Dimerization Is Essential for DNA Binding and Repression by the ArsR Metalloregulatory Protein of Escherichia coli*

Arsenical resistance (ars) operons produce resistance to trivalent and pentavalent salts of the metalloids arsenic and antimony in cells of Escherichia coli. The first gene in the operon, arsR, was previously shown to encode a homodimeric trans-acting metalloregulatory repressor protein. Dimerization of ArsR was investigated using the yeast two-hybrid system in which the ArsR protein was fused to theSaccharomyces cerevisiae GAL4 DNA-binding domain and GAL4 activation domain to produce chimeric proteins. Transcriptional activation of lacZ reporter indicated that dimerization of the ArsR is stable in yeast. The results indicated that residues 1–8 and 90–117 are not required for ArsR dimerization. The genes for a series of truncated ArsR proteins containing six histidine tags were constructed and the proteins purified. The mass of each recombinant protein, as determined by size exclusion chromatography, was consistent with the results from two-hybrid analysis. The results of β-galactosidase assays in vivo and gel mobility shift assays in vitro showed that dimers retained the ability to bind to the ars promoter and to respond to inducer, whereas monomeric ArsRs did neither. These results suggest that a core sequence of about 80 residues has all of the information necessary for dimerization, repression, and metal recognition.

The Escherichia coli chromosomal ars operon confers low level resistance to arsenicals and antimonials (1). The product of the first gene of the operon, the 117-residue ArsR protein, has been shown to be a trans-acting repressor that senses environmental As(III) and Sb(III) (2). This 13-kDa protein belongs to the ArsR family of metalloregulatory proteins that respond to a variety of metals including As(III), Sb(III), Cd(II), and Zn(II) (3). We have postulated that members of the ArsR family of repressor proteins should have at least three domains, a metal binding domain, a DNA-binding domain, and a dimerization domain (4). The ArsR sequence ELC 32 VC 34 DL has been proposed to form a portion of the metal binding domain, and a putative helix-turn-helix DNA-binding motif from residues 38 to 54 has been identified in ArsR (3). The goal of this study was to identify regions of ArsR required for dimerization.
Numerous studies have utilized the yeast two-hybrid system to analyze protein-protein interaction (5)(6)(7)(8)(9). The assay is based on the coexpression of two fusion proteins, each containing a protein or protein region fused to a domain of the GAL4 transcription factor. Through protein-protein interactions, the GAL4 DNA-binding and GAL4 activation domains are brought together to regulate expression of a reporter lacZ gene. The results of deletion analysis of the arsR gene using the yeast two-hybrid system show that the amino-terminal 8 residues and carboxyl-terminal 28 residues are not required for dimerization of ArsR. These results were confirmed in vitro by size exclusion chromatography of each purified truncated ArsR. In addition, only arsR genes that produced proteins capable of dimerization exhibited metalloregulation in vivo and only ArsRs that could dimerize bound to DNA in vitro.

MATERIALS AND METHODS
E. coli Strains, Plasmids, and Media-The bacterial strains and plasmids used in this study are described in Table I. E. coli cells were grown in LB medium at 37°C. Ampicillin (100 g/ml), kanamycin (80 g/ml), tetracycline (15 g/ml), or chloramphenicol (20 g/ml) were added as required. For protein expression 0.2 mM isopropyl-1-thio-␤-Dgalactopyranoside was used as inducer. Sodium arsenite or phenylarsine oxide were added at the indicated concentrations. All chemicals were obtained from commercial sources.
DNA Manipulation-Preparation of plasmid DNA was performed by using a Wizard DNA purification kit (Promega). Endonuclease digestions, DNA fragments separations and isolations, ligations, transformations, and Klenow fragment fill in were performed according to standard procedures (10) unless otherwise noted. The conditions for polymerase chain reaction were as described previously (2). Restriction endonucleases, T4 DNA ligase, Klenow fragment of DNA polymerase I and Taq polymerase were from Life Technologies, Inc. For DNA sequencing, double-stranded DNA was isolated with a plasmid mini kit from QIAGEN and then sequenced by the method of Sanger et al. (11) using an ALFexpress system and a Cy5 labeled sequence kit (Pharmacia Biotech Inc.).
Yeast Two-hybrid Analysis-Cells of yeast strain SFY526 were grown in YPD or the appropriate selective minimal medium. Competent cells were prepared and transformed as described (12). Portions of the transformation mixture were spread on selective plates, after which the plates were incubated at 30°C for 3-4 days. Yeast expression vectors pGBT9 containing the DNA-binding domain of the Saccharomyces cerevisiae GAL4 protein and pGAD424 with the GAL4 activation domain were from MATCHMAKER Two-hybrid System (CLONTECH Laboratories, Inc.). Full-length arsR was cloned in-frame in both pGBT9 and pGAD424 vectors by polymerase chain reaction (PCR) 1 mutagenesis using primers P2 and P13 (Table II) to place an EcoRI site at the 5Ј end and a BamHI site at the 3Ј end of arsR, producing plasmids pGBT9-R and pGAD424-R. Before cloning into pGBT9 and pGAD424 vectors, the PCR product was cloned into pGEM-T vector (Progema), and the absence of random mutations was verified by DNA sequencing. The same methodology was applied for all subsequent PCR cloning. Deletion mutants producing carboxyl-terminal truncations were constructed by PCR cloning EcoRI/BamHI fragments using oligonucleotide P2 as a forward primer for all mutants and the following oligonucleotides as reverse primer: P5 for pGAD424-R⌬62-117, P6 for pGAD424-R⌬81-117, P7 for pGAD424-R⌬84 -117, P8 for pGAD424-R⌬86 -117, P9 for pGAD424-R⌬88 -117, P10 for pGAD424-R⌬90 -117, and P11 for pGAD424-R⌬94 -117. Plasmids pGAD424-R⌬1-40 and pGAD424-* 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.
R⌬1-11 containing amino-terminal deletion mutants were generated by cloning EcoRI/BamHI fragments from PCR using either oligonucleotide P3 (for deletion of 1-11 amino acid residues) or P4 (for deletion of 1-40 amino acid residues) as forward primer and oligonucleotide P13 as reverse primer. In the case of deletion of residues 1-8, an MunI/BamHI DNA fragment from plasmid pGAD424-R was cloned into vector plasmid pGAD424 that had been digested with EcoRI and BamHI, generating plasmid pGAD424-R⌬1-8. Protein-protein interaction was analyzed by in vivo measurement of ␤-galactosidase activity using a filter assay (12). The yeast reporter strain SFY526 was grown on YPD medium or SD synthetic medium containing adenine hemisulfate and all required amino acids. To select for transformants containing pGBT9-derived plasmids, tryptophan was left out of SD medium. To select for transformants containing pGAD424-derived plasmids, leucine was left out of SD medium. Colonies were transferred to sterile Whatman No. 1 filters presoaked in 2 ml of Z buffer (60 mM Na 2 HPO 4 , 40 mM NaH 2 PO 4 , 10 mM KCl, 1 mM MgSO 4 , and 50 mM ␤-mercaptoethanol) (13) containing 0.8 mM 5-bromo-4-chloro-3-indolyl-␤-D-galactoside, permeabilized in liquid nitrogen, and placed onto a sterile Whatman No. 1 filter paper that had been soaked in the same 5-bromo-4-chloro-3-indolyl-␤-D-galactoside containing medium at 30°C. Positive colonies appeared within 10 h. Plasmids pVA3, encoding a murine p53/GAL4 DNA-binding domain hybrid, and pTD1, expressing an SV40 large T-antigen/GAL4 activation domain hybrid, were used as positive controls. The vectors alone (pGBT9 and pGAD424) were used as negative controls.
Expression and Purification of ArsR Proteins-Vector plasmids pET28a and pET28b (Novagen) were used to construct recombinant plasmids for expression of arsR deletions and genes with six histidine codon tags (His6-tag). Plasmid pETR, encoding His6-tagged wild type ArsR, was generated by ligation of an NdeI fragment from the plasmid pCR0 (2) into the NdeI site of vector plasmid pET28a. The orientation of arsR was confirmed by restriction endonuclease digestion. Except for construction of plasmid pETRN9, all arsR deletion mutants were produced by PCR mutagenesis, with several steps of subcloning through vectors pGEM-T and pT7-7 to achieve in-frame fusions. The correct reading frames were verified by DNA sequencing. For deletion genes encoding carboxyl-terminal truncations, the forward primer was the oligonucleotide P1, in which an NdeI site was introduced by altering A to C, and the reverse primers were P12 (for deletion of arsR codons 81-117), P8 (for deletion of codons 86 -117), P10 (for deletion of codons 90 -117), and P11 (for deletion of codons 94 -117). The four final PCR products were cloned into vector pET28b to generate plasmids pETRC80, pETRC85, pETRN89, and pETRC93. Plasmid pETRN41, expressing amino-terminal ArsR truncation of residues 1-40, was constructed by cloning a PCR fragment produced using oligonucleotide P4 as the forward primer and P13 as the reverse primer into vector pET28a. For the deletion of arsR codons 1-8, an MunI-HindIII fragment from plasmid pGAD424-R was cloned into vector pET28b that was digested with EcoRI and HindIII, generating plasmid pETRN9.
Cells of E. coli strain BL21 (DE3) bearing pETR series plasmids were grown overnight in LB medium containing kanamycin at 37°C. The culture was diluted to 100-fold with fresh, prewarmed medium and grown at 37°C. When the culture reached an A 600 of 0.8, expression of arsR genes was induced by addition of 0.2 mM isopropyl-1-thio-␤-Dgalactopyranoside for an additional 3 h. Induced cells were harvested by centrifugation and washed once with buffer A (20 mM Tris-HCl, pH 7.9, 5 mM imidazole, 0.5 M NaCl, and 5 mM ␤-mercaptoethanol). The pelleted cells were suspended in 5 ml of buffer A per g of wet cells and disrupted by a single passage through a French pressure cell at 20,000 p.s.i. Unbroken cells and membranes were removed by centrifugation at 150,000 ϫ g for 1 h. The soluble fraction was loaded onto a 1.2-cm diameter column filled to 8 cm with ProBond resin (Invitrogen) preequilibrated with buffer A. The column was washed with 10 volumes of  Gel Mobility Shift Assay-Gel mobility shift assay was performed as described previously (15). A DNA fragment containing the ars operator/ promoter was produced by PCR. The purified PCR product was digested with EcoRI to produce a 153-base pair fragment. The DNA was labeled with [␣-32 P]ATP and the Klenow fragment of DNA polymerase I and purified using a Wizard DNA clean-up kit (Promega).
Expression of lacZ Reporter Genes-The lacZ reporter plasmid pGBD⌬R2 was constructed to monitor the regulatory properties of arsR genes. An EcoRI-HindIII fragment from plasmid pGBD⌬R1 (16) was filled in with the Klenow fragment of DNA polymerase and ligated into plasmid pACYC184 that had been digested with EcoRI and PvuII, generating plasmid pGBD⌬R2.
Overnight cultures of E. coli strain BL21 (DE3) harboring both pGBD⌬R2 and pETR series plasmids were diluted 50-fold into 3 ml of fresh LB medium containing 80 g/ml kanamycin and 15 g/ml tetracycline. After 2 h of shaking at 37°C, cells were induced with 50 M sodium arsenite and grown for another 2 h. After centrifugation of 1 ml of cell culture, the pellet cells were suspended in 0.5 ml of Z buffer, following which the cells were permeabilized by adding 30 l of 0.1% SDS and 50 l of chloroform, with vortexing for 30 s. The reaction mixture contained 50 l of cell extract, 0.1 ml of 8 mg/ml o-nitrophenyl-␤-D-galactopyranoside, and 0.85 ml of Z buffer. ␤-Galactosidase activity was estimated from the release of nitrophenol at 420 nm at 37°C and was expressed in Miller units (13).

Assay of ArsR Dimerization
Using the Yeast Two-hybrid System-Intersubunit interactions were investigated using the yeast two-hybrid protein-protein interaction assay. When yeast reporter strain SFY526 was cotransformed with pGBT9-R (arsR cloned in-frame and carboxyl-terminal to the S. cerevisiae GAL4 DNA-binding domain) and vector plasmid pGAD424 or with vector plasmid pGBT9 and pGAD424-R (arsR cloned in-frame and carboxyl-terminal to the GAL4 activation domain), the colonies were white. However, when the plasmids pGBT9-R and pGAD424-R were coexpressed, blue colonies expressing lacZ were observed (Fig. 1). These results demonstrate that dimerization of ArsR is detectable with the yeast two-hybrid system.
The products of deletions of arsR from the 5Ј or 3Ј ends were examined for their ability to interact with a wild type arsR gene product (Fig. 1). All of the deletions were fused to the 3Ј end of the sequence for the GAL4 activation domain in the vector plasmid pGAD424. Plasmid pGAD424-R⌬90 -117, with deletion of codons 90 -117, and plasmid pGAD424-R⌬94 -117, with deletion of codons 94 -117, each produced blue colonies. However, deletion of an additional two codons (pGAD424-R⌬88 -117) abolished ␤-galactosidase activity, as did further deletions to codons 85, 83, 80, or 61 ( Fig. 1). Deletion from the 5Ј end of the gene were analyzed. Deletion of codons 1-11 (pGAD424-R⌬1-11) or 1-40 (pGAD424-R⌬1-40) also prevented transcriptional activation of lacZ. In contrast, deletion of only the first eight codons (pGAD424-R⌬1-8) resulted in lacZ expression when cotransformed with pGBT9-R (Fig. 1). These results suggest that amino-terminal residues 1-8 and carboxyl-terminal residues 90 -117 of ArsR are not required for dimerization. Properties of Purified ArsR Proteins-The wild type arsR gene and six deletion mutants were cloned into plasmids pET28a or pET28b (Novagen). The resulting gene products, ArsRH6, RN9H6, RN41H6, and RN89H6, contained the His6 tag at their amino terminus, whereas RC80H6, RC85H6, and RC93H6 contained the His6 tag at their carboxyl terminus. The proteins were produced by expression of the genes in E. coli BL21 (DE3), and each protein was purified to greater than 90% homogeneity by chromatography on Ni-NTA columns. The mass of each protein was determined by size exclusion chromatography. The apparent mass of ArsRH6, RN9H6, RN89H6, or RC93H6 was consistent with the predicted mass of the dimer for each (Table III). In contrast, the elution positions of RN41H6, RC80H6, or RC85H6 were consistent with the mass of a monomer.
In Vivo Regulatory Properties of arsR Deletions-To investigate the relationship between dimerization and function of ArsR, the repressive and metalloregulatory activities of arsR and deletion mutants were assayed in vivo using lacZ gene expression as a reporter. Eight plasmids, pETR, pETRN9, pETRN12, pETRN41, pETRC80, pETRC85, pETRN89, and pETRC93, were individually cotransformed with plasmid pBGD⌬R2, in which transcription of the lacZ gene was under control of the up-stream ars promoter, into E. coli strain BL21 (DE3). ␤-Galactosidase activity was measured in cells harboring both plasmids (Fig. 2). Expression of the gene for His6tagged ArsR repressed lacZ expression in the absence of inducer and induced in the presence of sodium arsenite. The metalloregulatory activities of the genes encoding RN9H6, RN89H6, and RC93H6 were similar to that of the wild type, although basal lacZ expression was somewhat higher. In contrast, lacZ expression in cells with the deletion mutants encoding RN41H6, RC80H6, and RC85H6 was constitutive. Because RN41H6, RC80H6, and RC85H6 are monomers and RN9H6, RN89H6, and RC93H6 are dimers, these results suggest that dimerization is essential for the repressive and metalloregulatory activities of the arsR.
In Vitro DNA Binding Activity of ArsR Repressors-Gel mo- FIG. 1. Yeast two-hybrid analysis of ArsR homodimerization. A gene fusion of the entire arsR gene to the sequence for the GAL4 DNA-binding domain (left) was co-expressed with the indicated sequences from arsR fused to the GAL4 activation domain (right) by co-transformation into yeast SFY526 cell. Expression of the lacZ reporter gene was scored by generation of blue color in filter assays of ␤-galactosidase activity. Each assay was performed in duplicate and repeated two or three times. bility shift assays were used to examine the DNA binding activity of His6-tagged wild type and truncated ArsRs (Fig. 3). Wild type and RN9H6, RN89H6, and RC93H6 each retarded the migration of the promoter DNA. However, none of monomeric truncated proteins RN41H6, RC80H6, or RC85H6 formed a retardation species with ars promoter DNA. An equimolar mixture of RN41H6 and RC80H6 similarly did not retard DNA migration. The affinity of the purified ArsRs for promoter was determined (Fig. 4). From a least-squares fit of the data, the K d of the His6-tagged wild type ArsR was calculated to be approximately 1 M. The affinity of RN9H6, RN89H6, and RC93H6 was each reduced approximately 4 -8fold. These results indicate that removal of the first eight residues from the amino terminus or residues 94 -117 from the carboxyl terminus prevents neither dimerization nor DNA binding but causes some reduction in affinity for the promoter. This decreased affinity for DNA is consistent with the increased basal level of expression observed in vivo (Fig. 2). Metal Binding Properties of ArsR Repressors-ArsR binds arsenite, antimonite, or the organoarsenical phenylarsine oxide through the cysteine thiolates of Cys-32 and Cys-34 (16). Cys-37 is able to participate in metal binding but is not required for induction. The effect of phenylarsine oxide, which is the most effective inducer of the ars operon (2), on the DNA binding properties of his-tagged wild type ArsR and truncated ArsRs RN9H6, RN89H6, and RC93H6 was determined in gel mobility shift assays (Fig. 5). Each protein dissociated completely from promoter DNA with 50 M phenylarsine oxide. Thus, removal of ArsR residues 1-8 or 90 -117 does not prevent metal recognition nor the ability of metal to dissociate the repressor from the promoter.

DISCUSSION
The 117-residue ArsR repressor purifies as a dimer (2), and the dimeric form binds to DNA (15). Is dimerization essential for regulation? We have shown previously that when gene fusions were constructed between arsR and blaM, fusions encoding ArsR-␤-lactamase chimerae retained metalloregulation when the fusion site was at arsR codons 92 or greater (2). In contrast, gene fusions to arsR codons 79 or less lost metalloregulation. However, the oligomeric state of the products of the gene fusions was not determined. In this study truncated ArsR proteins were produced with the objective of investigating the relationship between dimerization and function of the ArsR protein.
Three methods were used to analyze the results of truncations as follows: the yeast two-hybrid system, in vivo regulatory properties in E. coli, and biochemical analyses in vitro. The yeast two-hybrid system has been used to characterize proteinprotein interactions (17). This assay was used to determine regions of ArsR that could be deleted without preventing homodimerization (Fig. 1). A full-length arsR gene was fused to the sequence for the GAL4 DNA-binding domain, and fulllength and partial arsR sequences were fused to the GAL4 activation domain. Full-length ArsR and those with carboxylterminal truncations retaining residues 1-93 or 1-89 were able to interact with a full-length ArsR, whereas the products of deletions that retained the sequences for residues 1-87, 1-85, 1-83, 1-80, and 1-61 did not interact with the full-length ArsR. For amino-terminal truncation mutants, the product of deletion of codons for the first 8 amino acids of ArsR retained the ability to interact with a full-length ArsR. In contrast, removal of the first 11 or 40 residues of ArsR abolished interaction. Thus the dimerization domain of ArsR must be con- tained between residues 9 and 89.
To confirm the results from the two-hybrid analysis, a series of genes encoding His6-tagged full-length and truncated ArsRs were constructed. The gene products were expressed and purified. By size exclusion chromatography, ArsRH6, RN9H6, RN89H6, and RC93H6 proteins migrated as apparent dimers. Three other proteins, RN41H6, RC80H6, RC85H6, migrated as monomers (Table III). These results are consistent with results from the two-hybrid system, confirming that sequences required for dimerization are contained in a core of residues 9 -89.
If dimerization is required for the metalloregulatory activity of ArsR, then there should be a correlation between the aggregation state of the protein and its ability to repress in the absence of inducer and derepress in the presence of inducer. The in vivo regulatory properties of the wild type arsR and deletion mutants were examined using a lacZ reporter gene under control of the ars promoter. There was a clear correlation between the ability of an arsR gene to regulate transcription of the reporter gene in trans and the ability of its gene product to dimerize (Fig. 2). Similarly the in vitro DNA binding properties of the truncated ArsRs correlated with their ability to dimerize. RN9H6, RN89H6, and RC93H6, each of which eluted from gel filtration as a dimer, retained the ability to bind to the ars promoter (Fig. 3) and dissociated from the DNA upon binding of inducer (Fig. 5). In contrast, the monomeric RN41H6, RC80H6, and RC85H6 proteins lost their ability to bind to the ars promoter (Fig. 3). These data are all consistent with dimerization being essential for ArsR function.
The chromosomal ArsR is a member of the ArsR family of metalloregulatory proteins (3,4). Members of this family have variable lengths, with a core of similar residues and variable lengths on the amino and carboxyl termini. The smallest is ArsR of the archebacterium Methanococcus jannaschii, with only 89 residues (18). That homolog aligns with residues 2-90 of the E. coli ArsR, with all of the identities between ArsR residues 9 and 89 (Fig. 6). Similarly, the degree of similarity between the E. coli and plasmid R773 ArsRs, both of which consist of 117 residues, is 75% along its entire sequence but 86% within residues 9 -89 (not shown). This is consistent with the results of the present study that demonstrate that a core sequence of approximately 80 residues is sufficient for all of the regulatory properties of the ArsR repressor: dimerization, DNA binding, and metal recognition.