In Streptomyces coelicolor, the AbsA1–AbsA2 two-component system regulates the expression of multiple antibiotic gene clusters. Here, we show that the response regulator encoded by the absA2 gene is a negative regulator of these antibiotic gene clusters. A genetic analysis shows that the phosphorylated form of the AbsA2 response regulator (phospho-AbsA2), generated by the cognate AbsA1 sensor histidine kinase, is required for normal growth phase regulation of antibiotic synthesis. In the absence of phospho-AbsA2, antibiotics are produced earlier and more abundantly. Overexpression of AbsA1 also deregulates antibiotic synthesis, apparently shifting the AbsA1 protein from a kinase-active to a phospho-AbsA2 phosphatase-active form. The absA1 and absA2 genes, which are adjacent, are located in one of the antibiotic gene clusters that they regulate, the cluster for the calcium-dependent antibiotic (CDA). The absA genes themselves are growth phase regulated, with phospho-AbsA2 responsible for growth phase-related positive autoregulation. We discuss the possible role and mechanism of AbsA-mediated regulation of antibiotic synthesis in the S. coelicolor life cycle.
Streptomycetes are notable among prokaryotes for their fungal-like developmental cycles and synthesis of multiple antibiotics. Early in the growth of a colony, multinucleoidal vegetative hyphae extend through the growth medium, branching extensively to form a mycelial mat. Later, in response to poorly understood signals, the vegetative hyphae initiate a programme of multicellular differentiation. Morphological differentiation produces sporulating aerial hyphae on the colony surface (reviewed by Chater and Losick, 1997; Chater, 1998), whereas the temporally parallel but spatially distinct process of secondary metabolite (‘antibiotic’) production occurs in the substrate mycelium (reviewed by Champness and Chater, 1994; Chater and Bibb, 1997). Streptomycete antibiotic biosynthetic pathways involve multiple enzymes that are encoded in large clusters of genes. Each species typically contains several antibiotic gene clusters, and these are subject to a complex network of regulation.
Much of what is known about the regulation of antibiotic genes has come from genetic studies in Streptomyces coelicolor. Studies of S. coelicolor antibiotics have been facilitated by the ease of assaying the antibiotics. Two are pigments: actinorhodin (Act) and undecylprodigiosin (Red) are blue and yellow, respectively, at high pH; both are red at low pH. The other two S. coelicolor antibiotics, calcium-dependent antibiotic (CDA) and methylenomycin (Mmy), can be assayed in simple plate culture inhibition assays. Production of the S. coelicolor antibiotic pigments can easily be observed as being growth phase regulated in both plate and liquid cultures.
One level of regulation that was discovered in S. coelicolor, but is probably common to all streptomycetes, is so-called ‘pathway-specific regulation’, a mechanism in which a cluster-linked transcriptional regulator – usually an activator – regulates the expression of numerous polycistronic transcripts in an antibiotic gene cluster. In the cases of the S. coelicolor antibiotics, actinorhodin and undecylprodigiosin, which are particularly well characterized, the pathway-specific activators are ActII-ORF4 and RedD respectively (Takano et al., 1992; Gramajo et al., 1993). Both are OmpR-like DNA-binding proteins and are founding members of the SARP (for streptomycete antibiotic regulatory protein) family of regulators, which also includes many of the known cluster-linked regulators for other streptomycete antibiotics (Wietzorrek and Bibb, 1997). It has been demonstrated that the temporal regulation of expression of the antibiotic gene clusters results from growth phase-regulated expression of the pathway-specific activators (Takano et al., 1992; Gramajo et al., 1993; White and Bibb, 1997).
Less well understood is what regulates the pathway-specific activators. However, one such control involves the absA two-component system, which was discovered in a genetic analysis of global, or co-ordinate, antibiotic regulation. Mutants of absA were first identified because of their actinorhodin/undecylprodigiosin-minus, sporulation-plus phenotype; subsequently, they were also shown to be calcium-dependent antibiotic minus and Mmy minus (Adamidis et al., 1990). The phenotype for not making any of the four antibiotics was named Abs−, and further work showed a deficiency of actII-ORF4 and redD transcription in absA mutants (Aceti and Champness, 1998), explaining the Abs− phenotype, at least with respect to actinorhodin and undecylprodigiosin synthesis.
Classic genetic mapping showed that the Abs− phenotype was attributable to mutations in a locus, absA, located close to the only known cda mutation, but far from the red and act loci (Adamidis et al., 1990; Brian et al., 1996). Recent genomic sequencing of S. coelicolor has revealed that absA is associated with the cda gene cluster (http://www.sanger.ac.uk/Project/S_coelicolor/). Before genomic sequencing, the only defined segment of the cda cluster was peptide synthase-encoding DNA involved in the biosynthesis of CDA (Chong et al., 1998), which is a cyclic lipopeptide (Kempter et al., 1997). Now, it is apparent that absA lies in a 12 kb region between the peptide synthetase genes and a putative SARP-like regulator for the cda cluster, cdaR (Fig. 1). The function of absA as a regulator of multiple antibiotic clusters, while being genetically associated with one cluster, makes absA highly unusual among antibiotic regulators.
The absA-encoded two-component system is highly ‘orthodox’, including the features common to many of the better studied two-component systems (Brian et al., 1996). In typical two-component systems, a dimeric histidine kinase uses ATP to autophosphorylate, with one subunit transphosphorylating the other on a specific conserved histidine residue (reviewed by Stock et al., 1995). The phosphoryl group is then transferred to an aspartate residue on a cognate response regulator, modulating its activity as a transcriptional regulator. The absA1 gene is predicted to encode a histidine kinase, and the adjacent downstream gene, absA2, is predicted to encode a response regulator with a C-terminal helix–turn–helix (HTH) DNA-binding domain. Following the two-component paradigm, sequence conservation predicts that the AbsA1 protein would autophosphorylate at His-202, and the phosphoryl group would transfer to Asp-54 of AbsA2. AbsA2 is highly homologous to NarL of Escherichia coli, and the transmitter domain of AbsA1 is similar to the cognate kinase NarX (Grebe and Stock, 1999). Closely related two-component systems from Bacillus subtilis include DegS/DegU and ComP/ComA (reviewed by Msadek et al., 1995).
Marker rescue experiments (Brian et al., 1996) and subsequent sequence analysis (Anderson et al., 1999) of absA mutants located the mutations that were responsible for the Abs− phenotype to the transmitter domain of AbsA1. Below, we refer to these alleles as absA1*. Additional genetic experiments revealed that absA could also mutate to a phenotype essentially opposite to Abs−; this phenotype was characterized by early onset and an increased level of antibiotics (Brian et al., 1996). Antibiotic gene transcription was correspondingly increased in the overproducing mutants (Aceti and Champness, 1998). Two absA disruption mutations caused the overproduction phenotype, suggesting that the role of absA in antibiotic regulation was primarily negative (Brian et al., 1996).
We have undertaken a genetic dissection of the absA locus, described here. This work evaluates the role of phosphorylation in absA-mediated regulation of actinorhodin, undecylprodigiosin and CDA and establishes the genetic basis for the two opposing phenotypes observed in absA mutants. We also describe a transcriptional analysis of the absA genes, which reveals growth phase regulation, as well as autoregulatory behaviour of AbsA2. Together, the results of these experiments have implications for the mechanism by which AbsA signal transduction regulates antibiotics during the S. coelicolor life cycle.
Negative regulation of antibiotics by the AbsA2 response regulator and AbsA1 histidine kinase
In previous genetic studies, certain mutations that disrupted the absA locus in S. coelicolor caused a visible phenotype of early, enhanced production of the actinorhodin and undecylprodigiosin antibiotics; we refer to this phenotype as Pha (for precocious hyperproduction of antibiotics). Apart from the effect on antibiotic synthesis, the Pha phenotype includes a defect in morphology. Pha mutants produce aerial hyphae relatively sparsely, and their colony surfaces are notably crenulated (Brian et al., 1996).
Previous Pha mutant alleles were created by insertions into absA1, which is the upstream gene in a putative absA1–absA2 operon (Brian et al., 1996; Fig. 2). The phenotype in these mutants may have resulted from disruption of absA1 or from polar effects on the expression of absA2. To distinguish between these possibilities, we directly tested the function of absA2 by specifically disrupting the absA2 gene. A fragment internal to absA2 was cloned into the non-replicating plasmid pIJ963 to create pTBA500 (Fig. 2), which was then integrated into the absA locus of strain J1501. The resulting strain, C500, was absA1+absA2::pTBA500, with absA2 truncated upstream of the predicted HTH domain (Fig. 3). Disruption of absA2 in C500 caused a Pha phenotype (Fig. 4A), thereby demonstrating the involvement of absA2 in negative regulation of antibiotic production. Both repressor and activator functions are well documented for two-component response regulators; AbsA2 might function in antibiotic regulation as a repressor, or it might be an activator of a repressor.
To test specifically the function of absA1, we created an in frame deletion in the chromosomal copy of absA1 in J1501 (Fig. 5). This deletion removed all the known functional regions of the transmitter domain of histidine kinases, including the conserved H, N, D and G boxes. The phenotype of the absA1Δ530 strain C530 was Pha (Fig. 4A) and, moreover, was identical to that of C500. This result implicated AbsA1 phosphorylation in the negative regulation effected by AbsA2.
Genetic evaluation of the role of phosphorylation in AbsA2-mediated regulation
For most response regulators, phosphorylation of a conserved aspartate residue is essential for the regulatory functions of the proteins in vivo. Following this precedent, the AbsA2 regulatory activity would probably require that AbsA2 be phosphorylated; AbsA1 would probably be responsible for AbsA2 phosphorylation. The Pha phenotypes of C500 and C530 would be consistent with this scenario, but it was important to consider the additional factor that many of the characterized two-component system histidine kinases are bifunctional enzymes that possess both kinase and phosphatase activities; the phosphatase activity dephosphorylates the phosphorylated response regulator. In the case of AbsA1, the in frame deletion in C530 would remove AbsA2-specific phosphatase activity, as well as the kinase activity associated with the transmitter domain. Thus, in strain C530, phospho-AbsA2 may be present if AbsA2 can be phosphorylated by an alternative kinase or low-molecular-weight phosphate donor, and the C530 phenotype might be caused by a lack of the AbsA1 phosphatase and a resulting overabundance of phospho-AbsA2. In this case, the negatively regulating form of AbsA2 would be the unphosphorylated form.
In order to distinguish whether phospho-AbsA2 or unphosphorylated AbsA2 function as the negative regulator, we constructed several mutants with site-directed changes to the chromosomal absA2 gene (Fig. 2), altering the AbsA2 aspartate residue (D54) that is analogous to the conserved phosphorylated aspartate of response regulators. Separate gene replacements created strains C570, C516 and C532 with AbsA2 amino acid replacements D54E, D54A and D54N respectively. Similar aspartate substitutions have been shown to prevent phosphorylation of numerous response regulators of other species (Moore et al., 1993; Dahl et al., 1997; Alon et al., 1998). All three mutant strains exhibited the Pha phenotype (Fig. 4A; C516 not shown). Thus, these results supported the hypothesis that phospho-AbsA2 functions as the negative regulator. It is notable that the absA1Δ530 allele causes a phenotype similar to that seen in strains disrupted for absA2 or altered in the presumptive AbsA2 phosphorylation site. This suggests that, in vivo, AbsA2 is not subject to phosphorylation by alternative histidine kinases or low-molecular-weight phosphodonors.
If AbsA1 transfers a phosphate to AbsA2, thereby converting it into a negative regulator, then mutations in AbsA1 that prevent autophosphorylation should cause the Pha phenotype. The histidine residue in AbsA1 that corresponds to the site of phosphorylation in well-characterized members of the histidine kinase family is His-202 (Fig. 2). A site-directed mutation, H202L, was made in the chromosomal absA1 gene of J1501, creating strain C550. Strain C550 exhibited a Pha phenotype, a result consistent with a requirement for histidine kinase activity in negative regulation. However, the phenotype differed from that of C530 (ΔabsA1) in several respects. First, C550 visibly produced undecylprodigiosin earlier than actinorhodin, whereas C530 produced both antibiotics precociously. Secondly, hyperproduction of antibiotics, relative to strain J1501, never reached the levels seen for C530 (Fig. 4A). The reason for the weaker Pha phenotype of C550 is not clear at this time. We considered the possibility that AbsA1 (H202L) contained a second site of phosphorylation. However, we could find no evidence for such a second site using in vitro assays. A fusion protein containing the AbsA1 (H202L) transmitter domain fused to a maltose-binding protein did not demonstrate any autokinase activity in vitro, unlike a similar MPB fusion to wild-type AbsA1 (T. B. Anderson and W. C. Champness, unpublished).
Precocious hyperproduction of calcium-dependent antibiotic, undecylprodigiosin and actinorhodin in absA mutants
The map location of the absA locus within the cda gene cluster suggests that the role of the absA1–absA2 system in regulating this cluster could be different from its role in regulating the other antibiotic gene clusters under its control. Accordingly, we sought to determine whether Pha mutations affected synthesis of calcium-dependent antibiotic. To assess CDA activity, plugs from plate-grown cultures were tested for anti-Staphylococcus aureus activity. In the presence of added calcium, the lipopeptide CDA is active, damaging cell membranes (Lakey et al., 1983). For CDA assays, culture plugs were tested on plates with and without added calcium. CDA activity is typically detectable 12–20 h before the pigmented antibiotics appear (Adamidis et al., 1990; W. C. Champness, unpublished). In a 2 day time course, shown in Fig. 6, CDA activity was detected in Pha mutants at least 7 h earlier than in J1501. Pha mutants C550 and C530 are shown. Similar results were obtained with C550, C530 and C570 on R5 media (data not shown). These results showed that AbsA also negatively regulates CDA synthesis.
Over the course of cultivating Pha mutants, we have observed variability in how much earlier a given Pha mutant produces antibiotics compared with J1501. The acceleration of antibiotic production has ranged from at least 7 h to several days on different batches of media, including R5, SpMR and ONA. In addition, in quantitative assays of actinorhodin and undecylprodigiosin, the Pha-related overproduction has varied from fivefold to more than 60-fold (data not shown). An exploration of this phenomenon will be reported in more detail elsewhere.
Precocious hyperproduction of antibiotics resulting from AbsA domain overexpression
In some cases, overexpression of an unphosphorylated response regulator can mimic the regulation of target promoters that are normally effected by a phosphorylated response regulator (e.g. Webber and Kadner, 1997). To evaluate whether overexpression of unphosphorylated AbsA2 could regulate antibiotics, we introduced a high-copy clone of the absA2 (D54E) mutant allele (pTBA175; Fig. 2) into J1501 and C577S25, a strain deleted for absA2 and most of absA1 (Anderson et al., 1999). The pTBA175 plasmid included absA1+ and the absA promoter region. If unphosphorylated AbsA2 could negatively regulate antibiotics, we might have observed a delay of antibiotics in the Pha C577S25 strain. However, we observed no change in the Pha phenotype (data not shown), suggesting that negative regulation requires phosphorylation of AbsA2, even at high protein abundance.
When pTBA175 was introduced into J1501, the resulting phenotype was Pha, indicating an interference with normal AbsA-mediated regulation (Fig. 4B). To examine this phenomenon further, we evaluated the effects of overexpressing selected domains of the AbsA1 and AbsA2 proteins (Fig. 2). First, we excluded an effect of the absA promoter region by introducing plasmid pCB540; this plasmid did not alter the J1501 phenotype (Fig. 4B). Secondly, we observed that multiple copies of the entire absA locus in pCB520 produced no change in the Abs+ phenotype (Fig. 4B). Next, we evaluated a set of high-copy plasmids that expressed wild-type AbsA2 but carried phosphorylation-minus absA1 alleles; these included pCB530, carrying the in frame deletion absA1Δ530, and pTBA156, carrying the absA1 (H202L) allele. These produced no change in the Abs+ phenotype (Fig. 4B). In contrast, a Pha phenotype resulted from plasmids that lacked absA2+ but contained absA1 sequences. Two such plasmids were pCB220 and pTBA155. A pattern that emerged from these results was that an increase in gene dosage of absA2+, with or without an increase in absA1, did not alter antibiotic regulation. However, an increase in absA1 sequences without a corresponding increase in absA2+ deregulated antibiotic synthesis. One interpretation of these results is that a high absA1 gene dosage causes a shift in the ratio of AbsA1 kinase to phosphatase activity to favour the phosphatase activity and, with relatively low expression of AbsA2+, there may be insufficient AbsA2∼P accumulation to downregulate antibiotics. These results suggest that AbsA1 can negatively regulate AbsA2, very possibly through phospho-AbsA2 phosphatase activity.
High-resolution S1 nuclease mapping of the absA transcription start site
The absA1 and absA2 open reading frames (ORFs) are separated by only 17 nucleotides (nt) and are probably co-transcribed. To define the transcription start site for absA, high-resolution S1 nuclease mapping was performed. First, a polymerase chain reaction (PCR)-generated double-stranded DNA probe specific to the predicted promoter region for absA1 was used (see Experimental procedures). A 455 bp absA1 probe protected a single product of 291 nt (Fig. 7), identifying the transcription start site for absA1 as the A that is coincident with the first nucleotide of the putative translation start site. To evaluate co-transcription of absA1 and absA2, the region upstream of absA2 was probed with a 504 bp fragment. The S1 product showed no indication of independent promoter activity for absA2 (data not shown). These results indicated that absA1 and absA2 are expressed from a single, leaderless transcript. Leaderless transcripts are not uncommon in actinomycetes, as documented by Strohl (1992), who reported that 11 out of 139 promoters analysed produced leaderless transcripts.
Inspection of the sequence upstream of absA1 revealed a −10 region with the sequence TAGCGT (Fig. 7); this is similar to the consensus sequence proposed by Strohl (1992) for transcription from Streptomyces RNA polymerase that contains an Escherichia coli-like Eσ70, e.g. HrdB or HrdD (Buttner et al., 1990). There was, however, no recognizable consensus sequence in the −35 region.
Growth phase-dependent expression and autoregulation of absA
To evaluate the temporal profile of absA expression, RNA was isolated over a 54 h time course from cultures grown in liquid media. The media used, SpMR (see Experimental procedures), supported production of the actinorhodin and undecylprodigiosin pigments by strain J1501. Under these conditions, all strains formed mycelial clumps; however, biomass accumulation was comparable in all cultures and, as is generally observed, the antibiotics showed growth phase-dependent production kinetics, appearing only after a period of biomass accumulation. S. coelicolor does not sporulate when grown in liquid cultures, so temporal comparisons of antibiotic production and sporulation could not be made in this experiment.
S1 nuclease protection assays were performed on the absA promoter region from RNA isolated from J1501 cultures grown for 18 h, 30 h and 54 h. Antibiotics were not produced in the 18 h culture, but were visible in the 30 h J1501 culture. Figure 8 shows that the absA transcript was present in the 18 h culture and increased significantly in abundance from 18 h to 30 h. The transcript then remained at a constant level until 54 h. Comparisons of transcript abundance in different cultures were aided by the addition of a probe for the glk (glucose kinase) gene to each S1 assay. The absA signal increased approximately fivefold relative to the glk signal over the course of culture growth. The pattern of absA expression from duplicate cultures was consistent.
Figure 8 also includes S1 nuclease protection assays of RNAs isolated from two strains that are mutant for the absA locus. One strain was C542, an Abs− strain mutant in absA1 (i.e. an absA1* strain; Adamidis et al., 1990; Brian et al., 1996), as described above. The second was C570, the Pha strain carrying the D54E mutation in absA2 described in Fig. 2. The profile of absA expression was altered in both absA mutants. In C570, the absA transcript abundance was very low at all time points. In contrast, in C542, the absA transcript was several-fold more abundant than in J1501 at all time points.
The S1 protection assays revealed several aspects of absA regulation. First, the effects of the absA mutations indicated that absA expression is autoregulated. Secondly, the mutant effects on the absA transcript were the opposite of the previously observed effects on antibiotic transcripts: whereas the Abs− and Pha phenotypes were found to correlate with decreased or increased antibiotic gene transcription, respectively, the absA transcript was decreased in the Pha strain but increased in the Abs− strain. These results suggest that autoregulation by absA is positive, in contrast to absA negative regulation of antibiotics. Thirdly, the low level of absA transcript in C570, the absA2 (D54E) mutant, suggests that phospho-AbsA2 is the autoregulatory form of AbsA2, which is consistent with data from the genetic analysis that implicates phospho-AbsA2 as the antibiotic regulatory form. Finally, the absence of any growth phase-related increase in absA transcript in C570 suggests that phospho-AbsA2 was responsible for the growth phase regulation observed in J1501 and C542. Thus, the growth phase regulation of absA appears to result from phosphorylation-dependent AbsA2-mediated autoregulation. We have not determined at this time whether the absA autoregulation is direct or indirect.
In this paper, we have described a genetic and transcriptional analysis of the absA locus that further characterizes aspects of the mechanism of AbsA-mediated regulation of antibiotic production. Disruptions in the absA1 and absA2 genes demonstrated that the AbsA two-component system is a negative regulator of the multiple antibiotics produced by S. coelicolor, including calcium-dependent antibiotic, actinorhodin and undecylprodigiosin. In addition, gene replacements in the absA locus with alleles with altered putative sites of phosphorylation of AbsA1 or AbsA2 indicated that the phosphorylated form of AbsA2 is the negative regulator. As predicted from sequence conservation with other two-component systems, both the His at position 202 of AbsA1 and the Asp residue at position 54 of AbsA2 were required for normal regulation of antibiotic synthesis: each of the gene replacement strains tested attained an antibiotic-overproducing phenotype (Pha) consistent with a mechanism in which the phosphorylated form of AbsA2 is the active negative regulator of antibiotic synthesis. Our results did not prove directly that AbsA2∼P is the direct repressor of the antibiotic genes. It is possible that AbsA2∼P activates a gene for a repressor that is, in turn, responsible for the repression.
Without absA regulation, the timing of antibiotic production is advanced but, even in Pha cultures, a period of approximately 2 days passes before antibiotics appear. One interpretation of this observation is that the appearance of antibiotics in Pha cultures indicates the time at which the culture enters an antibiotic production-competent state, but the AbsA system normally imposes a delay on production. The heterogeneities in a growing mycelial biomass complicate distinctions of growth phases but, for the purposes of further discussion, we refer to the postulated ‘AbsA-repressed’ period as the ‘transition stage’.
We can envisage several models of how AbsA, as a signal transduction system, could modulate the production of antibiotics during culture growth. One model, which accommodates both genetic and transcriptional data, supposes the following. Early in growth, a culture is not competent for antibiotic synthesis; also, the absA genes are expressed at a low level. After a period of growth, the culture enters the ‘transition stage’. During this time, the signal that regulates AbsA may be present at significant levels. If AbsA1 is similar to many sensor kinase/phosphatases, it will require signal binding to activate the kinase function and exist in a phosphatase-dominant mode if the signal is absent (e.g. Williams and Stewart, 1997). Once the signal is present and AbsA1 is shifted to a kinase-dominant form, AbsA2∼P will accumulate and negatively regulate antibiotics, and also positively autoregulate, accounting for the AbsA2∼P-dependent, growth phase-related increase in absA transcript seen in J1501. Easing of AbsA repression may require that the signal be depleted or degraded, allowing AbsA1 to switch to the phosphatase form and dephosphorylate AbsA2, allowing antibiotic gene expression. At present, we have no information regarding the nature of the signal hypothesized to regulate AbsA1.
If the normal function of the AbsA system is in the negative regulation of antibiotics, what explains the Abs− phenotype in the mutants that first defined the absA locus? We hypothesize that these absA1* alleles lock the AbsA system into the negatively regulating mode, i.e. in which AbsA2 is phosphorylated. In support of this notion, the Abs− phenotype requires absA2+ (Anderson et al., 1999). Also, the mutations causing the Abs− phenotype lie in the transmitter domain of AbsA1 (Anderson et al., 1999), which possesses the kinase and phosphatase activities of histidine kinase proteins. The mutant AbsA1* proteins might be constitutively kinase-dominant forms, either lacking phosphatase capability or functioning as signal-independent kinases. The latter possibility would be most consistent with the increased level of absA transcript observed in C542, e.g. AbsA2∼P would be present even in young cultures lacking signal and would autoregulate. Another observation that could be explained by signal-independent AbsA1 kinase activity is the persistence of the Abs− phenotype over the life of mutant cultures; even colonies that grow for several weeks remain unpigmented. Further work will be required to determine the precise nature of the activities of the AbsA1* proteins.
An alternative model for signal regulation in the AbsA system is that AbsA1 is a kinase in the absence of signal and a phosphatase in the presence of signal, as some sensor kinase phosphatases are proposed to function (e.g. Pratt and Silhavy, 1995; Freeman and Bassler, 1999). In this case, the transition-stage culture would lack the signal-regulating AbsA1, and AbsA1-kinase activity would generate AbsA2∼P. Later, a signal would switch AbsA1 to the AbsA1 phosphatase mode, so that it could dephosphorylate AbsA2∼P, allowing antibiotic synthesis. We consider this model to be less compelling than the first because the AbsA2∼P-dependent transcription profiles are more simply explained if the AbsA1 kinase activity is activated by a transition-stage signal.
What purpose does AbsA regulation of antibiotics serve in the S. coelicolor life cycle? One relevant observation is the substantial perturbation of morphogenesis observed in Pha mutants: these mutants usually produce only sparse aerial hyphae. One possibility is that precocious antibiotic synthesis per se is deleterious to normal sporulation. Calcium-dependent antibiotic may be especially inhibitory (W. C. Champness, unpublished). Thus, it may be that S. coelicolor acquires competence for antibiotic production before the sporulation process has proceeded adequately, and the function of the AbsA system is to delay antibiotic production to allow optimal sporulation.
What factors establish the state that we have referred to as ‘antibiotic production competent’? Probable candidates include the genes that have been identified on the basis of mutant defects in antibiotic synthesis. Among these is a second gene found in Abs− mutant hunts, absB, which encodes the S. coelicolor homologue of RNase III (Price et al., 1999). Another large group of genes is known to regulate both antibiotic synthesis and the onset of sporulation. Some genes in this group are the bld genes, several of which encode regulators of gene expression (reviewed by Champness, 1999). Another is relA, which encodes (p)ppGpp synthetase (Martinez-Costa et al., 1996; Chakraburtty and Bibb, 1997). Also important are the components of γ-butyrolactone signalling pathways (Yamada et al., 1997; Ohnishi et al., 1999).
Additional antibiotic regulatory genes have been isolated on the basis of multicopy stimulation of antibiotic production. The best characterized of these are the AfsQ1/Q2 two-component system (Ishizuka et al., 1992) and the AfsR/K serine–threonine phosphoprotein/kinase pair (Matsumoto et al., 1994; Floriano and Bibb, 1996). Mutations in the afsQ1/Q2 genes cause no phenotype, but disruptions to the afsR/K locus conditionally reduce antibiotic synthesis, especially on high-phosphate media (Matsumoto et al., 1994; Floriano and Bibb, 1996). The multicopy effect of afsR/K has been shown to correlate with increased antibiotic pathway-specific activator transcription (Floriano and Bibb, 1996). Multicopy clones of the afsR/K locus can restore antibiotic synthesis to Abs−absA1* mutants (Champness et al., 1992) and so can overexpression of the AfsQ1 response regulator (Ishizuka et al., 1992). These observations imply that these genes can compete against the postulated persistent negative regulation imposed by absA1* alleles.
It is widely observed that phosphorylation of response regulators modifies their activities, probably causing conformational changes that affect promoter recognition or co-operative binding at the target promoter (Stock et al., 1995). However, the extent to which phosphorylation is required for DNA binding and transcriptional regulation in vivo varies for different response regulators. In the Nar system of E. coli (reviewed by Stewart and Rabin, 1995), which regulates nitrate/nitrite-responsive anaerobic respiratory pathways, phosphorylation of NarL is absolutely required for DNA binding and regulatory activity (Baikalov et al., 1996). Conversely, in E. coli UhpA-mediated regulation of sugar phosphate uptake, high-copy expression of the unphosphorylated UhpA D54N protein allowed phosphorylation-independent activation of the uhpT promoter (Webber and Kadner, 1997). If unphosphorylated AbsA2 was functional in vivo, high-copy expression of the absA2 (D54E) allele, on plasmid pTBA175 (Fig. 2), might have repressed antibiotic synthesis. As it did not (Fig. 4B), it appears that AbsA2 regulatory activity is strongly dependent on phosphorylation.
It is noteworthy to contrast AbsA2 with several other recently discovered antibiotic cluster-linked regulators that are closely related in sequence. One such protein is RedZ, a red cluster-linked activator of redD transcription (White and Bibb, 1997). The amino acid sequence of RedZ has end-to-end similarity with AbsA2, including the putative HTH region, with 27% identical residues overall. However, RedZ lacks the conserved aspartate residue that is normally the site of phosphorylation in response regulators (Guthrie et al., 1998). A homologue of RedZ, DnrN, is found in the daunorubicin gene cluster of Streptomyces peucetius, in which it regulates dnrI, which encodes a SARP pathway-specific regulator of the dnr cluster. Although the DnrN protein sequence has retained the conserved aspartate, other residues of the phosphorylation pocket are not conserved, and phosphorylation does not appear to be involved in DnrN function in vivo (Furuya and Hutchinson, 1996). Thus, DnrN and RedZ appear to serve as regulators in the unphosphorylated state. It is not known whether a modification other than phosphorylation regulates the activity of RedZ or DnrN.
Our results have shown that the AbsA two-component system is a negative regulator of the calcium-dependent antibiotic, actinorhodin and undecylprodigiosin. For the latter two antibiotics, absA negatively regulates the SARP pathway-specific activator genes (Aceti and Champness, 1998). It will be important to determine whether AbsA2 directly regulates the SARP genes and, in the cda cluster, whether AbsA2 regulates the SARP homologue, cdaR, or whether it regulates another, as yet unidentified, cda regulator or directly represses cda biosynthetic genes.
Streptomyces strains were cultured in YEME broth (Hopwood et al., 1985); for chromosomal DNA extraction, SpMR (Kendrick and Ensign, 1983) plates overlaid with cellophane disks were used. Cultures used for RNA extraction were grown in 50 ml of SpMR broth in 300 ml baffled flasks, inoculated with 1 × 108 spores and incubated at 30°C, 250 r.p.m., for 18, 30 or 54 h. Thiostrepton was used at 10 µg ml−1 in liquid culture or 50 µg ml−1 in agar. Hygromycin (Hyg) was used at 200 µg ml−1. E. coli was grown in L broth or L agar (Sambrook et al., 1989). Ampicillin was used at 50 µg ml−1.
Assay conditions for the calcium-dependent antibiotic were as described previously (Adamidis et al., 1990). Strains were grown on Oxoid nutrient agar (ONA) or R5 (Hopwood et al., 1985) and placed onto plates with or without added calcium [as Ca(NO3)2 to 12 mM]. Soft ONA or ONA plus Ca was seeded with CDA-sensitive Staphylococcus aureus and overlaid around the plugs. Plates were incubated overnight at 37°C. Actinorhodin and undecylprodigiosin determinations were as described previously (Adamidis et al., 1990).
Plasmid and DNA manipulations
Oligonucleotide primers (Table 1) were prepared by the Macromolecular Structure Facility at Michigan State University. Streptomyces plasmid preparations and transformations have been described by Hopwood et al. (1985). Streptomyces chromosomal DNA was isolated using the method of Pospiech and Neumann (1995). E. coli plasmid preparations were carried out by alkaline lysis (Sambrook et al., 1989) or using QIAprep spin columns (Qiagen). All replicative plasmids shown in Fig. 2 were constructed by first cloning the S. coelicolor absA region of interest into pBluscript-II SK+ (Stratagene) using standard cloning techniques (Sambrook et al., 1989). Inserts flanked by BamHI sites were then subcloned directly into the BglII site of pIJ702 (Hopwood et al., 1985), as in the cases of pCB220 and pTBA155. Inserts used to construct pCB520, pCB530, pCB540, pTBA156 and pTBA175 were first subcloned into pIJ2925 (Janssen and Bibb, 1993) and then excised as BglII fragments from the pIJ2925 polylinker for ligation into the BglII site of pIJ702. Plasmids pCB620 and pCB200, used for complementation experiments, carried the same inserts as pCB520 and pCB220, respectively, and were constructed in an identical manner using low-copy plasmid pIJ922 (Hopwood et al., 1985). Replicative ligations were transformed into S. lividans 1326; plasmids recovered from these transformants were then transformed into S. coelicolor J1501.
Table 1. Oligonucleotide primers used in this study.
A fragment internal to absA2 was amplified using polymerase chain reaction (PCR) from primers WC8 and WC9, both of which contained BglII restriction sites at their ends. The truncated region of AbsA2 encoded by the WC8/WC9 PCR product is illustrated in Fig. 3. PCR amplification was carried out in a 100 µl reaction volume with 100 ng of J1501 chromosomal DNA template under the buffer and thermal cycler conditions for absA2 amplification described by Anderson et al. (1999). The absA2 amplification product was purified on Wizard PCR preparatory columns (Promega) before and after digestion with BglII. The resulting 5′- and 3′-truncated absA2 fragment was cloned as a BglII fragment into the BamHI site of pIJ963 (Janssen and Bibb, 1993) to produce the integrative plasmid pTBA500 (Fig. 2). pTBA500 was passed through dam−, dcm−E. coli ET12567 before transformation into S. coelicolor J1501. Hygr resistance was used to select for single cross-over recombinants; these displayed the Pha phenotype. Plasmid integration was analysed using Southern hybridization with absA- and hyg-specific probes. Strain C500 was complemented to the wild-type phenotype by plasmid pCB620.
Construction of an in frame deletion in absA1 in C530
The integrative plasmid pTBA533 (Fig. 2) was constructed by first digesting pCB400 (pIJ2925 with a 2 kb absA1 BamHI fragment) with NaeI, removing 0.8 kb internal to absA1 (Fig. 5). The resulting 1.2 kb BamHI absA1 fragment was ligated into the BamHI site of pCB300 (pSK+ with a 1.8 kb BamHI–XhoI absA2-containing region) to produce pCB500. The 2.4 kb XhoI–PstI absA region of pCB500 was subcloned as a SacI–PstI fragment into pIJ2925 to create pCB501. The entire pCB501 insert was removed from the polylinker as a BglII fragment and ligated into BamHI-digested pIJ963 to produce the integrative plasmid pTBA533.
Initial attempts at gene replacement used a strain with an ermE replacement in absA (C430), but this strain transformed poorly (Brian et al., 1996). Therefore, gene replacements were created in strain J1501 as follows. pTBA533 was passed through dam−, dcm−E. coli DM-1 (Gibco BRL) before transformation into S. coelicolor J1501 and selection for Hygr resistance. Single cross-over recombinants verified by Southern blot analysis were propagated without Hyg to allow double cross-over curing of the plasmid. Single colonies were screened for double cross-over using PCR amplification from primers WC12/WC13, both internal to the in frame deletion, and WC16/WC26, producing a 1 kb product for absA1Δ530 versus a 1.8 kb product for absA1+, which was then analysed by Southern hybridization for loss of the hyg signal and a single XhoI signal of 2.4 kb for the absA probe. Final confirmation of the integrity and fidelity of the C530 absA1-deleted locus was obtained by sequencing the entire absA1 and absA2. Procedures for the amplification and sequencing of absA1 and absA2 are described elsewhere (Anderson et al., 1999). Strain C530 was fully complemented by both pCB200 and pCB620.
In each of the gene replacements described in this study, double cross-over replacement of the mutant allele required propagation of single cross-over transformants for numerous generations under non-selective conditions. It should be noted that, although cultures demonstrated Hyg sensitivity after only a few generations of growth in the absence of antibiotic, the hyg marker could be detected by Southern blot analysis. Similarly, PCR screening that suggested complete resolution by double cross-over was frequently contradicted by Southern blot analysis. Final confirmation of successful gene replacement required careful analysis by Southern hybridization and sequencing.
The absA2 (D54E) allele was generated using PCR amplification with mutagenic primers. Separate upstream and downstream absA2 fragments, with an overlapping region centred at the site of the D54E-encoded mutation, were amplified from pCB460 (pSK+ carrying a 3.9 kb BamHI–XhoI fragment with the entire absA locus) using primer pairs WC24/WC29 and WC15/WC28 (Table 1). The resulting GAC to GAG change also introduced an XhoI site into the PCR products. Thus, the D54E-containing fragments were digested with XhoI and an additional restriction enzyme, the site for which was present in the upstream or downstream region surrounding absA2: BamHI for the (WC24/WC29) product upstream of absA2 and PstI for the downstream (WC15/WC28) product. A three-way ligation between these fragments and pSK+BamHI–PstI produced pTBA160 containing a 1.2 kb insert with the entire absA2 (D54E) allele. Confirmation of the site-directed change was obtained by sequence analysis of the entire absA2 (D54E) allele from pTBA160. Subsequently, a 2 kb BamHI absA1 region was ligated into pTBA160 BamHI to create pTBA162. The 3.2 kb BamHI–PstI insert of pTBA162 containing the entire absA locus was removed as a XbaI–KpnI fragment from the pTBA162 polylinker for ligation into pIJ2925 to produce pTBA166. The same 3.2 kb insert was excised from the pTBA166 polylinker as a BglII fragment and ligated into pIJ963 BamHI to generate the integrative plasmid pTBA570 (Fig. 2). Strain C570 possessing a chromosomal absA2 (D54E) mutation was created through double cross-over gene replacement with integrative plasmid pTBA570. pTBA570 was demethylated as described above before transformation into S. coelicolor J1501 with Hyg selection for resistance. Single cross-over recombinants were propagated non-selectively and screened for plasmid curing by PCR amplification from primers WC30/WC28, which amplified a 1.2 kb product containing the entire absA2 allele. Only the absA2 (D54E) allele was susceptible to digestion with XhoI. Southern hybridization was performed on XhoI digests of chromosomal DNAs. Colonies that had successfully undergone double cross-over integration of the absA2 (D54E) allele showed no signal for the hyg probe and signals of 1.7 kb and 1.5 kb for the absA probe. The C570 absA locus was sequenced. Strain C570 was complemented to the wild-type phenotype by low-copy plasmid pCB620.
The D54A and D54N alleles of absA2 were created by PCR-mediated introduction of these mutations using the QuickChange site-directed mutagenesis kit (Stratagene). For D54A, a nucleotide change (GAC to GCC) was introduced with complementary primers WC65 and WC66. Complementary primers WC67 and WC68 produced a single mismatch (GAC to AAC) for D54N. The mutagenesis reactions were carried out on 50 ng of pTBA400 (pSK+ carrying a 1.8 kb BamHI–XhoI absA2-containing fragment) under the manufacturer-prescribed buffer conditions with the addition of 5% glycerol and 2.5% dimethyl sulphoxide (DMSO). The thermal cycler conditions were 95°C for 5 min, followed by 12 cycles of 95°C for 1 min, 65°C for 45 s and 72°C for 12 min. PCR products digested with DpnI were transformed into E. coli DH5α (Gibco BRL) to create pTBA410 [containing absA2 (D54A)] and pTBA430 [containing absA2 (D54N)]. Both the D54A and D54N mutations removed a TaqI restriction site at the mutagenized codon; transformants were amplified with primers WC35 and P11, digested with TaqI and analysed on 1.2% agarose gel. In each case, over 90% of the transformants tested screened positive for the mutation. Confirmation of the desired mutations was obtained by sequencing the absA1 and absA2 portions of the mutagenized plasmids pTBA410 and pTBA430. Integrative plasmid pTBA532 (Fig. 2), encoding the absA2 (D54N) mutation, was made by subcloning the 1.8 kb BamHI–KpnI fragment from the polylinker of pTBA430 into pIJ963. In order to construct integrative plasmid pTBA516 (Fig. 2), the pTBA410 insert was increased in size to 3.8 kb by cloning in a 2 kb BamHI fragment containing the upstream portion of absA1 to produce pTBA516. The 3.2 kb XhoI fragment of pTBA516 [containing absA2 (D54A)] was then cloned into the SalI site of pIJ2927 (Janssen and Bibb, 1993) to create pTBA414. This same 3.2 kb insert was then removed as a 3.2 kb BglII fragment from the polylinker of pTBA414 and cloned into pIJ963 BamHI to create the integrative plasmid pTBA516. Strains C516 and C532 were created using double cross-over gene replacements with integrative vectors pTBA516 and pTBA532, as described for C570. Initial screening for plasmid curing was performed by PCR amplification of TaqI restriction digest analysis of the WC35/P11 absA2 products. Chromosomal DNA that did not amplify absA2+ was digested with XhoI and analysed by Southern hybridization for loss of hyg and a signal of 3.2 kb for absA. The C516 and C532 absA loci were sequenced.
The absA1 (H202L) allele was generated using PCR overlap extension (Vallejo et al., 1995). Separate upstream and downstream absA1 fragments, with an overlapping region centred at the site of the H202L mutation, were amplified from pCB460 (pSK+ carrying a 3.9 kb BamHI–XhoI fragment with the entire absA locus) using primer pairs WC16/WC30 and WC14/WC26 (Table 1). Then, the full-length absA1 (H202L) allele was amplified by combining 5–10 ng of the upstream and downstream PCR products as a template, together with primers WC16 and WC26. PCR amplifications were carried out in 50 µl reactions using high-fidelity Pfu DNA polymerase (Stratagene). Buffer conditions and thermal cycler settings were as described previously (Anderson et al., 1999), except that the extension time was increased from 1 min to 4 min. The expected size product was agarose gel purified and digested with XhoI and BamHI to produce a 1.45 kb fragment that was ligated into pSK+ to create pTBA140. Confirmation of the site-directed change was obtained by sequence analysis of the absA1 XhoI–BamHI fragment from pTBA140. Subsequently, a 5 kb BamHI fragment (containing the 3′ region of absA1 and all of absA2) was ligated into pTBA140 BamHI to produce pTBA142. The 3.2 kb XhoI region [containing most of the absA1 (H202L) allele and all of absA2] was excised from pTBA142 and ligated into XhoI-digested pCB360 (pIJ2925 containing a 1.6 kb SacI–XhoI insert with the 5′ region of absA1 and upstream SCE8.17c) to create pTBA144. The entire 5.4 kb insert was removed from pTBA144 as a BglII fragment for ligation into pIJ963 BamHI to create the integrative plasmid pTBA150 (Fig. 2). Strain C550 was created by gene replacement with pTBA150. Screening for plasmid curing took advantage of a TaqI restriction site introduced by the H202L mutation. An internal region of absA1 was amplified from chromosomal DNA of putative C550 strains using primers WC12 and WC13, followed by TaqI digest analysis. Chromosomal DNAs failing to amplify absA1+ were digested with XhoI and analysed by Southern hybridization for loss of hyg and a signal of 3.2 kb for absA. The C550 absA locus was sequenced. Strain C550 was effectively complemented by both pCB200 and pCB620.
Streptomyces RNA isolation was carried out as described by Hopwood et al. (1985) using the preparation method for dot-blot and Northern blot analysis. Two independent isolations at 18, 30 and 54 h of growth were performed for S. coelicolor strains J1501 (hisA1 uraA1 strA1 SCP1− SCP2− Pgl−), C542 (absA1–542;Anderson et al., 1999) and C570 [absA2 (D54E) hisA1 uraA1 strA1 SCP1− SCP2− Pgl−]. Four 50 ml cultures were pooled for each 18 h RNA preparation, whereas two 50 ml cultures were pooled for the 30 h and 54 h samples. The concentration, purity and integrity of the RNA samples was evaluated by spectrophotometry and agarose gel electrophoresis. Isolation of E. coli RNA, for use as a negative control, was performed with an RNeasy RNA purification column (QIAGEN).
S1 nuclease protection assays
All experiments were performed using 50 µg of RNA and 60 000–100 000 c.p.m. of 32P end-labelled double-stranded DNA probe. The absA transcript time course analysis incorporated an absA1 probe together with a glk (glucose kinase; Angell et al., 1992) probe, which served as an internal standard for normalizing the quantity of RNA in each assay (Aceti and Champness, 1998). The 455 bp absA1 probe was generated using PCR with the WC64 forward primer and the 5′32P end-labelled WC20 reverse primer. The template for absA1 probe synthesis was pCB400, containing the 2 kb BamHI region of the absA locus cloned into pIJ2925. A 309 bp 32P end-labelled glk probe was also generated using PCR from the primers and template described by Aceti and Champness (1998). Primers (50 pM) were end labelled using the T4 polynucleotide kinase (Promega) forward reaction, as described by the manufacturer with minor modifications. Before initiating the end-labelling reaction, each primer was incubated with spermidine (10 mM final concentration) at 70°C for 10 min. Similarly, ethanol precipitation of the labelled oligos was facilitated by the addition of 2 µg of glycogen. The labelled oligo was divided between duplicate 50 µl PCR reactions. The reaction mix contained 20 pM each primer, 100 ng of template, 0.2 mM dNTPs, 1.5 mM MgCl2, 5% glycerol, 2.5% DMSO, 1% formamide and 1.25 U of Taq polymerase (Perkin-Elmer). Thermal cycler conditions were 95°C for 5 min, followed by 30 cycles of 95°C for 1 min, 65°C for 45 s, 72°C for 1 min and a final extension at 72°C for 10 min. The S1 nuclease protection assay was performed as described previously (Aceti and Champness, 1998). Replicate RNA isolates were tested in independent S1 experiments. Time course experiments included E. coli RNA as a negative control. In addition, the presence of excess probe was verified by treating an RNA sample with twofold concentrations of each probe and comparing their signals with the same sample treated with normal levels of probe. Results were analysed using electrophoresis on 6% polyacrylamide sequencing gels (Sambrook et al., 1989) and autoradiography. Transcript sizes were estimated by running 32P end-labelled φX174–HinfI molecular weight markers (Promega) on the same gel. To map the absA1 transcription start site, a sequencing ladder was generated from primer WC20 using the fmol DNA sequencing system (Promega) and compared with S1-treated 18 h C542 RNA hybridized to the absA1 probe.
The region upstream of the absA2 translation start site was probed with a 504 nt absA2 probe, protecting the region from 330 nt upstream to 174 nt downstream of the absA2 translational start, and generated by PCR using forward primer WC24 and 5′32P-labelled WC29 reverse primer. The PCR template was pCB460 (pSK+ carrying a 3.9 kb BamHI–XhoI insert with the entire absA locus). The absA2 probe was used in S1 nuclease protection assays with 18 h and 30 h C542 RNA.
We are grateful to The Sanger Centre for access to the Streptomyces coelicolor genome sequence. We thank Renqui Kong for expert technical assistance. This work was supported by grant MCB9604055 (to W.C.C.) from the National Science Foundation.