Binding and Transport of Metal Ions at the Dimer Interface of the Escherichia coli Metal Transporter YiiP*

YiiP is a representative member of the cation diffusion facilitator (CDF) family, a class of ubiquitous metal transporters that play an essential role in metal homeostasis. Recently, a pair of Zn2+/Cd2+-selective binding sites has been localized to two highly conserved aspartyl residues (Asp157), each in a 2-fold-symmetry-related transmembrane segment 5 (TM5) of a YiiP homodimer. Here we report the functional and structural interactions between Asp157 and yet another highly conserved Asp49 in the TM2. Calorimetric binding analysis indicated that Asp49 and Asp157 contribute to a common Cd2+ binding site in each subunit. Copper phenanthroline oxidation of YiiPD49C, YiiPD157C, and YiiPD49C/D157C yielded inter- and intra-subunit cross-links among Cys49 and Cys157, consistent with the spatial proximity of two (Asp49-Asp157) sites at the dimer interface. Hg2+ binding to YiiPD49C or YiiPD49C/D157C also yielded a Cys49-Hg2+-Cys49 biscysteinate complex across the dimer interface, further establishing the interfacial location of a (Asp49-Asp157)2 bimetal binding center. Two bound Cd2+ ions were found transported cooperatively with a sigmoidal dependence on the Cd2+ concentration (n = 1.4). The binding affinity, transport cooperativity, and rate were modestly reduced by either a D49C or D157C mutation, but greatly diminished when all the bidentate aspartate O-ligands in (Asp49-Asp157)2 were replaced by the monodentate cysteine S-ligands. The functional significance of these findings is discussed based on the unique coordination chemistry of aspartyl residues and a model for the translocation pathway of metal ions at the YiiP dimer interface.

The cytosolic zinc pool of a functioning Escherichia coli cell is thought to be Ͻ10 Ϫ12 M in free Zn 2ϩ (1). At such a low cytosolic free Zn 2ϩ level, zinc efflux pumps selectively bind and actively move zinc ions across the membrane using energy. The transport activities of two classes of zinc efflux transporters, namely the P-type ATPase ZntA (2) and proton-linked cation diffusion facilitators (CDF) 3 (3) control the overall efflux of excess Zn 2ϩ in E. coli. To maintain a rapid flow of zinc ions, zinc should bind, change the transporter conformation at the site of binding and subsequently leave the binding site as rapidly as possible. At present, little is known about the molecular architecture of the metal translocation pathway in any CDF protein, even less about the metal coordination chemistry underlying the dual functionalities of selective binding and rapid movement of metal ions in membrane transporters. Among the divalent cation transporters, a well characterized example is the Ca 2ϩ -ATPases (4 -6). Compared with calcium, zinc binds much more strongly, and once bound, zinc protein off-rates are in general five orders of magnitude slower (7). In this light, zinc transporters are anticipated to exploit unusual zinc coordination chemistry, because typical zinc binding in many structural or catalytic sites tends to make zinc ions almost permanent parts of the metalloproteins (8).
CDF is a ubiquitous family of metal transporters found in prokaryotes and eukaryotes (9). The E. coli CDF transporter YiiP has emerged as a prototype for structural and functional studies of atypical coordination chemistry for selective metal binding and transport. Structural analysis of YiiP has established that YiiP is a stable homodimer both in the membrane and in detergent micelles (10). Each monomeric subunit contains six transmembrane segments with both N and C termini located on the cytoplasmic side of the membrane (11). Functional analysis of YiiP and its homolog ZitB has indicated that these two CDF proteins are proton-linked antiporters that utilize the free energy derived from the downhill Hϩ influx to pump the cytosolic Zn 2ϩ out of the cells (12). Recently, an active metal binding site in YiiP has been localized to a highly conserved Asp 157 . Topological analysis has further established that Asp 157 is embedded within the hydrophobic core of TM5 (11). The local hydrophobic environment of Asp 157 suggests that additional negatively charged groups might be needed in collaboration with Asp 157 to form a divalent zinc binding site. A likely candidate could be another highly conserved aspartyl residue at position 49 in TM2. Phenotype analyses of two homologous CDF transporters, CzcD from Ralstonia metallidurans and ZitB from E. coli have demonstrated that mutating Asp 49 renders host cells hypersensitive to zinc, probably because of a loss of zinc efflux pumping activities (13). It is not clear, however, whether Asp 49 is directly involved in Zn 2ϩ binding and transport, and if so, whether Asp 49 and Asp 157 contribute to a common active binding site.
In the present study, we explored how protein tertiary folding and subunit assembly may contribute to the Zn 2ϩ /Cd 2ϩ translocation pathway in YiiP. Toward this end, we first characterized the functional role of Asp 49 by examining the effects of D49A and D49C mutations on metal binding and transport. Then, we established the physical proximity between Asp 49 and Asp 157 by disulfide cross-linking and Hg 2ϩ binding. Our results suggest that the four highly conserved aspartyl residues form a (Asp 49 -Asp 157 ) 2 binding center for binding and cooperative transport of two Cd 2ϩ ions at the dimer interface. Further mutation-function characterization identified two neighboring residues, Asp 45 and His 153 , contributing to the bimetal binding center. Because aspartyl residues are frequently observed in multimetal zinc enzymes in which aspartate carboxylates bridge adjacent metal ions by bidentate oxygen-metal coordination (14), we further examined the bridging role of (Asp 49 -Asp 157 ) 2 . Our experiments led to a model for the translocation pathway of metal ions at the YiiP dimer interface.

EXPERIMENTAL PROCEDURES
Site-directed Mutagenesis-Construction of the expression plasmid pYiiP-His and site-directed mutagenesis were performed as described previously (15). All DNA sequences were verified by sequencing of both strands.
Isothermal Titration Calorimetry (ITC)-Calorimetric titrations were carried out on a Microcal MCS titration calorimeter (Microcal) at 25°C as described previously (15). The dilution heat was measured by injecting a titrant into the degassed HPLC buffer and subtracted from the corresponding titration heat generated by injecting the titrant to an HPLC-purified YiiP or mutant sample. The titration data were fitted to a binding model consisting of either one or two sets of non-interacting binding sites using a nonlinear least-squares algorithm provided by the Microcal Origin software. The binding enthalpy change ⌬H, association constant K a , and the binding stoichiometry n were permitted to float during the least-squares minimization and taken as the best fit values. Protein concentrations were determined using BCA protein assay (Pierce) and amino acid analysis (W. M. Keck Biotechnology Resource, Yale University).
Reconstitution and Stopped-flow Transport Assay-Reconstitution and stopped-flow transport experiments were performed in an assay buffer containing 20 mM HEPES, 50 mM K 2 SO 4 , pH 6.5 as described previously (12). Briefly, HPLC-purified YiiP and mutants were reconstituted with E. coli polar lipids (Avanti Polar Lipids) at a protein/lipid molar ratio of 1:20,000. SDS-PAGE analysis of proteoliposome samples confirmed that approximately the same amount of YiiP or mutants was reconstituted in each experiment. Liposomes were made in parallel to proteoliposomes as a protein-free control sample. Liposomes or proteoliposomes were then loaded with 200 M fluorescence indicator fluozin-1 (Molecular Probes), and 1:1 mixed on a stopped-flow apparatus (KinTek Corp.) with the assay buffer containing CdSO 4 at an indicated concentration. For each data trace, three successive stopped-flow recordings at 525 nm (excited at 490 nm) were collected and averaged, and then normalized to the maximum fluorescence intensity induced by mixing proteoliposomes with 4 mM CdSO 4 in an assay buffer containing 3% n-octyl-␤-D-glucoside. Liposome traces were subtracted from the proteoliposome traces for baseline corrections, and the resultant kinetic trace was fitted to an exponential function using the data analysis software Sig-maPlot 4.0. The initial rate of the exponential rise was taken as the velocity of Cd 2ϩ influx, v i . Least-squares fit of v i as a function of [Cd 2ϩ ] yielded n, K 0.5 , and V max according to Equation 1, where V max is the maximum transport velocity, K 0.5 is the Cd 2ϩ concentration at half-maximal velocity, and n is the cooperativity of Cd 2ϩ transport. Chemical Cross-linking-Disulfide cross-linking was induced by copper phenanthroline (Cu(Phen) 3 ) oxidation (17). A Cu(Phen) 3 working solution (10 mM) was freshly prepared by 3:2:15 (v/v) mixing of 200 mM phenanthroline (dissolved in methanol), 100 mM CuSO 4 , and ddH 2 O. Overexpressing host cells from a 10-ml overnight cell culture were harvested by centrifugation, washed with a reaction buffer (20 mM HEPES, pH 7.5, 100 mM NaCl), and then ruptured by three passages through an ice-chilled microfluidizer (Microfluidics Co. Newton, MA) at 12,000 psi. The resulting membrane vesicles were collected by centrifugation at 140,000 ϫ g for 45 min, then washed, and resuspended in the reaction buffer. Oxidation reactions were initiated by adding Cu(Phen) 3 to a final concentration of 0.5 mM. After 30 min of incubation at room temperature, the reactions were terminated by adding 10 mM EDTA. Vesicles were washed again to remove EDTA, and then exposed to 1 mM NEM to block any remaining reactive cysteine residue. The Cu(Phen) 3 -treated proteins were solubilized using 1% DDM in 2 ml of solubilization buffer (20 mM HEPES, pH 7.5, 100 mM NaCl, and 20% glycerol). After a brief sonication (1 min) and 30 min of incubation on ice, the solubilization mixtures were pelleted again, and the supernatants were either directly subjected to immunoblot detection using a peroxidase-conjugated monoclonal clone His1 antibody (Sigma), or incubated with Ni 2ϩ -NTA resin for 30 min on ice. The affinity resin was washed with solubilization buffer in the presence of 50 mM imidazole and 0.05% DDM, and eluted using the same solution with an elevated imidazole concentration at 500 mM. The resulting purified proteins were analyzed by non-reducing SDS-PAGE on a 15% Tris-HCl polyacrylamide gel and stained with Coomassie Blue.
Protection of MPB Labeling by Metal Binding-Cells harboring the overexpression of His-tagged YiiP D49C , YiiP D157C , or YiiP D49C/D157C were pelleted and gently resuspended in a label-ing buffer (20 mM HEPES, 100 mM NaCl, 2 mM MgCl 2 , 10% sucrose, 0.25 mM TCEP, pH 7.5). The labeling reaction was initiated by adding maleimide polyethylene oxide (PEO) 2 biotin (MPB) to a final concentration of 1 mM to the cell suspensions in the presence of either 0.2 mM EDTA or a metal ion at 0.2 mM as indicated. Cells were incubated with MPB at room temperature for either 1 h or 5 min as indicated, and then 20 mM ␤-mercaptoethanol was added to quench the unreacted MPB. The resulting cells were pelleted again, washed free of MPB, and then MPB-exposed proteins were solubilized using 1% DDM and purified using the Ni 2ϩ -NTA resin as described above. The resultant purified proteins were subjected to SDS-PAGE in duplicate. One gel was stained with Coomassie Blue and the other was transferred to a nitrocellulose membrane using a Trans-blot semi-dry transfer cell (Bio-Rad) and exposed to a peroxidase-conjugated monoclonal anti-biotin clone BN-34 antibody (Sigma). MPB labeling was detected using a SuperSignal West Pico chemiluminescent substrate (Pierce).

Correlation between Cd 2ϩ
Binding and Transport-Asp 49 is a highly conserved aspartyl residue located in the TM2 of YiiP (Fig. 2B). To establish the functional role of Asp 49 in metal binding and transport, we examined the mutation-function correlation using two Asp 49 mutants, YiiP D49C and YiiP D49A . Cysteine and aspartate are Zn 2ϩ /Cd 2ϩ -liganding residues frequently exchangeable in many metalloproteins, whereas alanine is not a Zn 2ϩ /Cd 2ϩ -liganding residue. Cd 2ϩ binding to YiiP, YiiP D49A , and YiiP D49C were characterized directly by ITC as described under "Experimental Procedures." The integrated heat per mole of injected Cd 2ϩ was plotted as a function of the Cd 2ϩ /protein molar ratio (lower panel, Fig. 1A), according to the heat flow in response to a sequence of Cd 2ϩ titrations shown in the upper panel. The titration heat decreased as the Cd 2ϩ concentration increased progressively to saturated concentrations. Qualitatively, the midpoint of the YiiP binding isotherm was at 2.5 stoichiometric units, corresponding to 2.5 Cd 2ϩ binding sites per subunit. The midpoints for YiiP D49A and YiiP D49C titrations were around 1.5 and 2.5 stoichiometric units, respectively, indicating that YiiP D49A lost one stoichiometric site whereas YiiP D49C retained all the 2.5 binding sites. To quantitatively compare Cd 2ϩ binding, binding isotherms of YiiP and YiiP D49C were fitted with a two-site, whereas YiiP D49A with a one-site binding model. As shown in Table 1, the values a Binding parameters taken from a previous study for comparison (11).
of the association constant K a , stoichiometry n and binding enthalpy change ⌬H of site 1 and site 2 for YiiP and YiiP D49C were near identical within experimental errors. Fit of YiiP D49A binding isotherm yielded K a and n values in excellent agreement with those of site 1 of YiiP and YiiP D49C . The binding enthalpy change for YiiP D49A decreased to Ϫ8.5 kcal/mol from Ϫ5.8 kcal/mol for YiiP, indicative of a more favorable enthalpy for Cd 2ϩ binding to site 1. These results indicated that a D49A mutation specifically disrupted Cd 2ϩ binding to site 2, while leaving site 1 intact. We also examined Zn 2ϩ titrations of YiiP, YiiP D49A , and YiiP D49C and qualitatively obtained similar results (data not shown). However, as described in earlier studies, titrations of YiiP with Zn 2ϩ resulted in a biphasic binding isotherm, precluding a definitive deconvolution of the heat components (15).
To establish the functional role of Asp 49 in metal transport, we further examined the correlation between Cd 2ϩ binding to Asp 49 and Cd 2ϩ transport. HPLC-purified YiiP, YiiP D49A , and YiiP D49C were reconstituted into proteoliposomes loaded with a Cd 2ϩ -sensitive fluorescent indicator fluozin-1 for monitoring transmembrane influx of Cd 2ϩ as described under "Experimental Procedures." Upon rapid mixing YiiP proteoliposomes with an assay buffer containing various concentrations of Cd 2ϩ , the fluozin-1 fluorescence response increased progressively as the Cd 2ϩ concentration increased (Fig. 1B). A control experiment with liposomes under identical conditions only showed a background fluorescence response. Thus, the observed proteoliposome responses were derived from Cd 2ϩ influx mediated by the recon-stituted YiiP. As described previously (12), we also observed a linear relationship between the initial rate of fluorescence rise and the amount of reconstituted YiiP within a YiiP/lipid molar ratio up to 15/20,000 (data not shown). This observation indicated that the rate of Cd 2ϩ influx was linearly related to the YiiP transport activity. As shown in Fig. 1B, the overall fluorescence responses of YiiP D49C and YiiP proteoliposomes were comparable both in rate and amplitude within the Cd 2ϩ concentration range examined, while the responses of YiiP D49A proteoliposomes reduced to a background level (upper panel). The initial rate of YiiP transport was dependent of the Cd 2ϩ concentration in a sigmoidal fashion (lower panel), indicative of positive cooperativity. The transport cooperativity n, maximum transport velocity V max and the half-maximal concentration K 0.5 were obtained by least-squares fit of the concentration dependent data to Equation 1, yielding n values of 1.4 Ϯ 0.1 for YiiP and 1.1 Ϯ 0.1 for YiiP D49C . It appeared that the effect of a D49C mutation is anticooperative. Furthermore, the K 0.5 value was increased to 0.58 Ϯ 0.07 mM for YiiP D49C from 0.27 Ϯ 0.02 mM for YiiP, whereas the V max value was marginally decreased to12.1 Ϯ 0.6 s Ϫ1 for YiiP D49C from 14.4 Ϯ 0.4 s Ϫ1 for YiiP. The modest K 0.5 values increase was consistent with ITC binding measurement, showing an insignificant decrease of the association constant K a value of site 2 to 0.8 Ϯ 0.3 M Ϫ1 for YiiP D49C from 1.1 Ϯ 0.5 M Ϫ1 for YiiP. Taken together, ITC binding and stopped-flow kinetic analyses demonstrated the following mutation-function correlation: substitution of Asp 49 with a nonliganding Ala residue disrupted a Cd 2ϩ binding site, and consequently abolished the transport activity, while substitution of Asp 49 with a liganding Cys residue retained overall binding capacity and transport activity. Therefore, our result suggested that Asp 49 is a Cd 2ϩ coordination residue required for both metal binding and transport. The stoichiometry of Cd 2ϩ binding to Asp 49 was estimated according to the stoichiometry difference between Cd 2ϩ binding to YiiP and YiiP D49A , approximately yielding one Asp 49 site for each subunit, or a pair of symmetry-related Asp 49 sites for a YiiP homodimer.
Cd 2ϩ Binding to YiiP D49A/D157A -The localization of site 2 to Asp 49 mirrored an earlier finding that a D157A mutation specifically disrupted site 2 without affecting site 1 (Table 1) (11). These results raised a possibility that both Asp 49 and Asp 157 might contribute to a common binding site (site 2), although Asp 49 and Asp 157 are topologically located in TM2 and TM5, respectively (Fig. 2B). To further establish a functional connection between Asp 49 and Asp 157 , we examined the effect of a D49A/D157A double mutation on Cd 2ϩ binding. Compared with YiiP and YiiP D49A , YiiP D49A/D157A exhibited a binding isotherm characteristic of the YiiP D49A behavior, with a midpoint shifted leftward by one stoichiometric unit from the midpoint of the YiiP binding isotherm (Fig. 2A). Fitting the YiiP D49A/D157A binding isotherm to a one-site model yielded K a ϭ 7.5 Ϯ 0.6 M Ϫ1 , n ϭ 1.4 Ϯ 0.1, in excellent agreement with the binding parameters for site 1 of YiiP, YiiP D49A , and YiiP D157A ( Table 1). The fitted ⌬H value slightly decreased from Ϫ8.5 kcal/mol for YiiP D49A to Ϫ9.9 kcal/mol for YiiP D49A/D157A . Nevertheless, it was evident that the D49A/D157A double mutation disrupted site 2 while retaining the overall binding capacity of site 1. The same disruptive effect on site 2 by D49A and D157A and D49A/D157A mutations suggested that Asp 49 and Asp 157 belong to a common binding site (site 2).
Disulfide Cross-link-The functional connection between Asp 49 and Asp 157 suggested that tertiary folding of the YiiP polypeptide chain might bring together these two aspartyl residues to a common binding site. To establish a possible physical proximity between Asp 49 and Asp 157 , we examined whether copper phenanthroline oxidation could induce Cys 49 -Cys 157 disulfide linkage. The cross-linking experiments were performed on a YiiP D49C/D157C/C287S/C127V quadruple mutant. The two native Cys 287 and Cys 127 were removed to prevent their adventitious contributions to the cross-linking reaction. The YiiP C287S/C127V double mutation caused no detectable change in protein expression, Cd 2ϩ binding to site 2 and transport activity. 4 As shown in Fig. 3, YiiP D49C/D157C/C287S/C127V purified from membrane vesicles without copper phenanthroline exposure only yielded a single protein band on a 15% non-reducing gel (lane 1), corresponding to a monomeric YiiP species as indicated. In contrast, copper phenanthroline exposure yielded three additional protein bands with distinct mobility (lane 2). Adding 5 mM dithiothreitol (DTT) to the copper phenanthroline-treated protein sample prevented all aberrant mobility (lane 3), indicating that the observed band shifts reflected the formation of disulfide cross-links. Thus, the major lower protein species moving in front of the monomeric band was interpreted as a more compact monomeric species caused by an intrasubunit D49C-D157C cross-link as indicated. The two distinct upper protein bands in lane 2 were approximately at positions expected for a dimeric YiiP species (10). To further test the formation of dimeric cross-links, we examined whether copper phenanthroline oxidation could induce D49C-D49C or D157C-D157C intersubunit cross-link using YiiP D49C/C287S/C127V or YiiP D157C/C287S/C127V triple mutants, respectively. Both mutants were shown as a monomeric species on the gel without copper phenanthroline treatment (lanes 4 and 7), yet an additional dimeric species appeared after copper phenanthroline oxidation (lanes 5 and 8). The cross-linked dimers could be fully reduced to the monomeric form by DTT treatment (lanes 6 and 9). As expected, copper phenanthroline oxidation did not cause fast-moving protein species in front of the monomeric protein ( lanes 5 and 8), because a single cysteine residue in either triple mutant was not able to form an intrasubunit cross-link. Interestingly, the dimeric (YiiP D49C/C287S/C127V ) 2 in lane 5 moved slightly faster than (YiiP D157C/C287S/C127V ) 2 in lane 8. Thus, the observed mobility difference between these two dimeric species accounted for the two distinct upper (YiiP D49C/D157C/C287S/C127V ) 2 bands shown in lane 2. As a control experiment, copper phenanthroline treatment of YiiP C287S/C127V under the identical condition did not yield any cross-linked protein species (lane 11).
A relatively high concentration of copper phenanthroline (0.5 mM) was used in this study because membrane vesicles in the reaction mixture absorbed a significant amount of copper ions. The actual Cu(Phen) 3 concentration in the reaction mixture was found to be only 0.043 mM as estimated by a fluorescence metal ion indicator Phen Green-SK (Molecular Probes) according to a standard curve generated using a set of known Cu(Phen) 3 concentrations. We also performed cross-linking reactions using 0.05 mM copper phenanthroline with ϳ10-fold less membrane vesicles. Under this condition, the cross-linked products were only detectable by immunoblot detection using an antibody against the polyhistidine tag. Nevertheless, this less stringent oxidation condition yielded essentially the same results except that the dimeric (YiiP D49C ) 2   could not be distinctly resolved after proteins were transferred to the nitrocellulose membrane. To further validate the crosslinking result, we tested a collection of nine single cysteine variants. As shown in Fig. 3B, D45C, D49C, H153C, D157C, and Cys 287 were cross-linked by 0.5 mM copper phenanthroline while Cys 127 , S144C, D179C, and E200C were not able to be cross-linked under the same oxidation condition. Among the four non-reactive positions, Cys 127 is located in the middle of TM4, S144C is located in the connecting loop between TM4 and TM5, D179C is at a highly conserved position in the connecting loop between TM5 and TM6, and E200C is in the C-ter-minal hydrophilic domain. This set of positions provided negative controls for the five reactive positions. Among them, the observation of D49C-D49C and D157C-D157C intersubunit disulfide linkages suggested that both aspartyl residues are located at the dimer interface. The observation of D49C-D157C intrasubunit disulfide linkage, as shown in Fig. 3A, further indicated physical proximity between the Asp 49 and Asp 157 . Therefore, the four highly conserved aspartyl residues, two from each subunit, may be located within the disulfide bonding distance at the dimer interface. D45C and H153C were cross-linked, probably because of their closeness to Asp 49 and Asp 157 , respectively. Assuming that both TM2 and TM5 adopt a ␣-helical configuration, Asp 45 is located one helical turn from Asp 49 toward the outer leaflet, while His 153 one helical turn from Asp 157 toward the inner leaflet of the cell membrane. We also performed cross-linking reactions using 0.5 mM 1,1-methanediyl bismethanethiosulfonate (18) instead of 0.5 mM copper phenanthroline under otherwise identical conditions. Dimeric cross-links were observed for D45C and H153C single cysteine mutants, consistent with the copper phenanthroline results. In contrast, we did not observed dimeric D49C-D49C and D157C-D157C cross-links, suggesting that D49C and D157C are not accessible to the cross-linker whereas D45C and H153C are located at more accessible regions. Cys 287 is located in the C-terminal domain and the observation of the Cys 287 -Cys 287 intersubunit cross-link may reflect their physical proximity.
A D49C Mutation Generates a Biscysteinate Hg 2ϩ Binding Site at the Dimer Interface-A pair of cysteine residues may form a biscysteinate Hg 2ϩ -binding site if they are positioned within a disulfide bonding distance. To further establish the interfacial location of Asp 49 , we examined whether a D49C mutation generated a biscysteinate Hg 2ϩ binding site across the dimer interface. As shown in Fig. 4, Hg 2ϩ titrations of YiiP C287S yielded a binding isotherm with a stoichiometry of 1.0 Ϯ 0.1 ( Table 2). A D49C/C287S mutation caused a rightward shift of the binding isotherm, which could be described by a two-site binding model. The fitted parameters for the first site agreed well with that of YiiP C287S . The second site had a stoichiometry of 0.6 Ϯ 0.1 (Table 2), corresponding to an Hg 2ϩ binding site shared by two subunits in a YiiP homodimer. Thus, the D49C mutation generated a biscysteinate Hg 2ϩ -binding site at the dimer interface. In contrast, titrations of YiiP D49C/D157C/C287S exhibited a binding isotherm similar to that of YiiP D49C/C287S (Fig. 4), indicating that a D157C mutation did not generate additional Hg 2ϩ binding sites. This observation was consistent with our previous finding that Cys 157 is not accessible to Hg 2ϩ binding (11).  Table 2. Note that the arrow indicates rightward shifts of the YiiP D49C/C287S and YiiP D49C/D157C/C287S binding isotherms.
Probing the Local Environment of the (Asp 49 -Asp 157 ) 2 Bimetal Binding Center-To explore the local environment of the (Asp 49 -Asp 157 ) 2 cluster, we introduced D 3 C mutations and examined metal protection of thiol-specific labeling of MPB to the reporter cysteinyl residues. There are two native cysteines in YiiP, at position 127 and 287, respectively. Earlier studies have established that neither cysteine residue can be labeled from the extracellular side by the membrane impermeant MPB when the hydrophobic barrier of the plasma membrane is intact (11). Thus, the protection of different divalent metal ions was examined in intact cells with a 5-min exposure to MPB in the presence of either EDTA or one of the following divalent cations: Zn 2ϩ , Cd 2ϩ , Mg 2ϩ , Ca 2ϩ , Mn 2ϩ , Co 2ϩ , Ni 2ϩ , and Fe 2ϩ (Fig. 5A). Immunoblot detection of MPB labeling showed that D49C could be readily labeled in the presence of EDTA. Among the divalent cations tested, only Zn 2ϩ , Cd 2ϩ , and Fe 2ϩ reduced immunoblot signals to the background level. Coomassie Blue staining of a duplicated gel showed that the YiiP D49C loading on the gel was approximately the same for each lane, thus the observed immunoblot signal difference was due to the difference in MPB labeling. The selective protection of D49C by divalent cations suggested that Cys 49 is a selective binding residue for Zn 2ϩ , Cd 2ϩ , and Fe 2ϩ , consistent with the in vivo function of YiiP as a Zn 2ϩ and probably a Fe 2ϩ efflux pump (19). Like-wise, MPB labeling was performed with YiiP D157C , but a prolonged MPB incubation (1 h) was required (lane 3, Fig. 5B). The rate of MPB labeling to Cys 157 appeared to be 5-10-fold slower compared with that of Cys 49 (lane 1). This rate difference reflected the difference in solvent accessibility between Cys 49 and Cys 157 . We speculated that Asp 49 is located in a more openly accessible environment, probably in the entrance to the (Asp 49 -Asp 157 ) 2 cluster whereas Asp 157 is buried in an inner hydrophobic crevice. To test this hypothesis, we further examined Hg 2ϩ protection of MPB labeling to Cys 49 and Cys 157 . As shown in Fig. 5B, adding Hg 2ϩ from the extracellular side protected Cys 49 from MPB labeling (lane 2), but had no protective effect on Cys 157 (lane 4). The positional dependence of Hg 2ϩ protection was consistent with the Hg 2ϩ binding analysis, showing that a D49C mutation generated a Cys 49 ⅐Hg 2ϩ ⅐Cys 49 complex at the dimer interface while an additional D157C mutation did not yield additional Hg 2ϩ binding (Fig. 4). Interestingly, although Hg 2ϩ did not protect Cys 157 s, Hg 2ϩ binding to Cys 49 s protected Cys 157 s from MPB labeling (lane 6). This result further suggested that Hg 2ϩ binding to Cys 49 s may block the entrance to the inner Cys 157 s as illustrated in Fig. 5C.
Additional Coordination Residues Contribute to the (Asp 49 -Asp 157 ) 2 Bimetal Binding Center-Asp 49 and Asp 157 are the two most conserved residues in the CDF protein family (9). Their functional roles in metal binding and transport suggest that the (Asp 49 -Asp 157 ) 2 cluster is located within the translocation pathway. In the entire transmembrane spanning domain, only three residues, Asp 45 , Cys 127 , and His 153 can meet the topological and chemical expectations for a role as a metal coordination residue. Cys 127 is a variable residue and a C127V mutation caused no detectable change in Cd 2ϩ transport and binding to the (Asp 49 -Asp 157 ) 2 binding site. 4 Both Asp 45 and His 153 are somewhat conserved, each located one helical turn above or below the (Asp 49 -Asp 157 ) 2 binding center as depicted in Fig. 2B. Cd 2ϩ titrations of YiiP D45C/C287S or YiiP H153C/C287S revealed a binding isotherm that could be fitted to a two-site model (Fig.  6). Compared with YiiP, the D45C/C287S mutation caused little or no change to both binding sites while the H153C/C287S mutation reduced the K a value of site 2 by 7-fold without significantly affecting site 1 (Table 1). Interestingly, both D45C/ C287S and H153C/C287S mutations conferred NEM sensitivity to site 2. As shown in the top and middle panels of Fig. 6, incubating protein samples in 5 mM NEM for 30 min caused a loss of one binding component for both YiiP H153C/C287S and YiiP H153C/C287S . The fitted parameters for the remaining sites were consistent with that of site 1, indicating that NEM selectively disrupted site 2 (Table 1). Control experiments with YiiP or YiiP C287S showed no detectable NEM effect on the respective Cd 2ϩ binding isotherm, thereby localizing the NEM-sensitive residue to Cys 45 or Cys 153 . The effects of D45C and H153C mutations on Cd 2ϩ transport were examined by stopped-flow analysis of Cd 2ϩ influx into reconstituted proteoliposomes. As shown in the bottom panel of Fig. 6, both mutations minimally affected the transport kinetics. Fits of the Cd 2ϩ -concentration-dependent data yielded K 0. 5   reduced Cd 2ϩ influx to the background level. Taken together, our data suggested both Asp 45 and His 153 contribute to binding and transport of Cd 2ϩ .
Aspartyl Residues Bridge Two Metal Binding Sites at the Dimer Interface-Aspartyl residues are frequently found in metal binding clusters where they bridge a pair of metal ions by providing two oxygen ligands, each interacting with one of the two adjacent metal ions (14). If certain carboxylates in (Asp 49 -Asp 157 ) 2 act as bridging residues, replacing bidentate carboxylate oxygens with monodentate thiolate sulfurs is expected to disconnect the bridging interaction, leading to the disruption of the bimetal binding. As noted above, a D49C mutation reduced the cooperativity of YiiP transport from 1.4 to 1.1, with only marginal effect on V max and K 0.5 . Likewise, a D157C mutation also reduced the cooperativity of YiiP transport to 0.99 Ϯ 0.08 with modest effects on V max (13.5 Ϯ 0.9 s Ϫ1 ) and K 0.5 (0.50 Ϯ 0.08 mM) (data not shown). In sharp contrast, a D49C/D157C double mutation was found to selectively disrupt the bimetal binding center (site 2). As shown in Fig. 7A, Cd 2ϩ titrations of YiiP D49C/D157C yielded a binding isotherm shifted leftward by one stoichiometric unit as compared with those of YiiP and YiiP D49C . Fitting the YiiP D49C/D157C binding isotherm with onesite model yielded K a ϭ 1.4 Ϯ 0.2 M Ϫ1 , n ϭ 1.3 Ϯ 0.1, and ⌬H ϭ Ϫ5.5 Ϯ 0.1 kcal/mol, in agreement with the site 1 binding parameters of YiiP, YiiP D49C , and YiiP D157C (Table 1). Corresponding to the loss of site 2 in YiiP D49C/D157C , the fluozin-1 responses of YiiP D49C/D157C proteoliposomes were greatly diminished (Fig. 4B). The residual transport activity was apparently correlated with a weak Cd 2ϩ binding component charac-terized by a high titration heat plateau as the titration proceeded toward saturation (upper panel, Fig. 7A). Fitting the YiiP D49C/D157C concentration-dependent data to Equation 1 revealed that the double cysteine mutation converted Cd 2ϩ transport into a negative cooperative process (n ϭ 0.73 Ϯ 0.25) with significant alternations of both V max (2.7 Ϯ 1.8 s Ϫ1 ) and K 0.5 (3.53 Ϯ 6.28 mM). Because of a greatly reduced site 2 binding affinity in YiiP D49C/D157C , the maximum experimental Cd 2ϩ concentration (4 mM) was not sufficiently high to elicit a quasistatic-state response. The fitted parameters for YiiP D49C/D157C transport were of less certainty. Nevertheless, it was evident that a D49C/D157C double mutation caused severe impairments of the site 2 binding capacity, and consequently metal transport. Considering that cysteine substitution of either Asp 49 or Asp 157 only marginally compromised Cd 2ϩ binding, it appeared that two of the four aspartates in (Asp 49 -Asp 157 ) 2 may play bridging roles in bimetal binding and transport at the dimer interface.

DISCUSSION
The experiments described herein establish a strong correlation between YiiP transport and metal binding to two highly conserved aspartyl residues (Asp 49 and Asp 157 ). Direct metal binding and indirect metal protection analyses both demonstrated a Cd 2ϩ -Asp 49 interaction characteristic of general coordination chemistry of Cd 2ϩ binding. In sharp contrast to the disruptive effect of a D49A mutation on Cd 2ϩ binding and transport, a D49C mutation resulted in a functional transporter  Table 1. Bottom panel, kinetics of Cd 2ϩ transport. YiiP D45C/C287S was reconstituted at a molar ratio of 1:20,000 (protein/lipid). Stopped-flow traces were elicited by 1:1 mixing of proteoliposomes with a transport assay buffer containing 0, 0.125, 0.25, 0.5, 1, 2, or 4 mM CdSO 4 . The initial rate of Cd 2ϩ influx v i was calculated from the kinetic trace, and plotted as a function of [Cd 2ϩ ]. The solid line represents the best fit of the Cd 2ϩ concentration dependent data to Equation 1 as described under "Experimental Procedures." B, functional characterization of YiiP H153C/C287S . Experimental conditions were identical to those described in A except that 0.25 mM Cd 2ϩ was used for titrations of NEMmodified YiiP H153C/C287S . Thermodynamic binding parameters of YiiP H153C/C287S are summarized in Table 1. that retained full metal binding capacity and overall transport activity. The Asp O-and Cys S-donor are preferred Zn 2ϩ /Cd 2ϩ ligands, and frequently interchangeable in many polyhedral coordination systems. The overall functional convertibility between Asp and Cys is consistent with the coordination chemistry expected for a direct Cd 2ϩ -Asp 49 binding. Furthermore, although cysteine is a preferred coordination residue for soft metal ions, the thiol reactivity of Cys 49 responded to metal ions in a selective fashion. Among the divalent cations tested, only Zn 2ϩ , Cd 2ϩ , and Fe 2ϩ protected C49 from MPB labeling, whereas Co 2ϩ and Ni 2ϩ showed no protection under the identical MPB labeling condition. This observation suggests that the residue chemistry at position 49 is built into a coordination system where spatial arrangements of coordination residues impose geometrical constrain for selection of one metal ion against another similar one.
A quantitative comparison of Cd 2ϩ binding to YiiP and YiiP D49A led to the assignment of Asp 49 to one of the two Cd 2ϩ binding sites (site 2). This result mirrors an earlier finding that a D157A mutation selectively disrupts Cd 2ϩ binding to site 2 without affecting site 1 (11). Moreover, the D49A/D157A double mutation has the same disruptive effect specific to site 2, providing further evidence that Asp 49 and Asp 157 contribute to a common binding site. In corroboration with the Asp 49 -Asp 157 functional connection, the formation of the Cys 49 -Hg 2ϩ -Cys 49 biscysteinate binding, and Cys 49 -Cys 49 , Cys 157 -Cys 157 intersubunit and Cys 49 -Cys 157 intrasubunit disulfide cross-links established physical prox-imities among Cys 49 s and Cys 157 s across the dimer interface. Our results lead to the localization of active metal binding to the dimer interface of YiiP, thereby demonstrating for the first time that the functional unit of a representative CDF transporter is a homodimer.
ITC analysis indicated that Cd 2ϩ binds to (Asp 49 -Asp 157 ) in each subunit with a stoichiometry near unity, thus the (Asp 49 -Asp 157 ) 2 cluster forms a bimetal binding center at the dimeric interface. In general, this binding of metal ions in pair is opposed by electrostatic and entropic considerations, but the hydrophobic environment of (Asp 49 -Asp 157 ) 2 may act in favor of bimetal binding. The two bound Cd 2ϩ ions and four aspartyl carboxylates, each potentially carries a full negative charge, may allow an overall electrostatic balance in the middle of the transmembrane spanning domain. In support of the bimetal binding, Cd 2ϩ transport showed a sigmoidal dependence on the Cd 2ϩ concentration, a characteristic of positive cooperativity of Cd 2ϩ binding. It appears that the binding energy of the first Cd 2ϩ ligation is transduced to the neighboring binding site and increases its Cd 2ϩ affinity.
Coordination geometry consideration suggests that a total of eight ligand groups are needed to fill the coordination sphere of the bimetal binding center if each of two bound Cd 2ϩ adopts a tetrahedral coordination system. Besides Asp/Glu carboxylates, other potential Cd 2ϩ ligand groups are Cys thiolates and His imidazoles. Because Asp 157 has been mapped to the hydrophobic core of TM5 (11), only three coordination residues, by virtue of their topological locations, could potentially contribute additional ligand groups to the bimetal binding center. Among them, Cys 127 appears to be a functionless residue whereas cysteine substitutions of Asp 45 and His 153 confer NEM sensitivity to both Cd 2ϩ binding and transport. Asp 45 and His 153 are predicted one helical turn above Asp 49 and below Asp 157 , respectively. The topological locations of Asp 45 and Asp 153 may permit them to line up with the (Asp 49 -Asp 157 ) 2 bimetal binding center on the dimer interface. Indeed, copper phenanthroline and alkylthiosulfonates cross-linking of D45C and H153C both yielded intersubunit cross-links, in support of the interfacial locations of Asp 45 and His 153 .
The functional contributions of Asp 45 and His 153 may extend the (Asp 49 -Asp 157 ) 2 bimetal binding center to two peripheral pairs of Asp 45 s and His 153 s along the 2-fold axis of a YiiP homodimer. Each of these eight coordination residues appears to be indispensable for bimetal binding. An Asp 3 Ala mutation at positions 49 or 157, or NEM modification to D45C or FIGURE 8. Schematic bimetal binding center in the metal translocation pathway. The proposed bimetal binding center is composed of eight coordination residues, positioned along the dimer 2-fold-symmetry-axis (dashed line). YiiP adopts two distinct conformational states, alternately exposing six of the eight coordination residues for the binding of two Cd 2ϩ ions (filled circles) either from the extracellular or intracellular side as indicated. Solid lines represent eight coordinates in the bimetal binding center. Note that two bound Cd 2ϩ ions are bridged via bidentate interactions with either a pair of Asp 49 s or Asp 157 s, resulting in two interconvertible metal coordination modes as indicated by small arrows.