Proteins encoded by the conservon of Streptomyces coelicolor A3(2) comprise a membrane-associated heterocomplex that resembles eukaryotic G protein-coupled regulatory system

Authors


*E-mail ueda@brs.nihon-u.ac.jp; Tel. (+81) 466 84 3937; Fax (+81) 466 84 3935.

Summary

Streptomyces coelicolor A3(2) retains unique conserved operons termed conservons. Here, one of the conservons (cvn9), which encodes five proteins (A9-E9), was characterized. Mutants for cvnA9 and cvnA10 conditionally overproduced actinorhodin and performed precocious aerial growth, while a cvnE9 mutant showed the parental phenotype. Transcription of bldG, adpA and bldN was upregulated in the cvnA9 mutant. A9-D9 were detected in the insoluble fraction of cell-free extract of S. coelicolor by Western analysis. Biochemical analyses revealed that A9 has ATP-hydrolysing and adenine nucleotide-binding activities; D9 has GTP-hydrolysing and guanine nucleotide-binding activities; and E9 shows a typical spectrum similar to cytochrome P450. The comprehensive interaction assays demonstrated the occurrence of specific interactions between A9 and B9, A9 and C9, B9 and B9, B9 and D9, and C9 and D9. A9 associated with and dissociated from B9 (and C9) when ATP and ATP-γ-S were supplied in the reaction respectively. Similarly, D9 associated with and dissociated from B9 (and C9) when GTP and GTP-γ-S were supplied respectively. A9 and B9 were also shown for the occurrence as homocomplexes. Probably, Cvn9 proteins comprise a membrane-associated heterocomplex resembling the eukaryotic G-protein-coupled receptor system, which may serve as a signal transducer that connects to the bld cascade.

Introduction

The filamentous, Gram positive, soil bacteria of the genus Streptomyces perform complex morphological differentiation resembling that of filamentous fungi (Chater, 1993). Early in the life cycle, the organism grows as branching substrate hyphae. In response to nutritional limitation and various environmental signals, the substrate hyphae produce aerial mycelia, which culminate into spore chains by septum formation at regular intervals. Streptomyces is also characterized by the ability to produce a wide variety of secondary metabolites, including antibiotics and antitumour substances, that have important applications in pharmaceutical industries (Miyadoh, 1993). Due to the unique and useful characteristics, the genomic information from Streptomyces and related organisms has attracted a great deal of attention. To date, three Streptomyces genomes and one Nocardia and one Frankia genome have been completely sequenced.

Unveiling the whole genome of the organisms has revealed various notable properties. Among them, a common interesting feature is the presence of a unique operon that consists of four or five coding sequences (CDS). A sequence similarity search suggests that they encode (from 5′ to 3′) a putative membrane protein that contains coiled-coil domains and an ATPase domain (CvnA), a roadblock/LC7 family protein (CvnB), an unknown conserved protein (CvnC), a putative GTP/GDP-binding protein (CvnD) and cytochrome P450 (tentatively CvnE). While the last CDS for cytochrome P450 is present only in some operons, the other four CDSs are conserved in all the operons. Streptomyces coelicolor A3(2), the first Streptomyces strain whose genome has been completely sequenced, retains 13 copies of the operon (Bentley et al., 2002). Based on the highly conserved feature, Bentley et al. named the operon as ‘conservon’ and annotated the 13 operons as cvn1–13, and the CDSs in each operon as cvnA, cvnB, cvnC and cvnD from 5′ to 3′.

Previously, we reported the first genetic evidence for the involvement of a conservon in Streptomyces physiology (Komatsu et al., 2003). During the study on the regulation of morphological and physiological development in Streptomyces griseus, the streptomycin producer, we found that the introduction of a DNA fragment on a plasmid restored the aerial mycelium formation in a mutant for amfR, the central regulator for the onset of morphogenesis (Ueda et al., 1993; 1998; Ueda et al., 2005). Nucleotide sequencing revealed that the DNA fragment contained the rar operon (restoration of aerial mycelium in amfR mutant), which shows a marked similarity to the cvn9 operon of S. coelicolor A3(2). The activity of aerial growth induction strongly suggested that the operon has some role in the regulation of the onset of cellular differentiation. The original DNA fragment that showed the aerial mycelium-inducing activity carried rarA, the first CDS encoding a putative membrane protein. An rarA-disruption mutant of S. griseus showed precocious aerial mycelium formation and overproduced the secondary metabolites, which suggested that rarA is involved in the negative regulation of the onset of both morphological and physiological differentiation in S. griseus. We also showed that the disruption of cvnD9 caused a similar effect on the morphological and physiological development in S. coelicolor A3(2), which suggested that the operon plays a similar regulatory role in the two distinct Streptomyces (Komatsu et al., 2003).

Although the conserved operon structure strongly suggests that the constituents of the operon have a concerted function, the biochemical properties of each Cvn component have not yet been characterized. Here, we selected cvn9, the rar orthologue of S. coelicolor A3(2), as a model and examined the biochemical characteristics of the five components (CvnA9–E9). The results indicate that the Cvn9 proteins comprise a membrane-associated heterocomplex, which resembles the eukaryotic G protein-coupled regulatory system.

Results

Specific distribution of conservon in Actinomycetales

Table 1 summarizes the distribution of conservon in bacterial genomes. Till date, the presence of the operon has been shown only by the whole genome sequencing studies, except for rar, one of the operons of S. griseus that was identified in our previous study (Komatsu et al., 2003). The other bacteria that have been elucidated for their whole genome sequence do not retain conservon, although some bacteria retain homologues for cvnB and/or cvnC. This result demonstrates that conservon distributes specifically to the genera that belong to Actinomycetales (high G + C Gram-positive bacteria). In addition, it is evident that multiple copy-numbers are present in the genera that are characterized by the ability to perform complex morphological differentiation and secondary metabolite formation including Streptomyces and Nocardia.

Table 1.  Specific distribution of conservon in Actinomycetales genome.
OrganismaNumber of cvnb
Streptomyces coelicolor A3(2)13 (4)
Streptomyces scabies13 (4)c
Streptomyces avermitilis12 (2)
Streptomyces griseus9 (4)
Nocardia farcinica7 (1)
Thermobifida fusca5 (1)
Frankia sp. Ccl33 (1)
Rhodococcus sp. RHA11 (0)
Mycobacterium avium1 (0)
Mycobacterium tuberculosis CDC15511 (0)
Mycobacterium tuberculosis H37RV1 (0)
Mycobacterium bovis1 (0)
Mycobacterium sp. MCS1 (0)
Mycobacterium leprae1 (0)

Phenotypes of cvn mutants

Several disruption mutants for cvn genes were generated to study the effect of their inactivation on the phenotype of S. coelicolor A3(2). A cvnA9 mutant showed an identical phenotype as the cvnD9 mutant that was characterized previously (Komatsu et al., 2003); it conditionally overproduced the antibiotic actinorhodin (a blue-coloured pigment) and performed aerial mycelium and spore formation on the media supplied with glucose. The phenotype was evident when the medium was supplied with relatively high concentration of glucose (Fig. 1); on solid medium supplied with 2% glucose, the cvnA9 mutant yet performed actinorhodin production and aerial mycelium and spore formation while those in the parental strain were markedly repressed. Meanwhile, the phenotype was not clear when the mutant was grown on media supplied with maltose (Fig. 1A) or mannitol and the medium that was not supplied with sugar (data not shown). The introduction of a plasmid that contained cvnA9 restored parental phenotype in the cvnA9 mutant (data not shown). The growth rate of the cvnA9 was the same as that of the parental strain (data not shown). We also generated a mutant for another cvnA paralogue (cvnA10) and found that the mutant showed a similar phenotype as the cvnA9 mutant (Fig. 1). On the other hand, a cvnE9 mutant showed parental phenotype under any condition tested (data not shown).

Figure 1.

Phenotypes conferred by the inactivation of cvn genes in S. coelicolor A3(2).
A. Colony appearances of the mutants for cvnA9 and cvnA10 of S. coelicolor A3(2) grown on Bennett's solid medium added with 2% glucose (left) and 1% maltose (right). The colonies of the parental strain and cvnD9 mutant that was described previously (Komatsu et al., 2003) are also shown. The mutants produce actinorhodin (diffusible blue pigment) and partially form aerial mycelium and spores (rough and white colony surface).
B. Scanning electron micrographs of the above sporulating colonies of cvnA9 and cvnA10 mutants grown on Bennett's solid medium added with 2% glucose. Parental strain grew only substrate mycelia under the condition. Electron microscopic observation was carried out as described previously (Ueda et al., 1993). Bar, 1 μm. All patches were photographed after 4 days' culture at 28°C.

The above mutant phenotype made us to speculate that the inactivation of cvnA9 affect the expression of bld genes, which control the onset of both antibiotic production and morphological differentiation in S. coelicolor A3(2) via cascade regulation (Chater and Horinouchi, 2003). Therefore, we studied the transcription level of several bld genes by a low-resolution S1 protection analysis (Fig. 2); the result showed that the transcription of bldG, adpA (bldH) and bldN was upregualted in the early growth phase in the cvnA9 mutant, while the transcription of bldD appeared to be in the same level as in the parental strain. The effect of the mutation on the transcription of actII-ORF4, the regulatory gene specific for actinorhodin biosynthesis, was not clear due to its low expression level, but it appeared slightly upregulated in the early growth phase. These results implied that Cvn9 has a connection with the bld cascade.

Figure 2.

Transcription of the bld genes in the cvnA9 mutant. The transcripts of several bld genes in the parental and cvnA9 mutant strain were quantified by low-resolution S1 mapping. RNA was extracted from the cells grown on Bennett's/glucose solid medium for indicated periods. The S1-protected fragments for the promoter region of bldD encoding a global transcriptional repressor (101 bp) (Elliot et al., 1998), bldG encoding an anti-anti-sigma factor protein (98 bp for P1 and 139 bp for P2) (Bignell et al., 2000), adpA(= bldH) encoding a global transcriptional activator (119 bp for P1) (Takano et al., 2003b), bldN encoding an ECF sigma factor (154 bp) (Bibb et al., 2000), and actII-ORF4 encoding a specific transcriptional activator for actinorhodin biosynthesis (106 bp) (Gramajo et al., 1993) are shown. The order of the genes presented is based on the hierarchy proposed by Willey et al. (1993). The specific transcripts for P2 and P3 of adpA could not be detected in this experiment. The quality of mRNA was checked by the control assay for hrdB, a gene for house-keeping sigma factor.

Furthermore, in order to biochemically characterize the Cvn proteins, the cellular localization of Cvn9 proteins in the parental and cvnA9 mutant strains was studied by Western analyses (Fig. 3). CvnA9 (A9) protein was fractionated into the insoluble fraction of parental strain, which indicates that A9 is associated with the cell membrane as predicted from the hydrophobicity profile (Fig. 4A). No hybridization signal for A9 was detected with the cell extract of the cvnA9 mutant, which confirmed the true disruption as well as the specificity of antibody. B9 and C9 proteins were detected in the insoluble fractions of both the parental strains and the cvnA9 mutant. The result suggests that these proteins are also associated with membrane, although it is possible that the antibodies cross-reacted to other CvnB and CvnC proteins. On the other hand, D9 was detected only in the parental strain both in the soluble and insoluble fractions. It is not yet known why D9 was not detected in the cvnA9 mutant. It is possible that the protein is subject to rapid degradation without the correct function of A9.

Figure 3.

Immunological detection of Cvn9 proteins in the cell extract of S. coelicolor A3(2). The cell-free extracts of parental strain and a cvnA9 mutant (ΔcvnA9) of S. coelicolor A3(2), which were cultured for the indicated periods, were separated into cytoplasmic (soluble) and membrane (insoluble) fractions by ultracentrifugation (70 000 g for 1 h) and analysed by SDS-PAGE. The gel was examined for the presence of CvnA proteins by Western blotting.

Figure 4.

ATP-hydrolysing and -binding activities of CvnA9 and its truncated variants.
A. Hydrophobicity plot of CvnA9 analysed with an algorithm by Kyte and Doolitle (1982) with sliding window 20. The regions corresponding to the coiled-coil domains and ATPase homology domain are indicated. The corresponding positions of the intact (A9) and the truncated variants (A9a–e) used in this study are also shown. The A9 proteins were expressed and purified as GST-fused proteins.
B. Polyethyleneimine-cellulose thin layer chromatography (upper) and paper chromatography (lower) of 32P-labelled ATP that was reacted with GST-A9 protein and the truncated variants (A9a, A9b and A9c). The former analysed the reaction with [α-32P]-ATP to detect adenine nucleotides (AMP, ADP and ATP), and the latter analysed the reaction with [γ-32P]-ATP to detect free monophosphate (Pi) and diphosphate (PPi). The reaction was performed for the indicated periods, and the radioactive signal was detected by an image analyser. The standard positions indicated by arrowheads were determined by analysing authentic samples.
C. Phosphorylation assay. Each A9 recombinant was incubated with [γ-32P]-ATP and [α-32P]-ATP for the indicated periods and analysed by semi-native-PAGE followed by radioactive signal detection on an image analyser. The panels show the gel images at the corresponding positions to which the A9 proteins migrated.
D. ATP-binding assay by UV cross-linking. Each A9 protein was mixed with [α-32P]-ATP and UV-irradiated on ice for 5 min. Subsequently, the reactant was similarly analysed by SDS-PAGE.

ATP-hydrolysing activity of CvnA9

The above results strongly suggested that Cvn9 proteins comprise a membrane-associated complex and serve as a signalling apparatus that regulates the onset of physiological and morphological development in S. coelicolor A3(2). To study the details of its function, we performed biochemical analyses on each component.

A9 shows a weak similarity to histidine kinase. In addition, protein motif search using Pfam database (http://www.sanger.ac.uk/Software/Pfam/) suggested that the region corresponding to 262–375 aa of A9 is an ATPase domain (Fig. 4A). Therefore, the A9 recombinant protein was assessed for its reaction with ATP by using the 32P-labelled ATP.

As shown in Fig. 4B, the intact A9 recombinant showed an ATPase activity; incubation of the protein with [α-32P]-ATP and [γ-32P]-ATP generated radioactive ADP (upper panel) and free monophosphate (lower panel) respectively. The truncated A9 mutants (A9a–e see Fig. 4A) were also assessed for ATPase activity. A9a, which lacked the N-terminal 60 aa region, showed an ATPase activity higher than that of intact A9. On the other hand, the other four truncated mutants (A9b–e) did not show the activity (data not shown for A9d and A9e). It is possible that the N-terminal region is related to the affinity of A9 to the substrate.

The A9 proteins were also examined for autophosphorylation activity. As shown in Fig. 4C, no signal was detected with any of the A9 proteins when they were incubated with [γ-32P]-ATP. On the other hand, a marked radioactive signal was detected when A9a was incubated with [α-32P]-ATP. The wild-type A9 showed a smear signal after a prolonged incubation (30 min) with [α-32P]-ATP. The UV cross-linking treatment (Fig. 4D) showed the binding of 32P derived from [α-32P]-ATP to A9 and A9a. The results indicate that A9 does not perform autophosphorylation instead it plays the role of binding adenine nucleotides. The nucleotides were probably ATP and ADP, because no AMP was generated after the incubation of A9 with ATP (Fig. 4B).

Previously, we observed that the introduction of rarA, the cvnA9 orthologue of S. griseus, on a high-copy number plasmid restored aerial mycelium formation in an amfR mutant of S. griseus (Komatsu et al., 2003). Here, we introduced the cvnA9 of S. coelicolor and the truncated variants (see Fig. 4A) into the amfR mutant of S. griseus. The introduction of A9 and A9a region on pIJ702 restored aerial mycelium formation of the amfR mutant (the data are available as a supplementary Fig. S1). On the other hand, the introduction of A9b region did not show the restoration activity.

GTP/GDP-binding and GTPase activity of CvnD9

Sequence similarity search has indicated that D9 protein contains two domains that show similarities to the GTP/GDP-binding domains of eukaryotic G proteins (Komatsu et al., 2003). As shown in Fig. 5A, incubation of the D9 recombinant with [α-32P]-GTP and [γ-32P]-GTP generated radioactive GDP (upper panel) and free monophosphate (lower panel) respectively. The result indicates that D9 has a GTPase activity. The UV cross-linking treatment (Fig. 5B) demonstrated that 32P derived from [α-32P]-GTP bound D9. The radioactive signal was reduced by the addition of non-radioactive GTP or GDP to the reaction mixture, but the addition of other nucleotides did not affect the signal intensity. These results suggest that D9 specifically binds GTP and GDP.

Figure 5.

GTP-hydrolysing and -binding activities of CvnD9.
A. Polyethyleneimine-cellulose thin layer chromatography (upper) and paper chromatography (lower) of 32P-labelled GTP reacted with the D9 protein. The former analysed the reaction with [α-32P]-GTP to detect guanine nucleotides (GMP, GDP and GTP) and the latter analysed the reaction with [γ-32P]-GTP to detect free monophosphate (Pi) and diphosphate (PPi). The reaction was performed for the indicated periods, and the radioactive signal was detected by an image analyser. The standard positions indicated by arrowheads were determined by analysing authentic samples.
B. GTP-binding assay by UV cross-linking. The D9 protein was mixed with [α-32P]-GTP in the absence (lane 1) and presence of 0.2 mM of cold GTP (lane 2), GDP (lane 3), GMP (lane 4), ATP (lane 5), CTP (lane 6) and UTP (lane 7). After UV irradiation on ice for 5 min, the reactant was similarly analysed as described in Fig. 4C. The panel shows the gel images at the corresponding position to which the D9 protein migrated.

Spectral characteristics of CvnE9

The probable cytochrome P450 protein E9 has been annotated as CYP157C1 according to the website http://drnelson.utmem.edu/CytochromeP450.html. Lamb et al. (2002) previously described that this protein shows similarity to the known cytochrome P450s that are orphans with little clue as to function. The purified recombinant of E9 was examined for absorption spectra and was shown to possess the typical characteristics of the family monooxigenase. The absorbance spectrum of oxidized E9 between 300 and 600 nm (supplementary Fig. S2) showed the Soret peak at 418 nm; the reduced form had a broad peak at the same position. A reduced CO difference spectrum showed a maximum at 447 nm.

Comprehensive protein–protein interaction analyses

The conserved operon structure strongly suggests that the proteins encoded by the operon have a concerted function via protein–protein interaction. To assess the association between the Cvn proteins, a comprehensive two-hybrid analysis was performed. The CDSs for all the Cvn9 proteins as well as the truncated variants for A9 (A9c and A9e; Fig. 4A) were cloned onto the bait and target plasmids, both of which were cotransformed into an Escherichia coli host strain (see Experimental procedures). As summarized in Table 2, the marked β-galactosidase activities, which indicate specific interaction between the bait and the target proteins, were observed between A9c and B9, B9 and B9, B9 and D9, and C9 and D9. A9, A9e and E9 did not show marked interaction activity with any Cvn component.

Table 2.  Two-hybrid analysis between Cvn components.
Baitβ-Galactosidase activity (× 10−5ΔA410 min−1 mg−1)
A9B9C9D9E9
  1. Data shown are means ± SD from three separate experiments. Bold numbers are the activities suggesting the occurrence of specific binding. NT, not tested.

Target
A97.6 ± 0.712.6 ± 3.78.7 ± 1.89.3 ± 2.18.7 ± 0.5
A9cNT19.4 ± 2.94.3 ± 0.34.4 ± 0.44.6 ± 0.3
A9eNT9.5 ± 0.210.6 ± 0.89.8 ± 1.210.4 ± 0.6
B99.0 ± 2.031.9 ± 2.011.0 ± 1.519.6 ± 1.211.5 ± 2.7
C99.6 ± 0.013.4 ± 4.19.5 ± 1.651.9 ± 6.914.0 ± 0.9
D99.7 ± 0.016.6 ± 2.779.2 ± 1.36.9 ± 0.612.4 ± 3.2
E910.0 ± 0.713.4 ± 0.610.9 ± 1.610.9 ± 0.08.3 ± 1.2

Furthermore, to characterize the interaction between the Cvn components, a comprehensive pull-down analysis was performed with each of the purified recombinant protein. A bait Cvn protein, which was prepared as a GST-fused protein was incubated with a target Cvn protein, which was prepared as a non-tagged recombinant protein. The reactant was applied onto a glutathione Sepharose column, and the eluted fraction was examined for the presence of each Cvn protein by Western blotting. The analysis was performed for all pairs of different Cvn proteins. As a result, marked interaction was observed between A9 and B9, A9 and C9, B9 and D9, and C9 and D9.

The interaction of A9 with B9 as well as C9 was assessed for the dependence on mononucleotides (Fig. 6A). The binding of A9 to B9 and C9 occurred when ATP was supplied in the binding buffer. However, binding did not occur when ATPγS, an ATP analogue that is not hydrolysed by the ATPase activity, was supplied. The supply of ADP caused interaction between A9 and C9 but not between A9 and B9.

Figure 6.

In vitro protein–protein interaction analysis by pull-down assay.
A. Interaction of GST-A9 with B9 and C9. The bait (GST-A9) and the target (B9 or C9) proteins were incubated in the absence and presence of ATP, ATPγS, or ADP. After the binding and elution from glutathione Sepharose column, the proteins were applied to SDS-PAGE followed by Western blotting using a specific antibody (anti-A9, anti-B9 and anti-C9). The presence of signals for B9 and C9 indicates their specific interaction with A9.
B. Interaction of GST-D9 to B9 and C9. The bait (GST-D9) and the target (B9 or C9) proteins were incubated in the absence and presence of GTP, GTPγS, or GDP and analysed similarly as described above. The presence of signals for B9 and C9 indicates their specific interaction with D9.

Similarly, the binding of B9 and C9 to D9 was assessed for its dependence on guanine nucleotides (Fig. 6B). Both B9 and C9 bound D9 when GTP or GDP was supplied, while the binding did not occur when GTPγS was supplied. Binding that occurred without the addition of guanine nucleotides could be due to the occurrence of a few D9 recombinants in a GDP-bound form during the expression in E. coli.

Molecular weight measurement of CvnA9 and CvnB9 homocomplexes

The result of two-hybrid analysis suggested that A9 and B9 perform self-assembly. To examine the homooligomer formation, each purified recombinant protein was analysed for its apparent molecular weight. SDS-PAGE analyses of the protein samples after a chemical cross-linking treatment showed that A9 and B9 partially migrated as multicomplexes (Fig. 7), while the other Cvn proteins (C9, D9 and E9) migrated only as monomers (data not shown). A9, a 60 kDa protein in its monomer form, partially migrated as a dimer, a trimer and a tetramer or more higher homocomplexes after incubation with DMP. Similarly, B9, a 14 kDa protein, partially migrated as a dimer and a trimer. The same results were obtained by seminative PAGE analyses of the native A9 and B9 recombinant proteins (Fig. S3). For B9, gel filtration analysis showed its occurrence as a trimer (Experimental procedures and Fig. S3).

Figure 7.

SDS-PAGE analysis of chemically cross-linked CvnA9 and CvnB9. The A9 and B9 recombinant proteins were incubated with 0–2.0 mM of DMP and applied onto an SDS-PAGE gel followed by Western blotting using anti-A9 and anti-B9 antibodies respectively. The positions of molecular size standards indicated by arrowheads were determined by separately staining the lane for molecular weight marker.

Discussion

This study revealed that the four proteins encoded by one of the conservons of S. coelicolor A3(2) comprise a heterocomplex (Fig. 8); specific interactions occur between C-terminal A9 and B9, B9 and D9, and C9 and D9. The result of the in vitro pull-down experiment indicated that the interaction between A9 and C9 also takes place. The two-hybrid analysis failed to reveal the interaction of A9, possibly because the protein was not fully active in the host cell so as to reveal the true nature of the interaction. In addition to the interaction between the different components, A9 and B9 were shown to occur as a homocomplex. This raises a possibility that the A9-B9-D9-C9 heterocomplex further assembles into a homooligomeric structure via the self interaction of A9 and B9. The result of Western analysis (Fig. 3) showed that the Cvn proteins were mostly detected in the insoluble fraction. We think it most likely that the Cvn proteins exist in association with cell membrane based on the feature of CvnA9 as a membrane-associating protein, although we can not exclude the possibility that the proteins form or associate with a large macroprotein complex.

Figure 8.

A simple model for heterocomplex formation by Cvn9 proteins. The four Cvn proteins (CvnA9, CvnB9, CvnC9 and CvnD9) comprise a membrane-associated heterocomplex. The molecular ratio of each component in the heterocomplex is not yet known. By the ATP-binding activity of CvnA9, CvnB9 and CvnC9 dissociate from CvnA9. The GDP-bound CvnD9 associates with CvnB9 and CvnC9, while the GTP-bound CvnD9 dissociates from CvnB9 and CvnC9. CvnD9 has a GTPase activity. The correlation between the ATP-hydrolization by CvnA9 and the GTP-form formation in CvnD9 is not yet clear.

The pull-down experiment showed that the interaction of A9 with B9 and C9 is affected by adenosine phosphates. The inhibitory effect of ATPγS on the interaction indicates that the ATPase activity of A9 is involved in the interaction with both B9 and C9. The interaction between A9 and C9, which occurred in the presence of ATP and ADP but not in the presence of ATPγS, possibly depends on the nucleotide-binding state of A9; ATP-bound A9 dissociates from C9 and ADP-bound A9 associates with C9. The interaction that occurred in the presence of ATP is probably the result of conversion of ATP to ADP by the ATPase activity of A9. On the other hand, interaction between A9 and B9 did not occur when ADP was supplied to the reaction mixture. It suggests that the ATP-hydrolysing activity of A9, and not the ADP-binding activity, causes the interaction. A possibility is that phosphoryl transfer to B9 takes place and facilitates the interaction between A9 and B9, but we have been unsuccessful in detecting phosphorylated B9.

The pull-down experiment also showed that B9 and C9 associate with the D9 domain and the interaction depends on the presence of guanosine phosphates. Analogous to the interaction between A9 and C9, the result suggests that GDP binding of D9 promotes the association with B9 and C9, and GTP binding causes the dissociation of the complex (Fig. 8). The behaviour of D9 is reminiscent of the typical action of eukaryotic heterotrimeric G protein and associating surface receptor (G protein-coupled receptor; GPCR); the GDP-bound form of G protein associates with GPCR to form a membrane-associated heterocomplex, while the GTP-bound form dissociates from GPCR and in turn associates with an effector protein (Strader et al., 1994). The effector protein is activated by interaction with the G protein and performs further signalling and/or a new intracellular biochemical reaction. The GPCR system is widely distributed in eukaryotes. On the other hand, the major signal transducer common to prokaryotes is the two-component system, which consists of a sensor histidine kinase and a response regulator (Mizuno, 1998).

The well-characterized GTP-binding protein that widely occurs in prokaryotes is the protein known as Era (E. coliRas-like) family. Era is a cytoplasmic protein essential for cell viability. It has been shown that this protein is involved in ribosome assembly and plays an important role in growth rate-regulated checkpoint in E. coli (Britton et al., 1998). The Era family is highly conserved not only in prokaryotes but also in eukaryotes, which suggests its general role in the control of cellular growth. A homologue is also retained by S. coelicolor A3(2) (SCO2539; http://www.sanger.ac.uk/Projects/S_coelicolor/). Another GTP-binding protein known for its wide distribution is Obg/Gtp1 family, which is also a cytoplasmic protein essential for viability (Bourne et al., 1991). The studies on Bacillus subtilis (Welsh et al., 1994) and Streptomyces (Okamoto et al., 1997; Okamoto and Ochi, 1998) have suggested that the protein family also plays an important role in the regulation of morphological differentiation. In contrast to these, the wide but limited distribution of the conservon in Streptomyces and related bacteria implies that the role of the G protein complex is related to the physiological property specific to the group of organisms.

Presumably, the proteins encoded by other cvn operons similarly exist in association with the cell membrane of S. coelicolor A3(2) and related organisms. The phenotype of cvnA10 implies that Cvn10 has a similar role and function as Cvn9. What is the role of the conservon in these organisms? The behaviour of the Cvn proteins strongly suggests that the complex serves as a signal transducer, which receives an environmental signal and stimulates the corresponding intracellular function. Like the eukaryotic G protein system, the GTP-bound D9 domain may dissociate from the Cvn complex and in turn associate with an effector molecule to activate it. If this is the case, an important clue to deduce the role of a putative signalling system lies within the function of the effector protein. The cytochrome P450 encoded by the fifth CDS (E9) is a candidate for the effector protein, but no evidence was obtained that suggests the interaction of E9 protein with the other Cvn9 domains. Unlike cvnA9 and cvnD9, gene disruption of cvnE9 did not cause any effect on the phenotypes of S. coelicolor. Therefore, we currently speculate that E9 component is not directly involved in the Cvn function, although some relationship is still implied by the conserved feature. In Nocardia farcinica, one of the conservons (nfa53060–53110; http://nocardia.nih.go.jp) contains two tandem CDSs for P450 proteins; the translational initiation codon of these CDSs overlaps the stop codon of the preceding CDS, suggesting their translational coupling to the upstream Cvn proteins.

A possibility is that CvnA protein acts as a sensor for a specific ligand molecule and transmits the signal to the effector modules via the function of CvnBCD. The multiple copies of Cvn systems may correspond to the various environmental signals and/or unidentified hormone-like substances. The evidence obtained by the genetic study suggests that the Cvn9, and possibly Cvn10, mediate the signalling leading to the onset of morphological differentiation and secondary metabolite formation. A marked difference between the GPCR system and the two-component system is in the mode of regulation of the target function; the GPCR system activates the effector module post-translationally via protein–protein interaction, while the two-component system controls the transcription of the target gene via the binding of a phosphorylated response regulator to the promoter. In Streptomyces physiology, some unknown mechanisms could be effectively controlled by the former system, which enables quicker response of the target function to the corresponding signal than the latter system.

The result of transcriptional analysis performed in this study (Fig. 2) raises the possibility that the Cvn9 system regulates the onset of development in connection with the bld cascade. Based on the understandings for the hierarchical relationships among the bld genes (Chater and Horinouchi, 2003), we assume that the Cvn9 function relates to a transcriptional control for bldG and/or its hierarchically upstream regulators. bldG encodes an anti-anti-sigma factor protein that plays an essential role for the onset of antibiotic production and aerial mycelium formation in the early step of the cascade regulation (Bignell et al., 2000). It is not yet known how the expression of bldG is regulated. In addition, we also assume that the conservon system has a connection to the other eukaryotic regulatory mechanisms including tyrosine kinase and Ser/Thr kinase, which have been shown to play significant roles in Streptomyces physiology (Hong et al., 1993; Ueda et al., 1996; Zhang, 1996); previously, we identified a Ser/Thr kinase AfsK of S. griseus by a similar biological activity as that we observed with the rar opron in S. griseus (Ueda et al., 1996). Elucidating the role of the regulatory system will open up another aspect of the complex genetic and physiological properties of Streptomyces, the boundary organism between eukaryotes and prokaryotes.

Experimental procedures

Bacterial strains, plasmids and media

Streptomyces coelicolor A3(2) M130 was obtained from John Innes Centre, UK. The amfR mutant of S. griseus was described previously (Ueda et al., 1998). Escherichia coli JM109 (Maniatis et al., 1982) and Rosetta (DE3) pLysS (Novagen) were used as hosts for DNA cloning and expression of recombinant proteins respectively. The standard experimental conditions and materials used for the genetic manipulation of E. coli and Streptomyces were as described by Maniatis et al. (1982) and Kieser et al. (2000) respectively. For the details of plasmids and media composition, see Appendix S1. Oligonucleotide primers used for plasmid construction are shown in Table S1.

Genetic analyses

The cvn mutants were generated by the homologous recombination technique as described previously (Komatsu et al., 2003) by using disruption plasmids (Appendix S1). The introduction of cvnA9 and truncated variants into S. griseus amfR mutant was performed similarly as described previously (Komatsu et al., 2003) (see also Appendix S1).

S1 mapping

Methods for RNA preparation and S1 nuclease mapping were undertaken as described by Kelemen et al. (2001). Oligonucleotide primers used for probe construction are shown in Table S1. All the downstream primers of the S1 probes (Appendix S1) were labelled at their 5′-end with [γ-32P]-ATP using T4-polynucleotide kinase. The protected fragments were analysed on 6%-polyacrylamide gels. A 100 bp ladder marker (Takara) labelled with [γ-32P]-ATP using T4-polynucleotide kinase was used as a standard to estimate the transcript sizes in the low-resolution assay. Radioactivity detection in this study was performed by exposing samples to a Fuji imaging plate (Fuji Film), which was scanned by a STORM 860 image analyser (Amersham Biosciences).

Preparation of Cvn proteins and antibodies

The cell-free extract of S. coelicolor used to study the localization of Cvn proteins (Fig. 3) was prepared as follows. The cells of S. coelicolor cultured at 28°C for 3 days in J medium were washed three-times with 100 mM potassium phosphate buffer (pH 7.2), disrupted by mechanical cell presser (G1000, Gaulin-Rannie), and centrifuged at 6000 g for 0.5 h. The resultant cell-free extract was centrifuged at 70 000 g for 1 h to obtain soluble fraction (supernatant) and insoluble fraction (precipitate). Each fraction was subjected to Western analysis to detect each Cvn protein. For the preparation of soluble Cvn recombinants (A9–E9 and A9a–e; see text), each protein was expressed as a protein fused with glutathione-S-transferase (GST). The expression plasmids (Appendix S1) directed the expression of the Cvn proteins that carry an N-terminal GST in E. coli Rosetta (DE3) pLysS. Each E. coli strain harbouring the expression plasmid was first cultured overnight at 28°C in Luria–Bertani (LB) liquid medium. Subsequently, the seed culture was inoculated at 1% into 150 ml of 2 × YT medium prepared in a 500 ml baffled Erlenmeyer flask. After 4.5 h of cultivation at 28°C with rotary shaking (135 r.p.m.), isopropyl-β-d-thiogalactopyranoside (IPTG) was added at a final concentration of 0.1 mM, and the culturing was continued at 16°C for 16 h. For the expression of CvnE9, a cytochrome P450 protein, the main culture was supplemented with 1 mM 5-aminolevulinic acid, a haem precursor. The E. coli cells thus obtained were harvested by centrifugation; suspended in PBS buffer (Appendix S1) containing 1% Triton X-100, 5 mM dithiothreitol and 0.16% (w/v) protease inhibitor cocktail (Sigma, no. P8465); and disrupted with a mechanical cell presser. The cell-free extract was centrifuged at 70 000 g for 30 min, and the resultant supernatant was used for GST affinity chromatography. The supernatant was applied onto a 5 ml GSTrap FF column (Amersham Biosciences) with an ÄKTA FPLC system (Amersham Biosciences) according to the manufacturer's instructions. For the removal of the GST region, Precision protease (Amersham Biosciences) was added to the column according to the manufacturer's instructions. Both the GST-fused and GST-free forms of the Cvn recombinants were used for the biochemical assays. The GST-free A9c and C9 proteins were also used as antigens for the preparation of specific antibodies.

CvnB9, D9 and E9 were also expressed as recombinants carrying a C-terminal hexahistidine (6 × His) tag for use as antigens for antibody preparation. The E. coli Rosetta (DE3) pLysS cells that harbour the expression plasmids (Appendix S1) were cultured under the same conditions as above, except that the cultivations were performed at 37°C. The Cvn recombinants, which were obtained as an inclusion body, were purified by the standard protocol and used as an antigen. Protein concentrations were measured with a protein assay kit (Bio-Rad) using bovine serum albumin as the standard. Polyclonal antibodies were raised against the purified Cvn9 proteins by the standard protocol using mouse as a host.

Spectral characterization of E9

The absorption spectra of E9 were obtained by the method described by Omura and Sato (1964). The absorption of the purified E9 (0.1 mg) that was dissolved in 100 mM potassium phosphate buffer (pH 7.2) supplemented with (reduced form) and without (oxidized form) sodium dithionite was measured between the wavelengths of 300–600 nm. The absorption of the reduced protein was also measured immediately after exposure to carbon monoxide (CO-binding form). These spectra were recorded on a Shimadzu UV-1240 spectrophotometer.

Gel electrophoresis and Western analysis

The Cvn recombinant proteins were analysed by the standard method of SDS-polyacrylamide gel electrophoresis (SDS-PAGE) followed by Coomasie Brilliant Blue (CBB) gel staining or Western analysis (Maniatis et al., 1982). In the seminative PAGE used for the multimer formation of A9 and B9, protein samples were dissolved in the standard 5 × SDS-PAGE loading buffer (Appendix S1), and applied onto a native polyacrylamide gel (7%−15%). The specific detection of Cvn proteins was carried out by the standard Western analysis using the specific mouse polyclonal antibodies (primary antibody) and a commercial anti-mouse immunoglobulin (isolated from sheep) linked to horseradish peroxidase (the secondary antibody; Amersham Biosciences). The samples were developed using an ECL advance Western blotting kit (Amersham Biosciences) according to the manufacturer's instructions and scanned for the chemiluminescence on a luminoimage analyser LAS-3000mini (Fuji Film).

ATP-binding assay of A9

The reaction of A9 with ATP was examined using 32P-labelled ATPs. The reaction was carried out in 40 μl of phosphorylation buffer (Appendix S1) supplemented with 20 pmol A9 or its derivative and 10 μCi of [γ-32P]-ATP or [α-32P]-ATP. After incubation at 20°C for 60 min, the reactant was applied onto a seminative PAGE and subjected to radioactivity detection.

Nucleotide hydrolysis assay

The ATPase assay for A9 was carried out in 50 μl of NTPase buffer (Appendix S1) supplemented with 20 pmol A9 and 10 μCi [γ-32P]-ATP or [α-32P]-ATP. The GTPase assay of D9 was carried out using 32P-labelled GTP instead of ATP under the same conditions as above. The mixture was incubated at 28°C for appropriate periods. The reaction was terminated by adding 100 μl of ice cold 20 mM EDTA. A 0.5 μl portion of each sample was spotted onto a 3 mM chromatography paper (Whatman) and a polyethyleneimine-cellulose thin layer chromatography (TLC) plate (Merck). The paper chromatograph was developed with a mixed solvent [n-butanol:n-propanol:acetone:80% formic acid:30% trichloroacetic acid = 40:20:25:25:15 (v/v)] supplemented with 0.5 mg ml−1 EDTA and was subjected to radioactivity detection after drying as described previously. As standards, adequate amounts of cold monophosphate and diphosphates were similarly developed and visualized by spraying first with 0.1% FeCl2 in 80% ethanol and then with 1% sulphosalicylic acid in 80% ethanol. The polyethyleneimine-cellulose thin layer chromatograph was developed with 750 mM KH2PO4 (pH 3.65). After drying, the plate was subjected to radioactivity detection. The standard spots corresponding to GTP, GDP, GMP, ATP, ADP, or AMP were identified by UV shadowing.

Cross-linking

UV cross-linking was performed in 50 μl of UV cross-linking buffer (Appendix S1) supplemented with 20 pmol of purified GST-A9 or D9, 2 μCi of [α-32P]-ATP (for A9) or [α-32P]-GTP (for D9), and 4 pmol of cold competitor nucleotide (for D9). The reaction mixture was first incubated at 4°C for 5 min and then irradiated with UV (λ = 254 nm) for 5 min on ice. The proteins were analysed by SDS-PAGE using 9% (A9) and 15% (D9) polyacrylamide gels. After drying, the gels were subjected to radioactivity detection. Chemical cross-linking was performed by using dimethylpimelimidate (DMP; Pierce Chemicals) as a cross-linker. Each 0.1 nmol of the purified GST-free A9 and B9 was incubated with various concentrations (0.1–2.0 mM) of DMP in 50 mM HEPES buffer (pH 7.4) at 25°C for 1 h. Samples were analysed by SDS-PAGE using 9% (A9) and 15% (B9) polyacrylamide gels and Western blotting.

In vitro protein–protein interaction assay

Pull-down assay buffer (0.5 ml) (PDA buffer; Appendix S1) that was supplemented with 0.2 nmol of a GST-fused bait protein and 5 μl of glutathione Sepharose 4B (Amersham Biosciences) was incubated at 4°C for 30 min. Subsequently, the resin was collected by centrifugation and washed twice with 1 ml of PDA buffer. The pellet was resuspended in 0.5 ml of PDA buffer followed by addition of 0.1 nmol of a non-tagged target protein and 2 mM of an appropriate nucleotide (ATP, ADP and ATPγS for the reaction with A9; GTP, GDP and GTPγS for the reaction with D9). After incubating for 1 h at 28°C with gentle agitation by rotating, the resin was transferred to a MicroSpin column (Amersham Biosciences) and washed five times with 300 μl of PDA buffer.The proteins that were associated with the resin were eluted with 80 μl of elution buffer (Appendix S1). Subsequently, 20 μl of 5 × SDS loading buffer was added to the eluate and boiled for 5 min. The boiled protein sample (20 μl) was analysed by SDS-PAGE and Western blotting for the presence of target protein associated with the bait protein.

Two-hybrid system analysis

A two-hybrid system analysis using E. coli as a host (Bacteriomatch Two-Hybrid system Vector Kit; Stratagene) was performed according to the manufacturer's instructions. The bait and target plasmids (Appendix S1) were cotransformed into the host strain supplied by the manufacturer (XL-1 Blue MR derivative); the resultant transformants were cultured in LB medium and their β-galactosidase activities were measured by the standard method using o-nitrophenyl-β-d-galactopyranoside as a substrate (Takano et al., 2003a).

Acknowledgements

We thank Susumu Okamoto for critically reading the manuscript, Yasuo Ohnishi and Sueharu Horinouchi for the information about S. griseus genome sequence, Haruo Ikeda and Jun Ishikawa for the helpful discussions and Yuji Sawada and Tomo-o Watsuji for the technical assistance. This study was supported by the 21st Century COE Program of MEXT, Japan. M.K. was supported by a JSPS fellowship.

Ancillary