The Protein Complex Composed of Nickel-binding SrnQ and DNA Binding Motif-bearing SrnR of Streptomyces griseusRepresses sodF Transcription in the Presence of Nickel*

Nickel-responsive transcriptional repression of sodF, which codes for iron- and zinc-containing superoxide dismutase of Streptomyces griseus, was mediated through an operator (−2 to +15) spanning over the 5′ end (+1) of the transcript. Two open reading frames, SrnR (12,343 Da) and SrnQ (12,486 Da), with overlapping stop-start codons were identified downstream from sodF and found responsible for the repression of sodF. The deduced amino acid sequence of SrnR revealed a DNA binding motif and showed homology to the transcriptional regulators of ArsR family, whereas SrnQ did not show any similarity to any known proteins. When srnRQ DNA was maintained in trans in S. griseuson a multicopy plasmid, sodF transcription was highly repressed by nickel, but neither srnR nor srnQalone showed the effect. Consistently, the sodFtranscription of srnR-interrupted mutant was no longer repressed by nickel, which was complemented only with srnRQDNA. Nickel-dependent binding of SrnR and SrnQ to thesodF operator DNA was observed only when the two proteins were provided together. The maximum protein-DNA interaction was shown when SrnR and SrnQ were present in one-to-one stoichiometric ratio. The two proteins appear to constitute an octamer composed of four subunits of each protein. SrnR directly interacted with SrnQ, and the protein interaction did not require nickel. The conformation of SrnQ was changed upon nickel binding, which was in the ratio of one Ni2+ ion per protein molecule. A model is proposed in which SrnQ of the protein complex senses nickel and subsequently enhances the DNA binding activity of SrnR through the protein-protein interaction.

can be generated from incomplete reduction of dioxygen during respiration. The ROS may cause oxidative damage to DNA, protein, and lipid (1). The Fenton reaction generating hydroxy radicals by reduction and oxidation of Fe ions in the presence of ROS has been proposed as a mechanism for cell damage (2). Major defense system against ROS comprises superoxide dismutase (SOD), catalase, and peroxidase, which concertedly convert the ROS into O 2 and H 2 O (3). The SOD is distinguished into several types according to their active metals (2). The SOD containing either manganese (Mn-SOD) or iron (Fe-SOD) has been found in the cytoplasm of prokaryotic cells. The copperand zinc-containing SOD (CuZn-SOD) has been described in the periplasm of several bacteria (4 -6). In addition, two novel SODs containing nickel (Ni-SOD) or both iron and zinc (FeZn-SOD) have been characterized as cytoplasmic enzymes of Streptomyces griseus and Streptomyces coelicolor (7,8).
The Ni-SOD and FeZn-SOD are found as homotetramers of approximately 13-and 22-kDa subunits, respectively, in S. griseus (7) and S. coelicolor (8). The expression of sodN coding for Ni-SOD of S. coelicolor (9) and S. griseus 2 increased in the presence of nickel, whereas their FeZn-SOD (sodF) expressions were regulated through the repression of sodF transcription in the presence of nickel (10,11). Previously, we found that the nickel-responsive transcriptional repression of sodF of S. griseus was exerted through an inverted-repeat sequence (TTGCAN 7 TGCAA) overlapping the 5Ј end (ϩ1, G) of sodF transcript (11). In addition, nickel-dependent interaction between cell extracts and the sodF operator DNA was confirmed by gel-mobility shift assay (11), suggesting the presence of trans-acting regulatory protein(s) responsible for the repression. To the best of our knowledge, a nickel-responsive transcriptional regulator has not been described in Gram-positive bacteria, whereas a transcriptional regulator, NikR, has been found for the repression of nik operon coding for ATP-dependent nickel transport system of Escherichia coli when nickel is provided in excess (12). An inverted-repeat sequence in the promoter of nik operon was proposed as a potential NikRbinding site (13,14). No sequence homology was found between the inverted-repeat sequence of E. coli nik operon and the operator of S. griseus sodF gene.
In this study, we identified srnRQ coding for two small proteins mediating nickel-responsive transcriptional repression of S. griseus sodF (srn stands for sodF repression by nickel). They are located immediately downstream from sodF. SrnR and SrnQ appear to form an octameric complex composed of four subunits of each protein and then bind to the sodF operator in a nickel-responsive manner. We propose that DNA binding motif-bearing SrnR is a repressor, whereas nickelbinding SrnQ is a co-repressor. A hypothetical model for the * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AF1418663.
¶ To whom correspondence should be addressed: Dept. of Life Science, Sogang University, Mapo, Shinsu  repression of sodF transcription by the SrnR⅐SrnQ complex is presented.
Nucleotide Sequence Accession Number-Nucleotide sequence of 0.86-kb SmaI/SalI DNA (Fig. 1), which includes srnRQ of S. griseus, has been assigned to GenBank TM under the accession number AF1418663.
DNA Mutagenesis and Plasmid Construction-The initiation codon of SrnR and SrnQ were mutated by PCR including overlap extension as described previously (19). The 1.0-kb PstI/SalI DNA fragment containing srnRQ (Fig. 1) was cloned into pBS (Stratagene) to generate pBSRQ and used as a DNA template. The primer 5Ј-TGCACCCTTGGAAT-CAC-3Ј (A of SrnR ATG codon was replaced by T) was used for the disruption of initiation codon of SrnR, whereas the SrnQ initiation codon was mutated using the primer 5Ј-CCTCATAGTCAGGCCT-3Ј (AG was substituted for GA corresponding to the third (G) and first base (A) of the first and second codons of SrnQ, respectively, leaving the SrnR stop codon, originally TGA, as TAG). The PCR products were confirmed for the mutations through the DNA sequence analyses followed by digestion with PstI and SalI and cloning into PstI/SalI sites of pFD666 (20) to generate pFDR Ϫ Q and pFDRQ Ϫ carrying the mutations in srnR and srnQ, respectively. Plasmid pFDRQ, also a derivative of pFD666, contained the same 1.0-kb PstI/SalI DNA including the wildtype srnRQ (Fig. 1).
Plasmid pIJK365 derived from a low copy number plasmid pXE4 (21) has the sodF regulatory DNA from Ϫ365 to ϩ59 (ϩ1, 5Ј end of sodF transcript), which was transcriptionally fused to promoterless xylE DNA (11). pIM5in(ϩ8ϩ9) carries the same sodF DNA as in pIJK365 with the exception that a 5-bp nonspecific DNA was inserted between the left and right halves of the dyad symmetry (11). All plasmids contain transcription-translation stop ⍀ DNA (Sm r /Sp r ) (22) upstream from the sodF DNA to block any fortuitous transcriptional read-through originated from the vector DNA.
Detection of Catechol Dioxygenase Activity-S. griseus containing xylE fusion plasmids were cultured for 48 h on an YEME agar plate. Cells were broken by sonication, and catechol dioxygenase activity was measured with cell extracts as described previously (21).
Detection of SOD Activity, Preparation of SOD, and Western Immunoblot Analysis-Activity staining of SOD in a native polyacrylamide (10%) gel was performed as described previously (23). The FeZn-SOD of S. griseus was purified and used as an immunogen to raise an antiserum from mouse as described previously (7). The preparation of cell extracts, electrophoresis (SDS-PAGE, 15% polyacrylamide), and electroblotting of proteins were also performed according to the methods as described previously (7,24). A blot treated with FeZn-SOD-specific antibody was reacted with a 1/5000 dilution of goat anti-mouse IgG (Pierce) conjugated with horseradish peroxidase, which was visualized using an ECL detection system supplied by Amersham Biosciences (11). Protein was determined by a modified Lowry method using bovine serum albumin (Sigma) as a standard (25). The relative SOD activities and FeZnSOD protein levels between samples were quantified by scanning the gel and blot with Tina 2.0 program of a BIO-Imaging analyzer (Fuji).
Expression of SrnR and SrnQ in E. coli-pBSRQ has an insert of the 1.0-kb PstI/SalI DNA containing srnRQ (Fig. 1) in the same orientation as lac promoter of pBS. pBSR Ϫ Q and pBSRQ Ϫ harbored the same insert DNA as in pBSRQ with the exception of mutations in initiation codon of SrnR and SrnQ, respectively. Plasmids pBSRQ, pBSR Ϫ Q, and pBSRQ Ϫ were transformed into E. coli DH5␣. Expression of SrnR and SrnQ was induced by IPTG at 0.1 mM for 2 h after A 600 of E. coli culture reached 0.6 -0.7. SrnR and SrnQ were separated using Tricine-SDS-PAGE (26). The separating gel of 16.5% T, 6% C was overlaid with a 4% T, 3% C stacking gel (T denotes the total percentage concentration of both acrylamide and bisacrylamide, and C is the percentage concentration of bisacrylamide relative to the total concentration).
Gene Disruption-A partially digested BamHI fragment of 4.2 kb containing sodF and srnRQ (Fig. 7) was cloned into pKC1139 (Apr r ) (27), which has a temperature-sensitive replication origin to generate pKCR. The EcoN1 site within srnR of the plasmid was interrupted through ligation with 2.2-kb transcription-translation stop ⍀ DNA (Km r ) (28) following end flushing with Klenow fragment. The resulting plasmid pKCRKm was introduced into S. griseus, and transformants were selected on YMPD agar plate containing Km and Apr at 28°C. Single cross-over recombinants (Km r and Apr r ) were obtained after incubation of the transformant at 37°C. The spores of single cross-over then were grown in YMPD broth without Apr for 72 h and plated on to the same medium containing Km. Kanamycin-resistant colonies thus obtained were examined for their sensitivity to Apr, and finally, three colonies showing Km r and Apr s , which were regarded as srnR-interrupted mutants, were selected among approximately 50 colonies. All of the mutants were confirmed for their chromosomal arrangements of disruptions by Southern hybridization with probes labeled with [␣-32 P]dCTP using DNA-labeling beads (Amersham Biosciences), and one srnR disruptant, R45, was chosen for further analyses.
RNA Isolation, Quantification, and Northern (RNA) Hybridization Analysis-Total RNA was extracted from S. griseus as described previously (29). RNA quantification, electrophoretic separation of denatured RNA, blot transfer, labeling of strand-specific RNA probe using [␣-32 P]CTP, and hybridization with the probe were performed as described previously (11,30). Transcript levels were quantified by densitometer scan of the resulting autoradiograms with the BIO-Imaging analyzer.
Overexpression and Purification of SrnR and SrnQ Using Fusion to Glutathione S-Transferase (GST)-To clone srnR and srnQ into GST fusion plasmid pGEX-5X-3 (Amersham Biosciences), initiation and stop codons of the genes were modified to have BamHI and XhoI sites, respectively. An N-terminal primer, 5Ј-AACATGCACGGATCCAAT-CACGCGCC-3Ј (mutated sequence underlined unless otherwise noted), with the SrnR start codon changed into BamHI site (in boldface) was used in PCR with a C-terminal primer, 5Ј-GAGGCCTGATCTCGAG-GCGTCGCC-3Ј, containing XhoI site (in boldface) mutated from the SrnR stop codon. For srnQ, an N-terminal primer, 5Ј-GGCGACGCCG-GATCCTCAGGCCTCG-3Ј (BamHI site in boldface), and a C-terminal primer, 5Ј-CGCGCGGCCGTCTCGAGCCGGGTAGATC-3Ј (XhoI site in boldface), were used. pBSRQ was used as a template. The PCR products were digested with BamHI and XhoI and cloned into the expression vector pGEX-5X-3 to generate pGEXR and pGEXQ for the expression of SrnR and SrnQ, respectively. Sequence analyses confirmed the in-frame insertion into the plasmid as well as no other mutation in the amplified DNA.
When E. coli DH5␣ containing either pGEXR or pGEXQ grew up to A 600 between 0.6 and 0.7, IPTG was added at 0.1 mM for overexpression of GST-SrnR and GST-SrnQ proteins. The cells were harvested at 2 h after IPTG induction, disrupted by sonication, and centrifuged at 4°C to obtain cell-free extracts. The GST fusion proteins of ϳ39 kDa were identified by SDS-PAGE (12% polyacrylamide). The fusion proteins were purified from the cell lysate using glutathione-Sepharose 4B (Amersham Biosciences) followed by elution and concentration of native SrnR and SrnQ after the removal of GST by Factor Xa (New England Biolabs) digestion. The purified proteins were assessed for the purity and size by SDS-PAGE (15% polyacrylamide) (data not shown). The purified SrnR and SrnQ were used as immunogens to raise antisera as described previously (7).
Gel-mobility Shift Assay-The DNA fragments to be run for the assay were labeled with [␥-32 P]ATP using T4 polynucleotide kinase (Promega). The DNA probes (ϳ10 4 cpm) were incubated at 25°C with the varying amounts of the purified proteins of SrnR and/or SrnQ in the binding reaction buffer as described previously (11,31). After a 10-min incubation with or without nickel, the reaction mixture was examined using 5% non-denaturing polyacrylamide gel as described previously (11). The results provided in this work are typical ones of at least three to four independent experiments showing virtually the same patterns.
Far-Western Immunoblot Analysis-Interaction between SrnR and SrnQ was analyzed as described previously (32). Native or SDS-heatdenatured (95°C for 5 min in 0.2% SDS) SrnR (2 and 5 pmol) was prepared in TDGT buffer (50 mM Tris-HCl, pH 7.9, 0.15 mM NaCl, 1 mM dithiothreitol, 5% glycerol, and 0.05% Tween 20) with or without nickel followed by spotting onto nitrocellulose membrane. The nickel was used twice as much as the protein concentration. The blots were dried at room temperature for 1 h. After soaking in phosphate-buffered saline (24) containing 5% nonfat milk at room temperature for 30 min, blots were incubated in the same buffer containing native SrnQ (15 nM) for 1 h. After washing with phosphate-buffered saline three times, the membranes were treated with SrnQ-specific mouse antibodies in the same buffer, reacted with goat anti-mouse IgG conjugated with horseradish peroxidase, and detected with an ECL detection system (Amersham Biosciences).
Analytical Gel-filtration Chromatography-The complex of SrnR⅐SrnQ was identified through analytical Sephadex G-100 (Sigma) column (1.2 ϫ 20.0, inner diameter ϫ height (in cm)). The column buffer was 50 mM Tris-HCl, pH 7.5, containing 100 mM KCl (33). SrnR and SrnQ (60 g each) were incubated at 25°C for the protein interaction in the binding reaction buffer, which is the same as that used in the gelmobility shift assay. The incubation continued with or without nickel (20 M) for 10 min, and the samples were subjected to the column chromatography, which was run at room temperature. The A 280 of the eluted fractions was measured, and SrnR and SrnQ in elutes were quantified using Tricine-SDS-PAGE (26).
CD Spectroscopy-Either SrnR or SrnQ (15 M) was incubated at 25°C in 10 mM Hepes, pH 7.6, containing 100 mM NaCl (14) with or without nickel (20 M). The CD spectra of the proteins were obtained in wavelength range between 200 and 250 nm at 0.1-nm intervals with J-720 CD spectropolarimeter (Jasco). Five spectra with a resolution of 1 nm, a scan speed of 50 nm/min, and a response time of 1 s were averaged. The mean residue ellipticity was calculated using the molecular mass of each protein.
Nickel Binding Assay-Nickel binding of SrnQ and SrnR was assayed by atomic absorption spectrophotometry following equilibrium dialysis. 2-ml solution of SrnQ or SrnR (6.4 M) was dialyzed against 0.5-liter buffer (10 mM Hepes, pH 7.6, 100 mM NaCl) (14) containing different amounts of NiCl 2 at 4°C for 48 h. Unbound nickel was then removed by dialysis in ϳ500-fold volume of the 10 mM Hepes, pH 7.6, at 4°C for 6 h. Metal content was determined with an AA-880 mark II atomic absorption (flame) spectrophotometer from Thermo Jarrel Ash. Each reading reported by the instrument was an average of three determinations. Bovine serum albumin was used as a control.

Nickel-dependent Expression of S. griseus FeZn-SOD Was
Severely Repressed by the Multicopy Presence of the 1.0-kb DNA Downstream from sodF-The activity and protein level of FeZn-SOD were found much more repressed by nickel when the 1.0-kb PstI/SalI DNA (Fig. 1) downstream from sodF was maintained in multiple copies in cells (Fig. 2, pFDRQ).
S. griseus harboring pFDRQ containing the sodF downstream DNA was grown on YEME agar plates supplemented with nickel up to 5 M, and the FeZn-SOD activity and the protein level were measured at 48 h after inoculation. Although the FeZn-SOD expression of S. griseus having vector DNA, pFD666, decreased in proportion to nickel concentration (Fig. 2, pFD666), the presence of pFDRQ in repeated analyses always lowered the enzyme expression to a greater extent to display the activity ( Fig. 2A) and the protein level ( Fig. 2B) that are barely detectable at 5 M nickel.
The sodF Downstream DNA Contains Two ORFs, SrnR and SrnQ-A sequence analysis of the sodF downstream DNA revealed two ORFs, which are oriented in the same direction as sodF, and started at 160 and 501 bp downstream from the SodF stop codon, respectively (Fig. 1). The first ORF (SrnR) consisted of 114 deduced amino acids with molecular mass of 12,343 Da, whereas the second ORF (SrnQ) revealed 110 amino acids of 12,486 Da. The initiation codon of SrnQ overlaps the stop codon of SrnR. A comparison of amino acid sequence between residues 12 and 82 of SrnR with data base revealed its homology to the transcriptional regulators of ArsR family (Fig. 3): ϳ61% similarity (33% identity) to HlyU, which up-regulates the transcription of hemolysin gene of Vibrio cholerae (34); 60% similarity (36% identity) to ArsR as a transcriptional repressor of the arsenical resistance genes of Staphylococcus xylosus (35); 59% similarity (36% identity) to the NolR involved in negative control of the expression of nodulation genes of Rhizobium meliloti (36); 57% similarity (40% identity) to MerR acting as an activator-repressor protein involved in mercury resistance of Streptomyces lividans (37); and 52% similarity (42% identity) to a putative ArsR found in genome data base of S. coelicolor M145 at Sanger center (www.sanger.ac.uk/Projects/S_coelicolor/). Around the middle of SrnR was found a putative helixturn-helix motif as depicted in Fig. 3. No protein displaying homology to SrnQ was found. SrnQ has remarkably high content (26%) of arginine.

Both SrnR and SrnQ Are Required for the Repression of FeZn-SOD Expression in the Presence of Nickel-
To determine which ORF(s) of the sodF downstream DNA mediates the enhanced repression of FeZn-SOD in the presence of nickel, SrnR and SrnQ were mutagenized through PCR. As illustrated in pFDR Ϫ Q (Fig. 1), the initiation codon of SrnR was changed into TTG resulting in SrnR Ϫ . The SrnR mutation might have polar effect on the expression of downstream SrnQ, so the 213-bp XmaIII-EcoNI DNA corresponding to the peptide from residues 9 -78 of SrnR was in-frame deleted to generate pFD(⌬R)Q (data not shown). In addition, plasmid pFDRQ Ϫ was constructed to have a substitution of ATA for the initiation codon of SrnQ, but the SrnR stop codon, originally TGA, was still kept as TAG (Fig. 1).
The FeZn-SOD expression of S. griseus harboring either pFDR Ϫ Q or pFDRQ Ϫ in trans was compared with that of the control cells containing pFD666 in the presence of nickel (Fig.  2). No significant difference in FeZn-SOD expression was observed. S. griseus having pFD(⌬R)Q showed the same results as those of the cells harboring pFDR Ϫ Q (data not shown), suggesting no polar mutation for the expression of SrnQ by the SrnR Ϫ mutation of pFDR Ϫ Q. Thus, the results clearly indicate that SrnR and SrnQ are required together for the in trans effect of the downstream DNA in multicopies.
SrnR and SrnQ Are Expressed in E. coli-It was examined whether srnR and srnQ code for the proteins of deduced sizes. The proteins were expressed through IPTG induction from E. coli harboring pBSRQ, pBSR Ϫ Q, and pBSRQ Ϫ . The orientation of srnRQ in these plasmids was the same as that of lac promoter, but their reading frames were shifted from that of the LacZ ␣-polypeptide of the plasmid. The Tricine-SDS-PAGE of the cell extracts of E. coli containing pBSRQ revealed two discrete polypeptides of ϳ12 kDa only after IPTG induction (Fig. 4, lane 4). The slow-moving band was regarded as SrnQ, because it was still detected with pBSR Ϫ Q (Fig. 4, lane 6). The other fast-moving one was concluded to be SrnR because it was observed with pBSRQ Ϫ (Fig. 4, lane 8). The sizes of the expressed polypeptides were also similar to the deduced values of SrnR (12,343 Da) and SrnQ (12,486 Da). We did not believe that E. coli RNA polymerase(s) recognized the promoter for srnR and srnQ, because the polypeptides were not expressed when srnRQ were oriented in opposite direction to the plasmid lac promoter (data not shown).
Multicopy Effect of srnR and srnQ on FeZn-SOD Expression Is Exerted at the Transcriptional Stage through the sodF Operator-It was determined whether the effects of srnR and srnQ in multicopies shown above was exerted at the level of sodF transcription, and if so, whether the sodF operator was involved. The sodF::xylE fusion plasmids, pIJK365 (the wildtype operator) (Fig. 5) and pIMin(ϩ8ϩ9) (the mutated operator) (Fig. 5), were derived from pXE4, so that they can be compatible with pFD666 or its derivative, pFDRQ. The intro-  2. FeZnSOD expression of S. griseus containing srnR and srnQ in trans. S. griseus harboring pFD666, pFDRQ, pFDR Ϫ Q, or pFDRQ Ϫ was grown on YEME agar plates supplemented with nickel (0, 0.1, or 5 M) for 48 h. Cell extracts were prepared through sonication. A, activity staining of SOD was done with 20 g of protein/lane in 10% native polyacrylamide gel. B, Western immunoblot analysis of FeZnSOD was accomplished with 7 g of protein/lane.

FIG. 3. Sequence comparison of S. griseus SrnR with bacterial homologues.
Deduced amino acid sequence from residues 12-82 of SrnR was shown, and the sequence alignment was obtained using the ClustalW, version 1.80, program. Fully conserved residues are indicated with asterisks, whereas similar amino acids are marked with dots. The helix-turn-helix motif is indicated. The size of each protein is denoted by residue number in parenthesis, and the residue locations are shown with numbers on each side of the sequence. duction of pFDRQ into cells, which had harbored pIJK365 (Fig.  5, left panel), reduced ϳ72% XylE activity in the presence of nickel (5 M) compared with that without nickel addition, whereas the vector DNA pFD666 (Fig. 5, left panel) resulted in only 14% decrease of activity in the same comparison. The result suggested that the nickel-responsive repression by srnR and srnQ in multicopies is regulated at the level of sodF transcription. Even without nickel addition, pFDRQ lowered activity (Fig. 5, left panel, Nickel 0), which was interpreted as being due to the effect of overexpressed SrnR and SrnQ interacting with the residual nickel in YEME medium. On the other hand, pIM5in(ϩ8ϩ9) did not show any nickel-responsive repression by the presence of pFDRQ (Fig. 5, right panel). The results indicated that the enhanced repression by srnRQ in multicopies should be exerted through the sodF operator.
The nickel-dependent repression by pFDRQ was further confirmed by analyzing the transcription of chromosomal sodF. S. griseus containing pFDRQ, pFDR Ϫ Q, and pFDRQ Ϫ were grown in YEME broth until exponential phase (A 600 ϭ 0.3), and nickel was added up to 0.1 or 5 M. Cells were harvested at times indicated and analyzed for the sodF transcript (Fig. 6). The transcript level was largely decreased at 0.1 M nickel and barely detectable at 5 M nickel when pFDRQ was contained in trans. The reduction of the transcript level by pFDRQ was also observed without nickel addition, which is consistent with the lower XylE activity of pIJK365 in the presence of pFDRQ (Fig.  5, left panel, Nickel 0). The nickel-responsive repression by pFDR Ϫ Q and pFDRQ Ϫ was not much different from that shown by pFD666 (Fig. 6). In addition, the half-life of sodF transcript of the cells containing pFD666 was ϳ13 min irrespective of the nickel treatment, which was not changed at all by the presence of srnR and srnQ (data not shown). Thus, srnRQ in multicopies highly repressed the sodF transcription in a nickel-dependent way.
Chromosomal Interruption of srnR Resulted in Derepressed Transcription of sodF Even in the Presence of Nickel-The chromosomal copy of srnR was interrupted with transcriptiontranslation stop ⍀ DNA (Km r ) (28) to confirm the in vivo effects of overexpressed SrnR and SrnQ. The interruption at EcoNI site of srnR (Fig. 7A), we believe, also resulted in a lack of srnQ expression because no XylE activity was observed with S. griseus containing srnQ::xylE fusion construct with its 5Ј-DNA limited to the EcoNI site (data not shown), illustrating the dependence of srnQ expression on the regulatory DNA upstream from srnR. The genomic Southern analysis of srnRinterrupted mutant R45 (Fig. 7A) revealed the presence of 3.7-kb BamHI bands hybridized to both srnR (Fig. 7B, left panel) and the ⍀ DNA (Fig. 7B, right panel), confirming the correct arrangement of chromosomal interruption. Mutant R45 showed the FeZn-SOD activities and sodF transcription (Fig. 7, C and D, respectively), which were no longer repressed by nickel. As expected, the mutation phenotype was complemented only when pFDRQ (Fig. 2) was introduced into R45 (data not shown). The results were consistent with the observed effects by overexpressed SrnR and SrnQ.
Nickel-dependent Binding of SrnR and SrnQ to the sodF Operator Was Observed Only When Both Proteins Were Provided-SrnR and SrnQ were overexpressed as GST fusion proteins and purified from E. coli as described under "Experimental Procedures." The purified SrnR and SrnQ showed a molecular mass of ϳ12 kDa that were identified by SDS-PAGE (data not shown). The 113-bp StyI-BalI (Fig. 5) DNA containing either wild-type sequence (Fig. 8A) or the mutated operator (Fig. 8B,  ϩ5 insertion) was used for gel-mobility shift assay. The wildtype DNA was retarded only when SrnR and SrnQ were provided together in the nickel-containing reaction mixture. The retarded band was intensified as nickel concentration increased (Fig. 8A). However, neither SrnR nor SrnQ alone shifted the DNA, even in the presence of nickel (Fig. 8A). On the other hand, the DNA including the mutated operator did not show any retardation (Fig. 8B), being consistent with the XylE activities of pIM5in(ϩ8ϩ9) shown in Fig. 5. Thus, SrnR and SrnQ appear to exert the repressive effect in a nickelresponsive way through binding to the operator. Because SrnR has a DNA binding motif, it is expected that SrnR may directly interact with the operator. However, SrnQ should be present together for the repression.
The gel-mobility shift of the operator DNA was examined with the two proteins in different ratios. No retardation was consistently observed in the absence of one of the two proteins (Fig. 9, lanes 1 and 7). The amount of the binding complex reached the maximum level when SrnR and SrnQ were provided simultaneously at 10 M each (lane 4). The signal inten-sity of the binding complex (C) was ϳ2-fold higher than that of lanes 3 and 5 in which one of the two proteins was maintained at half-concentration (5 M). Thus, the maximum interaction of FIG. 7. Characterization of srnR  disruptant R45. A, chromosomal disruption of srnR by ⍀ transcription-translation stop DNA (Km r ) (28). B, genomic Southern hybridization analysis of R45 and single cross-over (SCO) recombinant for comparison with wild type (WT). The chromosomal DNA was digested with BamHI and probed with PCR-amplified srnR DNA (left panel) or probed with 2.2-kb transcription-translation stop ⍀ DNA (Km r ) (right panel). C, the extracts from wild-type and R45 mutant, which had been grown in the presence of nickel (0, 5, or 20 M), were used for the activity staining of SOD as described in Fig. 2. D, Northern (RNA) hybridization analysis of sodF transcript of wild type and R45. Total RNA was isolated from the cells harvested at 6 h after treatment of culture with nickel (0, 5, or 20 M). The same RNA probe as used in Fig. 6 was used.    (32) were spotted onto nitrocellulose membrane in the absence (Nickel Ϫ) or the presence (Nickel ϩ) of nickel. The nickel concentration was twice as much as the protein concentration. Spots 1 contained 2 pmol of SrnR, whereas spots 2 and 3 had 5 pmol of the proteins. The blots were incubated with native SrnQ (15 nM) followed by reaction with anti-SrnQ antiserum. Native and denatured SrnQ (panels A and B, spots 3, respectively) were used as positive controls, whereas the native and denatured bovine serum albumin (panels A and B, BSA, respectively) were included as negative controls. BSA, bovine serum albumin.
SrnR and SrnQ to the sodF operator appears to require both proteins in 1:1 ratio.
SrnR and SrnQ Directly Interact with Each Other, and the Interaction Does Not Require Nickel-Both native (Fig. 10A, spots 1 and 2) and SDS-heat-denatured SrnR (Fig. 10B, spots 1 and 2) were spotted on nitrocellulose membrane, and the blots were incubated with native SrnQ followed by treatment with anti-SrnQ mouse antiserum, which was then detected using anti-mouse IgG conjugated with horseradish peroxidase. The two proteins in native forms interact with each other by direct binding, even in the absence of nickel (Fig. 10A, NickelϪ, spots  1 and 2). Indifference in signal intensity between the applications of 2 and 5 pmol (spots 1 and 2, respectively) of SrnR onto the membrane was probably attributed to the limiting concentration of SrnQ (15 nM) used in binding. Denatured SrnR did not interact with SrnQ irrespective of nickel treatment (Fig.  10B, spots 1 and 2). The control immunoblots with SrnQ in native and SDS-heat-denatured forms showed positive signals by anti-SrnQ antiserum as expected (Fig. 10, A and B, spots 3, respectively). Bovine serum albumin did not show any reaction (Fig. 10, A and B, BSA), and no cross-reactivity was observed between SrnR and anti-SrnQ antiserum (data not shown). The analyses with SrnQ on the membrane followed by sequential reactions with SrnR, anti-SrnR antiserum, and anti-mouse IgG-horseradish peroxidase displayed the same results (data not shown). From the results, it was evident that the interaction between the two proteins does not require nickel and that the protein-protein interaction can take place before binding to the operator DNA.
SrnR and SrnQ Constitute a Hetero-octameric Complex Composed of Each Protein in 1:1 Ratio-The formation of SrnR⅐SrnQ complex was confirmed through gel-filtration chromatography of Sephadex G-100, and its molecular weight was measured using the mass calibration curve drawn with the standard proteins (Fig. 11A). The chromatographic resolution of either SrnR or SrnQ alone showed a peak, which was estimated around 10 -12 kDa (Fig. 11B, c and d). Nickel treatment to each protein prior to gel loading showed the same results (data not shown), which implied that SrnR or SrnQ by itself, does not form homomultimeric complex. The two proteins were incubated with or without nickel and then subjected to the column chromatography. The elution profiles revealed an identical peak corresponding to ϳ95 kDa regardless of the nickel treatment (Fig. 11B, a and b). The Tricine-SDS-PAGE of the 95-kDa fractions showed the two protein bands with equal intensity, which co-migrated with SrnR and SrnQ (Fig. 11C). Thus, SrnR and SrnQ appear to form an octameric complex composed of four subunits of each protein.
Nickel Binds to SrnQ-The CD spectra of SrnQ were largely changed after nickel treatment (Fig. 12A, left panel). However, the spectral patterns of SrnR did not show any significant difference with respect to nickel (Fig. 12A, right panel). The spectral change of SrnQ by nickel disappeared following incubation of the protein mixture with EDTA (50 mM) for 3 h at room temperature (data not shown). The results suggested that the nickel binding to SrnQ causes the conformational change of the protein.
Atomic absorption spectrophotometry of SrnQ following equilibrium dialysis against nickel showed the metal-binding capacity of the protein that saturates at ϳ0.80 atom of nickel/ molecule with an apparent K d of 0.65 M. Thus, the stoichiometric balance of at least 1:1 is expected for Ni 2ϩ ion bound to SrnQ. Other metals such as Zn 2ϩ and Cd 2ϩ were found in the ratio Ͻ0.04 metal atom/protein molecule, and no binding signal was detected with other metal ions including Ca 2ϩ , Mg 2ϩ , Cu 2ϩ , Mn 2ϩ , and Co 2ϩ (data not shown). In addition, SrnR did not reveal any metal-binding capacity (data not shown). Thus, the nickel specifically binds to SrnQ, which appears to be in a stoichiometric balance of 1:1 (Ni 2ϩ /SrnQ). DISCUSSION A highly aerobic organism, S. griseus contains Ni-SOD and FeZn-SOD for protection against oxidative damages by superoxide radicals. Ni-SOD activity increased after nickel treatment, 2 whereas FeZn-SOD expression is repressed by nickel at the stage of transcription (10,11). The transcriptional repression was exerted through an operator of sodF (11). The antagonistic production of the two SODs has been proposed as a regulatory circuit to keep the total SOD activity constant (10). In this work, we found two ORFs, SrnR and SrnQ, which are responsible for the nickel-responsive transcriptional repression of S. griseus sodF. When SrnR and SrnQ were maintained in multicopies in S. griseus, the Ni-SOD activity was not changed much, indicating that sodN expression is regulated differently from that of sodF.
The calculated pI of SrnQ, fairly rich (26%) in arginine, is 12.8, whereas that of SrnR is 7.0. The physiological implication of the basic pI of SrnQ is not known. The constitutive expression of srnRQ independent of the presence of nickel was identified using srnR::xylE transcriptional fusion construct. 2 This result explains the observation that the binding complex with the sodF operator DNA was still detected in gel-mobility shift assay when nickel was added to the cell extracts prepared from a nickel-deficient culture of S. griseus. 2 The gel-mobility shift band (Fig. 8) was compared with those observed with the cell extracts of S. griseus. The retarded band with the purified proteins exactly co-migrated with one (fastermoving band) of the two binding complexes with the cell extracts (11). The slower-moving band was not detected, even with up to five times more the amount of the purified proteins shown in Fig. 8, 2 so the band might reflect additional binding of other proteins from cell extracts. The probe DNA contains the sodF promoter as well.
The results shown in this work clearly suggest that SrnR interacts with SrnQ to form an octameric complex, which appears to be composed of four subunits of each protein (Fig. 13). The overlapping stop-start codons of SrnR and SrnQ may provide a balanced translation to maintain 1:1 stoichiometry of the two proteins. The protein interaction does not require nickel. Our current data explain that SrnQ of the complex is probably a co-repressor of SrnR, because it can hold nickel at a ratio of one Ni 2ϩ /polypeptide and the ligand binding appears to change the conformation of SrnQ, which may in turn enhance the DNA  (11). Nickel ion was shown as closed circles. R and Q denote SrnR and SrnQ, respectively. binding activity of SrnR through the protein-protein interaction. The complex then binds to the sodF operator for transcriptional repression (Fig. 13). A similar observation has been reported for the interaction between BarA repressor and BarX co-repressor of Streptomyces virginiae (38 -40). They are needed together for binding to the operator site of barB-varS operon, which codes for a putative DNA-binding protein and virginiamycin S-specific transport protein, respectively. BarX was suggested to induce DNA binding ability of BarA by direct interaction.
A nickel-sensing transcriptional repressor NikR of E. coli has been reported as a direct sensor of nickel ion to negatively regulate nikABCDE expression by binding to an operator consisting of two 5Ј-GTATGA-3Ј half-sites related by dyad symmetry and separated by 16 base pairs (14). Roughly, a one-nickel ion was proposed to bind to each NikR subunit of the acting dimeric complex (41). No homology between the primary structures of SrnR and NikR was found.
Taken together, the experimental results presented here provide an interesting example of transcriptional repression by the protein complex composed of the DNA binding motif-bearing repressor and nickel-binding co-repressor. The detailed nature of the protein-protein interaction between the repressor and co-repressor remains to be determined.