As(III) and Sb(III) uptake by GlpF and efflux by ArsB in Escherichia coli.

The toxicity of the metalloids arsenic and antimony is related to uptake, whereas detoxification requires efflux. In this report we show that uptake of the trivalent inorganic forms of arsenic and antimony into cells of Escherichia coli is facilitated by the aquaglyceroporin channel GlpF and that transport of Sb(III) is catalyzed by the ArsB carrier protein; everted membrane vesicles accumulated Sb(III) with energy supplied by NADH oxidation, reflecting efflux from intact cells. Dissipation of either the membrane potential or the pH gradient did not prevent Sb(III) uptake, whereas dissipation of both completely uncoupled the carrier protein, suggesting that transport is coupled to either the electrical or the chemical component of the electrochemical proton gradient. Reciprocally, Sb(III) transport via ArsB dissipated both the pH gradient and the membrane potential. These results strongly indicate that ArsB is an antiporter that catalyzes metalloid-proton exchange. Unexpectedly, As(III) inhibited ArsB-mediated Sb(III) uptake, whereas Sb(III) stimulated ArsB-mediated As(III) transport. We propose that the actual substrate of ArsB is a polymer of (AsO)(n), (SbO)(n), or a co-polymer of the two metalloids.

Arsenic, one of the most prevalent toxic metals in the environment, derives primarily from geochemical origins but also from man-made sources. Consequently, nearly every organism has intrinsic or acquired mechanisms for arsenic detoxification (1). The arsenical resistance operon (arsRDABC) of the conjugative R-factor R773 confers resistance to inorganic As(III) and Sb(III) in Escherichia coli. The arsenic transport system exhibits a dual mode of energy coupling depending on the subunit composition (2). When both ArsA and ArsB are present, they form the As(III)-translocating ArsAB ATPase, which is independent of the electrochemical proton gradient (3). In contrast, in the absence of ArsA, ArsB catalyzes As(III) extrusion coupled to electrochemical energy, which suggests that ArsB is a uniporter that extrudes the arsenite anion in response to the positive exterior membrane potential (4). This dual mode of energy coupling led us to propose that the ArsAB pump evolved by association of a secondary carrier with a soluble ATPase (5).
Over a decade ago, we proposed that other primary ATP-coupled pumps such as ATP-binding cassette transporters evolved in similar ways (5,6).
ArsB is the most widespread determinant of arsenic resistance in bacteria and archaea, yet its transport properties are not well characterized. It is a member of the ion transporter superfamily (7), with 12 membrane-spanning segments and a membrane topology that is similar to many carrier proteins (8).
To date, it has been shown to transport only As(III) (4). Here we report for the first time that ArsB transports inorganic Sb(III) in E. coli, and we describe the relationship with As(III) transport. Considering that the pK a of Sb(III) is 11.8 and As(III) is 9.2, at cytosolic pH the concentration of the oxyanion of either metalloid is negligible. Thus, it is unlikely that ArsB could be an electrophoretic anion uniporter. Instead, we demonstrate here that ArsB is a trivalent metalloid/H ϩ antiporter. ArsBcatalyzed uptake of Sb(III) into everted membrane vesicles coupled to either the pH gradient or membrane potential components of the electrochemical proton gradient. Sb(III)/H ϩ exchange was monitored using a fluorescent reporter, acridine orange, for the pH gradient and oxonal V for the membrane potential. The nature of the substrate of ArsB was explored by competition and co-transport experiments with Sb(III) and As(III). Given their pK a values, the physiologically relevant species of Sb(III) and As(III) in a solution of neutral pH are hydroxides, but the nature of the true substrate of ArsB is not clear. As(III) inhibits ArsB-mediated Sb(III) transport, yet Sb(III) stimulates ArsB-mediated As(III) transport. From these results and the results of As(III) and Sb(III) co-transport experiments, we propose that the true substrate of ArsB is a polymer of As(III) or Sb(III), or a co-polymer of As(III) and Sb(III).

EXPERIMENTAL PROCEDURES
Materials-Restriction enzymes and nucleic acid-modifying enzymes were purchased from Invitrogen and New England Biolabs, Inc. Carrier-free 73 AsO 4 3Ϫ was obtained from Los Alamos National Laboratories. 125 SbCl 5 was produced by PerkinElmer Life Sciences. The pentavalent forms of the isotopes were reduced to 73 As(III) and 125 Sb(III) by the method of Reay and Asher (9). All other chemicals were obtained from commercial sources.
Strains, Plasmids, and Media-E. coli strains and plasmids used in this study are listed in Table I. E. coli strains harboring the indicated plasmids were grown in Luria-Bertani medium (10) at 37°C with 100 g/ml ampicillin, 35 g/ml chloramphenicol, or 50 g/ml kanamycin as required. Protein expression was induced by addition of 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside.
Resistance Assays-For assays of resistance to arsenite and antimonite, cultures were grown overnight at 37°C with shaking. The cells were diluted 100-fold into fresh, prewarmed medium with the indicated concentrations of As(III) in the form of sodium arsenite or Sb(III) in the form of potassium antimonyl tartrate and incubated at 37°C with shaking for an additional 6 h. Growth was estimated from the absorbance at 600 nm.
Transport Assays-For uptake assays in intact cells, cultures were grown to A 600 nm of 1 at 37°C with aeration in Luria-Bertani medium. The cells were harvested, washed, and suspended in a buffer consisting of 75 mM HEPES-KOH, pH 7.5, containing 0.15 M KCl and 1 mM MgSO 4 , brought to a concentration of 80 mg of wet cells/ml at room temperature. To initial the transport assay, 0.1 ml of cells was diluted in 1 ml of the same buffer at room temperature containing 10 M sodium arsenite and 0.4 Ci of 73 As(III) or 10 M potassium antimonyl tartrate and 0.4 Ci of 125 Sb(III). Samples (0.1 ml) were withdrawn at the indicated times, filtered through 0.2-m pore diameter nitrocellulose filters (Whatman), and washed with 15 ml of the same buffer, all at room temperature. The filters were dried and quantified by liquid scintillation counting.
Transport assays using everted membrane vesicles were performed as described (4). Everted membrane vesicles were prepared essentially as described previously (4). Unless otherwise noted, the reaction mixture contained 0.3 mg of membrane protein in a final volume of 0.6 ml of a buffer consisting of 75 mM HEPES-KOH, pH 7.5, containing 0.1 M K 2 SO 4 , 0.25 M sucrose, and 1.25 Ci of either 73 As(III) or 125 Sb(III). Assays were initiated by the addition of 5 mM NADH, final concentration. The concentration of As(III) was adjusted by the addition of buffered sodium arsenite, and buffered potassium antimonyl tartrate was used to adjust the total Sb(III) concentration. Transport assays using inductively coupled mass spectrometry (ICP-MS) 1 were performed by the same procedure with a PerkinElmer ELAN 9000 except that nonradioactive metalloids were used, and the total volume was increased to 1.8 ml. At the indicated times, 0.3-ml samples were withdrawn and filtered through 0.2-m pore size nitrocellulose filters (Whatman). After filtration, the filters were washed with 10 ml of the same buffer and air-dried. For radioactive assays, the radioactivity was quantified by liquid scintillation counting. For ICP-MS measurements, the filters were digested with 0.3 ml of concentrated HNO 3 (69 -70%) (EM Science) overnight at room temperature. The dissolved filters were incubated for 10 min at 70°C, allowed to cool to room temperature, and diluted with 5.25 ml of high pressure liquid chromatography grade water (Sigma) to produce a final concentration of HNO 3 of approximately 4%. Standard solutions were made in the range of 0.5-150 ppb in 4% HNO 3 using arsenic and antimony standards (Ultra Scientific). Initial rates were determined from a linear regression of time points at 0.5, 1, 2, 3, and 4 min. Kinetic data were analyzed by nonlinear regression using Sig-maPlot version6.1. Protein content was determined with a BCA protein assay kit (Pierce), using bovine serum albumin as a standard.
Measurement of pH Gradient Formation by Acridine Orange Fluorescence-Formation of ⌬pH was estimated from the quenching of acridine orange fluorescence (11). The reaction mixture consisted of 20 mM HEPES-KOH, pH7.2, containing 0.1 M KCl, 2.5 mM MgSO 4, 2 M acridine orange, and 0.13 mg of membrane protein in a volume of 2 ml. Quenching was initiated at room temperature by addition of 5 mM NADH (final concentration). Fluorescence was measured in stirred cuvettes with an Aminco AB2 spectrofluorometer, with excitation at 492 nm and emission at 527 nm.
Measurement of Membrane Potential Formation by Oxonol V Fluorescence-Formation of ⌬ was estimated from the quenching of oxonol V fluorescence as described (12). The reaction mixture consisted of 20 mM HEPES-KOH, pH7.2, 2.5 mM MgSO 4 , 5 M oxonol 595 (Aldrich), and 0.13 mg of membrane protein in a volume of 2 ml. Quenching was initiated at room temperature by addition of 5 mM NADH (final concentration). After 5-6 min, potassium tartrate or potassium antimonyl tartrate was added at the indicated concentrations. Fluorescence was measured in stirred cuvettes with an Aminco AB2 spectrofluorometer with excitation at 589 nm and emission at 616 nm.

Uptake of As(III) and Sb(III) in E. coli-We have previously
shown that disruption of the aquaglyceroporin GlpF confers resistance to Sb(III) (13). This disruption was interpreted as a loss of the uptake pathway for Sb(III); however, transport of Sb(III) has never been directly demonstrated in E. coli. Moreover, the GlpF disruption did not gain resistance to As(III), so it was unclear whether GlpF is an uptake pathway for As(III). Uptake of either 73 As(III) or 125 Sb(III) was measured in the E. coli strains AW3110 or AW10, in which the chromosomal arsRBC operon was deleted (4,14), and OSBR1, which was created from AW3110 by inserting TnphoA into glpF (13). Because AW3110 lacks arsB, it was unable to extrude metalloids and accumulated 125 Sb(III) (Fig. 1A) or 73 As(III) (Fig.  1B). Disruption of glpF greatly reduced the level of uptake of both metalloids. This clearly demonstrates that GlpF is the major uptake pathway for both As(III) and Sb(III). Aquaglyceroporins are channels that facilitate the movement of neutral substrates such as glycerol and other polyols but not ions (15). Considering that the pK a value of trivalent arsenic and antimony is 9.2 and 11.8, respectively, there is essentially no arsenite or antimonite anion at physiological pH levels. To be substrates of aquaglyceroporins, the metalloids would be expected to be the neutral hydroxides As(OH) 3 or Sb(OH) 3 , which are the inorganic equivalents of polyols (13,16). However, as discussed in more detail below, other possibilities exist, so, for the purposes of this study they are designated As(III) and Sb(III).
ArsB Confers Sb(III) and As(III) Resistance-Intracellular As(III) and Sb(III) are toxic to most cells, including E. coli (1). E. coli has a chromosomal arsB gene in the three-gene arsRBC operon that confers moderate levels of resistance to these metalloids (14). Plasmids such as R773 have five-gene arsRDABC operons that confer high levels of resistance (17). The R773 and chromosomal ArsBs share 90% identity at the amino acid level, and the R773 arsB gene complemented both Sb(III) ( Fig. 2A) and As(III) (Fig. 2B) hypersensitivity resulting from deletion of the chromosomal arsB gene. Expression of arsA in trans increased resistance. These results demonstrate that ArsB can function either alone or as a complex with ArsA to confer resistance to Sb(III), and importantly, they demonstrate the interchangeability of the chromosomal and plasmid ArsB carriers.
ArsB Catalyzes 125 Sb(III) Uptake in Everted Membrane Vesicles-ArsB has been shown to transport As(III) (4), but Sb(III) transport has not been demonstrated. Uptake into everted membrane vesicles prepared from cells expressing the R773 ArsB is the equivalent of efflux from cells. Everted membrane vesicles from strain AW10 (⌬ars::cam) expressing arsB from plasmid pKMB1 accumulated 125 Sb(III) when NADH was used as a respiratory substrate (Fig. 3). No uptake was observed without NADH or in vesicles from cells with vector only. Addition of the uncoupler carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP) produced a rapid loss of accumulated Sb(III). These results demonstrate that Sb(III) uptake is coupled to the electrochemical proton gradient. Similar results were obtained if 125 Sb(III) was prepared with unlabeled potassium antimonyl tartrate or SbCl 3 (data not shown), showing that only the oxidation state of antimony is relevant. The effects of oxyanions on ArsB-catalyzed 125 Sb(III) uptake into everted membrane vesicles were examined (Fig. 4). Among the oxyanions tested, the sodium or potassium salts of AsO 4 3Ϫ , Sb(OH) 6 Ϫ , PO 4 3Ϫ , NO 3 Ϫ , NO 2 Ϫ , SO 3 2Ϫ , SeO 3 2Ϫ , and BO 3 Ϫ , or tartrate had little effect on 125 Sb(III) transport. Only As(III) (added as sodium arsenite) inhibited transport; therefore ArsB appears to be specific for Sb(III) and As(III).
ArsB Catalyzes Sb(III)/H ϩ Exchange-We previously proposed that ArsB is a uniporter for the arsenite anion, AsO 2 Ϫ , which was thought to be the form of As(III) in solution (4). This idea is consistent with the uncoupling effect of FCCP on NADH-driven As(III) uptake in everted membrane vesicles that shows that ArsB-catalyzed As(III) uptake is driven by the electrochemical proton gradient, which is acid and positive interior in these vesicles (Fig. 3). However, as discussed above, the metalloid substrates of ArsB are not anions at physiological pH levels. An alternative for coupling transport of a neutral solute into the acid and positive interior of the everted vesicles is by exchange with a cation or proton, but dependence on a specific cation such as Na ϩ , K ϩ , Mg 2ϩ , or Ca 2ϩ was not observed (data not shown). To examine the possibility that ArsB is an electrophoretic Sb(III)/H ϩ antiporter, the effects of a permeant anion (SCN Ϫ ) and weak base (NH 4 ϩ ) on Sb(III) uptake into everted membrane vesicles were measured (Fig.  5). In an attempt to limit the use of exotic isotopes, the amount of antimony taken up in this and other assays, as noted, was determined by ICP-MS. When SCN Ϫ and NH 4 ϩ were added together, they uncoupled Sb(III) transport from NADH oxidation as well as did FCCP. However, neither SCN Ϫ , which dissipates the positive interior ⌬ but does not dissipate ⌬pH, nor NH 4 ϩ , which dissipates the acid interior ⌬pH but does not dissipate ⌬, by themselves had any effect on NADH-driven accumulation of Sb(III). These results indi- cate that either ⌬ or ⌬pH alone is sufficient to energize Sb(III) transport. Because dissipation of only ⌬ does not inhibit transport, ArsB cannot be an electrophoretic anion uniporter. Instead, these results are consistent with metalloid-proton exchange.
The effect of Sb(III) on ⌬pH was examined. Solute/proton exchange can be assayed by the effect of the co-transported species on ⌬pH using a fluorescent, weak base (18,19). The fluorescence of the weak base acridine orange is quenched on formation of a pH gradient in everted membrane vesicles (Fig.  6) (11). Addition of 10 mM (NH 4 ) 2 SO 4 completely reversed fluorescence quenching, consistent with dissipation of ⌬pH by a weak base (data not shown). Addition of Sb(III) in the form of potassium antimonyl tartrate reversed quenching in a concentration-dependent manner; potassium tartrate alone had no effect. Membranes from cells lacking an arsB gene showed no fluorescence enhancement, demonstrating that this assay measures ArsB activity.
The effect of Sb(III) on ⌬ was examined. Membrane potential formation in everted membrane vesicles can be visualized from the quenching of the permeant dye oxonol (12). Addition of NADH quenched oxonol fluorescence (Fig. 7). Sb(III) rapidly reversed quenching in a concentration-dependent manner. There was no effect of Sb(III) on oxonol fluorescence in membranes from cells without arsB. Potassium tartrate slowly depolarized the membrane, but the same effect was observed in membranes from cells with or without arsB. Addition of 10 mM KSCN rapidly and com- ArsB Catalyzes Co-transport of As(III) and Sb(III)-Everted membrane vesicles prepared from cells of E. coli strain AW10 expressing ArsB from plasmid pKMB1 exhibited NADH-coupled 73 As(III) uptake compared with cells with vector plasmid pKK223-3 (Fig. 8). Comparing the rates of uptake at a metalloid concentration of 0.1 mM, the rate of 125 Sb(III) uptake (31.3 nmol/min/mg protein) (Fig. 3) was approximately 35-fold higher than that of 73 As(III) (0.9 nmol/min/mg protein) (Fig. 8). When Sb(III) was added together with 73 As(III), the rate of uptake was stimulated 10-fold (9.1 nmol/min/mg protein) (Fig. 8).
In contrast, As(III) inhibited the uptake of 125 Sb(III) (Fig. 4). To examine this apparently paradoxical result in more detail, the rates of As(III) and Sb(III) uptake were determined simul-taneously using ICP-MS (Fig. 9). In this experiment, the concentration of one metalloid was fixed at five different concentrations: 10, 50, 100, 500, and 1000 M. At each of these concentrations of one metalloid, the concentration of the other metalloid was varied between 0 and 1000 M. At every concentration of As(III), Sb(III) stimulated uptake of As(III), and at every concentration of Sb(III), As(III) inhibited uptake of Sb(III). When the two metalloids were present at approximately equal concentrations, the rates of uptake of the two were approximately the same. The most parsimonious explanation for these results is that ArsB catalyzes co-transport of the two metalloids.
However, co-transport does not explain the converse effect of one metalloid on the rate of uptake of the other. If it were simply co-transport, with separate sites on ArsB for As(III) and Sb(III), then As(III) would be expected to stimulate Sb(III) transport as Sb(III) stimulates As(III) transport. If As(III) and Sb(III) were simply alternate substrates for the same site on ArsB, then Sb(III) would be expected to compete with the uptake of As(III) as As(III) competes for Sb(III) uptake. Moreover, the effect of As(III) on Sb(III) uptake does not appear to be simple competitive inhibition. When the concentration dependence of Sb(III) uptake was analyzed as a function of As(III), increasing sigmoidicity was observed (Fig. 10). In the absence of As(III), the data from two separate experiments could be reasonably fitted to the Michaelis-Menten relationship, generating apparent K m and V max values of 43 M and 182 nmol/mg protein/min, respectively. When the data were analyzed using the Hill relationship, the apparent K m value was 44 M, with a V max of 172 nmol/mg protein/min and a Hill coefficient of 1.6 ( Table II). As the concentration of As(III) increased, the apparent K m increased, the V max decreased, and the Hill coefficient in- creased to a value of 2.5 at 1000 M As(III). These results strongly suggest some sort of interaction of As(III) and Sb(III) associated with transport by ArsB. DISCUSSION Reflecting the pervasiveness of environmental arsenic (20), ArsB is a ubiquitous transport protein found in the genomes and plasmids of most bacteria and archaea (1). ArsB is unusual in that it is either a secondary carrier coupled to the electrochemical proton gradient or the translocation subunit of the As(III)/Sb(III)-translocating ArsAB ATPase (2, 3). Based on this novel dual mode of energy coupling, we had proposed that not only the ArsAB pump but other solute-translocating ATPases such as the F 0 Ϫ F 1 and ATP-binding cassette transporters evolved from the association of carriers or channels with soluble ATPases (5,6).
Yet, both the mechanism and substrate of ArsB are probably different from those that were previously conceived (4). Based on a dependence on the electrochemical proton gradient, we had proposed that ArsB is a uniporter that catalyzes electrophoretic efflux of the arsenite anion out of cells in response to the outside positive membrane potential. In this report we demonstrate for the first time translocation of Sb(III) by ArsB. Because the pK a value of inorganic trivalent antimony is 11.8, the concentration of the antimonite anion at a cytosolic pH level of 7.5 (21) is 4 orders of magnitude lower than the total Sb(III) concentration. The paucity of the intracellular antimonite anion would make a uniport mechanism improbable; rather, As(III) and Sb(III) are protonated neutral molecules at cytosolic pH level, and mechanisms that couple efflux of a neutral substrate to the electrochemical proton gradient must be considered. Extrusion of a neutral molecule from cells into the acid and positive exterior could be accomplished by exchange with a cation. Because dependence on an inorganic cation was not observed, protons are a reasonable alternative. This hypothesis was tested in two ways. First, the electrochemical proton gradient was applied as only a membrane potential or only a pH gradient. Either was capable of supporting Sb(III) uptake into everted membrane vesicles. These results are inconsistent with either a uniporter for a neutral solute, which would catalyze only facilitated diffusion, or a uniporter for an anion, which would be able to couple only to the membrane potential and not the pH gradient. Second, Sb(III)/H ϩ exchange can be inferred by the dissipation of ⌬pH concomitant with the addition of Sb(III). That exchange is electrophoretic is shown by the ability of Sb(III) to dissipate ⌬. The most reasonable explanation for these results is that ArsB is a metalloid-proton antiporter. From the reported topological determination of ArsB (8), two glutamate and four aspartate residues can be predicted to be located in transmembrane domains of ArsB, some of which may be involved in H ϩ translocation.
What is the nature of the trivalent metalloid substrate of ArsB? We have recently shown by extended x-ray absorption fine structure spectroscopy (EXAFS) that in solution at a neutral pH the predominant arsenic species is As(OH) 3 , 2 and by analogy the antimony species would be Sb(OH) 3. Indeed, we postulate that the trihydroxides are the forms of the metalloids that are translocated by GlpF and the eukaryotic aquaglyceroporin channels (16).
On the other hand, As(OH) 3 and Sb(OH) 3 are not likely to be the substrates of ArsB because these forms cannot explain the interactions observed between As(III) and Sb(III): 1) As(III) inhibits uptake of Sb(III); 2) Sb(III) stimulates As(III) uptake; 3) the rates of uptake of the two metalloids are approximately equal when the two are present at roughly equivocal concentrations; and 4) the kinetics of Sb(III) uptake become increasingly sigmoidal in the presence of As(III). One possibility is that ArsB oligomerizes, with subunit-subunit interaction; another possibility is that ArsB has separate binding sites for As(III) and Sb(III). Neither possibility easily explains the stimulation of uptake of one substrate by the other and yet reciprocal inhibition of the first by the second.
A third possibility is that the substrate polymerizes analogously to phosphate, pyrophosphate, and polyphosphate. In fact, trivalent As(III) is known to readily form oxo-bridged polymers. The crystal structure of arsenious oxide, As 4 O 6 , is a six-membered (As-O) 3 ring with the fourth As(III) coordinated to the three axial oxygens (22). In addition, a search of the Cambridge Structural Database identifies 109 oxo-bridged As-O-As compounds, nearly all of which are cyclic, including 10 with hexose-like six-membered (AsO) 3 rings. (Note that polymerization of (AsO) n creates even-numbered rings, and six-and eight-membered rings are the most common in the data base, whereas pentose-like five membered rings are not formed.) If these are physiologically relevant forms, why are they not visible in the EXAFS spectra? First, EXAFS is remarkably accurate in determining bond length, but it is quite insensitive to the presence of minor species. If the As(OH) 3 equilibrium with polymeric forms favors the monomeric form, the polymeric forms would probably not be observed. Second, an EXAFS assay was not performed because Sb(III) requires higher energy x-rays than were available, so it is possible that oxobridged Sb(III) species would be observed. Admittedly, this proposition relies on the existence of solution structures that have not yet been identified. However, we have recently found that As(III) is transported by most of the hexose transporters in Saccharomyces cerevisiae and have proposed that the substrate is a hexose-like six-membered (As-O) 3 ring (24). In yeast transport of As(III) via hexose, permeases are inhibited by hexoses, and glucose transport is inhibited by As(III). In contrast, ArsBmediated transport of either As(III) or Sb(III) is not inhibited by a 1000-fold excess of glucose, mannose, galactose, or fructose (data not shown). Thus, ArsB is a specific metalloid carrier and not a sugar carrier.
A proposal for the substrate of ArsB that explains the present results similarly involves six-membered rings composed of As(III), Sb(III), and co-polymers of the two metalloids (Fig. 11). If the equilibrium favors rings containing Sb(III) over those containing As(III), the apparent preference for Sb(III) in ArsB catalysis may in actuality be simply the higher concentration of the (SbO) 3 substrate compared with the (AsO) 3 substrate at the same total amount of metalloid. The apparent stimulation of As(III) by Sb(III) is simply the mass action effect of Sb(III) in formation of the true substrate of ArsB. Similarly, the apparent inhibition of Sb(III) uptake by As(III) is also the result of mass action. The results of the co-transport experiment (Fig. 9) appear to require 1:1 co-transport, but these results could also be explained by formation of the two co-polymer forms (Fig. 11B) that would appear as an average of 1:1 co-transport. The most parsimonious explanation for the apparent cooperativity that As(III) imposes on Sb(III) uptake is that the rate-limiting step in uptake is formation of the substrate. As the concentration of As(III) increases, formation of the preferred (SbO) 3 is impeded. The increase in the Hill coefficient with increasing concentration of As(III) to values approaching 3 is consistent with a progression of six-membered rings from (SbO) 3 to (AsO)(SbO) 2 to (AsO) 2 (SbO) to (AsO) 3 . Thus, rather than interactions of sites within the ArsB carrier, we propose that cooperativity is imparted by interaction of substrates.  11. Model of metalloid transport by GlpF and ArsB. A, in cells of E. coli, As(OH) 3 (or Sb(OH) 3 ) uptake is facilitated by the GlpF channel. The ArsB antiporter exchanges a six-membered, oxo-bridged metalloid ring in exchange for a proton. Exchange of the neutral metalloid with positively charged H ϩ couples efflux to the electrochemical proton gradient. B, the true substrate of ArsB is generated by the equilibrium of three As(OH) 3 with the six-membered (AsO) 3 ring. This oxo-bridged ring is identical to the six-membered (AsO) 3 ring found in arsenious oxide (As 4 O 6 ) (22). Similarly, other substrates of ArsB are proposed to be generated in solution by the equilibrium of three Sb(OH) 3 with a six-membered (SbO) 3 ring, and mixtures of As(OH) 3 and Sb(OH) 3 in various ratios are in equilibrium with the two mixed rings shown.