The role of absC, a novel regulatory gene for secondary metabolism, in zinc-dependent antibiotic production in Streptomyces coelicolor A3(2)

Authors


*E-mail andrew.hesketh@bbsrc.ac.uk; Tel. (+44) 1603 450757; Fax (+44) 1603 450778.

Summary

The availability of zinc was shown to have a marked influence on the biosynthesis of actinorhodin in Streptomyces coelicolor A3(2). Production of actinorhodin and undecylprodigiosin was abolished when a novel pleiotropic regulatory gene, absC, was deleted, but only when zinc concentrations were low. AbsC was shown to control expression of the gene cluster encoding production of coelibactin, an uncharacterized non-ribosomally synthesized peptide with predicted siderophore-like activity, and the observed defect in antibiotic production was found to result from elevated expression of this gene cluster. Promoter regions in the coelibactin cluster contain predicted binding motifs for the zinc-responsive regulator Zur, and dual regulation of coelibactin expression by zur and absC was demonstrated using strains engineered to contain deletions in either or both of these genes. An AbsC binding site was identified in a divergent promoter region within the coelibactin biosynthetic gene cluster, adjacent to a putative Zur binding site. Repression of the coelibactin gene cluster by both AbsC and Zur appears to be required to maintain appropriate intracellular levels of zinc in S. coelicolor.

Introduction

In an era when it is widely recognized that a crisis is emerging because of the emergence and spread of antibiotic resistance among microbial pathogens (Spellberg et al., 2004; Davies, 2006; Boucher et al., 2009), streptomycetes remain one of the richest potential natural sources of antibiotics, as well as providing a wealth of other important pharmacologically active molecules. Streptomyces species are ubiquitous in soil and sediments, where they grow as a mycelium capable of differentiating to yield spores for survival and dispersal. Soil is a complex, sometimes nutritionally rich but highly variable environment, and streptomycetes have evolved appropriate adaptive sensory and regulatory networks to respond to the challenges it presents.

Ever since the first commercial production of a streptomycete-derived drug, streptomycin from Streptomyces griseus in 1943 (Schatz et al., 1944), attempts have been made to understand and manipulate the developmental programme and the physiological and environmental signals that regulate the biosynthesis of antibiotics and other secondary metabolites. Roles for many different families of regulatory proteins have been identified, with both extracellular and intracellular signalling molecules acting as important elicitors (reviewed in Bibb, 2005). All of the genes required for biosynthesis of a particular secondary metabolite are generally clustered together at a single genetic locus in the producing organism. The cluster also frequently contains pathway-specific regulatory gene(s) that control the onset of biosynthesis. Genome sequencing of different Streptomyces species has revealed the genetic potential to produce upwards of 10 and as many as 30 different secondary metabolites per organism (Omura et al., 2001; Bentley et al., 2002; Ohnishi et al., 2008), but only a minor fraction of the compounds are made under typical laboratory screening conditions. A major current challenge is to understand the regulatory basis of the lack of expression of such ‘cryptic’ gene clusters and to use this knowledge to release the full biosynthetic potential of these organisms.

The identification and characterization of regulatory factors that control the expression of more than one secondary metabolic gene cluster has also been an important area of study. Production of two pigmented antibiotics – the blue-coloured actinorhodin (Act), and the red-pigmented undecylprodigiosin (Red) in the model streptomycete S. coelicolor has facilitated the identification of such ‘pleiotropic’ regulatory systems (Lawlor et al., 1987; Adamidis et al., 1990; Adamidis and Champness, 1992; Fernandez-Moreno et al., 1992a; Floriano and Bibb, 1996; Chakraburtty and Bibb, 1997; Aceti and Champness, 1998; Price et al., 1999; Takano et al., 2001; Rigali et al., 2006). Two of these, involving dasR and relA, have been shown to link the control of antibiotic biosynthetic pathways to the nutritional status of the mycelium.

The GntR-family regulator DasR inhibits production of both Act and Red by binding to operator sequences upstream of the pathway-specific regulatory genes of each of the biosynthetic gene clusters, thus repressing their transcription. Repression is relieved in the presence of glucosamine-6-phosphate, the abundance of which indicates levels of extracellular N-acetyl glucosamine, an important environmental source of carbon and nitrogen (Rigali et al., 2006; 2008). In contrast, the ribosome-bound RelA synthesizes the intracellular signalling molecule (p)ppGpp in response to amino acid insufficiency. This highly phosphorylated guanine nucleotide plays an important role in triggering production of three antibiotics (Act, Red and the calcium-dependent antibiotic) in S. coelicolor, probably via a direct effect on transcription of pathway-specific regulatory genes (Chakraburtty and Bibb, 1997; Hesketh et al., 2001; 2007; Sun et al., 2001). Binding of (p)ppGpp to RNA polymerase may allow it to adopt a configuration that favours transcription initiation at the promoters of these genes, or to select a particular σ factor that preferentially recognizes their promoter sequences. The two-component regulatory system PhoRP, which regulates genes involved in phosphate uptake and metabolism, has also been shown to exert an important influence over expression of the Act and Red biosynthetic gene clusters (Sola-Landa et al., 2003; Rodríguez-García et al., 2007). However, from characterization of the PhoP DNA operator sequence and the results of in vitro studies, it does not appear that the regulator binds directly to any pathway cluster genes, and the effect on antibiotic production may be due to cross-talk with another pleiotropic regulator, AfsR (Sola-Landa et al., 2005; Rodríguez-García et al., 2007; Santos-Beneit et al., 2009).

Zinc, a common element in the earth's crust, is routinely encountered by soil-dwelling microorganisms. After iron, ionic zinc is the next most abundant transition metal ion in living cells and plays an indispensable role in cellular biochemistry as a catalytic or structural cofactor for a wide variety of metalloproteins. Indeed, zinc binding proteins constitute 4–10% of the proteomes of both bacteria and eukaryotes: in S. coelicolor the fraction is estimated to be 5.5% (Andreini et al., 2006). However, like most metals, zinc can be toxic if accumulated to excess, and its intracellular concentration is controlled within safe limits by homeostatic regulatory mechanisms. In bacteria this is largely achieved by tight control of import and export mechanisms via the zinc-responsive regulatory protein Zur (Patzer and Hantke, 1998; Brocklehurst et al., 1999; Gaballa et al., 2002). Here we show that production of Act and Red in S. coelicolor is significantly influenced by the concentration of Zn2+ supplied to cultures. We also report the isolation and characterization of a new genetic locus, absC (antibiotic synthesis C), that exerts pleiotropic but zinc-dependent effects on antibiotic biosynthesis. We show that this is mediated via activation in an absC deletion mutant of the gene cluster for production of coelibactin, a putative siderophore-like compound also reported to be regulated by Zur in response to Zn2+ levels (M. Paget, pers. comm.). AbsC appears to repress the coelibactin biosynthetic operon directly, binding to an operator sequence adjacent to the putative Zur binding site.

Results

Zinc influences production of the antibiotics Act and Red in S. coelicolor

To determine whether antibiotic production in S. coelicolor is affected by the level of Zn2+, the plasmid-free derivative M145 was grown on the casamino acid supplemented minimal medium SMMS (Kieser et al., 2000) containing 0–175 μM zinc sulphate (Fig. 1). The mutant strains M510 and M511 (Floriano and Bibb, 1996), defective in production of Red and Act, respectively, were also analysed to assess effects on production of the individual pigmented antibiotics. The concentration of zinc added to the medium clearly influenced production of both antibiotics. Act production was optimal between 3.5 and 35 μM added zinc, but was difficult to detect when no zinc was added or at the highest concentration used, 175 μM. Red production was similarly poor at 175 μM zinc, and growth was also slightly inhibited under these conditions. In contrast to Act, Red production started earliest with no added zinc, but after 4 days was similar over the 0–35 μM range. It was also evident that there was sufficient Zn2+ present in SMMS, presumably derived from media components (e.g. water, agar and casamino acids) to support growth when no additional zinc sulphate was supplied.

Figure 1.

The concentration of ionic zinc affects production of the pigmented antibiotics Act and Red during growth of S. coelicolor on the supplemented minimal medium SMMS. Spores of strain M145, M510 (red-) and M511 (act-) were applied to SMMS prepared with the indicated concentrations of zinc sulphate as individual drops containing about 2 × 105 colony forming units. The spore drops were allowed to dry and the plates incubated at 30°C. Images were recorded after (A) 2 days and (B) 4 days.

Identification of absC, a novel regulator for antibiotic production, through complementation of a S. coelicolor M145 mutant collection with a genomic DNA library

Previous screening for pleiotropic mutants in S. coelicolor had been carried out using the auxotrophic strain J1501 and glucose minimal medium (Adamidis et al., 1990; Champness et al., 1992). As the mutant phenotypes of several known pleiotropic regulatory genes (e.g. relA and afsR) are much more striking on the modified minimal medium SMMS (Takano et al., 1992) than on glucose minimal medium, we undertook a screening effort for antibiotic-deficient mutants of M145 using SMMS. After excluding mutants that were also affected in morphological differentiation, or only partially antibiotic-deficient, and also mutants complemented by those clones from the genomic library of M145 DNA that either contained known regulatory genes or complemented more than one mutant, one strain, M801, remained in which antibiotic production could be restored with clone pIJ8449 from the library. Complementation still occurred with pIJ8453, an integrative vector carrying a 3 kb EcoRI-XbaI subfragment of the original 11 kb library insert (Fig. 2). Sequencing revealed that the 3 kb segment contained most of SCO5403 (without the first 9 nt of its protein-coding sequence), SCO5404, SCO5405 and a partial sequence of SCO5406 (without the first 293 nt of its protein-coding sequence). To identify the ORF responsible for restoration of antibiotic biosynthesis in M801, further subclones were constructed in pSET152 containing either part of SCO5403 and all of SCO5404 (pIJ8457), or SCO5405 with the incomplete SCO5406 ORF (pIJ8458; Fig. 2). Of these, only pIJ8458 restored antibiotic production to M801. As the mutant was only defective in antibiotic synthesis and sporulated normally, it was classified as an abs mutant (for antibiotic synthesis-deficient, Adamidis et al., 1990) and SCO5405 was consequently designated absC. PCR amplification of SCO5405 from M801 and subsequent nucleotide sequencing revealed a single-nucleotide substitution resulting in an arginine to tryptophan change at position 91 (data not shown). An unmarked in-frame deletion mutant, M804, in which most of the SCO5405 coding sequence had been removed, was constructed. M804 exhibited the same deficiency in antibiotic production as M801, although complementation of the mutation using pIJ8458 led to only partial restoration of antibiotic production (see Experimental procedures). M801 also exhibited reduced levels of production of the calcium-dependent antibiotic of S. coelicolor in a Bacillus mycoides overlay assay (Floriano and Bibb, 1996; data not shown).

Figure 2.

Complementation of absC, a mutant of S. coelicolor M145 pleiotropically defective in production of the pigmented antibiotics Act and Red. (A) illustrates the ORFs present in the 3 kb EcoRI-XbaI genomic DNA fragment in pIJ8453 that complements the absC mutant phenotype (see text). Subcloning of the pIJ8453 insert shows that SCO5405 is sufficient for complementation, as shown in (B), where introduction of pIJ8458 restores pigment production to parental levels in M801 while the vector control pSET152 has no effect.

absC encodes a MarR homologue that is highly conserved in streptomycetes

absC encodes a protein of 158 amino acids with a molecular mass of 18 311 Da. AbsC contains a Pfam match to PF01047, the helix–turn–helix motif of MarR family proteins (http://www.sanger.ac.uk/Software/Pfam/), and shares 43% amino acid identity with the MarR homologue MmcW of Streptomyces lavendulae, which regulates production of the antitumour antibiotic mitomycin C (Mao et al., 1999). It is also 22% identical to MhqR from Bacillus subtilis, which regulates expression of enzymes that confer resistance to 2-methylhydroxyquinone and catechol (Toewe et al., 2007). AbsC is highly conserved in all streptomycetes for which a genome sequence has been published, showing 84–87% identity to putative orthologues from Streptomyces avermitilis (Omura et al., 2001), S. griseus (Ohnishi et al., 2008) and Streptomyces scabies (the S. scabies genome sequencing project at http://www.sanger.ac.uk/Projects/S_scabies) (data not shown).

The defect in Act and Red production in M804 (ΔabsC) is zinc-dependent

To determine whether the involvement of absC in the regulation of antibiotic biosynthesis was linked to a specific nutritional depletion we compared pigmented antibiotic production in M145 and M804 on a range of media with different levels of nutrients. The absC phenotype of M804 was shown to be zinc-dependent; the mutant was completely defective in production of both Act and Red on SMMS supplemented with low concentrations of zinc (0.35 μM or less), but produced both pigments at close to parental levels with 3.5 μM or higher levels of zinc (Fig. 3). However, production in the mutant strain was not completely restored to levels seen in M145 at the higher zinc concentrations, showing a notable delay in production (data not shown).

Figure 3.

The defect in antibiotic production in the absC deletion mutant M804 is zinc-dependent, being most pronounced when the concentration of ionic zinc is low but nearer wild-type levels on SMMS containing 3.5 μM or more added zinc. Spores of M145 and M804 (ΔabsC) were applied to SMMS prepared with the indicated concentrations of zinc sulphate as described in Fig. 1, and the phenotype was recorded after 7 days of incubation.

The defect in Act and Red production in M804 (ΔabsC) is linked to abnormal expression of the pathway-specific activator genes controlling their biosynthesis

Production of Act is absolutely dependent on the pathway-specific activator protein encoded by actII-ORF4 (Fernandez-Moreno et al., 1991), while Red biosynthesis is activated via the redZ and redD regulatory genes, with levels of RedD ultimately controlling transcription of the red biosynthetic operons (White and Bibb, 1997). To determine the effect of absC deletion on the expression of these pathway-specific activator genes, RNA was isolated from liquid-grown cultures of M145 and M804 (ΔabsC) at three time points corresponding to late-exponential phase (T1), late-transition phase (T2) and early stationary phase (T3). Accumulation of Act and Red in the cultures was determined spectrophotometrically, and transcript abundance measured via quantitative RT-PCR (qRT-PCR) analysis of the isolated RNA (Fig. 4). Although there was some variation between the triplicate experiments, Act and Red production was consistently and reproducibly lower in the ΔabsC mutant cultures (Fig. 4A and B). This was reflected in the level of expression of the regulatory genes actII-ORF4 and redD, which was notably lower at both the T1 and T2 time points. In contrast, redZ expression was lower only at T1 in the mutant strain. At T3, transcription of all the regulators in the mutant strain was at least as high as that observed in the parent, and achieved levels intermediate between those seen at T1 and T2 in the parental strain. However, this was evidently not sufficient to elicit a recovery in antibiotic titres in the ΔabsC strain later in growth (data not shown).

Figure 4.

In SMM (0.35 μM ionic zinc), production of Act and Red is markedly reduced in the ΔabsC mutant and coincides with a reduction and/or delay in expression of the pathway-specific activator genes redZ, redD and actII-ORF4. (A) and (B) indicate Act and Red production normalized to total protein levels as determined in triplicate cultures at the following times: late exponential (T1, 16 h); late transition (T2, 27 h); early stationary (T3, 43 h). The average value is shown, the shaded column representing M145 and the white one M804 (ΔabsC). Error bars indicate standard deviations. RNA isolated at these times was subjected to qRT-PCR analysis to determine abundance of transcripts from the Act and Red pathway-specific regulatory genes, and the results, normalized to the internal control gene SCO4742, are shown in (C). Triplicate cultures were analysed, and the mean copy number values are shown on the y-axis. Error bars indicate standard deviations.

DNA microarray analysis reveals changes in expression of multiple genes involved in zinc homeostasis in the ΔabsC mutant, including constitutive overexpression of the coelibactin biosynthetic cluster

DNA microarray analysis of RNA isolated from the triplicate cultures at all three time points described in Fig. 4 was performed to determine the global effects of absC deletion on patterns of gene transcription. Statistical analysis using two-way anova revealed 166 genes whose expression was significantly different between the two strains at the 5% probability level (Supporting information S1). Strikingly, this included 24 genes previously reported as being under the control of the zinc-responsive regulator Zur (Owen et al., 2007; Shin et al., 2007; M. Paget, pers. comm.), including all 17 genes from the secondary metabolite gene cluster assigned to coelibactin production (Bentley et al., 2002; Table 1). qRT-PCR analysis of the samples confirmed that SCO7682, encoding the non-ribosomal peptide synthetase (NRPS) of the coelibactin gene cluster, was upregulated 6.3-, 11.2- and 53-fold in the T1, T2 and T3 samples respectively (data not shown). The organization of the coelibactin gene cluster is shown in Fig. 5, with a divergent promoter between SCO7681 and SCO7682 putatively driving expression of most if not all of the biosynthetic genes, while SCO7676–80 encode a ferredoxin homologue and four genes with putative transport functions. Potential Zur protein binding motifs (Owen et al., 2007; Shin et al., 2007) are present in the DNA sequences upstream of SCO7676 and in the intergenic region between SCO7681 and SCO7682, suggesting that production of coelibactin is at least in part regulated in response to zinc. Coelibactin has not been structurally characterized, but is a non-ribosomal peptide derivative predicted to have siderophore-like properties (Bentley et al., 2002). The transcript abundance ratios listed in Table 1 indicate that the biosynthetic genes SCO7681–92 are 1.7- to 10.9-fold upregulated in the absC mutant at all time points measured, consistent with the trend observed upon qRT-PCR analysis, while SCO7676–80 and the other genes listed are generally upregulated more than twofold only at T3.

Table 1.  Transcriptome analysis reveals 24 Zur-regulated genes that are significantly differently expressed as a result of mutation in absC, including all 17 genes from the putative coelibactin secondary metabolite gene cluster (SCO7676–SCO7692).
GeneProductP-value anovaFold changea (M804 ΔabsC/M145)
T1T2T3
  • a. 

    Average, normalized transcript abundance in M804 divided by that in M145. Values higher than twofold are highlighted in bold.

SCO0473Putative solute binding lipoprotein0.04881.21.73.7
SCO0475ABC transporter protein, integral membrane subunit0.03181.51.85.1
SCO2505Putative ABC-transporter metal binding lipoprotein0.01141.82.36.3
SCO2506Putative metal transport ABC transporter0.02021.81.94.3
SCO2507Putative metal transport ABC transporter0.007891.92.25.1
SCO3429Putative 50S ribosomal protein L280.03611.71.32.0
SCO7676Putative ferredoxin0.001352.11.82.7
SCO7677Putative secreted solute binding protein0.003162.52.15.3
SCO7678Putative metal transport integral membrane protein0.003151.81.71.9
SCO7679Putative transport system integral membrane protein0.03411.71.71.9
SCO7680Putative ABC transporter ATP binding protein0.01261.61.41.4
SCO7681Putative AMP binding ligase0.0001012.62.82.2
SCO7682Putative non-ribosomal peptide synthase3.82E-055.04.15.5
SCO7682bPutative non-ribosomal peptide synthase3.82E-053.32.73.3
SCO7683Putative non-ribosomal peptide synthase3.82E-054.94.05.3
SCO7683bPutative non-ribosomal peptide synthase8.27E-055.26.410.9
SCO7684Conserved hypothetical protein1.63E-055.13.96.9
SCO7685Conserved hypothetical protein1.63E-056.56.19.1
SCO7686Putative cytochrome P4502.00E-055.36.19.3
SCO7687Putative thioesterase3.82E-055.76.36.5
SCO7688Conserved hypothetical protein5.76E-054.85.55.1
SCO7689Putative ABC transporter ATP binding protein5.76E-055.27.67.4
SCO7690Putative ABC transporter ATP binding protein0.001213.12.82.5
SCO7691Putative lyase8.27E-053.02.93.3
SCO7692Hypothetical protein SC4C2.278.27E-053.62.71.7
Figure 5.

The coelibactin secondary metabolite gene cluster. SCO7681–92, encodes the non-ribosomal peptide synthase genes (SCO7682, 7683), other enzymes (SCO7681, 7686, 7687, 7691), ABC transporters (SCO7689, 7690) and proteins of unknown function (SCO7684, 7685, 7688, 7692). SCO7676 encodes a ferredoxin homologue, and SCO7677–80 encodes a putative metal transport system. The latter genes are predicted to be transcribed in a single operon, as are SCO7682–90 (http://scocyc.streptomyces.org.uk:14980/SCO/). The putative Zur binding motifs located upstream of SCO7676 and in the intergenic region of the divergent SCO7681 and SCO7682 genes are marked with circles.

Inhibition of Act and Red production in M804 ΔabsC depends on a functional coelibactin cluster

To determine if the high level of expression of the coelibactin gene cluster in the absC deletion mutant M804 was responsible for the deficiency in Act and Red production, an insertion mutation in SCO7682 encoding the NRPS was introduced into M804 using a mutant cosmid obtained from a library generated by in vitro transposon mutagenesis (Bishop et al., 2004). The ΔabsC SCO7682::Tn double mutant strain M905 showed normal antibiotic production on SMMS medium lacking zinc, or supplemented with 17.5 μM zinc sulphate (Fig. 6A). The suppressive effect of the SCO7682 mutation indicated that increased production of coelibactin in the absC mutant strain depressed Act and Red production. qRT-PCR analysis of transcription of SCO7682 in liquid cultures of M145 and M804 (ΔabsC) revealed that coelibactin gene expression was 10-fold higher in the absC mutant in the absence of added zinc, where Act and Red production was abolished (Fig. 6B). Figure 6B also indicates that when zinc was added at concentrations that are permissive for Act and Red production in the absC mutant, e.g. 25 μM, coelibactin expression was markedly repressed 100-fold or more in both strains, presumably via activity of the zinc-responsive regulator Zur. The zinc-repressed levels of coelibactin expression were still higher in the absC mutant than in the parent strain, consistent with dual regulation of the clusters by both Zur and AbsC. No effect on antibiotic production was observed when the SCO7682::Tn coelibactin mutation alone was introduced into M145 (data not shown).

Figure 6.

A. The defect in pigmented antibiotic production at low zinc ion concentrations in the ΔabsC mutant is restored when coelibactin production is abolished by insertional mutation in the SCO7682 NRPS gene, while all three strains produce comparable amounts at higher concentrations of zinc. Spores of the indicated that strains were inoculated onto SMMS agar plates as in Fig. 1, dried, then incubated for 7 days.
B. Mutation in absC increases expression of the coelibactin cluster by about 10-fold at low zinc concentrations, but expression is repressed 100-fold following addition of 25 μM zinc. Expression was determined by measuring SCO7682 transcript abundance by qRT-PCR (see Experimental procedures). The 0 μM figure was obtained from cultures grown to mid-exponential phase (OD450 about 0.5) in SMM lacking zinc. These cultures were then treated by addition of zinc sulphate to 25 μM, and sampled again after 30 min to give the 25 μM reading. Duplicate cultures were analysed, and mean copy number values normalized to an internal control gene, SCO4742, are given. Standard deviations are shown in brackets.

Induction of absC expression at low zinc concentrations represses expression of the coelibactin gene cluster, and restores pigmented antibiotic production to the ΔabsC mutant strain

AbsC (SCO5405) is a member of the MarR family of regulators, the vast majority of which function as repressors. Their repression of cognate promoters is usually relieved by binding of a small-molecule ligand. To assess the in vivo impact of overexpressing absC in S. coelicolor, a strain was constructed in the ΔabsC mutant background that carries a single copy of absC under the control of the thiostrepton-inducible promoter tipAp. This was achieved using pIJ8643 (see Experimental procedures), a derivative of pIJ8600 that integrates into the S. coelicolor chromosome at the ΦC31 phage attachment site (Sun et al., 2001). The resultant M804 derivative M1020 remained defective in pigmented antibiotic production on SMMS containing no added zinc when grown in the absence of thiostrepton, but produced the antibiotics normally when expression of absC was induced by the presence of 25 μg ml−1 thiostrepton (Fig. 7A). Strain M991 (M804 carrying only the integrated pIJ8600 vector) remained defective in Act and Red production under both the induced and uninduced conditions. To analyse the short-term effect of inducing absC expression on transcription of the coelibactin cluster, and on the Act and Red pathway regulatory genes actII-ORF4 and redD, M1020 was grown to mid-exponential phase in liquid culture, and transcription of absC induced by addition of thiostrepton. Half the volume of the cultures was left untreated, and RNA samples were prepared from both the induced and untreated cultures immediately before and 60 min following treatment. qRT-PCR analysis showed that absC expression was increased around 35-fold by thiostrepton induction (Fig. 7B). Over the same 60 min period, transcription of the coelibactin NRPS gene SCO7682 was reduced approximately threefold, while actII-ORF4 and redD transcription remained essentially unchanged. Expression of SCO7677 from the putative coelibactin transport cluster operon (see Fig. 5) was reduced approximately twofold.

Figure 7.

A. Induction of absC expression at low zinc concentrations restores pigmented antibiotic production in the ΔabsC mutant background. Strains M991 [M804 (ΔabsC) carrying pIJ8600] and M1020 [M804 (ΔabsC) carrying pIJ8463] were grown for 5 days on SMM lacking zinc sulphate, and containing 0 or 25 μg ml−1 thiostrepton, for the un-induced and induced conditions, respectively.
B. Following induction in liquid culture, absC represses expression of the coelibactin cluster but has no effect on the Act or Red pathway regulators. Cultures of M1020 (M804 (ΔabsC) carrying pIJ8463) were grown to mid-exponential phase (OD450 about 0.5) in SMM lacking zinc, and then divided into two aliquots. One aliquot was treated with 25 μg ml−1 thiostrepton to induce absC expression while the second was an untreated control. RNA samples were taken 60 min after treatment, and abundance of transcripts determined by qRT-PCR. The results are expressed as the ratio of transcript abundance detected in the induced cultures relative to the untreated control. Triplicate cultures were analysed, and mean copy number values normalized to an internal control gene, SCO4742, are given. Standard deviations are shown in brackets.

AbsC binds in vitro to the divergent promoter region controlling SCO7681 and SCO7682 gene expression, and protects a site adjacent to the putative Zur binding motif

The upregulation of transcription of the coelibactin gene cluster in the ΔabsC mutant (Table 1), coupled with the observed repression of SCO7677 and SCO7682 transcription following controlled induction of absC expression (Fig. 7B), suggested that AbsC might act as a direct repressor of transcription at the promoters controlling these genes. To identify potential AbsC binding sites, electrophoretic mobility shift assays (EMSAs) were performed using purified AbsC protein and radiolabelled DNA probes corresponding to the divergent promoter sequence between SCO7681 and SCO7682, and to the 440 nt upstream of the SCO7677 translational start site, which also incorporates SCO7676 and its putative promoter (Fig. 8). AbsC retarded the SCO7681/SCO7682 intergenic DNA, but had no effect on the SCO7676/SCO7677 probe. Although the kinetics of binding have not been studied, the shift observed with the SCO7681/SCO7682 probe appears moderately weak, as it could be competed away not only by an excess of unlabelled probe DNA (Fig. 8B, lane 1), but also by unlabelled SCO6158 promoter DNA known not to interact with AbsC (data not shown) and used here as a negative control (Fig. 8B, lane 2). However, DNase I footprinting studies indicated that the interaction between AbsC and the DNA in the divergent promoter between SCO7681 and SCO7682 occurred at a specific site, and was not non-specific binding (Fig. 9). The protected site spanned 46 nt and was located immediately adjacent to the putative Zur binding site (Fig. 9E). Mutation of specific nucleotides within the identified AbsC binding region markedly reduced interaction of the DNA with AbsC protein, providing further evidence for the specific nature of the interaction (Fig. 9F). Changing either the TT or AA dinucleotides located in the centre of the AbsC protected region −98,99 nt and −88,89 nt upstream respectively, from the SCO7682 translational start site (indicated in bold in Fig. 9E) to GG markedly diminished binding of AbsC to the probe DNA.

Figure 8.

AbsC binds to the DNA sequence of the intergenic promoter region between SCO7681 and SCO7682 in the coelibactin cluster, but not to a sequence containing the putative promoter(s) controlling SCO7676 and SCO7677.
A. Results from an EMSA assay in which lanes marked with ‘+’ correspond to 32P-labelled DNA probes (4 nM) incubated with purified AbsC protein (280 nM), while those marked ‘−’ contained labelled DNA but no protein. The 7676 + 7 probe corresponds to a DNA sequence 109 nt upstream of the SCO7676 gene (which is 46 nt downstream of SCO7675) to 1 nt upstream of the SCO7677 gene, and includes the putative Zur binding site in the SCO7676 gene promoter. The 7681 + 2 gene probe comprises the entire intergenic region between SCO7681 and SCO7682.
B. A repeat of the EMSA with the 7681 + 2 probe shown in (A), where lane 1 corresponds to the ‘+’ lane but with the addition of 60 ng unlabelled SCO7682 probe DNA, and lane 2 the same but with 60 ng unlabelled non-specific DNA (SCO6158) of a similar length to the 7681 + 2 probe.

Figure 9.

DNase I footprinting experiments identify an AbsC binding site in the SCO7681–SCO7682 divergent promoter region, immediately adjacent to the putative Zur binding site. Intergenic probe fragments were uniquely labelled at the 5′ end of the top (for A and B) or bottom (C and D) strands, and 10 nM incubated with the indicated amounts of AbsC protein. The incubations were analysed by EMSA (A and C), and by DNase I footprinting (B and D). Regions of protected sequence in the footprints are marked, and their exact positions determined by generating G + A (GA) sequencing ladders from the probes. The sequences corresponding to the footprinted regions are shown by shading in (E), and slightly overlap the putative Zur binding site (boxed sequence). Arrowed lines indicate the location of the primers used to generate the 200 nt DNA probe used. (F) shows that mutation of the TT or AA dinucleotides in the AbsC protected region −98,99 nt and −88,89 nt upstream, respectively, of the SCO7682 translational start site (indicated in bold in panel E) to GG relieves binding of AbsC to the probe DNA. The EMSA was performed as described in the legend to Fig. 8, with probe DNA containing a wild-type AbsC binding region (WT, from pIJ12000); probe DNA where the −98,−99 nt TT is mutated to GG (TT/GG, from pIJ12001), or where the −88,−89 nt AA is mutated to GG (AA/GG, from pIJ12002), in the presence (+) or absence (−) of AbsC protein.

Expression of SCO7681 and SCO7682, but not SCO7676 or SCO7677, in low-zinc conditions is dependent on absC rather than zur

To assess the role of absC in controlling the SCO7681/SCO7682 divergent promoter in vivo, in the absence of any effects of regulation by the zinc-dependent regulator Zur, zur was deleted from both M145 and M804 (ΔabsC). Transcription of the putative coelibactin cluster operons in all four strains was analysed by qRT-PCR, using RNA samples from cultures grown to mid-exponential phase in liquid SMM to which no zinc sulphate had been added (Fig. 10). These zinc-deficient conditions were used to minimize any repression by Zur in the zur+ strains. Transcription of SCO7676 and SCO7677 in the putative coelibactin transport operon(s) was largely independent of strain genotype, varying only one- to twofold on deletion of absC, zur or both genes. This was similar to the control gene SCO2505, the first gene from the high-affinity zinc transport operon SCO2505-2507 previously shown to be regulated by Zur in response to ionic zinc (Owen et al., 2007; Shin et al., 2007). This supports the results of the EMSA analysis in Fig. 8A, which indicated that AbsC had no direct influence on the regulation of SCO7676 or SCO7677. Expression of SCO7681 and SCO7682 was similarly unaffected following deletion of zur, but was markedly upregulated in the ΔabsC mutant or ΔabsCΔzur double mutant strains: approximately 10- to 15-fold for the NRPS gene SCO7682, and 2.5- to threefold for the putative salicylate synthase gene SCO7681. This is entirely consistent with AbsC repressing transcription of both genes via binding to the operator site identified in the divergent promoter region (Fig. 9). Confirmation that Zur can repress expression of all five genes in the zur+ strains was achieved by adding 25 μM zinc sulphate to the M145 and M804 (ΔabsC) cultures and quantifying transcript abundances after 30 min. A 50- to 500-fold decrease in transcription of SCO2505, SCO7676, SCO7677 and SCO7682 was observed, while transcription of SCO7681 was reduced 6- to 22-fold (Fig. S1). Similar treatment of strains carrying a zur deletion produced at most a twofold reduction in expression of the genes.

Figure 10.

Expression of the coelibactin biosynthetic genes SCO7681 and SCO7682 at low zinc concentrations depends on absC and not zur. The strains indicated were grown to mid-exponential phase (OD450 = 0.5) in SMM lacking zinc sulphate, and RNA was isolated for qRT-PCR analysis. Duplicate cultures were analysed, and the mean copy number values, normalized to an internal control gene SCO4742, are shown on the y-axis. Error bars indicate standard deviations. Note that the y-axis scale for SCO7681 is approximately 10-fold lower than for the other genes, which are shown on the same scale.

M1018 (Δzur ΔabsC) is defective in production of Act and Red over a wider range of zinc concentrations than M804 (ΔabsC)

The qRT-PCR analysis above indicated that in a ΔzurΔabsC double mutant, expression of the coelibactin NRPS gene SCO7682 was derepressed, and was always at least 100-fold higher than in the parental strain, irrespective of the level of zinc in the culture medium. This differed from the ΔabsC mutant strain, which exhibited similar changes in expression only in media lacking zinc. As the defect in Act and Red production in the ΔabsC mutant was zinc-dependent and attributable to upregulation of coelibactin gene cluster expression (Figs 3 and 6), pigmented antibiotic production in the ΔzurΔabsC double mutant strain was predicted to be inhibited at both high- and low-zinc concentrations. Consistent with this, Fig. 11 shows that M1018 (ΔzurΔabsC) produced Act and Red poorly compared with M145 and M804 (ΔabsC) on SMMS supplemented with 17.5 μM or higher levels of zinc sulphate. M1016 (Δzur) was similar to M145 on these media. The deficiency in M1018 depended on coelibactin production as Act and Red synthesis was restored in the triple mutant strain M1019, which carried an additional mutation in SCO7682. However, in medium with 3.5 μM added zinc, M1018 (ΔzurΔabsC) produced the pigmented antibiotics at levels similar to those observed in M145. Furthermore, the defect in production at 0.35 μM or lower added zinc was not as severe as that observed in the ΔabsC mutant strain. Thus, there is not always a direct correlation between coelibactin production and inhibition of Act and Red biosynthesis.

Figure 11.

A strain deleted for both absC and zur is reduced in pigmented antibiotic production at both high and low zinc concentrations, and this defect depends on a functional coelibactin gene cluster. Spores of the indicated strains were applied to SMMS prepared with the indicated concentrations of zinc sulphate as described in Fig. 1, and the phenotype was recorded after 7 days incubation.

Discussion

Zinc influences antibiotic production in S. coelicolor

The nutritional environment has long been known to influence production of Act and Red in S. coelicolor, and systems for monitoring cellular amino acid availability via relA, and N-acetyl glucosamine via dasR, have been described (Chakraburtty and Bibb, 1997; Hesketh et al., 2001; 2007; Sun et al., 2001; Rigali et al., 2006; 2008). High-phosphate concentrations have also been reported to have an inhibitory effect on antibiotic biosynthesis, although the molecular basis for this is less well defined (Sola-Landa et al., 2003; 2005; Rodríguez-García et al., 2007). From the current work it is clear that the availability of zinc can now be added to this list of nutritional effects, with Act production in particular being optimal over a narrow range of zinc concentrations (Fig. 1). Metal ions are very reactive, and their destructive presence in the bacterial cytoplasm is avoided by exquisite control of uptake/export mechanisms, and by protein chaperones (reviewed for B. subtilis in Moore and Helmann, 2005). Zinc plays a vital role in cellular metabolism as a structural or catalytic cofactor for a wide range of metalloproteins, and a recent bioinformatic study predicted 5.5% of the proteome of S. coelicolor (421 proteins) to be capable of binding zinc (Andreini et al., 2006). These include many essential proteins, including those required for the transcriptional and translational machinery (RNA polymerase β′ subunit; ribosomal proteins; aminoacyl-tRNA synthetases), and for DNA replication.

Is zinc required as a cofactor for antibiotic biosynthesis?

Antibiotic biosynthetic gene clusters are generally only expressed after exponential growth has ceased, and positive feedback regulation has been proposed (Wang et al., 2009) to boost cluster expression upon initial detection of the product of the pathway. When extracellular zinc is in limited supply, most of the intracellular zinc is already likely to be associated with zinc binding proteins that are essential for growth and constitutively expressed. Consequently, any zinc-requiring enzymes produced at the end of rapid growth under conditions of zinc limitation may well be deprived of the cofactor, and unable to perform their catalytic function. Interestingly the actinorhodin polyketide cyclase/dehydratase enzyme encoded by actVII (SCO5090) and essential for Act production (Fernandez-Moreno et al., 1992b) is predicted to require zinc for its activity. Impaired ActVII function could result in lack of product formation with the resultant failure to activate any feedback induction system. Such a scenario could account for the reduced levels of actII-ORF4 and redDZ transcription in the absC mutant (see below), although none of the proteins encoded by the red cluster have yet been predicted to require zinc.

An absC mutant exhibits a zinc-dependent defect in antibiotic production

Mutation of absC (SCO5405) caused a zinc-dependent pleiotropic defect in production of both the Act and Red antibiotics in S. coelicolor (Fig. 3). Global analysis of differences in the pattern of gene transcription between the parental and mutant strains revealed that many of the genes significantly upregulated as a result of absC deletion were also controlled by the zinc-responsive regulatory protein Zur, indicating some degree of zinc starvation in the absC mutant. Only 166 genes, about 2% of the transcriptome, were found to be significantly differently expressed between the strains, and the basis for the zinc starvation is not yet understood. As the high-affinity zinc uptake system znuABC (encoded by SCO2505–7) is not downregulated in the mutant, the implication of the observed starvation response is that proteins with zinc export or chaperone functions are upregulated. No obvious candidates for this were observed in the list of differentially expressed genes.

The absC phenotype is dependent on coelibactin, a putative zincophore

Strikingly, all the genes from a single genetic locus, SCO7676–SCO7692, were significantly upregulated in the ΔabsC strain. This locus is predicted to be responsible for the production of an as yet uncharacterized non-ribosomal peptide product that has been named coelibactin and is predicted to possess siderophore-like properties (Bentley et al., 2002). The presence of two putative Zur binding motifs in the promoter regions of the gene cluster suggests that it may have a role in sequestering zinc rather than iron, and expression of the cluster has been shown to be zinc-dependent (Fig. 6B and Fig. S1; M. Paget, pers. comm.). Coelibactin may therefore more accurately be described as a putative zincophore, following the naming proposed for zinc binding metabolites in Patzer and Hantke (1998). The defect in Act and Red production during growth of the ΔabsC mutant on zinc-deficient media could be restored by disruption of the coelibactin cluster (Fig. 6), suggesting that overproduction of coelibactin has an inhibitory effect on production of the antibiotics. This explanation is also favoured over a more direct influence of AbsC on expression of the clusters as controlled induction of absC expression restored production in non-permissive growth conditions, but had no immediate effect on expression of the activator genes in the clusters (Fig. 7). Transcription of the coelibactin NRPS gene SCO7682 was, however, reduced threefold in the same experiment. Using a DNA-affinity capture approach, Park et al. (2009) recently identified AbsC as one of the proteins binding in vivo to biotinylated DNA probes containing actII-ORF4 and redD promoter sequences, inferring that AbsC may directly regulate the cluster activator genes. While this may occur under certain growth conditions, the absC induction experiment in this study provided no support for this notion. Moreover, we have been unable to demonstrate convincing binding of AbsC to DNA probes containing either the redD or actII-ORF4 promoters using our EMSA conditions, and a putative AbsC binding motif derived from the transcriptome analysis in this study is not present in either promoter region (data not shown).

Coelibactin affects antibiotic production at the level of transcription

The expression of genes encoding the activator proteins responsible for switching on the Act and Red pathways was downregulated in culture conditions that caused elevated transcription of the coelibactin cluster (Fig. 4). Whether this is the result of effects at the level of transcription, or is feedback from post-translational influences on product formation, as discussed above, is unclear. It will be possible in the future to address these questions if purified coelibactin becomes available, and the overproducing mutant strains described in this study may facilitate this goal. In addition to inhibition of pigmented antibiotic production, coelibactin-dependent defects in morphological differentiation were also observed in the ΔzurΔabsC mutant (and have also been observed in a Δzur mutant (M. Paget, pers. comm.)). It is interesting to speculate that coelibactin may therefore have a signalling function, diminishing sporulation and antibiotic production in the mycelium of streptomycetes growing close to the producing organism. The ΔabsC mutant strain exhibited both constitutive upregulation of the coelibactin cluster and signs of zinc starvation in stationary phase when compared with the parental strain, and it is possible that the latter is a consequence of the former. However, the zinc starvation could equally well result from changes exerted via other members of the AbsC regulon, and if coelibactin indeed functions as a zincophore this seems the more likely explanation.

AbsC directly regulates the coelibactin cluster

In vitro studies using purified AbsC protein and radiolabelled DNA probes identified an AbsC binding site in the divergent promoter region controlling expression of the putative coelibactin biosynthesis genes SCO7681–SCO7690 (Figs 8 and 9). The AbsC-protected site overlaps the putative Zur binding site. qRT-PCR analysis provided evidence for in vivo repression of the divergent promoter by AbsC (Figs 7 and 10), supporting the microarray data that indicated upregulation in the absence of AbsC. Overall, the results suggest a direct role for AbsC in regulating expression of the biosynthetic genes in the coelibactin cluster, but not of SCO7676-7680, which encode a putative coelibactin transport system. The observed twofold repression of the latter genes following induction of absC expression (Fig. 7) may be an indirect consequence of reducing transcription of the biosynthetic component of the cluster.

Coelibactin and zinc as regulators of antibiotic production in S. coelicolor?

We have attempted to convey the observations made in this study in a model presented in Fig. 12. Coelibactin biosynthesis in S. coelicolor is apparently very tightly regulated, directly controlled by two repressor proteins, Zur and AbsC, which appear to function independently. In conditions when Zur binding is relieved by low intracellular zinc ion concentrations, repression by AbsC can still limit expression of the cluster (as seen in Fig. 10). Conversely, elevated levels of coelibactin expression in the absence of AbsC binding can be repressed by Zur if zinc availability in the cells is increased (Fig. 6B and Fig. S1). The DNA binding activity of AbsC, a MarR-family regulator, is likely to be controlled in vivo by interaction with a low-molecular-weight compound making coelibactin expression dependent on the intracellular concentrations of ionic zinc and the as yet unidentified AbsC ligand. This level of control perhaps reflects the potent effects coelibactin and zinc can have on both antibiotic production and morphological differentiation in S. coelicolor.

Figure 12.

A model illustrating the interrelationship of AbsC, Zur, zinc coelibactin and antibiotic production in S. coelicolor. Lines ending in filled circles represent inhibition/repression. Dashed lines are used for hypothetical links that have yet to be experimentally established. The question mark indicates an as yet unidentified small molecule ligand for AbsC.

Experimental procedures

Bacterial strains and plasmids

The S. coelicolor A3(2) strains (Table 2) were manipulated as described previously (Kieser et al., 2000). For routine subcloning Escherichia coli strain DH5α (Sambrook et al., 1989) was grown and transformed according to Sambrook et al. (1989). ET12567 (dam dcm hsdS-; MacNeil et al., 1992) containing the RP4 derivative pUZ8002 (Flett et al., 1997) was used to propagate unmethylated DNA and to introduce it into S. coelicolor by conjugation. E. coli BW25113/pIJ790 (Gust et al., 2003) was the host for λRED-mediated PCR-targeted mutagenesis, and E. coli DH5α/BT340 was the host for FLP recombinase-mediated deletion of disruption cassettes (Datsenko and Wanner, 2000).

Table 2. S. coelicolor A3(2) strains used in this study.
StrainRelevant genotypeSource or reference
M145PrototrophKieser et al. (2000)
M510ΔredDFloriano and Bibb (1996)
M511ΔactII-ORF4Floriano and Bibb (1996)
M801absC point mutantThis study
M804ΔabsCThis study
M904SCO7682::TnThis study
M905ΔabsC SCO7682::TnThis study
M991ΔabsC pIJ8600This study
M1016Δzur::hygThis study
M1018ΔabsCΔzur::hygThis study
M1019ΔabsCΔzur::hyg SCO7682::TnThis study
M1020ΔabsC pIJ8463This study

Growth conditions for S. coelicolor strains

Mannitol soya agar (Kieser et al., 2000) was used to make spore suspensions, for viable counts of spore suspensions, and for plating out conjugations with E. coli ET12567 pUZ8002. SMMS agar for assessing pigmented antibiotic production was prepared according to Kieser et al. (2000), but the amount of zinc sulphate added was varied as stated in the text and figures. All strains were incubated at 30°C. For growth in liquid culture, strains were cultivated with vigorous agitation at 30° in SMM as previously described (Kieser et al., 2000). Briefly, spores (about 1010 cfu ml−1) were germinated in 2× YT medium (Kieser et al., 2000) for 7 h at 30°C. Germlings were harvested by centrifugation (5 min at 4000 g), resuspended in SMM, and briefly sonicated to disperse any aggregates before inoculation into 50 ml SMM in 250 ml siliconized flasks containing coiled stainless steel springs. Each flask received the equivalent of 5 × 107 cfu. The zinc content of SMM was varied as stated in the text by adding the appropriate amount of zinc sulphate. Where not stated for either SMMS or SMM, the published concentration of 0.35 μM was used. For induction of expression of absC in strains containing vector pIJ8463, thiostrepton was added to the medium at the concentration stated.

Creation and screening of a random mutant library of strain M145

M145 spores were mutagenized by UV light or N-methyl-N-nitro-nitrosoguanidine (NTG) then plated for single colonies on SMMS (Kieser et al., 2000) and screened for Act and Red production. Four mutants (two UV-derived and two NTG-derived) were obtained that were deficient in both Act and Red production on SMMS but that sporulated normally, irrespective of the inoculum size. Complementation of these four putative pleiotropic mutants was attempted by replica-plating a pre-existing M145 genomic DNA library constructed in pIJ698 (Ryding et al., 1999) and maintained in S. coelicolor J1501, onto confluent lawns of spores of each of the four mutants. A total of 27 library clones showed restoration of pigment production, 13 to two or more of the mutants, and 14 to a single mutant. The 27 library clones were further replicated onto lawns of existing antibiotic-deficient mutants to identify complementation of known gene(s), eliminating seven from further consideration. Of the 20 remaining candidates, nine were eliminated because they complemented more than one antibiotic-deficient mutant. Of the 11 uniquely complementing clones, only one, pIJ8449, restored antibiotic production to parental levels in mutant NTG341 (later designated M801). This result was confirmed by extracting plasmid DNA from the library clone and introducing it into protoplasts of M801 by transformation (Kieser et al., 2000). To facilitate further genetic manipulations, the insert in pIJ8449 was subcloned into the shuttle vector pHJL401 (Larson and Hershberger, 1986) as an 11 kb XbaI fragment. The resultant plasmid, pIJ8450, also restored antibiotic production to M801.

Complementation of mutant M801 by subcloning the pIJ8450 insert

To identify more precisely the gene(s) that restored Act and Red production in M801, the 11 kb XbaI fragment in pIJ8450 was subcloned into the integrative vector pSET152 (Bierman et al., 1992). A 3 kb EcoRI-XbaI fragment was first cloned into pIJ2925 (Janssen and Bibb, 1993) between the EcoRI-XbaI sites in the multiple cloning region. The insert was then removed as a BglII fragment and inserted into the BamHI site of pSET152 to yield pIJ8453. Conjugation of pIJ8453 into M801 via E. coli ET12567/pUZ8002 confirmed that it was able to complement the mutant phenotype, and the 3 kb insert was sequenced using the ABI Cycle Sequencing Ready Reaction Kit such that both strands were covered with overlapping sequences. The pIJ8453 insert contained two complete (SCO5404 and SCO5405) and two partial (SCO5403 and SCO5406) ORFs (see Fig. 2) and in order to identify which was responsible for the restoration of antibiotic production further subcloning was performed. The insert in pIJ8453 was cut to separate SCO5404 and SCO5405 from each other using PstI and BclI restriction sites located between the convergent genes. A 2.4 kb XbaI-PstI fragment carrying SCO5404 (and the partial SCO5403 ORF) was cloned into similarly cut pIJ2925, excised as a BglII fragment and ligated into BamHI-digested pSET152 to yield pIJ8457. Similarly, a 1.2 kb BclI-EcoRI fragment carrying SCO5405 (and the partial SCO5406 ORF) was cloned into pIJ2925 cut with BamHI and EcoRI, and the insert removed as a BglII fragment and ligated into BamHI-digested pSET152 to yield pIJ8458. pIJ8458 fully complemented the point mutant strain M801, but led to only partial restoration of antibiotic production in the M804 deletion mutant. This is presumably due to lower than wild-type levels of expression of the complementing gene at the integrated locus of the deletion mutant strain. Complete complementation of M801 (Fig. 2B) may reflect the presence of more than one copy of pIJ8458 in the point mutant strain, or the influence of additional mutations arising from the random mutagenesis.

Construction of pIJ8463 for controlled expression of absC

To construct a plasmid in which absC expression is under the control of the thiostrepton-inducible tipAp promoter, absC was amplified by PCR using primers 5′-catatggagaccgagacggccactc and 5′-ggatccactcagggtcgtccccgctgc engineered to provide an NdeI site at the ATG start codon (underlined) and a BamHI site following the stop codon (underlined) respectively. The 475 nt fragment was subcloned into vector pGEMT (Promega) and its sequence verified. The fragment was then excised by NdeI-BamHI digestion and ligated into similarly cut pIJ8600 (Sun et al., 1999) to produce pIJ8463.

Construction of mutant strains

The mutant strains constructed in this study are listed in Table 2. The majority were made using the PCR-targeting approach according to the general method detailed in Gust et al. (2003). The procedure for making each mutant is presented below.

M804 (ΔabsC).  An in-frame deletion mutant in absC (SCO5405) was constructed by first making a marked strain in which most of the coding region (all but the first and last 12 amino acids) had been replaced by an apramycin resistance cassette. Oligonucleotide primers 5′-TCAGGGTCGTCCCCGCTGCCCGCGCAGGTGTTCCGCGATTGTAGGCTGGAGCTGCTTC and 5′-ATGGAGACCGAGACGGCCACTCGCTGGCTGACCGATACGATTCCGGGGATCCGTCGACC were used to amplify the apramycin resistance cassette in pIJ773 (Gust et al., 2003), and the resultant PCR product used to target cosmid St8F4 (Redenbach et al., 1996) in E. coli BW BW25113/pIJ790. The mutated cosmid was transferred to S. coelicolor M145 by conjugation via ET12567/pUZ8002, and the desired mutants, products of double crossovers, were identified by screening for colonies that were apramycin resistant but kanamycin sensitive. The St8F4 absC::aac(3)IV mutant cosmid generated above was introduced into E. coli DH5α/BT340 to excise the disruption cassette by FLP recombinase, producing an unmarked St8F4 ΔabsC cosmid. This was introduced into the S. coelicolorΔabsC::aac(3)IV mutant by transformation via ET12567, selecting for single crossover transformants that were both apramycin and kanamycin resistant. One such strain was taken through a round of non-selective growth on mannitol soya agar, and spores were plated for single colonies and screened for loss of both apramycin and kanamycin resistance. Correct in-frame and unmarked deletion of absC in the sensitive colonies was confirmed by PCR and by Southern blotting using probe DNA generated from cosmid St8F4 by random oligo-priming. The verified strain was designated M804.

M1016 (Δzur::hyg).  A mutant in which the zur (SCO2808) coding region was replaced by a hygromycin-resistance cassette was made from M145 as above, but using primers 5′-GGCCCGGCAAGCCGCGAAGACGTGAGGAGGAATCCAGTGATTCCGGGGATCCGTCGACC and 5′-TACAGGGGCCCGGGGCACAGCCCCGGCCCCACCGGCTCATGTAGGCTGGAGCTGCTTC and template pIJ10700 (S. O'Rourke, unpubl. data) to generate the targeting cassette, with cosmid StC121 (Redenbach et al., 1996) as the target. Candidate hygromycin-resistant but kanamycin-sensitive mutants were verified by PCR and Southern blotting.

M1018 (ΔabsC Δzur::hyg).  A double mutant strain carrying both an in-frame deletion of absC and a replacement/deletion of zur was constructed as for strain M1016 above, but using strain M804 ΔabsC as the recipient of the mutated cosmid.

M904 (SCO7682::Tn), M905 (ΔabscC SCO7682::Tn), M1019 (ΔabsC Δzur SCO7682::Tn).  Coelibactin-minus mutant strains were constructed using transposant SC4C2.2.E03_04110412I3.seq (StrepDB) from a library of in vitro transposon mutagenized S. coelicolor cosmids (Bishop et al., 2004). This represents cosmid St4C2 with a transposon insertion at position 20874 nt, corresponding to a site 279 nt into the SCO7682 coding region. The mutant cosmid was introduced into S. coelicolor strains M145, M804 or M1016 by conjugation via ET12567/pUZ8002, and the desired mutants, products of double crossovers, identified by screening for colonies that were apramycin-resistant but kanamycin-sensitive.

Actinorhodin and undecylprodigiosin assays

Actinorhodin and undecylprodigiosin were quantified spectrophotometrically (Bystrykh et al., 1996; Kieser et al., 2000). Values were normalized to the total protein content of each culture sample, as determined using the Bio-Rad DC protein assay kit (Bio-Rad).

RNA isolation and DNA microarray analysis

RNA was isolated from liquid cultures according to Hesketh et al. (2007). Purified total RNA (10 μg) was processed into labelled and fragmented cDNA for hybridization to Streptomyces diS_div712a Affymetrix GeneChip arrays as previously described (Hesketh et al., 2007). The labelled fragmented cDNA was hybridized to the GeneChips in a Hybridization Oven model 640 (Affymetrix) according to protocols provided by the manufacturer. The GeneChips were washed and stained with streptavidin-phycoerythrin using GeneChip fluidics workstation model 450, and then scanned with a Gene Array Scanner, Model 3000 7G. Following scanning of the arrays, the data quality was verified using a variety of tools including the ‘affyPLM’, ‘affy’ and ‘simpleaffy’ packages for the statistical computing environment R (R Development Core Team, 2005), quality control methods available within GeneSpring 9 (Agilent), and data from report files generated in the Affymetrix Genechip Operating Software. To identify differentially expressed genes, array data were first imported into GeneSpring 9 normalizing using the Robust Multichip Average algorithm of Irizarry et al. (2003). The data were then filtered to remove genes deemed to be expressed at a level below reliable detection by determining those with a raw signal value below a defined background cut-off value of 20 in all samples. The results were further filtered to remove genes not significantly changing under the conditions of the experiment by identifying those with normalized expression values between 0.8 and 1.2 (1.5-fold change limit) in all conditions. The filtered data (2634 genes) were then subjected to two-way anova to identify genes significantly altered under the experimental conditions. This was performed using the parametric test option with a false discovery rate of P < 0.05, and assuming variances to be equal. P-values were corrected using the Benjamini and Hochberg false discovery rate multiple testing correction procedure. Details of the statistical calculations used in the software can be accessed through the manufacturer's manual.

qRT-PCR analysis

The qRT-PCR analyses were performed using 5 μg of total RNA as starting material as described in Hesketh et al. (2007). All determinations were performed in triplicate, and the results were analysed using Opticon 2 Monitor software (MJ Research). All values were ultimately normalized to an endogenous control gene, SCO4742, selected from S. coelicolor microarray data as being constantly and constitutively expressed in the conditions used. Normalizing to hrdB, commonly used as the control gene in S1 nuclease transcription studies, produced similar results to SCO4742 for all experiments where tested. Control samples from cDNA synthesis lacking reverse transcriptase gave values comparable to background in all cases, indicating that the RNA samples were not contaminated with genomic DNA. Primer pairs used to quantify expression of the genes analysed were as follows: for actII-ORF4, 5′-gaactccggtccctggtaat and 5′-cccagttcgtcggacagtat; for redD, 5′-acccagcctgtacaacttcg and 5′-gatcgatacgggtcccaata; for redZ, 5′-tcaccgaagtcaatgccata and 5′-gtcttgccctgggtcagtaa; for SCO7676, 5′-ttcgaccaggacgaggag and 5′-gaagagtgagcgctccaga; for SCO7677, 5′-gaccctcgacttcttcaacg and 5′-atgtcgattcccttgagcac; for SCO7681, 5′-acacgttgcgtggctactac and 5′-ggttgatctggtccttgagc; for SCO7682, 5′-ttgatgtccctggtcggta and 5′-gtgaggtcctggaaggtgac; for SCO2505, 5′-gccaccaaggtcttcttcac and 5′-ctgctggagttccttgatcc; for SCO4742, 5′-actgcgtacagcgtggaaac and 5′-cagccgtgccatcagttc.

EMSA analysis

Purified AbsC protein for use in in vitro studies was kindly supplied by Clare Stevenson, having been prepared according to Stevenson et al. (2007). DNA probes were prepared by PCR as detailed below, and 100 ng radiolabelled by incubation with [γ-32P]-ATP (740 KBq; 111 TBq mmol−1) and T4 polynucleotide kinase. EMSA incubations were typically performed in 20 μl volumes containing 280 nM AbsC and 4 nM labelled probe, in the presence of 20 mM Tris-HCl (pH 8.0) 100 mM KCl, 5 mM MgCl2; 1 mM EDTA, 0.5 mM DTT, 10% v/v glycerol, 0.1 mg ml−1 BSA and 0.05 μg ml−1 poly(dI-dC). Following incubation at room temperature for 25 min, the binding reaction mixtures were separated on a 5% polyacrylamide gel, and bands visualized using a Fujifilm FLA-7000 phosphorimager. The following EMSA probes were used: (i) 7676 + 7 (Fig. 8) was prepared by PCR using primers 5′-GTCGGGGAGAGTATTGGTGC and 5′-GCTTCTCCTTGACGTTCTGC and cosmid St4C2 (Redenbach et al., 1996) as template. The 440 nt product covers the sequence from 109 nt upstream of the annotated SCO7676 start codon to 1 nt upstream of that of SCO7677, and includes the putative Zur binding motif in the SCO7676 promoter. (ii) 7681 + 2 (Fig. 8), also made by PCR from St4C2 but with primers 5′-GGGTTACGCCTCTCATTCAT and 5′-CGGCGTCTCCTTCCACTGTC, producing a 177 nt product covering the intergenic region between SCO7681 and SCO7682 in its entirety. (iii) SCO6158 (Fig. 8) was used as a cold probe only and was prepared using primers 5′-GCTGTGCTTCACGGGTCCT and 5′-GTACGGGCGACAGGCATG with cosmid St1A9 (Redenbach et al., 1996), producing a 200 nt product. (iv) 7681 + 2 WT (Fig. 9) was synthesized using primers 5′-GGGTTACGCCTCTCATTCAT and 5′-TTTCATATGGGTCCTCCGGATCCTCGAGTGTCCGCGCCTCGACGACTGCTTCAT with pIJ12000 as template producing a 188 nt product covering the intergenic region between SCO7681 and SCO7682 in its entirety, but possessing 28 nt of engineered sequence at its 3′ end. (v) 7681 + 2 TT/GG (Fig. 10) was made using the same primer pair as for the 7681 + 2 WT probe, but with pIJ12001 as template. The resultant 188 nt probe therefore contains mutations changing the TT dinucleotide located −98 to −99 nt upstream of the SCO7682 start codon to GG. (vi) 7681 + 2 AA/GG (Fig. 10) was also synthesized using the same primer pair as for the 7681 + 2 WT probe, but with pIJ12002 as template. The resultant 188 nt probe therefore contains mutations changing the AA dinucleotide located −88 to −89 nt upstream of the SCO7682 start codon to GG.

DNaseI footprinting

To define the location of the AbsC binding site in the SCO7681–SCO7682 intergenic region, DNaseI footprinting studies were performed using probes generated by PCR using primers FP7682f (5′-CCGATTCGGCTTCGGACCAG) and FP7682r (5′-TGTCCGCGCCTCGACGACTG) with cosmid St4C2 as template. To generate uniquely end-labelled probes the PCR was performed with one labelled primer and one unlabelled primer. For the top strand probe, the FP7682f primer was radiolabelled prior to use in the PCR by incubation with [γ-32P]-ATP (740 KBq; 111 TBq mmol−1) and T4 polynucleotide kinase. Similarly, to produce a probe for the bottom strand the FP7682r primer was radiolabelled prior to the PCR reaction. Binding reactions were performed in the same buffer as described for the EMSA assays, but 50 μl volumes were used containing 10 nM of probe DNA and a range of AbsC concentrations from 0 to 25 nM (see Fig. 9). After incubation at room temperature for 25 min, 10 μl of the reaction mixtures was removed for EMSA analysis as described above. The remaining 40 μl was incubated for a further 5 min before adding 1 μl 100 mM CaCl2 and 1.2 units of FPLC pure DNaseI (GE Healthcare). The reactions were mixed thoroughly and allowed to digest at room temperature for 60 s before quenching by addition of 140 μl stop solution comprising 192 mM sodium acetate, 32 mM EDTA, 0.14% SDS and 70 μg ml−1 yeast tRNA. The reactions were extracted with phenol-chloroform (300 μl) before ethanol-precipitating the DNA. Pellets were resuspended in formamide loading buffer and separated on a 6% sequencing gel along with G + A Maxam Gilbert sequencing ladders generated from the labelled probe used. Separations were visualized using a Fujifilm FLA-7000 phosphorimager.

Site-directed mutagenesis of plasmid-borne SCO7681–SCO7682 intergenic DNA sequence

To generate EMSA probes carrying defined mutations in the SCO7681–SCO7682 intergenic DNA sequences, plasmid PCR templates were first created as follows. To make a plasmid carrying the M145 parental sequence, the intergenic region was amplified from cosmid St4C2 by PCR using the oligonucleotide primers 5′-CTAGGTACCCACGGCGTGCCGATTCGGCTTCGGACCAGGT and 5′-TTTCATATGGGGCCTCCGGATCCTCGAGTGTCCGCGCCTCGACGACTGCTTCAT generating a 246 nt product. The PCR product was cloned directly into the pGEMT easy vector (Promega), and designated pIJ12000 following confirmation of its sequence. To change the TT dinucleotide located −98 to −99 nt upstream of the SCO7682 start codon to GG, site-directed mutagenesis was performed on pIJ12000 using the Quikchange kit from Stratagene according to the manufacturer's instructions, employing mutagenic primers 5′-CGCCGAGTCCCGTCGTGTGAGGACGATGTGAAC and 5′-GTTCACATCGTCCTCACACGACGGGACTCGGCG. Following confirmation of the change by sequencing, the plasmid was designated pIJ12001. Similarly, to change the AA dinucleotide located −88 to −89 nt upstream of the SCO7682 start codon to GG, the mutagenic primers 5′-GTCCCGTCGTGTGATTACGATGTGGGCTATAGAAATGATAATCATTTCTAT and 5′-ATAGAAATGATTATCATTTCTATAGCCCACATCGTAATCACACGACGGGAC were used. The verified plasmid was designated pIJ12002.

Acknowledgements

We thank Janet White and Oliver Harris for technical assistance in creating the M145 mutant library. We thank Clare Stevenson for supplying purified AbsC protein, and Paul Dyson for mutated cosmid SC4C2.2.E03_04110412I3, and are grateful to David Hopwood, Keith Chater and David Lawson for critical reading of the manuscript. This work was funded by grants to J.I.C. from the Biotechnology and Biological Sciences Research Council. H.K. was supported by the Deutsche Forschungsgemeinschaft.

Ancillary