The Streptomyces produce a plethora of secondary metabolites including antibiotics and undergo a complex developmental cycle. As a means of establishing the pathways that regulate secondary metabolite production by this important bacterial genus, the model species Streptomyces coelicolor and its relatives have been the subject of several genetic screens. However, despite the identification and characterization of numerous genes that affect antibiotic production, there is still no overall understanding of the network that integrates the various environmental and growth signals to bring about changes in the expression of biosynthetic genes. To establish new links, we are taking a biochemical approach to identify transcription factors that regulate antibiotic production in S. coelicolor. Here we describe the identification and characterization of a transcription factor, designated AtrA, that regulates transcription of actII-ORF4, the pathway-specific activator of the actinorhodin biosynthetic gene cluster in S. coelicolor. Disruption of the corresponding atrA gene, which is not associated with any antibiotic gene cluster, reduced the production of actinorhodin, but had no detectable effect on the production of undecylprodigiosin or the calcium-dependent antibiotic. These results indicate that atrA has specificity with regard to the biosynthetic genes it influences. An orthologue of atrA is present in the genome of Streptomyces avermitilis, the only other streptomycete for which there is a publicly available complete sequence. We also show that S. coelicolor AtrA can bind in vitro to the promoter of strR, a transcriptional activator unrelated to actII-ORF4 that is the final regulator of streptomycin production in Streptomyces griseus. These findings provide further evidence that the path leading to the expression of pathway-specific activators of antibiotic biosynthesis genes in disparate Streptomyces may share evolutionarily conserved components in at least some cases, even though the final activators are not related, and suggests that the regulation of streptomycin production, which serves an important paradigm, may be more complex than represented by current models.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Mycelial bacteria of the genus Streptomyces and their relatives are of great medical and commercial importance through their production of some 70% of clinically useful antibiotics, in addition to many other therapeutic agents such as antihelminthics and anticancer agents. Moreover, such substances are only a subset of the vast range of secondary metabolites produced by this group (for review, see Demain, 1998). Many, if not most, Streptomyces species are capable of producing more than one secondary metabolite (for review, see Challis and Hopwood, 2003) and the timing of the production of secondary metabolites and the quantities generated are exquisitely sensitive to growth and environmental conditions (for review, see Martin and Demain, 1980). Secondary metabolites are typically, although not invariably, produced when the period of most rapid growth gives way to slower growth and eventually the production of aerial hyphae and spores (for review, see Demain and Fang, 2000). Furthermore, production may be affected by the nature and levels of the carbon and nitrogen source, the availability of phosphate (Martin and Demain, 1980; Martin, 2004), elevated temperature (Deeble et al., 1995) and plasmid carriage (Thomas et al., 1991). It has also been associated with the accumulation of small signalling molecules such as the stringent factor ppGpp (Hesketh et al., 2001; Jin et al., 2004) and a family of γ-butyrolactone microbial hormones (for reviews, see Horinouchi, 2002; Chater and Horinouchi, 2003). Where such effects have been studied in detail, they appear largely to reflect changes in the level of transcription of genes encoding activators that are pathway specific and found with the corresponding biosynthetic genes as part of a cluster (e.g. Vujaklija et al., 1991; Takano et al., 1992; Gramajo et al., 1993; for review, see Chater and Bibb, 1997). It has also been proposed that certain pathway-specific activators are regulated at the post-transcriptional level by the availability of the bldA leucyl-tRNA, which recognizes a rare codon (Leskiw et al., 1991). However, while it is clear that the bldA tRNA is required for the translation of these genes (Fernandez-Moreno et al., 1991), it has not been shown that the reported accumulation of the active form (Leskiw et al., 1993) provides a molecular switch for regulating gene expression.
As a means of establishing the regulatory pathways that control antibiotic production by this important bacterial genus, the model species Streptomyces coelicolor and its relatives have been screened to identify regulatory mutants. This has led to the identification of numerous genes in S. coelicolor that affect the production of antibiotics, e.g. the absA1,2 (Adamidis et al., 1990; Brian et al., 1996; Anderson et al., 2001), afsR,K (for review, see Horinouchi, 2003) and cutR,S (Chang et al., 1996) two-component systems, the scbAγ-butyrolactone synthase and associated scbR transcriptional repressor (Takano et al., 2001), the hrdB (afsB) sigma factor (Horinouchi et al., 1983; 1989; Aigle et al., 2000) and the absB RNase III homologue (Adamidis and Champness, 1992; Price et al., 1999). Mutations that in addition affect morphological development have also been identified, e.g. relA and C, which are involved in ppGpp production (Strauch et al., 1991; Ochi et al., 1997; Sun et al., 2001), the adpA (bldH) araC-like transcriptional regulator (Nguyen et al., 2003) and the bldA tRNAleu (Leskiw et al., 1991; 1993), which is necessary for the translation of adpA (Nguyen et al., 2003; Takano et al., 2003) as well as the pathway-specific activators of the biosynthetic genes of actinorhodin and undecylprodigiosin (Fernandez-Moreno et al., 1991), two pigmented and chemically unrelated antibiotics produced by this species (Chater and Bibb, 1997). Remarkably, despite the successful characterization of numerous genes, there is still no overall understanding of the regulatory network that integrates the various environmental and growth signals to bring about the production of antibiotics in S. coelicolor. However, as the initiation of secondary metabolism in most cases has been shown to be associated with increased transcription of a pathway-specific activator, the question of what regulates secondary metabolism can be addressed by identifying and characterizing transcription factor(s) that regulate the promoters of the pathway-specific regulators. Here we describe the characterization and identification of a transcription factor that is required for maximum transcription of actII-ORF4, the pathway-specific regulator of the actinorhodin biosynthetic genes. Disruption of the corresponding gene was found to reduce the production of actinorhodin, but had no detectable effect on the production of undecylprodigiosin or the calcium-dependent antibiotic. We also show that this transcription factor binds in vitro to a site upstream of the promoter of strR, a transcriptional activator unrelated to actII-ORF4 that is the final regulator of streptomycin production in Streptomyces griseus. Our results are discussed with regard to current models for the regulation of antibiotic production and the role of evolutionarily conserved components.
An activity in S. coelicolor binds to two sites flanking the promoter of actII-ORF4
We determined that a DNA-binding protein in S. coelicolor binds with specificity to the promoter region of actII-ORF4 using an electrophoretic mobility shift assay (EMSA) (Fig. 1). A 420 bp DNA fragment that encompassed the known promoter and 5′ end of actII-ORF4 and the 3′ end of the actII-ORF3 gene (Fig. 1A), which is located upstream, was incubated with an ammonium sulphate (AS; 0–40%) fraction from S. coelicolor D132. This strain, which is believed to be unaltered from the original A3(2) (D.A. Hodgson, pers. comm.), had been grown in minimal Streptomyces medium (MSM) as described previously (Potter and Baumberg, 1996). Two complexes formed with the 420 bp actII-ORF4 promoter fragment, which had been radiolabelled (Fig. 1B, lanes 1–3): the slower migrating complex was only detectable at the highest protein concentration used (lane 3). Furthermore, the proportion of radiolabelled fragment that was bound was reduced significantly by the addition of a 10-fold excess of unlabelled actII-ORF4 DNA (lane 5). The addition of an equal amount of unlabelled actII-ORF4 DNA also reduced binding, although not to the same extent (compare complex II, lanes 3 and 4). In contrast, the addition of unlabelled promoter regions of the vdh (valine dehydrogenase; lanes 6 and 7) and actI-III (lanes 8 and 9) genes, which are unlikely to be under the same regulation as actII-ORF4, did not have a substantial effect on binding to the actII-ORF4 promoter. Combined, these results suggested that the binding to the promoter region of actII-ORF4 occurs at two sites and has specificity. The actII-ORF4 promoter binding (AII4B) activity was also found at a lower level in a 40–60% AS fraction, but not in a 60–80% AS fraction (data not shown).
To identify the regions that are bound by the AII4B activity, we determined the effect of deleting segments within the actII-ORF4 promoter fragment (Fig. 2). The 420 bp fragment was cut with restriction enzymes to generate a set of subfragments that were shortened from the end encoding the 5′ end of actII-ORF4 gene. These were then purified and used as substrates in additional EMSAs that incorporated competition reactions. Similar to what was found for the 420 bp fragment (Fig. 2A), two discrete complexes were detected using a 387 bp DdeI subfragment (Fig. 2B). In contrast, 310 bp HphI, 161 bp AluI and 150 bp BsgI subfragments (Fig. 2C, D and E respectively) produced only single complexes (lane 3) that were not affected by the addition of unlabelled actI-III promoter region (lanes 4 and 5), but were reduced by addition of unlabelled actII-ORF4 promoter region (lanes 6 and 7). Binding to a 56 bp Eco0109 subfragment was not detected (Fig. 2F). Collectively the above results suggested that sequences between positions 56–150 bp (Region 1) and 310–387 bp (Region 2) within the actII-ORF4 promoter region are required for binding (Fig. 2G).
To determine whether the two regions identified as being required for binding were also sufficient, two DNA fragments representing Regions 1 and 2 were produced by polymerase chain reaction (PCR). The need to produce matched primers for PCR meant that the precise end points of these fragments were nucleotides 54 and 146 and nucleotides 299 and 399 respectively. EMSA analysis of binding to Regions 1 and 2 (Fig. 3) produced a single complex with each of the two fragments (lanes 2 and 3). In addition, unlabelled actII-ORF4 promoter region reduced the proportion of labelled DNA that was shifted in each of the binding reactions (lanes 4 and 5). The proportion of labelled Region 1 that was bound was also reduced by the addition of unlabelled Region 2 (Fig. 3A, lanes 6 and 7), and by the addition of unlabelled Region 1 (Fig. 3A, lanes 8 and 9). Similarly, binding to Region 2 could be competed by Region 1 as well as Region 2 (Fig. 3B, lanes 6–9). In contrast, the amount of labelled DNA that was shifted was not reduced by the addition of unlabelled actI-III or vdh promoter regions (lanes 10 and 11 and 12 and 13 respectively). Examined together, the results of the above assays indicate that a single DNA binding activity in the 0–40% AS fraction specifically recognizes both Region 1 and Region 2 within the actII-ORF4 promoter region. Moreover, compared with Region 1, a greater proportion of Region 2 was bound by the All4B activity at equivalent protein concentrations (cf. Fig. 3A with B, lanes 2 and 3) and Region 2 was found to be a better competitor than Region 1 independent of the specific fragment that was labelled (cf. lanes 6 and 7 with lanes 8 and 9 in Fig. 3A and B). Taken together these results indicate that the AII4B activity has a higher affinity for Region 2 than Region 1.
The sites of binding within Regions 1 and 2 of the actII-ORF4 promoter were then identified using phenanthroline copper footprinting of complexes within EMSA gels (Sigman et al., 1979; Papavassilou, 2001). As shown in Fig. 4, single footprints (areas of reduced cleavage) were detected on each DNA strand within both Region 1 and Region 2 (Fig. 4A and B respectively). The densitometric analysis of each strand was used to determine as precisely as possible the extent of each footprint. The footprint within Region 1 extends on both strands between positions 82 and 100 within the 3′ coding region of actII-ORF3; whereas the footprint within Region 2 extends on both strands between positions 326 and 351 within the 5′ coding region of actII-ORF4 (see Fig. 1). Examination of the sequences within the footprints suggested the presence of a motif that contains imperfect inverted repeats of the hexanucleotide GGAAT(G/C) sequence separated by three base pairs. The location of binding sites of transcriptional regulators within coding regions is not unusual, e.g. Bacillus subtilis PhoP, which regulates the phosphate starvation response, binds at sites internal to the 5′ coding regions of at least three genes it activates (Liu et al., 1998; Paul et al., 2004).
Identification of the activity that binds the actII-ORF4 promoter region
Having established that S. coelicolor has a factor that can bind with specificity to sites flanking the promoter of actII-ORF4, our next objective was to purify the AII4B activity sufficiently to allow its identification. We were able to isolate the majority of the AII4B activity in a 20–40% AS fraction. This was dialysed and fractionated further using anion-exchange (Q-Sepharose) chromatography. Fractions containing binding activity were pooled, loaded onto a column derivatized with Region 2 DNA and eluted using a gradient of lithium chloride. The yield and fold purification at each step is summarized in Table 1. After EMSA analysis, the remainder of the fraction eluted from the DNA-affinity column that contained the peak of AII4B activity was precipitated and analysed by SDS-PAGE (Fig. 5A). Many polypeptides were visible after staining with Coomassie blue; therefore, to identify candidates for site-specific binding to the actII-ORF4 promoter, we removed slabs of acrylamide corresponding to individual bands and incubated with trypsin to release peptides that were then analysed by mass spectrometry. The peptide mass fingerprints were compared with the predicted tryptic digest fragments of entries in protein databases using ProFound™ (Genomic Solutions) and Mascot™ (Applied Biosystems) software. In this way, two putative DNA-binding polypeptides from S. coelicolor were identified, a 30 kDa putative transcriptional regulator (SCO4118) and a 20 kDa probable single-stranded DNA-binding protein (SCO3907). However, from what is known about the biochemistry of its Escherichia coli orthologue (Meyer and Laine, 1990), the latter was excluded as it is unlikely to bind double-stranded DNA at specific sites.
Table 1. Partial purification of activity that binds the actII-ORF4 promoter region.
Total protein (mg)
To establish that the putative transcriptional regulator was the source of sequence-specific binding to the promoter region of actII-ORF4, the open reading frame of SCO4118 was amplified by PCR, expressed using the pET system in E. coli (Studier et al., 1990) and the recombinant gene product, which was tagged at the N-terminus with a His6 oligopeptide, was purified to near homogeneity using immobilized metal affinity chromatography (Fig. 5B). We found that this recombinant protein was able to bind the promoter region of actII-ORF4, but not that of vdh (Fig. 5C). The above results were consistent with those of our binding studies using partially purified fractions from S. coelicolor(Figs 1–4) and identified SCO4118 unambiguously as the source of the AII4B activity. We have designated the corresponding gene atrA (actinorhodin-associated transcriptional regulator). From estimating the protein concentration at which 50% of the Region 2 and actII-ORF4 DNA was bound by AtrA, we derived an apparent equilibrium dissociation constant (Kd′) of ∼150 nM assuming, as indicated by size exclusion chromatography (data not shown), that AtrA functions as a dimer. This level of in vitro affinity is similar to that reported for other specific protein: nucleic acid interactions that are considered to be of biological significance such as the binding of the bacteriophage T4 transcriptional activator, MotA, to the T4 middle promoter (Sharma et al., 1999) and the binding of the B. subtilis repressor/activator, AhrC, to the promoters of arginine catabolic genes (Miller et al., 1997).
The atrA gene is located within the central core of the linear chromosome of S. coelicolor (co-ordinates 4524394–4523501; Bentley et al., 2002) at a distance from known or putative clusters encoding secondary metabolite genes. It encodes a product of 297 amino acids. The atrA gene is flanked on its 3′ side by a gene that encodes a possible membrane protein of unknown function and is transcribed in the same direction (SCO4117), and on its 5′ side by two genes that are transcribed divergently from atrA and encode a possible NADH dehydrogenase (SCO4119) and a conserved hypothetical protein (SCO4120) of unknown function (Fig. 6). Although atrA and the gene encoding the possible membrane protein are encoded on the same strand, they are not necessarily co-transcribed: the intergenic region of 183 bp that separates these genes contains an inverted repeat that may encode a transcriptional terminator (data not shown).
The phenotypic effects on S. coelicolor of disrupting the atrA gene
To investigate the contribution of atrA to the regulation of actinorhodin production, we used a disruption cassette to replace much of the coding region of this gene in M145, the sequenced strain of S. coelicolor. The resulting strain was designated L645 (Fig. 6). AII4B activity was not detected in extracts from this strain (data not shown), suggesting that AtrA is the major activity that binds the actII-ORF4 promoter under the growth and experimental conditions used here (see Experimental procedures). Also included as controls in this study were L645 and M145 containing atrA encoded on pGU103, a SCP2*-based plasmid, and L645 into which had been introduced an atrA-containing construct pGU205 that does not replicate autonomously in streptomycetes. Instead pGU205 integrates via homologous recombination with segments that flank the disruption cassette in the coding region of atrA (Fig. 6). Two different recombinants were studied. L650 was produced via a single cross-over with pGU205 that resulted in a wild-type allele being integrated adjacent to the disrupted copy, whereas L655 was produced via a double cross-over that resulted in the disruption cassette being replaced with the wild-type atrA gene (Fig. 6). The latter strain should be equivalent to M145.
The phenotype of L645 was compared with wild-type M145, L650 and plasmid-containing strains by inoculating patches on R5 plates. After incubation at 30°C for 3 days, blue pigment corresponding to actinorhodin was readily detectable in the agar surrounding M145. In contrast, no halo was detected around L645 at that time (data not shown). By day 4, however, a faint blue halo of pigment surrounded L645: the intensity and size of the blue halo around M145 was indistinguishable from that recorded after 3 days (Fig. 7). Actinorhodin production was restored to normal levels when the mutant allele in L645 was replaced with a wild-type copy to generate strain L655 and when a wild-type copy of atrA was introduced into L645 either as part of the pGU103 plasmid or through the integration of pGU205 via a single cross-over. When combined, these results show that actinorhodin production is substantially reduced in L645 and that this phenotype is due to the disruption of atrA rather than polar effects on downstream genes or a mutation in another locus.
We also investigated whether disruption of atrA affects the production of the two other well-characterized antibiotics synthesized by S. coelicolor. After incubation for 2 days on R5 plates the mycelium of strain L645 was as red as that of M145 and L655 (Fig. 7B) indicating that atrA is not required for the production of red-pigmented undecylprodigiosin. Similarly, using Bacillus mycoides as an indicator strain, we found that strain L645 in the presence of calcium produces a zone of growth inhibition that is indistinguishable from those produced by M145 and L655 (Fig. 7C). Combined, these results show that unlike most, if not all, of the previously described transcription factor genes that are not physically associated with antibiotic gene clusters, atrA has specificity with regard to the biosynthetic genes it influences. Scanning electron microscopy of the mycelial surface of L655 during morphological development on R5 plates did not reveal any obvious defects in the production of aerial hyphae or spores (data not shown).
The effect of the atrA knockout on transcription of actII-ORF4
The relative level of actII-ORF4 mRNA in samples isolated from M145, L645 and L655 after incubation at 30°C on R5 plates for different periods of time was analysed using quantitative reverse transcription polymerase chain reaction (RT-PCR) (see Experimental procedures). Time points were analysed that represented distinct phases in the development of S. coelicolor. After 44 h, the mycelium of all three strains was pink in colour indicative of the start of undecylprodigiosin production. By 50 h, the mycelium of M145 and L655 contained small areas that were blue indicative of the start of actinorhodin production. In contrast, the mycelium of L645 showed no signs of blue-pigment production and was uniformly red in colour. After 64 h, the agar beneath and surrounding M145 and L655 was dark blue in colour whereas there was only slight blue pigmentation beneath L645 at this stage. After 160 h, the phenotype of L645 was less distinct from that of M145 and L655. The agar of all three was blue in colour and the mycelium was dark purple with patches of white indicative of the start of morphological development. Transcriptional analysis revealed, as reported previously by others (Gramajo et al., 1993), that the level of the actII-ORF4 transcript peaked in M145 when actinorhodin was first detected in the medium: between 44 h and 64 h there was an increase in transcription of actII-ORF4 of ∼30-fold (Fig. 8). The transcription profile of L655 was very similar to that of M145. In contrast, the level of the actII-ORF4 transcript was significantly lower in L645 even though it also peaked at 64 h. At 50 h when areas within the mycelium of M145 showed the first signs of the production of blue-pigmented actinorhodin, the level of actII-ORF4 mRNA was ∼4.5-fold lower in L645 compared with M145. Moreover, this result was reproducible in experiments using RNA isolated from a different set of mycelial patches that were grown independently of the set that was used to generate the results presented here (see Fig. S1 in Supplementary material). The RT-PCR results are consistent with atrA encoding a transcriptional activator that is required for the normal level of transcription of actII-ORF4.
Binding to the promoter of the strR promoter region of S. griseus
To assess whether homologues of atrA may be involved in controlling the expression of antibiotic genes in other Streptomyces species, we used electrophoretic mobility shifts to assay binding to a fragment that encompassed the promoter region of strR(Fig. 9A), the transcriptional activator of the streptomycin biosynthetic genes in S. griseus (Horinouchi and Beppu, 1994; Retzlaff and Distler, 1995; Beyer et al., 1996). The strR gene was chosen as it represents the final step of one of the best-characterized regulatory networks controlling secondary metabolite production in streptomycetes (Chater and Horinouchi, 2003). Using extracts from S. coelicolor, we detected two complexes both of which could be competed strongly by the addition of segments of the actII-ORF4 promoter region that contained binding sites for AtrA, but not by the addition of unlabelled actI-III or vdh promoter regions (data not shown). To investigate the relationship with the sites in the promoter of actII-ORF4, we mapped one of the two binding sites in the strR promoter. We first used competition assays (Fig. 9B) to narrow down binding to a 278 bp fragment (designated C), which contained the strR promoter. This fragment was then subdivided further and a 105 bp subfragment was identified (designated E) that formed a single complex when incubated with S. coelicolor fractions containing AtrA. By footprinting this complex in situ, we mapped the binding site (Fig. 9C), to a position 118–140 bp upstream of the known transcription start site of strR. This position is located between the known binding sites for AdpA (Yamazaki et al., 2004) and contains imperfect hexameric repeats that are similar in terms of sequence and spacing to those found in the binding sites that flank the actII-ORF4 promoter. Fragments containing this binding site can compete effectively with those in the promoter region of actII-ORF4 for AtrA (data not shown); moreover, a double-stranded oligonucleotide corresponding to the footprinted region in the strR promoter region is bound in vitro by recombinant AtrA purified from E. coli (Fig. 9D). The affinity of AtrA for this substrate is similar to its affinity for Region 2 of the actII-ORF4 promoter.
AtrA and pathways that regulate the production of antibiotics
Our studies have identified for the first time a transcription factor (designated AtrA) that binds within the promoter region of actII-ORF4 (Figs 1–5), the pathway-specific activator of the actinorhodin genes in the model species S. coelicolor (Fernandez-Moreno et al., 1991; Gramajo et al., 1993). By continuing to identify and characterize transcription factors that regulate the expression of genes already shown through genetic screens to be involved in the control of antibiotic production, we hope to establish regulatory links that will lead to an overall understanding of the network that integrates the various environmental and physiological signals to bring about the production of secondary metabolites. Disruption of the atrA gene (Fig. 6) resulted in a reduction in the rate of production of actinorhodin (Fig. 7) and reduced transcription of actII-ORF4 (Fig. 8). The simplest model to explain these results is that AtrA binds the promoter of actII-ORF4 in vivo thereby activating transcription. The reduced amount of actinorhodin that is detected in the absence of functional atrA (Fig. 7) may reflect the basal level of transcription from the actII-ORF4 promoter. In this model, ActII-ORF4 would accumulate more slowly in the atrA mutant, but eventually over an extended period of time reach a threshold at which transcription of the act biosynthetic genes is activated. Interestingly, disruption of atrA appears to have no substantial effect on the production of undecylprodigiosin or the calcium-dependent antibiotic (Fig. 7). To our knowledge in most previous cases where undecylprodigiosin production has been examined in other mutants that affect actinorhodin production via the alteration of genes not physically associated with the act cluster, e.g. absA (Ryding et al., 2002), absB (Price et al., 1999), hrdB (Aigle et al., 2000), bldA (Lawlor et al., 1987), afsR (Horinouchi et al., 1990), cutRS (Chang et al., 1996), scbA and scbR (Takano et al., 2001), it has been found to be similarly affected. The ppGpp synthase gene (relA) is an exception under certain growth conditions where deletion of this gene still blocks the production of actinorhodin, but not undecylprodigiosin (Martinez-Costa et al., 1996; Chakraburtty and Bibb, 1997). As the regulatory mechanisms that transmit environmental and physiological signals are largely uncharacterized in streptomycetes (Chater and Bibb, 1997), it will be informative to now determine the role of atrA in mediating the changes in the transcription of actII-ORF4 that have been correlated with the disruption of genes such as relA, and changes in growth conditions.
As transcription factors require time to accumulate to a level that is sufficient to relay a signal, the finding that atrA is not required for the activation of the undecylprodigiosin biosynthetic genes may provide a simple explanation, at least in part, for the observation that this antibiotic is produced ahead of actinorhodin. Another plausible explanation for the delay in actinorhodin production is that atrA relays a stimulatory signal that is produced later in the transition between vegetative growth and morphological development. In any case, the further characterization of the atrA gene may lead to a better understanding of the temporal regulation of different antibiotics. This aspect of gene regulation may allow streptomycetes to prioritize the allocation of limited resources to produce defences in an order that corresponds to the likely timing of attack by different competitors in their natural environment. Although atrA does not appear to be required for the production of undecylprodigiosin or the calcium-dependent antibiotic, it remains to be determined whether it has a role in controlling the expression of one or more of the cryptic secondary metabolite gene clusters that have been revealed by genome sequencing (Bentley et al., 2002).
To investigate whether atrA orthologues may have a role in regulating the transcription of at least some antibiotic pathway-specific activators in disparate Streptomyces, we chose to analyse the strR gene of S. griseus as it represents the final step of the best-characterized pathway regulating secondary metabolite production in streptomycetes (Chater and Horinouchi, 2003). The strR gene encodes a pathway-specific activator unrelated to ActII-ORF4 that regulates the streptomycin biosynthetic genes (Horinouchi and Beppu, 1994; Retzlaff and Distler, 1995; Beyer et al., 1996). The transcription of strR is dependent on activation by AdpA (Ohnishi et al., 1999), which is induced, perhaps as part of a quorum-sensing programme, by a γ-butyrolactone called ‘A-factor’ that binds and thereby inactivates ArpA (A-factor receptor protein), a transcriptional repressor of the adpA gene. The effects of A-factor extend beyond streptomycin production as AdpA is also an activator of genes required for morphological differentiation (Ohnishi et al., 1999; Yamazaki et al., 2000; 2003a,b; Kato et al., 2002). Our finding that AtrA from S. coelicolor is able to bind in vitro to two sites within the promoter region of strR, one of which was mapped (Fig. 9), raises the possibility that atrA orthologues may have a role in regulating the transcription of at least some antibiotic pathway-specific activators in disparate Streptomyces, even when the final activators are unrelated to actII-ORF4. An orthologue of atrA almost certainly exists in Streptomyces avermitilis, the only other streptomycete for which there is currently a complete genome sequence (Ikeda et al., 2003): The SAV4110 gene in S. avermitilis encodes a product that is similar in sequence to AtrA (68% overall sequence identity) and is in a locus where there is a high degree of synteny with the corresponding locus in S. coelicolor (data not shown). Our results also suggest that the regulation of streptomycin production, which serves an important paradigm, may be more complex than represented by current models in which strR transcription is dependent only on activation by AdpA (Ohnishi et al., 1999). Consistent with the latter notion is the previous finding by others that multiple activities in S. griseus bind in vitro with specificity to the promoter region of strR (Vujaklija et al., 1993).
The relationship of AtrA to the TetR family of transcription factors
The gene that corresponds to atrA (SCO4118) has been annotated as encoding a possible TetR-family transcriptional regulator (Bentley et al., 2002) because its product matches an entry (Accession No. PF00440) in the Protein families (Pfam) database (Bateman et al., 2004) that corresponds to the DNA-binding domain of E. coli TetR and related proteins. The affinity of TetR for its cognate operators is dramatically reduced in the presence of tetracycline as the result of a conformational change induced by the binding of tetracycline. AtrA does not match the Pfam entry (PF02909) that corresponds to core domain of E. coli TetR, which contains dimerization interfaces and the binding site for tetracycline (Hinrichs et al., 1994; Kisker et al., 1995; Orth et al., 1998). Combined, the above findings suggest that while the recognition of DNA by AtrA is likely to have elements in common with TetR any effector molecule that might bind AtrA is not necessarily related to tetracycline. When we identified AtrA as the source of the activity that binds the actII-ORF4 promoter, we tested the specificity of this interaction further by incorporating as a control the known binding site of ScbR, which is also a member of the TetR family: binding of AtrA to this additional control was not detected (Fig. 9D).
Many proteins related to TetR function as repressors; however, it is not unprecedented for proteins with TetR-like DNA-binding domains to function as transcriptional activators. Examples of the latter are LuxR and HapR (Jobling and Holmes, 1997), which are considered the master regulators of genes controlled by quorum sensing in Vibrio harveyi and V. cholerae respectively (for review, see Miller and Bassler, 2001), and BotR, which is required for the expression of botulinum neurotoxin genes in Clostridium botulinum A (Marvaud et al., 1998a). An orthologue of BotR in Clostridium tetani has been shown to regulate positively transcription of the tetanus toxin gene (Marvaud et al., 1998b). HapR is also able to function as a repressor and negatively regulates transcription of aphA, a transcriptional activator that is required for the development of virulence in V. cholerae (Kovacikova and Skorupski 2002). The ability of transcription factors to act as both activators and repressors in a gene-specific manner is not unusual, at least in E. coli, and in most cases the difference in effect can be correlated with the position(s) of the binding site (Gralla and Collado-Vides, 1996). We do not exclude therefore the possibility that atrA may also serve as a repressor of genes that have yet to be identified.
DNA binding and possible mechanisms of AtrA regulation
The binding sites of AtrA in the actII-ORF4 promoter region are centred relative to the known transcriptional start site at −162 and +86 bp (Gramajo et al., 1993) within the coding regions of actII-ORF3 and actII-ORF4 respectively (see Fig. 2). The actII-ORF4 promoter is recognized in vivo by RNA polymerase (RNAP) containing σHrdB (Aigle et al., 2000), which appears to be the functional equivalent of σ70 (Brown et al., 1992; Buttner and Lewis, 1992), the principal sigma factor in E. coli. After cataloguing and analysing approximately 150 E. coli promoters, Gralla and Collado-Vides (1996) concluded that activators of σ70 promoters tend to bind in a 50 bp region between −30 bp and −80 bp that allows specific contacts to be made with RNAP. The arrangement of the binding sites for AtrA relative to the known transcription start sites of actII-ORF4 is therefore unusual, at least in terms of the E. coli model. Although E. coli activators are known that bind to ‘remote sites’ (i.e. those that do not overlap the binding site of RNAP between −70 bp to +40 bp, approximately), these tend to function by interacting with activators that contact bound RNAP (Gralla and Collado-Vides, 1996). Possible mechanisms that could explain the activation of actII-ORF4 transcription by AtrA bound at a distance include the looping of DNA between the two sites to allow another protein, as yet unidentified, to contact RNAP, and the competition for binding with a repressor or a DNA-bending protein that has a negative influence on transcription from the actII-ORF4 promoter. During the partial purification of AtrA it was found that other activities within S. coelicolor can bind to the promoter region of actII-ORF4 ( Jolly, 2002); however, these activities remain to be characterized. It is also possible that activation by AtrA binding does not require additional protein factors and acts via a change in the local chromatin conformation per se, e.g. DNA bending, which has been shown to modulate transcription initiation (Perez-Martin et al., 1994). Implicit in all of these models, the binding of AtrA to the site internal to the 5′ coding region of actII-ORF4 is dynamic such that RNAP can pass this point once transcription is initiated.
Growth of cultures
Streptomyces coelicolor M145 and its mutants were cultivated on solid R5 medium as described in Kieser et al. (2000) for the purpose of phenotypic characterization. YEME medium was used to grow mycelium for the purposes of extracting DNA and preparing protoplasts. The activity that binds the actII-ORF4 promoter was purified from S. coelicolor D132 grown on MSM (Doull and Vining, 1989) to a density that previous work had shown corresponds to the transition between the exponential and stationary phases of growth (Potter and Baumberg, 1996). Protoplast manipulations and transformations were carried out as described in Kieser et al. (2000). To prevent the degradation of plasmids by the methyl-specific restriction system of S. coelicolor, the dam dcm hsdS E. coli strain ET12567 (MacNeil et al., 1992) was used as the donor in conjugative mating experiments, and as the source of plasmids for transformations as described by Ryding et al. (1999). E. coli strain XL1-Blue (Bullock et al., 1987) was used for plasmid construction and routine subcloning.
Assay of antibiotic production
Actinorhodin, undecylprodigiosin and the calcium-dependent lipopeptide antibiotic were assayed essentially as described in Kieser et al. (2000). To assess the relative levels of actinorhodin produced by patches of mycelium on agar, the plates were inverted and exposed to ammonia fumes for 5 min from a drop placed on the inner surface of the lid (Rudd and Hopwood, 1979). Levels of undecylprodigiosin, which is actually a mixture of several prodiginines (Tsao et al., 1985), were assayed by adding to mycelial patches a drop of 0.1 M acetic acid to develop fully the red colour of this antibiotic. Calcium-dependent antibiotic (CDA) was assayed by growing patches on Oxoid nutrient agar for 2 days at 30°C, overlaying each plate (15 ml of agar) with soft nutrient agar (5 ml) containing B. mycoides, an indicator strain, and 60 mM Ca(NO3)2, and continuing the incubation overnight at 30°C. A control was included in which calcium was omitted from the overlay.
Construction of the atrA knockout strain
This was achieved by replacing the coding region of atrA with the apramycin-resistance cassette [aac(3)IV] using the PCR-targeted method of Gust et al. (2003). The primers used for the amplification of the aac(3)IV cassette were 5′-TCT CCG GGG GGA GAC GTC ATT ACC GGG GGA TTG TCT ATG ATT CCG GGG ATC CGT CGA CC and 5′-GAA GGA GAT ACG GGC CCC CGA CGA CGT CGC GTC TCA TGT AGG CTG GAG CTG CTT C. Each has a 39 nt segment at the 5′ end that is specific for the atrA gene. The 3′ segments used for the amplification of the cassette are underlined. The cassette was then introduced by electroporation into E. coli BW25113 (Datsenko and Wanner, 2000) that contained a cosmid encoding the atrA gene, and apramycin-resistant recombinants were selected. The resulting cosmid was introduced into E. coli ET12567 containing pUZ8002, an oriT-minus derivative of the conjugative RP4 plasmid, before being mixed with S. coelicolor M145 (Redenbach et al., 1996). In addition to encoding aac(3)IV, the cassette contains an oriT that allows the transfer of the disrupted cosmid into S. coelicolor by intergenic conjugation (Gust et al., 2003). A knockout mutation in atrA was selected by first isolating exconjugants that were apramycin resistant, and then screening for those that had lost the cosmid as the result of a second cross-over and were consequently sensitive to kanamycin. The presence of the knockout was confirmed by the amplification of a 2.4 kb fragment using primers 5′-CTT GAA GCC GAC GCC GTG GTC and 5′-CGC CGT CAC CGA GTG AAC CGT, which flank the atrA gene. Amplification of the wild-type allele of atrA using this primer pair generated a 1.9 kb fragment.
Complementation and replacement of the atrA knockout mutation
For complementation, a 1.8 kb fragment encoding atrA was amplified by PCR from M145 chromosomal DNA using primers 5′-GGA TCC TCG GTG GCG AGG GTG GTG A and 5′-GGA TCC CCG CCC GTC AAC GCT CAT C and Pwo proofreading DNA polymerase (Roche). The fragment was incubated with BamHI (sites underlined in primers) and inserted into the corresponding site of pIJ904, which has an SCP2* replicon and was derived from pIJ903 (Kieser et al., 2000) by the insertion into the unique BamHI site of a polylinker isolated from pJOE814 (Altenbuchner et al., 1992) using BamHI and BglII. The presence of the atrA fragment was confirmed by sequencing. The resulting construct, called pGU103, was introduced after passage through E. coli ET12567 into the atrA knockout strain, called L645, and M145. As a negative control, pIJ904 was also introduced into these two strains.
To replaced the knockout atrA mutation with a wild-type allele, a 3.1 kb fragment containing the coding region of atrA and ∼1 kb of flanking sequence on either side was amplified using primers 5′-GGA TCC GCG GCG GCG CAG TTC GGT GAC and 5′-GGA TTC GCG CGG CTG GCC AGG ACA CGC as described above. The fragment was incubated with BamHI, which in addition to cutting at both ends (underlined nucleotides in primers) also cut a site 80 bp downstream from the first of the two primers listed above. The resulting 3.0 kb fragment was inserted into the corresponding site of pMAhyg, a vector that replicates autonomously in E. coli but not streptomycetes (Paget, 1994). The presence and orientation of the atrA fragment was confirmed by sequencing. After passage through E. coli ET12567, the resulting construct, called pGU205, was introduced into L645 and hygromycin-resistant transformants were selected. A series of PCR assays were then used to confirm that pGU205 had integrated adjacent to the disruption cassette within the chromosomal atrA gene. Spore preparations from individual isolates were then screened by replica plating for isolates in which a second cross-over had resulted in the replacement of the mutant allele in the chromosome of L645 with wild-type atrA, and consequently the loss of resistance to both hygromycin and apramycin. Replacement of the disruption cassette with wild-type atrA was confirmed by the amplification of a 1.9 kb fragment using primers that flank the atrA gene (see above).
Electrophoretic mobility shift assays
Electrophoretic mobility shift assays were performed using a modified version of the method of Vujaklija et al. (1993). DNA and protein were mixed to give a final volume of 20 µl and 1× TGEK buffer; 10 mM Tris-Cl (pH 7.9), 10% (v/v) glycerol, 0.1 mM EDTA and 50 mM KCl containing 2 µg of sonicated Herring sperm DNA. After incubation at 30°C for 20 min, samples were run in native 5% (80:1) acrylamide: bis-acrylamide gels with, as the re-circulated running buffer, 50 mM Tris-Cl (pH 8.5), 0.4 M glycine and 2 mM EDTA. Depending on whether the DNA, which was generated by PCR, was labelled at the 5′ end with 32P using T4 polynucleotide kinase (Sambrook and Russell, 2001) or with fluorescein incorporated during primer synthesis, complexes within the gel were visualized using either a phosphorimager (Fuji BAS-1000) after drying onto Whatman® chromatography paper or a fluorescence imager (Bio-Rad FX Pro) respectively. The primer pairs used to generate the various DNA probes were actII-ORF4 (420 bp), 5′-AAG CTT CTC GAT GTC GGC CGG TGG ATG TGG and 5′-GGA TCC AGC ACC CCC GAT CCC ACC ACC TCG; vdh (398 bp), 5′-AAG CTT CGA TCA CCG AGG ACA CCT TGG GCA and 5′-GGA TCC GGA GTG GAG GGC GAT CAC GGC CTT; actI-III (419 bp), 5′-AAG CTT GAA GAA CGA GAT CCG GCG CGT CGC and 5′-TGA TCA CCC CTC CTT ACC GAG CCT GCG GGC; Region 1 (93 bp), 5′-GAG GAC CCT TCC GAG GAC CCA GCC and 5′-TTT TGC TGG ATT TTA CCG AGA GGC; and Region 2 (101 bp), 5′-TTG GGA CGT GTC CAT GTA ATC ACC and 5′-TCG TGC CGC CTG AGG AGC AGC AGC; strR-D (668 bp) 5′-GGC CTC CAG CCC ATA GAA AAG GAC and 5′-GGA ATC GGA GGG AAG CAA TGA TCC.
In-gel phenanthroline copper footprinting
After resolving complexes using native PAGE, footprinting was performed using essentially the method of Papavassilou (2001). The gel (15 × 15 cm) was transferred to a glass dish containing 200 ml of 10 mM Tris-Cl (pH 8.0). A 1,10-phenanthroline-cupric ion complex was produced by mixing 1 ml of 40 mM 1,10-phenanthroline monohydrate (dissolved in absolute ethanol) with 1 ml of 10 mM cupric sulphate. This was added to 18 ml of double-distilled H20, mixed by vortexing, added to the buffer containing the gel, and the dish gently rocked to aid even dispersal of the ion complex. Cleavage of the DNA was initiated by adding 20 ml of a 1% (w/v) 3-mercaptopropionic acid solution. After incubating for an appropriate length of time, which was determined experimentally, 20 ml of 30 mM 2,9-dimethyl-1,10-phenanthroline (dissolved in absolute ethanol) was added to block DNA cleavage. The DNA was then purified from the gel and analysed using 8% (w/v) polyacrylamide, sequencing-type gels. G ladders were generated using the method of Maxam and Gilbert (1980). The AII4B activity for these experiments was partially purified by anion-exchange (Q-Sepharose) chromatography in addition to AS fractionation (see below).
Partial purification of the AII4B activity
EMSAs employing Region 2 of the actII-ORF4 promoter region were used to follow the purification of DNA binding activity. Cultures of S. coelicolor D132 were grown in 1.5 l of MSM medium (Doull and Vining, 1989) containing carboxy-methyl-cellulose (Potter and Baumberg, 1996) and glass beads (Doull and Vining, 1989) to promote dispersed growth. After incubation for 38 h at 30°C, 250 r.p.m., the mycelium was harvested by centrifugation (850 g for 10 min), resuspended in 40 ml of ice-cold TGEK buffer (see above), and then re-harvested to remove traces of the medium. The mycelium was then resuspended in 40 ml of ice-cold TGEK buffer containing 6 mM MgCl2, DNase I (40 µg ml−1), 0.1 mM dithiothreitol (DTT) and 0.5 mM phenylmethanesulphonyl fluoride (PMSF), and lysed by mechanical disruption using a French pressure cell (SLM Instruments) at 25 000 p.s.i. The lysate was cleared by centrifugation (100 000 g for 30 min at 4°C) and protein in the supernatant was fractionated using AS. Proteins that precipitated between 20% and 40% AS saturation were resuspended in 2 ml of ice-cold TGEK buffer containing 0.1 mM DTT and dialysed extensively against the same buffer overnight at 4°C using a microdialysis chamber (BRL). Fractions precipitated between 0% and 40% AS saturation were used in the initial characterization of the binding activity (Figs 1 and 2). Any protein that did not suspend was removed by spinning in a microcentrifuge at 15 000 g, 4°C for 15 min. The sample as three separate batches was then was loaded onto a 5 ml Q-Sepharose column (Amersham Biosciences), washed with 25 ml of TGEK, and then eluted using a 50-ml gradient from 50 mM to 1 M KCl in TGEK buffer. A flow rate of 1 ml min−1 was used for the above steps. The AII4B activity eluted reproducibly between 240 and 550 mM KCl. Fractions in this range were dialysed against TGEK, pooled and then concentrated around 10-fold using MacroSep ultrafiltration device with a 10 kDa cut-off membrane as described by the vendor (Flowgen). The next step was DNA-affinity chromatography. Region 2 of the actII-ORF4 promoter region was amplified using a primer that was biotinylated at the 5′ end, and ∼300 µg was mixed with 1 ml of UltralinkTM Immobilized Streptavidin Gel following the vendor's instructions (Pierce). After washing, it was calculated that ∼180 µg of Region 2 was immobilized on the gel. This was used to prepare a 1 ml column onto which was loaded the pooled fractions from the ion-exchange chromatography at a flow rate of 0.2 ml min−1. Before loading, however, the pool was mixed with 100 µg of sheared calf thymus DNA to reduced non-specific binding to the DNA immobilized on the column. The column was washed with 7 ml of TGEK and protein eluted from the column using a 20 ml gradient of 0–2 M LiCl in TGEK buffer. Fractions of 1 ml were collected and dialysed against TGEK buffer. Assay of the fractions revealed that although a high proportion of the AII4B activity was in the flow through, around 14% was eluted between 0.7 M and 1.4 M LiCl. The remaining protein in each fraction was then dialysed against ammonium bicarbonate, freeze dried and analysed by SDS-PAGE.
Production and purification of recombinant AtrA
Production and purification of recombinant AtrA was performed using the pET expression system (Studier et al., 1990). A DNA fragment encoding the predicted 297 amino acid residues of AtrA (Bentley et al., 2002) was generated using a pair of primers with the following sequences: 5′-CGC ATA TGC ATG TTC AGG ATT CTC AT (start codon is in italics, and NdeI site is underlined) and 5′-GGA TCC TCA TCA CAC CGG CCG CGA CCG CAG (anti-stop codon is in boldface, and BamHI site is underlined). The PCR fragment was cut with NdeI and BamHI and inserted between the corresponding sites in the expression vector pET16b (Novagen) generating a construct we designated pJAS407. The recombinant atrA gene encodes a product that is tagged at the N-terminus with a His6 oligopeptide, which facilitates purification by immobilized metal affinity chromatography (Porath et al., 1975). Ni-NTA™ spin columns were used to purify recombinant AtrA as described by the vendor (Qiagen). The final protein concentration was determined by measuring the absorbance of samples at 280 nm. The extinction coefficient used for AtrA tagged at the N-terminus was 13 940 M−1 cm−1 as calculated using the ExPASy Molecular Biology Server tool (http://ca.expasy.org/). The purity of the preparation was greater than 95% as judged by SDS-PAGE analysis.
Quantitative PCR analysis of transcript levels
RNA was isolated using a modified Kirby mix, phenol/chloroform extraction and DNase I treatment (Kieser et al., 2000) from mycelium scrapped from cellophane (AA Packaging, Lancashire, UK) laid on the surface of agar plates. The modified Kirby mix contained sodium dodecylsulphate instead of sodium-triisopropylnaphthalene sulphonate, cells were lysed by vortexing with glass beads, and a NaCl concentration of 150 mM was used in the precipitation of nucleic acid. After DNase I treatment, samples were quantified by measuring the absorbance at 260 nm, and analysed by agarose gel electrophoresis to check that ribosomal RNAs were intact and there was no obvious chromosomal DNA contamination. RNA samples (750 ng) were then reverse transcribed using SuperScript™II and random hexamers (Amersham Biosciences) as described by the vendor of the enzyme (Invitrogen). To control quantitatively for chromosomal DNA contamination, each sample was also incubated with a reaction mixture that lacked reverse transcriptase. The reactions after heat inactivation of the reverse transcriptase were then diluted serially in water containing 10 ng µl−1 yeast tRNA that had been extracted with phenol/chloroform. The tRNA was added to minimize non-specific interaction of low concentrations of reverse-transcribed RNA with plastic ware. Samples of each of the reversed-transcribed RNAs (and dilutions) were then analysed using an iCycler® thermal cycler (Bio-Rad Laboratories). Each reaction (25 µl) contained 0.1–10 ng of reverse-transcribed RNA depending on dilution, 1× FailSafe™ PCR Premix J (Epicentre), 0.3 µM of both a forward and reverse primer, 0.8× SYBR® Green I, 20 nM Fluorescein and 1 U of Hotstart Taq polymerase (ABgene). The primer pairs for the amplification of segments of 5S rRNA (92 bp), actII-ORF4 mRNA (131 bp) and hrdB mRNA (149 bp) were 5′-GCG TCC TAC TCT CCC ACA GG plus 5′-AGG GGA AAC GCC CGG TTA; 5′-TCA GGC GGC ACG AGG TGG T plus 5′-CCG GTG CTC CCC CAG CAG; and 5′-GCG GGC GTC GTC TCC ATG C plus 5′-TGC GAG CGC GAG GGG TGA C respectively. The size of each amplicon is provided in parenthesis. The PCR reactions conditions were 95°C for 15 min followed by 45 cycles of 95°C for 30 s, 63°C for 30 s, 72°C for 30 s. A step during which fluorescence was measured for 10 s was also included at the end of each cycle: the incubation temperature was 80°C for the 5S rRNA amplicon and 85°C for the others, which were longer and consequently had a higher melting temperature. The relative amounts of transcript in different samples were determined from the number of cycles (C) it took reactions to reach a common threshold (t), and an estimate of the efficiency of amplification of each transcript. The latter was derived by determining the Ct values for samples of a serial dilution of an RNA preparation that had been reverse transcribed. The efficiency of amplification was calculated by determining the gradient of the line of best fit through plots of Log2(dilution factor) against the Ct values. The analysis of each transcript in each sample was performed using three PCR reactions. Ct values were determined for each of the reactions and then expressed as the difference from the Ct value of one of the PCR reactions corresponding to the earliest sample harvested from the M145 strain. The relative difference in the level of a transcript in each of the other samples was calculated by expressing the average amplification efficiency multiplied by 2 (to give the maximum theoretical increase in product at each cycle), to the power of the difference in the Ct values. This generated three values for each transcript in each RNA samples from which the mean value and a standard deviation were calculated. The values for actII-ORF4 were normalized to that of an internal control, hrdB mRNA or 5S rRNA as indicated. Samples that had not been reverse transcribed were also assayed to ensure that chromosomal DNA contamination did not complicate the interpretation of the transcriptional analysis.
We thank our colleague P.G. Stockley for valuable discussion. This work was supported by Grant 24/G18095 and a quota studentship from the BBSRC. During much of this work, K.J. McDowall was the recipient of a Royal Society University Research Fellowship.