Interaction of ATP binding sites in the ArsA ATPase, the catalytic subunit of the Ars pump.

The ArsA ATPase is the catalytic subunit of the Ars pump that catalyzes arsenical extrusion in Escherichia coli, thus providing resistance. The active form of ArsA is a homodimer with four nucleotide binding sites, two from each monomer. The codons for Gly-15 in the N-terminal consensus nucleotide binding sequence and Gly-334 in the C-terminal sequence were individually mutated to cysteine codons. Cells expressing an arsAG334C mutation retained arsenite resistance, while an arsAG15C mutation resulted in substantial reductions in arsenite resistance, transport, and ATPase activity. Selection for suppression of the G15C mutation that restored arsenite resistance yielded an A344V substitution. Ala-344 is located adjacent to the C-terminal nucleotide binding sequence. The second site mutation did not suppress the loss of resistance resulting from G18D, G20S, or T22I substitutions in the N-terminal nucleotide binding site. Cells expressing the G15C/A344V double mutant regained arsenite extrusion. These results suggest a spatial proximity of Gly-15 and Ala-344 and support a model for interaction of the nucleotide binding sites in ArsA.

The ArsA ATPase is the catalytic subunit of the Ars pump that catalyzes arsenical extrusion in Escherichia coli, thus providing resistance. The active form of ArsA is a homodimer with four nucleotide binding sites, two from each monomer. The codons for Gly-15 in the Nterminal consensus nucleotide binding sequence and Gly-334 in the C-terminal sequence were individually mutated to cysteine codons. Cells expressing an arsA G334C mutation retained arsenite resistance, while an arsA G15C mutation resulted in substantial reductions in arsenite resistance, transport, and ATPase activity. Selection for suppression of the G15C mutation that restored arsenite resistance yielded an A344V substitution. Ala-344 is located adjacent to the C-terminal nucleotide binding sequence. The second site mutation did not suppress the loss of resistance resulting from G18D, G20S, or T22I substitutions in the N-terminal nucleotide binding site. Cells expressing the G15C/A344V double mutant regained arsenite extrusion. These results suggest a spatial proximity of Gly-15 and Ala-344 and support a model for interaction of the nucleotide binding sites in ArsA.
Biological resistances to heavy metals and antibiotics are frequently conferred by active transport systems (Dey and Rosen, 1995a). The arsenical resistance (ars) operon of Escherichia coli plasmid R773 encodes a transport system for the extrusion of trivalent arsenicals and antimonials out of the cells; lowering of the intracellular concentration of the toxic metalloid salts produces resistance . This ATP-coupled oxyanion pump is composed of two types of subunits. ArsA is the catalytic subunit, with As(III)/Sb(III)-stimulated ATPase activity. The integral membrane ArsB subunit is both the membrane anchor for ArsA and the anion-translocating sector of the pump (Dey et al., 1994).
The ArsA ATPase has homologous N-terminal (A1) and Cterminal (A2) halves, indicating an evolutionary gene duplication and fusion (Chen et al., 1986). The catalytically active form of the soluble enzyme is a homodimer (Kaur and Rosen, 1993), and formation of a three-coordinate complex among three spe-cific cysteine thiolates  and the effector, Sb(III) or As(III), favors dimerization (mei- Hsu et al., 1991;Bhattacharjee et al., 1995). The flow of information from the effector binding site to the Mg 2ϩ ATP sites allosterically activates the ATPase activity (Zhou et al., 1995), and the concomitant hydrolysis of ATP supplies chemical energy for pumping . This allosteric control ensures that the pump does not hydrolyze ATP without coupled anion translocation.
The 63-kDa ArsA ATPase contains two consensus ATPbinding motifs (Walker et al., 1982), one each in the A1 and A2 halves. The glycine-rich sequence of the consensus binding sites for the ␥-phosphate of ATP (P-loop) of ArsA is G 15 KGGVGKTSIS 25 and G 334 KGGVGKTTMA 344 , respectively, in the A1 and A2 halves (Chen et al., 1986). The results from genetic complementation (Kaur and Rosen, 1993) and biochemical reconstitution (Kaur and Rosen, 1994a) suggested that each catalytic unit in the active dimer consists of an A1 domain from one monomer and an A2 domain from the other monomer. To test this idea, we adopted second site suppressor analysis, which has been utilized effectively to study intergenic subunit-subunit (Omote et al., 1994) and intragenic domain-domain (Harris et al., 1991;Iwamoto et al., 1993) interactions. In this approach, if a second site mutation suppresses a defined first mutation, the two altered amino acid residues may be spatially proximate (although it should be emphasized that this is not the only way that intragenic complementation can occur). Based on this hypothesis, Iwamoto et al. (1993) mutagenized the codon for the first glycine residue in the P-loop of the ␤ subunit of the H ϩ -translocating ATPase to a cysteine codon and then isolated revertants of the ␤G149C mutant.
In this study, we used site-directed mutagenesis to alter the first glycine residues in the A1 and A2 P-loops. Gly-149 of the ␤ subunit of the F 1 ATPase corresponds to Gly-15 and Gly-334 of the two P-loops in ArsA. These were altered individually or in combination to cysteine residues. The G334C mutant exhibited a wild type phenotype and was not studied further. However, both the G15C and G15C/G334C mutants exhibited moderate sensitivity to arsenite. An arsenite resistant revertant of the G15C mutation was found to have a single substitution of A344V. The G15C/A344V revertant had wild type levels of arsenite transport in vivo. The single A344V mutant also showed wild type properties. The A344V mutation suppressed a G15C mutation but not G18D, G20S, or T22I mutations, each of which has been shown to prevent binding of ATP to the A1 P-loop (Karkaria et al., 1990). These results suggest spatial proximity of Gly-15 and Ala-344 and support a model in which the two types of ATP binding sites interact to form a catalytic unit.

EXPERIMENTAL PROCEDURES
Materials-All restriction enzymes and nucleic acid modifying enzymes were obtained from Life Technologies, Inc. Oligonucleotides were synthesized in the Macromolecular Core Facility of Wayne State University School of Medicine. 73 AsO 4 3Ϫ was purchased from Los Alamos Laboratories, Los Alamos, NM. All other chemicals were obtained from commercial sources.
DNA Manipulations-The conditions for plasmid isolation, DNA restriction endonuclease analysis, ligation, and transformation have all been described (Sambrook et al., 1989).
Oligonucleotide-directed Mutagenesis-Mutations in the arsA gene were introduced by site-directed mutagenesis using the Altered Sites TM In Vitro Mutagenesis System (Promega). Plasmid pALTER-AB containing the arsA and arsB genes was used as the template . The mutagenic oligonucleotides used and the respective changes (underlined) introduced were as follows: G15C, CGCCTCCTT-TACACGTAAAAAACAG; G334C, CCCACGCCACCTTTACACATCA-GCA TAATCAG. To introduce a HindIII site after the stop codon of the arsA gene, the primer used was CGTGATGACAATAAGC (GC replaced a single T) TTACCCAGC. The identity of the mutations was confirmed by DNA sequencing each mutant gene. Double-stranded plasmid DNA was prepared using the QIAGEN DNA Purification System. Sequencing was performed using the Pharmacia Cy5-labeled autosequence kit (Pharmacia Biotech Inc.) and ALFexpress apparatus by the method of Sanger et al. (1977).
Selection of Revertants-Hydroxylamine was used to mutate plasmid DNA by a modification of the method of Humphreys et al. (1976). Five g of DNA from plasmid pG15C were incubated with 0.4 M NH 2 OH in 2 ml of a buffer consisting of 0.1 M KPO 4 , pH 6.0, containing 1 mM disodium EDTA at 37°C for 10 h. The DNA was diluted 10-fold with a buffer consisting of 10 mM Tris-HCl, pH 8.0, 0.1 M NaCl, and 1 mM EDTA and precipitated with ethanol (Sambrook et al., 1989). Plasmid DNA was transformed into cells of E. coli strain JM109. Plasmid DNAs from single colonies showing resistance to 5 mM NaAsO 2 were sequenced, resulting in the isolation of pG15C/A344V (arsA G15C/A344V B). Other A1 P-loop/A344V double mutants (Table I) were constructed by molecular cloning of the A1 P-loop mutations into pA344V.
Quantitation and Cellular Localization of Altered ArsAs-Cultures of cells containing the appropriate plasmids were grown in LB medium at 37°C to an A 600 nm ϭ 0.8, followed by induction with 0.1 mM IPTG for 2.5 h. One ml of cells was pelleted and suspended in 0.5 ml of SDS sample buffer. After boiling for 5 min, the proteins from 2 l of sample were separated by sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE) using 8% polyacrylamide gels (Laemmli, 1970). Immunoblotting was performed utilizing a chemiluminescence assay (DuPont NEN) and exposed to x-ray film at room temperature as described earlier (Dey et al., 1994).
To determine the cellular location of the expressed proteins, 100 ml of induced cells were pelleted by centrifugation and washed with 100 ml of 10 mM Tris-HCl, pH 7.5, containing 0.1 M KCl. The cells were suspended in 5 ml of a buffer consisting of 10 mM Tris-HCl, pH 7.5, containing 2 mM disodium EDTA, 0.5 mM dithiothreitol, and 20% (v/v) glycerol and lysed by a single passage through a French Press at 20,000 psi, followed by immediate addition of 2.5 l of the serine protease inhibitor diisopropylfluorophosphate per g wet weight of cells. Insoluble protein bodies were pelleted at 10,000 ϫ g for 10 min and suspended in the original volume of the same buffer. Cytosol was produced by removing membranes at 150,000 ϫ g for 1 h. Samples (30 l) were mixed with 10 l of 4 ϫ concentrated SDS sample buffer and boiled for 5 min. One l of each sample was analyzed by Western blotting. 73 AsO 2 Ϫ1 Transport in Cells-73 AsO 2 Ϫ1 was prepared by reduction of 73 AsO 4 3Ϫ (Reay and Asher, 1977). 73 AsO 2 Ϫ1 uptake by whole cells reflects the efflux activity of the Ars pump, where lack of uptake is equivalent to active extrusion (Rosen and Borbolla, 1984;Dey and Rosen, 1995b). Cells bearing plasmids were grown overnight at 37°C in 5 ml of supplemented TEA medium containing 10 g/ml tetracycline. The culture was diluted into 50 ml of prewarmed supplemented TEA medium. At an A 600 nm ϭ 0.6, expression of the arsA and arsB genes was induced by addition of 0.1 mM IPTG for 1 h. The cells were harvested, washed, and suspended in 0.5 ml of TEA medium, all at room temperature. To initiate arsenite transport, 40 l of cells (approximately 1 mg of protein) was diluted into 0.6 ml of TEA medium containing 20 mM glucose, 0.1 mM NaAsO 2 , and 1.25 Ci of 73 AsO 2 Ϫ1 . Portions (0.1 ml) were withdrawn at intervals, filtered through nitrocellulose filters (0.45 m pore diameter; Whatman), and washed with 5 ml of TEA medium. The filters were dried, and radioactivity was quantified by liquid scintillation counting.
Purification and Assay of the ArsA ATPase-Proteins were purified from the cytosol of 2.5 liters of induced cultures as described previously (Hsu and Rosen, 1989). Each ArsA was judged to be Ͼ95% homogeneous by Coomassie Blue staining of samples separated by SDS-PAGE. The concentrations of purified ArsAs were determined using the method of Lowry et al. (1951). ATPase activity was measured using a coupled assay (Vogel and Steinhart, 1976), as described previously (Hsu and Rosen, 1989). This study

RESULTS
Isolation of a Primary Mutant-Mutations in the codons for residues in the A1 and A2 consensus nucleotide binding sequences, G 15 KGGVGKTSIS 25 and G 334 KGGVGKTTMA 344 , respectively, have been previously shown to reduce or eliminate arsenite resistance (Karkaria et al., 1990;Kaur and Rosen, 1992). Substitutions in the A1 P-loop (G18D, G18R, G20S, T22I (Karkaria et al., 1990) and K21E (Kaur and Rosen, 1992)) or A2 P-loop (G337R, K340E, and K340N (Kaur and Rosen, 1992)) each abolished the ability of the Ars pump to extrude arsenite, and the purified altered proteins were each catalytically inactive. From those results it was concluded that Gly-18, Gly-20, and Gly-337 in the two P-loops are required for resistance, transport, and ATPase activity.
The roles of Gly-15 and Gly-334 have not been studied previously. The equivalent glycine residue in the P-loop of the ␤ chain of the E. coli F 1 ATPase was changed to a cysteine with loss of function (Iwamoto et al., 1993). Using site-directed mutagenesis, the codons for Gly-15 and Gly-334 were changed to cysteine codons separately and in combination, producing ArsA derivatives G15C, G334C, and G15C/G334C. Cells harboring the mutated arsA genes and wild type arsB gene were phenotypically characterized for arsenite resistance (Fig. 1). Cells expressing the wild type arsAB genes could tolerate 5 mM NaAsO 2 ; cells containing only vector were unable to grow when the concentration of arsenite was more than 1 mM. Cells expressing an arsA G334C mutation exhibited the same resistance phenotype as the wild type. Cells expressing an arsA G15C mutation or an arsA G15C/G334C double mutation had an intermediate level of resistance.
Properties of the G15C ArsA-The G15C ArsA was purified, and its ATPase activity was determined (Table II). The enzyme retained a small amount of catalytic activity, with about 6% of the V max of the wild type. The defect was clearly manifested in the affinity for ATP, which was reduced by nearly two orders of magnitude. The ability to be allosterically activated by antimonite was comparatively little affected, consistent with the fact that the cysteine residues that form the effector binding site  are unaltered in the G15C enzyme. The fact that the arsA G15C gene conferred partial resistance and that the G15C enzyme retained some catalytic activity suggested that it might be able to be complemented by another mutation. Consequently, the arsA G15C mutation was used for isolation of intragenic suppressors.
Selection of a Second Site Suppressor of the G15C Mutation-Intragenic suppressors were isolated following chemical mutagenesis of plasmid pG15C (arsA G15C B) DNA with hydroxylamine. Cells were selected for growth in medium containing 5 mM sodium arsenite. Only a single isolate was identified as having a single point mutation in the arsA gene; when the entire arsA gene was sequenced the codon for Ala-344, GCT, was found to have been mutated to GTT, creating a valine substitution. Note that residue 344 immediately follows the highly conserved A2 P-loop sequence. The phenotype of cells expressing the arsA G15C/A344V double mutation was identical to that of the wild type ( Fig. 2A). To examine the effect of the A344V substitution in the absence of G15C, an arsA gene containing just the A344V mutation was constructed. Cells expressing the arsA A344V mutation alone exhibited a wild type phenotype.
To address the question of the allelic specificity of the A344V suppression, the effect of the A344V substitution on other A1 P-loop mutants was examined. Three other mutants in the A1 P-loop, G18D, G20S, and T22I, were combined with the A344V mutation. In none was the A344V substitution able to suppress the phenotypic effect of the P-loop mutations (Fig. 2B), indicating the suppression of A344V on the G15C allele was specific.
Expression and Cellular Location of Altered ArsAs-The steady state level of wild type and altered ArsAs in cells was estimated by Western blot analysis using antiserum to wild type ArsA. Each altered protein was produced in approximately the same amount and migrated with the same mobility as wild type ArsA (Fig. 3A). The cellular location of the altered proteins was determined. Although ArsA is functionally a component of the membrane-bound pump, it is found in the cytosol when expressed in high amounts (Rosen et al., 1988). The G15C protein was similarly found predominantly in the cytosol (Fig.  3B). In contrast, the G15C/A344V and A344V proteins were found nearly exclusively as insoluble aggregates. Although the instability of G15C/A344V and A344V enzymes prevented biochemical characterization, from the in vivo results, it is likely that they would have properties more similar to those of the wild type than to the G15C enzyme. 73 AsO 2 Ϫ1 Transport-Cells with an active Ars pump exhibit a low net accumulation of arsenite (Mobley and Rosen, 1982;Dey and Rosen, 1995a). In contrast, cells without the pump are unable to extrude arsenite and exhibit a much higher level of accumulation. Uptake of 73 AsO 2 Ϫ1 was assayed in cells expressing the various mutant arsA genes (Fig. 4). Cells without an ars operon (vector only) accumulated arsenite, while cells expressing a wild type ars operon had a 10-fold lower steady state level of arsenite accumulation. The steady state level of arsenite accumulation in cells expressing the arsA G15C B genes was intermediate, but indicates that the arsA G15C gene product is partially functional in vivo. Cells expressing the ArsA A344V B or ArsA G15C/A344V B genes exhibited the same or lower accumulation compared to wild type. These results correlate well with FIG. 1. Resistance to arsenite in cells expressing wild type, arsA G15C , and arsA G334C genes. Overnight cultures of E. coli strain JM109 bearing wild type and mutant ars plasmids were diluted 100fold into fresh LB medium containing varying concentrations of sodium arsenite. A 600 nm was measured after 8 h of growth at 37°C. Cells had the following plasmids: f, pA334C (arsA A334C B); å, pALTER-AB (arsAB); ࡗ, pG15C/G334C (arsA G15C/G334C B); q, pG15C (arsA G15C ); Ⅺ, vector plasmid pALTER TM -1. the resistance phenotypes of the mutants (Fig. 1) and further support the supposition that ArsA G15C/A344V and ArsA A344V enzymes have substantial ATPase activity in vivo.

DISCUSSION
In a previous study (Kaur and Rosen, 1993), the sequences for the A1 and A2 halves of ArsA were subcloned into compatible plasmids. While neither alone was sufficient to confer resistance, their co-expression restored resistance. Genetic complementation was also observed between arsA genes with point mutations in the two halves of the arsA gene and in cells with plasmids carrying combinations of the arsA1 or arsA2 subclones and point mutations. Complementation was only observed when one plasmid contained a wild type arsA1 sequence and the other a wild type arsA2 sequence. In a subsequent study, subclones of the arsA gene were used to produce the individual A1 and A2 polypeptides (Kaur and Rosen, 1994b). Neither purified polypeptide alone exhibited ATPase activity, nor did simple mixing of the polypeptides restore activity. However, when the two halves of the enzyme were denatured and renatured together, an active ATPase complex could be reconstituted. From the combination of genetic complementation and biochemical reconstitution, a model was proposed in which the ArsA dimer has two catalytic units, each composed of an A1 domain from one monomer and an A2 from the other monomer. However, none of those results provided direct evidence that the A1 and A2 nucleotide binding domains are located near each other in the quaternary structure of the enzyme.
The successful use of second site suppressors to investigate interacting domains in the F 1 F 0 -ATPase (Omote et al., 1994, Iwamoto et al., 1993Harris et al., 1991) suggested that a similar approach might provide useful information on the interactions of the subunits of the Ars ATPase. Of particular relevance is the suppression of an E. coli F 1 ␤G149C mutation by ␤G172D, ␤S174F, ␤E192V, and ␤V198A (Iwamoto et al., 1993), each of which was subsequently shown to be located near the site of binding of the ␥-phosphate of ATP (Abrahams et al., 1994). The suppression of a F 1 ␤S174F mutation by an ␣R296S mutation (Omote et al., 1994) is of particular interest, demonstrating that residues at the interface of the ␣ and ␤ subunits must interact to form a catalytic unit, a premise supported by structural information gained from the solution of the crystal structure of the F 1 -ATPase (Abrahams et al., 1994).
to cysteine residues. The G334C substitution appeared to be neutral, but the G15C substitution resulted in an enzyme that had lost Ͼ90% of its V max while retaining its allosteric regulation. The effect of the mutation in vivo was partial loss of arsenite resistance and transport. The partial phenotypes suggested that the G15C mutation would be a good candidate for suppressor analysis. After several rounds of screening arseniteresistant clones, only a single intragenic suppressor in the codon for Ala-344 was isolated. It is not known whether other resistant clones contained mutations in other genes, nor is it clear why more intragenic suppressors were not isolated. However, the A344V substitution appears to produce specific suppression of the G15C mutation and not other similar mutations in the codons for other residues in the A1 P-loop. At this point, the biochemical basis of suppression is not known. Attempts to isolate the G15C/A344V and A344V proteins were not successful. Although cells expressing the genes for those proteins appear to be fully resistant and have wild type transport properties, the enzymes appear to be unstable. Nearly all of the protein is found as insoluble aggregates, and the small amount of soluble enzyme rapidly loses activity. A single amino acid substitution of Ala 3 Val was similarly reported to have a significant effect on the conformation of H ϩ -translocating ATPase (Takeyama et al., 1990). Unsuccessful attempts were made to decrease the amount of expression or stabilize the altered ArsAs, including lower inducer concentration, shorter induction time, and expression in a lon Ϫ ompT Ϫ strain. Fusions were constructed between the mutant genes and a 5Ј sequence for six histidine codons, but the gene products similarly formed insoluble aggregates. Attempts to reconstitute active ATPases from insoluble aggregates using our previously described methods (Kaur and Rosen, 1994b) were also unsuccessful. It appears that the single Ala 3 Val substitution is sufficiently destabilizing that the half-life of the altered proteins is extremely short. However, cells that continually synthesize the protein have a constant supply of active enzyme and exhibit a wild type phenotype.
The results of previous genetic complementation between mutations in the sequences for the A1 and A2 P-loops suggested that the two domains interact, and that the interaction is between subunits rather than within a single subunit (Kaur and Rosen, 1993). The isolation of a genetic suppressor in the codon for a residue near the A2 P-loop of a mutation in the A1 P-loop is the first direct evidence that the two sites interact. This leads to a model in which the interface of the two subunits form a catalytic unit composed of the A1 site of one subunit and the A2 site of the other (Fig. 5). With information about the regions involved in this interface provided by the present study, more directed mutagenic studies will be designed to test this hypothesis.