Identification of the Zn2+ Binding Site and Mode of Operation of a Mammalian Zn2+ Transporter*

Vesicular zinc transporters (ZnTs) play a critical role in regulating Zn2+ homeostasis in various cellular compartments and are linked to major diseases ranging from Alzheimer disease to diabetes. Despite their importance, the intracellular localization of ZnTs poses a major challenge for establishing the mechanisms by which they function and the identity of their ion binding sites. Here, we combine fluorescence-based functional analysis and structural modeling aimed at elucidating these functional aspects. Expression of ZnT5 was followed by both accelerated removal of Zn2+ from the cytoplasm and its increased vesicular sequestration. Further, activity of this zinc transport was coupled to alkalinization of the trans-Golgi network. Finally, structural modeling of ZnT5, based on the x-ray structure of the bacterial metal transporter YiiP, identified four residues that can potentially form the zinc binding site on ZnT5. Consistent with this model, replacement of these residues, Asp599 and His451, with alanine was sufficient to block Zn2+ transport. These findings indicate, for the first time, that Zn2+ transport mediated by a mammalian ZnT is catalyzed by H+/Zn2+ exchange and identify the zinc binding site of ZnT proteins essential for zinc transport.


Vesicular zinc transporters (ZnTs) play a critical role in regulat-
ing Zn 2؉ homeostasis in various cellular compartments and are linked to major diseases ranging from Alzheimer disease to diabetes. Despite their importance, the intracellular localization of ZnTs poses a major challenge for establishing the mechanisms by which they function and the identity of their ion binding sites. Here, we combine fluorescence-based functional analysis and structural modeling aimed at elucidating these functional aspects. Expression of ZnT5 was followed by both accelerated removal of Zn 2؉ from the cytoplasm and its increased vesicular sequestration. Further, activity of this zinc transport was coupled to alkalinization of the trans-Golgi network. Finally, structural modeling of ZnT5, based on the x-ray structure of the bacterial metal transporter YiiP, identified four residues that can potentially form the zinc binding site on ZnT5. Consistent with this model, replacement of these residues, Asp 599 and His 451 , with alanine was sufficient to block Zn 2؉ transport. These findings indicate, for the first time, that Zn 2؉ transport mediated by a mammalian ZnT is catalyzed by H ؉ /Zn 2؉ exchange and identify the zinc binding site of ZnT proteins essential for zinc transport.
Zn 2ϩ ions are essential for a very large number and variety of cellular functions (1) but are also potentially toxic. Zinc homeostasis is therefore dynamically maintained by a variety of transporters and other proteins distributed in distinct cellular compartments. Zinc transport is mediated by two major protein families: the Zip family, which mediates Zn 2ϩ influx, and the ZnTs, 3 which are primarily linked to Zn 2ϩ sequestration into intracellular compartments and are, thereby, involved in lowering cytoplasmic Zn 2ϩ concentrations (2).
The ZnT family of proteins consists of 10 known members that play versatile roles in many physiological and pathophysiological aspects of zinc homeostasis (3). For example, sequestration of zinc into synaptic vesicles is mediated by ZnT3, with changes in ZnT3 expression being linked to gender-specific susceptibility to Alzheimer disease (4). Another example, zinc deficiency in maternal milk has been linked to mutation in ZnT4 in mice (5). Finally, ZnT8 mediates zinc transport into insulin secretory vesicles and is a marker for the genetic polymorphism associated with type 2 diabetes in humans (6).
In the present work, we have focused on ZnT5, since it exemplifies both the complexity and physiological importance of ZnT transport activity. ZnT5 was previously shown to localize to the Golgi apparatus and to mediate the Zn 2ϩ transport that is essential for proper folding of Zn 2ϩ binding proteins within this compartment (7,8). The phenotypes of ZnT5 knock-out mice underscore the importance of this transporter in multiple contexts. Among the most striking are decreased bone density, reduced weight, and fatal, male-specific, cardiac bradyarrhythmia (9). ZnT5 has been also shown to mediate Zn 2ϩ uptake (10). It has also been suggested that ZnT5 subunits assemble with those of ZnT6 and that this hetero-oligomerization is required for the activation of the Zn 2ϩ -dependent alkaline phosphatases (11). Oligomerization is also essential for functional complementation of yeast Zn 2ϩ transporters (12). It is unclear, however, whether such hetero-oligomerization is essential for ZnT5 catalytic activity or merely up-regulates Zn 2ϩ transport mediated by ZnT5. The manner in which mammalian ZnTs, among them ZnT5, mediate Zn 2ϩ transport into organelles, however, is not known.
Clues about the functional mechanism of mammalian ZnTs arise from studies on their bacterial, yeast, and plant homologues. The bacterial Zn 2ϩ transporter, ZitB, was shown to catalyze H ϩ /Zn 2ϩ exchange when reconstituted into proteoliposomes (13). The recently determined structure of the YiiP protein suggested that the catalytic Zn 2ϩ binding site contains four residues that are spatially arranged to coordinate Zn 2ϩ (14). Studies performed on yeast and plant ZnT homologues have demonstrated that Zn 2ϩ transport is dependent on the electrochemical gradient generated by the vacuolar ATPase (15,16). Notably, however, no direct H ϩ /Zn 2ϩ exchange has yet been demonstrated for these transporters, presumably because of leakiness of the purified vesicles to H ϩ . Thus, whether the yeast or plant transporters directly mediate H ϩ /Zn 2ϩ exchange or are merely utilizing the vesicular H ϩ gradient for an alternate mode of transport is still unclear.
Focusing on ZnT5 as our model protein, the primary aims of the current study were to 1) elucidate the functional mechanism by which ZnTs mediate Zn 2ϩ sequestration and 2) iden-tify the catalytic domains and residues that form the Zn 2ϩ binding site in mammalian ZnTs.
Generation of Mutants-Site-directed mutagenesis was performed on the hZnT5 double-stranded plasmid (containing the hZnT5 gene; accession number AF461760), using the QuikChange site-directed mutagenesis kit (Stratagene) according to protocols provided by the manufacturer. The following primers were utilized: D599A, CTACATGTTTTGGCAGC-TACTCTTGGCAGCATTG; D599E, CTACATGTTTGGCA-GAGACTCTTGGCAGCATTG; H451A, GATCTCGGATG-GATTCGCCATGCTTTTTGACTGC. The oligonucleotide sequences used for mutation sequencing encompassed a 391-bp fragment for Asp 599 and a 398-bp fragment for His 451 , which contains the mutations. A cassette encompassing the entire hZnT5 ORF or containing the mutations was subsequently transferred into the original plasmid.
Fluorescent Imaging of Ion Transport-Zn 2ϩ transport was determined in cells loaded with either 3.5 M Fura-2 AM (TEFLabs), 0.5 M FluoZin-3 AM, or 5 M Newport Green AM (Invitrogen), using an imaging system as described previously (21). To verify that the fluorescence changes were related to intracellular Zn 2ϩ , the cell-permeable heavy metal chelator N,N,NЈ,NЈtetrakis-(2-pyridylmethyl)-ethylenediamine (TPEN; 20 M) was used. Zn 2ϩ sequestration into the intracellular compartments was determined by staining cells with ZinPyr-1 (5 M), a Zn 2ϩ -sensitive dye that preferentially accumulates within the trans-Golgi compartment (TGN) (22). Changes in vesicular pH were determined in HEK-293T cells expressing pHluorins, which are targeted to the trans-Golgi region (23). TGN-pHluorin plasmid was kindly provided by Terry Machen (University of California, Berkeley) (24). For this purpose, cells were co-transfected with 0.25 g of the pHluorin plasmid in a 60-mm culture dish. Emission images of cells expressing pHluorin were collected through a 510-nm long pass filter during sequential excitation at 410 and 470 nm.
Analysis of the fluorescent signals was performed by normalizing the fluorescence (or the ratio of the fluorescent signal) to an initial base line averaged over the first 10 s acquired in each experiment and was averaged over at least 30 cells. For the pHluorin measurements, the Golgi pH was calibrated using a high K ϩ Ringer's solution set to pH values of 6 -8 in the presence of nigericin, as previously described (23). Initial rates of H ϩ or Zn 2ϩ transport were determined using this paradigm by calculating either dpH i /dt or d[Zn 2ϩ ] i /dt, respectively. The initial rates were normalized as indicated in each experiment, and their averages Ϯ S.E. are shown in the bar graphs.
Changes in intracellular pH were triggered using the ammonium prepulse paradigm, as described previously (25). Briefly, cells were superfused with Ringer's solution containing NH 4 Cl (30 mM, replacing the equivalent 30 mM NaCl), which was subsequently replaced with NH 4 Cl-free solution, thus triggering intracellular acidification.
Immunoblot Analysis-Cells were extracted using 300 l of boiling denaturative lysis buffer (1% SDS, 10 mM Tris-HCl, pH 8)/100-mm plate and transferred to ice. A protease inhibitor mixture (Boehringer Complete protease inhibitor mixture; Roche Applied Science) was then added, and protein concentrations were determined using the modified Lowry procedure (26) and stored at Ϫ80°C. SDS-PAGE and immunoblot analyses were performed as previously described (27), using the ZnT5 and the HA antibodies at a dilution of 1:2000 and 1:1000, respectively.
Three-dimensional Structural Modeling of ZnT5-The ZnT5 sequence was sent to the 3D-Jury metaserver (28), which suggested the bacterial Zn 2ϩ transporter, YiiP (14), as a high confidence homologue. The alignment proposed by the server was further edited manually to accommodate experimental results, which indicate a cytoplasmic position for the histidine-rich loop (residues 543-590). It should be emphasized that this manipulation did not involve any of the key residues discussed in this paper. The final model was generated by the homology model program of the MESHI package (29).
Statistical Analysis of Data-Data analysis was performed using the SPSS software (version 14.0; SPSS Inc., Chicago, IL). All results shown are the means Ϯ S.E. of at least three individual experiments (n Ն 3) with averaged responses of 40 -100 cells in each experiment. t test p values of Ͻ0.05 were considered significant following Levene's test for equality of variances. Significance of the results is indicated as follows: relative to control (*) or relative to ZnT5 (#). *, p Ͻ 0.05; **, p Ͻ 0.01.

Cytoplasmic Zn 2ϩ Removal Rates in Cells Expressing
ZnTs-To determine the expression levels and cellular localization of heterologously expressed ZnT5 and ZnT6, we conducted Western blot and immunocytochemical analyses of HEK293-T cells transfected with the indicated plasmids. Transfection of HEK293-T cells was followed by increased expression of ZnT5 and ZnT6 polypeptides at their expected molecular masses (18,30) of ϳ55 and ϳ47 kDa, respectively (Fig. 1a). We also found that both ZnT5 and ZnT6 migrated in a doublet form, which is consistent with expression of mature and immature forms following overexpression in HEK293 cells (31). To determine the cellular distribution of ZnT5 and ZnT6, cells transfected with either plasmid were co-transfected with pHluorins (23). As shown in Fig. 1b, an overlap was observed between the expression pattern of ZnT5 or ZnT6 and pHluorins, indicating, in agreement with previous studies (18,30), that heterologously expressed ZnTs localize to the trans-Golgi compartment of HEK293-T cells.
Because the above and previous results (18,30) show that ZnT5 and ZnT6 are expressed in intracellular compartments, our experimental paradigm, aimed at monitoring Zn 2ϩ sequestration by ZnT5, consisted of loading cells with cytoplasmic Zn 2ϩ and recording its subsequent transport by these proteins. We hypothesized FIGURE 2. Enhanced cytoplasmic Zn 2؉ transport in cells co-expressing the ZnT proteins. a, experimental paradigm for monitoring cytoplasmic Zn 2ϩ transport. HEK293-T cells, transfected with the indicated plasmids, were loaded with Fura-2. At the time indicated by an arrow, the cells were superfused with Ringer's solution containing 200 M Zn 2ϩ and 2 M pyrithione, which was followed by a rise in intracellular Zn 2ϩ . Cells were then superfused with Ringer's solution without Zn 2ϩ or pyrithione to remove the residual pyrithione from the membrane. Note that [Zn 2ϩ ] i remained occluded within the cells. Cytoplasmic Zn 2ϩ transport rates were monitored during the perfusion with the Zn 2ϩ -and pyrithione-free Ringer's solution for the time interval indicated by the bar. To ascertain that the signal is related to Zn 2ϩ , the membrane-permeable metal chelator, TPEN, was added to the superfusing solution at the end of the experiment (at the time indicated by the arrow). b, representative traces of Fura-2 signal normalized to initial fluorescence (R 0 ), in cells expressing the indicated transporters. The time period marked by the bar in a was used for determining the rate of Zn 2ϩ transport. The histogram shows averaged rates of cytoplasmic Zn 2ϩ removal normalized to the rates of control cells (mean Ϯ S.E., n ϭ 10). c, cells preloaded with the intracellular Zn 2ϩ -specific dye, Newport Green, subjected to the same paradigm described in a, and representative traces of changes in intracellular Zn 2ϩ , from the time period indicated by the bar in a. Shown are averaged rates of cytoplasmic Zn 2ϩ removal obtained from Newport Green-loaded cells (mean Ϯ S.E., n ϭ 8) (p value relative to control (*) or relative to ZnT5 (#); *, p Ͻ 0.05; **, p Ͻ 0.01).
that ZnT5 and ZnT6 activity will result in a decrease in cytoplasmic [Zn 2ϩ ] levels concomitant with a reciprocal increase in its level in intracellular compartments. HEK293-T cells, expressing either ZnT5 or ZnT6 or co-expressing ZnT5 and ZnT6 (ZnT5ϩ6) were preloaded with Fura-2, which is a ratiometric dye sensitive to Zn 2ϩ (32,33). Loading of cells with Zn 2ϩ was subsequently achieved by superfusion of Ringer's solution containing Zn 2ϩ (200 M) in the presence of the Zn 2ϩ ionophore, pyrithione (2 M), as indicated by the rise in Fura-2 fluorescence (Fig. 2a). Cells were subsequently washed with Ringer's solution without Zn 2ϩ or pyrithione. No rapid change in fluorescence was observed in the control cells following perfusion with pyrithione-free Ringer's solution, indicating that Zn 2ϩ ions were trapped within the cells. Finally, application of the cell-permeable Zn 2ϩ chelator, TPEN (20 M), was followed by reduction of the fluorescent signal back to base line, indicating that the Fura-2 signal is indeed related to changes in [Zn 2ϩ ] i (Fig. 2a). We then compared the rate of cytoplasmic fluorescent signal decrease during the wash period (marked with a bar in Fig. 2a) in cells expressing ZnT5, ZnT6, or ZnT5ϩ6 versus control cells (transfected with an empty vector). The averaged normalized rate of cytoplasmic Zn 2ϩ transport, compared with control cells, was about 3-fold higher (316 Ϯ 70% of control, p ϭ 0.001) in cells co-expressing ZnT5ϩ6 or 2-fold fold higher (182 Ϯ 26% of control, p ϭ 0.04) in cells expressing ZnT5 alone. In contrast, no significant change in fluorescence was observed in ZnT6-expressing cells following preloading with Zn 2ϩ , suggesting that ZnT6 by itself is not involved in reducing the cytoplasmic Zn 2ϩ level. Because Zn 2ϩ transport mediated by ZnT6 was indistinguishable from that of control cells, we have subsequently focused on monitoring the activity of ZnT5 and ZnT5ϩ6. To further verify that the change in cytoplasmic fluorescent signal was related to Zn 2ϩ , we performed the same protocol using the specific Zn 2ϩ -sensitive fluorescent dye, Newport Green. As shown in Fig. 2c, and consistent with the data presented in Fig. 2b, the rate of cytoplasmic Zn 2ϩ removal in cells co-transfected with ZnT5ϩ6 was enhanced by ϳ4-fold (390 Ϯ 70%, p ϭ 0.02) compared with controls. Taken together, these results suggest that expression of ZnT5, but not of ZnT6, is followed by accelerated removal of cytoplasmic Zn 2ϩ . Co-expression of ZnT5 and ZnT6, moreover, markedly enhanced the rate of cytoplasmic Zn 2ϩ removal.
We next sought to determine whether cytoplasmic Zn 2ϩ removal, mediated by ZnT5ϩ6, is associated with vesicular sequestration of Zn 2ϩ into the Golgi network. Cells were superfused with Zn 2ϩ , using the same experimental paradigm described above and subsequently stained with the Zn 2ϩ -sensitive dye, Zinpyr-1, that preferentially accumulates within the trans-Golgi apparatus or Golgi-associated vesicles (22). As shown in Fig. 3, the vesicular Zn 2ϩ accumulation was about 3.5-fold (337 Ϯ 59% of control, p ϭ 0.001) higher in cells expressing ZnT5ϩ6 compared with control cells (transfected with the vector alone) and 3-fold (290 Ϯ 58% of control, p ϭ 0.021) higher in the ZnT5-expressing cells. Application of TPEN was again followed by a reduction of the fluorescent signal to levels monitored in control cells (p ϭ 0.368). These results, taken together with the vesicular targeting of ZnT5 and ZnT6 and the enhanced cytoplasmic removal of Zn 2ϩ , support the conclusion that expression of ZnT5 or coexpression of ZnT5ϩ6 leads to enhanced cytoplasmic Zn 2ϩ removal and its accumulation in Golgi vesicles.
Rates of Vesicular H ϩ Transport in ZnT-expressing Cells-If the driving force for vesicular Zn 2ϩ transport is the H ϩ gradient, then changes in vesicular H ϩ gradient would be expected to affect cytoplasmic Zn 2ϩ removal. Cells were subjected to NH 4 Cl prepulse, which was reported to activate the V-type ATPase activity (34), and subsequently rates of ion transport were monitored. We initially asked if enhanced Zn 2ϩ transport mediated by ZnT5ϩ6 would be monitored by this experimental paradigm. HEK293-T cells were loaded with Fura-2 and treated with Zn 2ϩ pyrithione, following the NH 4 Cl prepulse. The rate of cytoplasmic Zn 2ϩ removal was subsequently compared in ZnT5 or ZnT5ϩ6-expressing cells versus controls (Fig. 4a). The rate of cytoplasmic Zn 2ϩ removal, following this procedure, was enhanced by about 2-fold (184 Ϯ 24% of control, p ϭ 0.018) in ZnT5-expressing cells and by 3-fold (314 Ϯ 90% of control, p ϭ 0.002) following expression of both ZnT5 and ZnT6. Rates of Zn 2ϩ transport were similar to those obtained in the absence of the NH 4 Cl prepulse. This may be related to the fact that the NH 4 Cl prepulse preceded Zn 2ϩ loading, and due to the leaky nature of this compartment the resulting increase in the pH gradient was at least partially dissipated. Nevertheless, the increase in Zn 2ϩ efflux rates in cells expressing ZnT5 or ZnT5ϩ6 versus control was maintained using this protocol. To determine the role of the vesicular pH gradient, the V-type H ϩ pump inhibitor, bafilomycin (100 nM), was applied to cells coexpressing ZnT5ϩ6, and rates of cytoplasmic Zn 2ϩ removal were monitored. Using this experimental paradigm, in HEK293-T cells expressing the Golgi-targeted pH sensor, pHluorin, we have found that bafilomycin enhanced the alkalinization rate in the Golgi (see Fig. 4b, inset), indicating that the V-type ATPase was inhibited by bafilomycin. In cells loaded with Fura-2, a reduction in the rates of Zn 2ϩ transport of ϳ2-fold (46 Ϯ 8% of control, p ϭ 0.024) was observed in the presence of bafilomycin, indicating that the vesicular pH gradi-ent is required for maintaining Zn 2ϩ transport mediated by ZnT5ϩ6 (Fig. 4b).
Although the results of the above experiment support a role for the H ϩ electrochemical gradient in intracellular Zn 2ϩ transport, they do not demonstrate direct coupling between H ϩ and Zn 2ϩ transport. If ZnTs mediate coupled H ϩ /Zn 2ϩ exchange, then vesicular pH would be expected to alkalinize faster in ZnT5ϩ6-expressing cells. To address the hypothesis, we next monitored Zn 2ϩ -dependent changes in vesicular pH. We compared the rates of vesicular alkalinization in cells transfected with the TGN-targeted pHluorins co-expressed with ZnT5ϩ6 or with an empty vector. In the absence of Zn 2ϩ loading, no significant difference in the vesicular pH change was observed in cells expressing ZnT5ϩ6 or in control cells (Fig.  5a). Importantly, following preloading with Zn 2ϩ , rates of vesicular alkalinization increased by 2-fold (220 Ϯ 54% of control, p ϭ 0.004) in cells expressing ZnT5 or by 3-fold (310 Ϯ 50% of control, p ϭ 0.0001) in cells expressing ZnT5ϩ6 versus control. No significant Zn 2ϩ -dependent alkalinization was apparent in cells expressing only ZnT6 (p ϭ 0.645) compared with control cells (Fig. 5b), indicating that expression of this transporter alone is not sufficient to facilitate Zn 2ϩ -coupled H ϩ transport.
Taken together, the results presented here indicate that transport of Zn 2ϩ by ZnT5 is mediated through H ϩ /Zn 2ϩ exchange and is powered by the vesicular H ϩ gradient. Similar to the results obtained for Zn 2ϩ transport, co-expression of ZnT5 with ZnT6 enhanced H ϩ transport, but the latter was not essential. Moreover, when expressed alone, ZnT6 was catalytically inactive.
The coupling of H ϩ and Zn 2ϩ transport is illustrated in Fig.  5c, which shows the simultaneous removal of cytoplasmic zinc and vesicular alkalinization in cells expressing ZnT5, suggesting that Zn 2ϩ transport mediated by ZnT5 is catalyzed by H ϩ /Zn 2ϩ exchange.
Modeling of the ZnT5 Zinc Binding Site and Mutational Analysis of Zinc-coordinating Residues-We next sought to identify the catalytic domain that mediates Zn 2ϩ transport by ZnT5. The structure of ZnT5 has not yet been determined; nor has that of any close homologue. However, the fold recognition metaserver, 3D-Jury (28), suggested that YiiP, a bacte- rial Zn 2ϩ transporter, is a potential, reliable homologue to the C-terminal region of ZnT5, which encompasses about 50% of ZnT5 (residues 411-732 were analyzed). The overall sequence homology between the ZnT5 construct and YiiP is rather low (i.e. 15%), although the pattern of putative transmembranal helices is conserved, as are several key membraneembedded residues. Among these, most notable are three of the four charged residues that coordinate the binding of Zn 2ϩ ions in the bacterial transporter. Asp 45 of the bacterial binding site is replaced in ZnT5 by a histidine residue, which has higher selec-tivity for Zn 2ϩ . Based on the YiiP structure, a model for the C-terminal domain of ZnT5 was constructed (Fig. 6a). This homologybased ZnT5 model suggested the putative binding site for Zn 2ϩ (Fig.  6b), supporting the notion that this ZnT5 fragment is the catalytic domain. Thus, the working hypothesis that guided our subsequent experiments was that residues His 451 and Asp 455 in helix 11 and His 595 and Asp 599 in helix 14 are involved in Zn 2ϩ coordination as they are on the bacterial YiiP protein.
If these residues compose the binding site, then replacement of even one of the residues would be expected to interfere with this process and hence to inhibit ZnT5-mediated Zn 2ϩ transport. To test this hypothesis, the Asp 599 residue was mutated to either Ala (D599A) or Glu (D599E). Expression analysis of the WT and mutants showed that the amount of WT plasmid required to achieve expression levels similar to that of the mutants was about 2-fold lower for the ZnT5 D599E mutant (Fig. 7a). The amount of plasmid used for the expression of the D599E mutant was adjusted accordingly for the functional analysis. Cellular distribution of mutant and WT ZnT5 was determined using confocal microscopy on cells co-expressing pHluorins in a manner to similar that described in the legend to Fig. 1. As shown in Fig. 7b, the cellular distribution of WT ZnT5 and the Asp 599 mutations was similar and overlapped with the distribution of pHluorins, indicating that the mutants are also targeted to the trans-Golgi.
We next assessed whether the mutations described above affect Zn 2ϩ transport, by monitoring the rate of cytoplasmic Zn 2ϩ transport, using the paradigm shown in Fig. 2a, in cells loaded with FluoZin-3. As shown in Fig. 7c, cytoplasmic Zn 2ϩ removal was blocked in cells transfected with the D599A ZnT5 mutant (21 Ϯ 14% of the WT ZnT5, p ϭ 0.01). When Asp 599 was replaced with glutamate, however, a similar rate of Zn 2ϩ transport activity was observed (119 Ϯ 29% of the WT ZnT5). We then asked if substitution of another residue, which, according to our model, coordinates Zn 2ϩ , also leads to inhibition of cytoplasmic Zn 2ϩ transport. As shown in Fig. 7d, although the Averaged normalized rates of vesicular pHluorin fluorescence change are shown at the right (mean Ϯ S.E., n ϭ 10). b, rates of vesicular pH in cells co-expressing either ZnT5ϩ6, ZnT5 alone, or ZnT6 alone were determined by measuring the rate at the time period following Zn 2ϩ loading. Averaged normalized rates are shown on the right (mean Ϯ S.E., n ϭ 8). c, representative traces of cytoplasmic Zn 2ϩ imaged with Fura-2 (as in Fig. 2a) and vesicular pH imaged with pHluorins (as in b), following loading with Zn 2ϩ (mean Ϯ S.E., n ϭ 7) are shown. Note the reciprocal change in Zn 2ϩ and pH, indicating that the transport of these ions is coupled (p value relative to control (*) or relative to ZnT5 (#); *, p Ͻ 0.05; **, p Ͻ 0.01). expression level of the ZnT5 H451A mutant is similar to that of WT ZnT5, cytoplasmic Zn 2ϩ transport in cells expressing this mutant was blocked (20 Ϯ 15% of the WT ZnT5, p ϭ 0.01). Finally, we asked if these residues are also responsible for the sequestration of Zn 2ϩ into the Golgi, by monitoring the change in vesicular Zn 2ϩ . Cells were superfused with a Zn 2ϩ -containing Ringer's solution and were stained with Zinpyr-1, as described in the legend to Fig. 3. Consistent with the Zn 2ϩ efflux results, vesicular Zn 2ϩ sequestration was blocked in cells expressing the D599A and H451A mutants and was indistinguishable from that monitored in cells transfected with vector alone (Fig. 7e). These results suggest that Asp 599 and His 451 are critical residues of the Zn 2ϩ binding site, as predicted by the YiiP-based homology model.

DISCUSSION
A major challenge in analyzing the activity of ZnTs is their intracellular localization, which limits direct access for the manipulating ion gradients. To overcome this, we have devised an experimental paradigm based on intracellular trapping of Zn 2ϩ using the ionophore, pyrithione. This approach enabled us to image changes in ion concentrations (Zn 2ϩ and H ϩ ) in the cytoplasmic and Golgi compartments. Focusing on ZnT5 and ZnT6, we demonstrated, for the first time, that vesicular ZnTs catalyze H ϩ /Zn 2ϩ exchange. This conclusion is based on the following findings. 1) Cytoplasmic removal of Zn 2ϩ into intracellular compartments occurs parallel to alkalinization of the Golgi in cells expressing ZnT5 alone or co-expressing it with ZnT6; 2) Zn 2ϩ transport is attenuated following pretreatment of the cells with the vacuolar V-type ATPase inhibitor, bafilomycin. This latter observation argues against a mechanism in which a primary pump mediates the Zn 2ϩ transport and in favor of a secondary active mechanism underlying the activity of ZnT proteins. This finding is also consistent with previous studies suggesting that sequestration of vesicular Zn 2ϩ transport, presumably mediated by ZnT3, is attenuated in the presence of bafilomycin (35). 3) Vesicular H ϩ efflux from the trans-Golgi network is accelerated in the presence of cytoplasmic Zn 2ϩ only in cells expressing ZnT5. It may also be argued that the Zn 2ϩ -dependent alkalinization of the Golgi observed by us is mediated by the modulation of the vesicular H ϩ channel. The latter is a major conductive pathway for H ϩ in the Golgi, and Zn 2ϩ is a low affinity inhibitor of this channel (36). This scenario is unlikely, however, based upon the following. 1) Zn 2ϩ -dependent alkalinization was found only in ZnT-expressing cells. Zinc-dependent alkalinization was not found in control cells or even in cells expressing only ZnT6, linking the akalinization to the expression of ZnT5 rather than to loading of cells with Zn 2ϩ . 2) Because Zn 2ϩ is an inhibitor of the channel, loading the cells with this ion is expected to block the H ϩ conductive pathway and to reduce the H ϩ leak. This would result in enhanced acidification of the Golgi. We found, in contrast, that loading of Zn 2ϩ into cells expressing ZnT5 leads to akalinization of the Golgi. Thus, our results are not consistent with a model in which modulation of the H ϩ conductive pathway occurs but agree well with one in which H ϩ /Zn 2ϩ exchange is mediated by the ZnT.
The pH dependence of vesicular ZnTs has potentially far reaching physiological implications, since it predicts that a large vesicular H ϩ electrochemical gradient will be followed by enhanced Zn 2ϩ sequestration. An intriguing example is the glucose-dependent acidification of insulin secretory vesicles (37). Although the exact physiological role of this acidification is unknown, it has been suggested that it may enhance the stabilization of insulin hexamers (38). The pH dependence of Zn 2ϩ transport mediated by ZnT5, as described here, provides a plausible mechanism by which an increased H ϩ electrochemical gradient across the membrane of insulin secretory vesicles will trigger enhanced Zn 2ϩ sequestration, followed by an accelerated rate of insulin hexamer formation.
If the vesicular ZnTs are H ϩ /Zn 2ϩ exchangers, then cytoplasmic pH may also affect Zn 2ϩ sequestration/release from the vesicles. Consistent with this, it was demonstrated previously that acidification of astroglia led to a dramatic rise in cytoplasmic Zn 2ϩ , attributed to the enhanced dissociation of Zn 2ϩ from metallothionines (39). Our results suggest that the rise in cytoplasmic Zn 2ϩ may also be the result of inhibition or even reversal of Zn 2ϩ transport mediated by vesicular ZnTs.
The Functional Implications of the Oligomeric Structure of ZnT Proteins-ZnTs, particularly ZnT5 and ZnT6, are capable of forming hetero-oligomers. Whether this organization is catalytically essential or only helps enhance ZnT activity is unclear. In agreement with our study, heterologous expression of ZnT5 was sufficient to trigger Zn 2ϩ transport in vesicles prepared from hZnT5-transfected HeLa cells (18). Endogenous ZnT6 expression could have contributed to ZnT5 activity; however, the simplest explanation is that ZnT5 can support Zn 2ϩ transport. We also found that neither Zn 2ϩ nor H ϩ transport is mediated by ZnT6, a finding consistent with a previous study suggesting that ZnT6 lacks the catalytic domain required for Zn 2ϩ transport (19). In the chicken cell line, DT40 mutant, and in the yeast Zrg17/Msc2 Zn 2ϩ transporter mutant, however, the co-expression and oligomerization of ZnT5 and ZnT6 is essential for activation of the Zn 2ϩ -dependent phosphatase, TNAP, and for the functional complementation of yeast strains lacking the Zn 2ϩ transporter, Zrg17/Msc2 (11,12). The failure of ZnT5 alone to support TNAP activation or to complement yeast strains may indicate that the slower Zn 2ϩ transport mediated by ZnT5 alone, as compared with ZnT5ϩ6, is insufficient to support these processes. Alternatively, hetero-oligomerization may be required for either targeting of these transporters to distinct subcompartments of the Golgi (40) or for maintaining complex stability (41), issues that our assay could not resolve.
The Catalytic Domain of ZnTs-Little is known about the organization of the catalytic domain and particularly the residues composing the binding site for ZnTs. Among the domains previously investigated, most notable is a His-rich region assumed to be a Zn 2ϩ binding region, which is conserved among all members of the ZnT (i.e. except ZnT6) and ZIP proteins (11,14,42). Although deletion of this domain in ZnT5 resulted in a marked attenuation of the activation of Zn 2ϩ -dependent TNAP, it is not clear if this domain is involved in Zn 2ϩ binding or transport. In ZIP proteins, this His-rich domain is important for the Zn 2ϩ -induced turnover of this protein (43), suggesting that it may have an additional post-translational role modulating the half-life of Zn 2ϩ transporters. Furthermore, studies on plant ZnT homologues have indicated that deletion in the His-rich region results in a dramatic increase in the rate of Zn 2ϩ transport (16).
Our results represent an important step toward identifying the catalytic domain of ZnTs. Despite the modest overall homology between the bacterial YiiP transporter and ZnT5, we have demonstrated that a striking overlap exists in their structure, particularly at the membrane domain. Of particular inter-est is the putative spatial organization of four amino acid residues (three of which are highly conserved) that according to our model and functional analysis form the Zn 2ϩ binding site of ZnT5 (44). The catalytically inactive ZnT6 (11) lacks two of the Zn 2ϩ -coordinating residues, providing further support for the importance of this catalytic core in Zn 2ϩ transport.
Coordination of Zn 2ϩ by carboxyl and imidazole groups, compared with thiolates, may allow a faster dissociation rate of Zn 2ϩ during the conformational changes underlying the transport cycle. This is consistent with the rate of Zn 2ϩ transport mediated by ZnT proteins (18,45,46). In addition, these coordinating residues may also contribute the H ϩ ions necessary for the counter-H ϩ /Zn 2ϩ exchange. Such a mechanism has been described for Ca 2ϩ /H ϩ exchange by the sarcoplasmic reticulum Ca 2ϩ pump (47).
In the present study, we have demonstrated in intact cells, for the first time, that the mammalian Zn 2ϩ transporter, ZnT5, mediates H ϩ /Zn 2ϩ exchange. In addition, based on modeling and on function analyses, we propose that Zn 2ϩ is coordinated by two Asp and two His residues, which form the core of the mammalian Zn 2ϩ transport site.