of substrate induction is exemplified by the fact that despite DasR binding to almost all of the genes of the chi regulon, transcription of the genes was not notably changed in dasR mutant cells due to the lack of chitin as inducer. Indeed, strong induction of the chitinolytic system by chitin was observed in soil-grown S. coelicolor cultures, as well as enhanced antibiotic production [33,74]. The expression levels of all these genes were up-regulated by chitin in soil cultures. This suggests that there are other mechanism(s) that overrule the inhibitory effect of DasR on gene expression.
Antibiotic production is typically extensively controlled, which ensures correct growth phase-dependent expression. Taking actII-ORF4 as an example, this gene is controlled by more than 10 transcriptional regulators. Besides DasR these include among others ActII-ORF4, AfsR, AtrA, SCO6008 and the stringent response mediated through ppGpp during growth cessation ( [40,75,76,77]; reviewed in [7,78]). As mentioned above, our genome-wide DNA-binding experiments demonstrated that DasR directly controls gene expression of the biosynthetic pathways for Act, Red, CDA and Cpk biosynthesis and also that this control is effectively relieved by the addition of GlcNAc. This surprising control of all antibiotics produced by S. coelicolor M145 (which lacks the plasmid-borne methylenomycin biosynthetic cluster) may have evolved to ensure global control of natural product formation that is modulated by changes in nutritional conditions. The relatively small changes in gene expression of the antibiotic-related genes seen under the chosen experimental conditions are a consequence of the choice of media, which only allow minor induction of antibiotic production, but were chosen here to allow proper comparison between wild type and mutant, which requires similar (and synchronized) growth and development; however, major changes in antibiotic production are seen under conditions of feast (rich media) and famine (MM with agar as the sole carbon source) [28].

In vivo response of DasR to the induction by GlcNAc
We have previously shown that glucosamine-6P (GlcN-6P) acts as an effector of DasR in vitro [27]. This is not unexpected, as GlcN-6P stands at the crossroads of (GlcNAc) n degradation, GlcNAc transport and intracellular metabolism, glycolysis, nitrogen and lipid metabolism, as well as peptidoglycan synthesis [79,80], and many genes of these pathways are subject to control by DasR. ChIP-on-chip analysis of MM-and R5-grown cultures before and after the addition of GlcNAc revealed the response of DasR to this inducer in vivo. The relief of DasR binding to its core regulon following the addition of GlcNAc to cultures provides in vivo evidence that DasR binding indeed undergoes major changes following binding of GlcNAcderived metabolites. In terms of antibiotic production, the same relief of binding was seen for the pathway-specific activator genes redD and redZ for prodiginine production, and actII-ORF4 and atrA for actinorhodin production, as well as biosynthetic genes for the cryptic polyketide Cpk and the calcium-dependent antibiotic CDA. This provides experimental in vivo proof that the activation of antibiotic production by GlcNAc is directly mediated through DasR in S. coelicolor and also shows that this goes further than the control of actII-ORF4 and redZ (which were previously identified as direct targets in vitro [28]).

Connection to other developmental regulons
Our data indicate that DasR exerts developmental control by targeting the expression of a number of key regulatory genes for the control of early developmental processes, including bldKc, bldM, bldN, wblA, wblC and crp. Of these, bldKc, bldN and wblC were identified as direct targets by ChIP-on-chip analysis, forming Class II targets as none of them is preceded by a dre element. Of these, binding to bldN was only seen for one experiment, and care has therefore to be taken as to whether this reflects true binding. Interestingly, however, the bldN gene product σ bldN is required for transcription of the chp, rdl and ramS genes [81], and the observed increase in bldN transcription in dasR null mutants apparently connects to the observed increase in chp, rdl and ramS expression during sporulation.
In terms of cross-talk with other developmental regulators, AtrA and DasR apparently counteract each other, with activating and repressing activity, respectively, on nagE2 and actII-ORF4. Their gene products form core members in the signalling cascade from GlcNAc accumulation outside the cell to the activation of actinorhodin production, with NagE2 importing the inducer and ActII-ORF4 as terminal activator. In contrast, there was surprisingly little overlap between the highly pleiotropic regulatory networks of BldD and DasR. Regulators of later sporulation events do not feature among the direct DasR targets, while, conversely, the primary BldD regulon contains a plethora of such developmental regulatory genes [73], suggesting that these two proteins act in different time (and perhaps also spatial) domains. Our data suggest that bldD may be a Class II target for DasR, with binding upstream of bldD verified independently in samples obtained after 54 h of growth on MM agar plates as well as from liquid-grown R5 cultures. Transcription of bldD was not changed in the dasR deletion mutant on MM mannitol, but under these conditions both the bldD and the dasR mutants sporulate. Whether and if so when DasR controls bldD transcription awaits further analysis.
In conclusion, our work provides detailed insights into how DasR links nutritional control to antibiotic production and development and shows how its affinity for different binding sites changes in vivo in response to the extracellular signalling nutrient GlcNAc (Fig 4). The canonical binding of DasR to Class I targets is observed both in vivo and in vitro, while the unconventional binding to Class II targets presumably requires cooperative binding with another protein(s), which has hitherto not been identified. Alternatively, these sites may require conformational changes in the DNA, or even a combination of the two. DasR mediates a higher order level of control, and relief of, or induction of, binding to its target sites does not necessarily result in an immediate change in gene expression. For that, the environmental conditions are an important determining factor-such as the presence of chitin for induction of chi gene expression-and further experimental data need to be obtained from genome-wide studies under different growth conditions. For example, in approaches to activate poorly expressed (or even dormant) antibiotic biosynthetic genes, relief of DasR-binding may well be an important step, but better understanding of the triggers that must be in place to effectively activate their biosynthetic gene clusters should aid their discovery and exploitation.

Experimental procedures
Bacterial strains and growth conditions E. coli JM109 [82] and ET12567 [83] were used for routine cloning procedures and for extracting non-methylated DNA, respectively. S. coelicolor M145 was obtained from the John Innes Centre strain collection and is the parent for the dasR null mutant GAM29, which had its entire coding region replaced by the apramycin resistance cassette as described previously [27]. All media and routine Streptomyces techniques are described in the Streptomyces manual [83]. R5 (regeneration media) is a rich media with glucose as carbon source and high concentration of sucrose (10% w/v) to allow protoplast regeneration, while NMMP is a standard minimal media buffered with phosphate. Cells of E. coli were grown in Luria-Bertani broth (LB) at 37°C. SFM (mannitol soya flour) agar plates were used to prepare spore suspensions. Strains were cultivated on MM agar plates with mannitol (1% w/v) (covered with cellophane discs) or in liquid R5 or NMMP supplemented with mannitol (1% w/v). For induction experiments, liquid cultures were grown at 30°C until mid-logarithmic phase (OD 600 approximately 0.4 for NMMP cultures and 0.8 for R5 cultures), followed by the addition of GlcNAc (50 mM). For the ChIP-on-chip studies on MM mannitol solid medium mycelium was harvested after 24 h and 54 h.

Plasmids and constructs
For the ChIP-on-chip studies a recombinant dasR gene was synthesised that allows production of DasR containing a C-terminal triple FLAG-tag epitope (GenScript, La Jolla, USA). The DNA fragment contained the dasR gene along with 306 bp promoter region, with the stop codon of the gene replaced by the sequence 5'-ATGGACTACAAGGACCACGACGGCGACTACAAG GACCACGACATCGACTACAAGGACGACGACGACAAGTAGACCAGAGCCCGCTCAC CCGGCCCCAGATTGCGGTTGAAGTCC-3' (stop codon underlined). In this way, the DasR protein was extended with the amino acid sequence MDYKDHDGDYKDHDIDYKDDDDK. The fragment was cloned into the integrative vector pSET152 [84]. This resulted in vector pGAM29, which expresses DasR-3xFLAG for use in ChIP-on-chip experiments.
Chromatin immunoprecipitation, gDNA labeling and array hybridizations S. coelicolor M145 containing the integrative vector pGAM29 expressing 3xFLAG-tagged DasR and the dasR null mutant GAM29 were grown in duplicate (see Bacterial strains and growth conditions). Cultures were treated with formaldehyde (final concentration 1%, 20 min, 30°C) in order to crosslink proteins to DNA. Subsequently glycine (final concentration of 0.5M) was added to quench the formaldehyde and cultures were incubated for further 5 minutes at 30°C. Mycelium was harvested by centrifugation, washed twice with 1xPBS and stored at -70°C. Chromatin isolation, labeling and hybridization were performed as described previously [38,86]. Biological replicates of DasR-3FLAG-IP chromatin were labeled with either Cy3 or Cy5-dCTP in a dye swap experimental design and co-hybridised with the mock 'no-antibody' IP chromatin on 4 × 44K whole genome Streptomyces arrays manufactured by Agilent Technologies [38].

RNA isolation, cDNA synthesis and labeling
Total RNA was purified using the Kirby-mix protocol [83]. DNaseI treatment was used to fully remove any traces of DNA. Before use, the RNA preparations were checked for their quality and integrity on the Agilent 2100 Bioanalyzer (Agilent Technologies). cDNA synthesis and labeling was performed as previously described [38,86].

Microarray expression data and ChIP-on-chip data analysis
For the time-course gene expression studies S. coelicolor M145 and dasR mutant (GAM29) were cultivated on MM mannitol agar plates overlaid with cellophane discs. Samples were taken at 24 h, 30 h, 36 h, 42 h and 54 h. Two biological replicates were analysed for all samples and for each strain. cDNA incorporating Cy3 dCTP was synthesized as described in http:// www.surrey.ac.uk/fhms/microarrays/Downloads/Protocols/Strep_hyb_protocol_1005.pdf, from 10 μg of total RNA extracted from each time point. Labeled cDNA was co-hybridized with genomic DNA, used as a common reference, labeled with Cy5-dCTP as described in http://www.surrey.ac.uk/fhms/microarrays/Downloads/Protocols/Strep_hyb_protocol_1005. pdf. The fluorescently labeled samples were hybridized onto whole genome custom Streptomyces microarrays 4 x 44K as described in ref. 35. For the ChIP-chip experiment, chromatin samples were either immuno-precipitated using a monoclonal anti-FLAG tag M2 antibody (Sigma cat. Number F1804)(sample) or incubated in absence of antibody (mock IP control). When mycelium from solid medium was used in the ChIP-chip experiment, the mycelium was harvested quickly with a spatula and immersed in 1% formaldeyde solution for 20 min at 30°C before adding glycine to 0.5 M final concentration to quench the formaldeyde. In the case of liquid cultures, the cells were treated with formaldeyde by adding the reagent directly to the culture at a final concentration of 1% and continuing the incubation for 20 min before adding the glycine as quencher. After the formaldeyde treatment the protocol followed was as described in [38].
Expression array data were normalized using the Limma package [87]. Briefly, data were normalized within each array using the global median value and then normalized across all arrays using the 'scale' function. The filtered data sets were then analysed using Rank Products analysis [88] via the web-based implementation of Rank-Prod [89] called RankProdIt [90]. Differentially expressed genes were identified as having a pfp (probability of false prediction) less than or equal to 0.15, equal to a false discovery rate of approximately 15%; this represents a previously validated threshold for gene expression analysis [91]. Some additional genes were also identified that demonstrated a >1.5 fold change in expression between wild-type and the dasR mutant. ChIP-on-chip data, allowing identification of genes/regions directly bound by DasR were analysed essentially as described previously [38]; the arrays were 'across-array' normalised only (using the Limma package) and poor quality probes (determined by Agilent Feature Extraction software version 9.1) were removed from downstream analysis. Probes with significant enrichment for DasR at a particular time-point were identified using a Mann-Whitney t-test to compare the log 2 (WT-DasR antibody/mock-antibody) and log 2 (Mutant-DasR antibody/mock-antibody) ratios for each time-point (using the R package 'multtest'). A probe was scored as significant if the associated t-statistic was greater than 2 which, given the number of replicates in this study, corresponds to a false discovery rate of less than 10%; this was considered to be a suitable initial threshold given our application of additional subsequent thresholds and manual inspection. Additional thresholds applied to highlight putative DasR binding events for manual inspection were: (1) a log 2 (WT-DasR antibody/mock-antibody)log 2 (Mutant-DasR antibody/mock-antibody) enrichment ratio of >0.3, where 0.3 is the point at which the distribution of these WT vs Mutant ratios depart from the typical Gaussian curve; (2) the presence of two significantly enriched DasR-ChIP probes within a 3.5 kb window. These DasR ChIP-enrichment regions containing putative DasR binding sites were then manually inspected to identify associated genes. Here, clear peaks within the 3.5 kb region were located by visual inspection of profile data using a genome browser.   Table. Genes whose expression was altered more than two-fold in the dasR null mutant as determined by microarray experiments. (PDF)