Mutagenesis and Derivatization of Human Vesicle Monoamine Transporter 2 (VMAT2) Cysteines Identifies Transporter Domains Involved in Tetrabenazine Binding and Substrate Transport*

The vesicle monoamine transporter (VMAT2) concentrates monoamine neurotransmitter into synaptic vesicles. Photoaffinity labeling, chimera analysis, and mutagenesis have identified functionally important amino acids and provided some information regarding structure and ligand binding sites. To extend these studies, we engineered functional human VMAT2 constructs with reduced numbers of cysteines. Subsets of cysteines were discovered, which restore function to an inactive cysteine-less human VMAT2. Replacement of three transmembrane (TM) cysteines together (net removal/ replacement of three atoms) significantly enhanced monoamine transport. Cysteine modification studies involving single and combination cysteine mutants with methanethiosulfonate ethylamine revealed that [ 3 H]di-hydrotetrabenazine binding is > 90% inhibited by modi- fication of two sets of cysteines. The primary target (responsible for ; 80% of inhibition) is Cys 439 in TM 11. The secondary target (responsible for ; 20% of inhibition) is one or more of the four non-TM cysteines. [ 3 H]Di-hydrotetrabenazine protects against modification of Cys 439 by a 10,000-fold molar excess of methanethiosulfonate ethylamine, demonstrating that Cys 439 is either at the tetrabenazine binding site, or conformationally linked to tetrabenazine binding. Supporting a direct effect, the position of tetrabenazine-protectable Cys inhibition was assessed in the presence and absence of 5 m M ATP, 5 m M Mg 2 1 , 4 m M Cl 2 (conditions used for uptake experiments). MTSEA sensitivity is independent of the presence of a proton gradient. C , the effect of adding b -mercaptoethanol or DTT on recovery of [ 3 H]TBZOH binding after inhibition by MTSEA. After treatment of COS vesicles containing WT human VMAT2 with 2 m M MTSEA (as described under “Experimental Procedures”), incubation with 100 m M DTT for . 15 min at 30 °C had no reversal effect, whereas 100 m M b -mercaptoethanol (under the same conditions) had a small but statistically significant effect. The small effect of b -mercaptoethanol is consistent with sulfhydryl reversal at the secondary, but not the primary ( Cys 8 ), MTSEA target site. D , the effect of adding b -mercaptoethanol (separate experiment from panel C ) or cysteamine on recovery of [ 3 H]TBZOH binding after inhibition by MTSEA. After treatment of COS vesicles containing WT human VMAT2 with 2 m M MTSEA (as described under “Experimental Procedures”), incubation with 100 m M cysteamine for . 15 min at 30 °C had no reversal effect, whereas 100 m M b -mercaptoethanol (under the same conditions) had a small but statistically significant effect. The small effect of b -mercaptoethanol is consistent with sulfhydryl reversal at the secondary, but not the primary ( Cys 8 ), MTSEA target site. target of MTSEA in human VMAT2, is located in a very hydrophobic environment, on a pep- tide that is photolabeled by [ 125 I]TBZ-AIPP, on a domain found important for high affinity TBZ binding from chimera experiments, and is positioned centrally with regard to known mutations which affect tetrabenazine binding. Possible MTSEA secondary targets and 9 are also near mutations reported to affect TBZ binding, and Cys 10 may be near the N-terminal [ 125 I]TBZ-AIPP photolabeled peptide.

The vesicle monoamine transporter (VMAT2) 1 is a protonmonoamine antiporter, which concentrates monoamine neuro-transmitters into synaptic vesicles. VMAT2 transports a wide range of substrates including serotonin, dopamine, and norepinephrine and is inhibited by a number of drugs including reserpine, tetrabenazine (TBZ), and ketanserin. Aspects of VMAT2 biology and physiology have been recently reviewed (1)(2)(3). Briefly, bovine VMAT2 has been biochemically purified (4), and the initial cDNA clones of a VMAT were from rat PC12 cells (5) and a rat cDNA library (6). Recombinant rat VMAT2 can be expressed at high levels in Sf9 cells and purified (7). The gene structure and promoter regions of both human (8) and mouse (9) VMAT2 have been identified, and knockout mice have been generated. VMAT2 knockout mice are born, but homozygotes die within a few days after birth (10). Heterozygotes live into adulthood, but display altered sensitivity to amphetamine, cocaine, and the neurotoxin 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (11,12). A percentage of heterozygotes die suddenly from what appears to be cardiac arrythmias due to prolonged QT intervals (11,13).
Based on amino acid sequence, the prediction is that VMAT2 has 12 transmembrane (TM) helices, a structural commonality with a number of other transporters including plasma membrane monoamine reuptake transporters and P-glycoprotein. A number of mutations of rat VMAT1 and VMAT2 have been tested for their effects on function (14,15), including chimera experiments to probe the domains and residues responsible for binding affinity differences between VMAT2 and the related isoform VMAT1 (16 -19). These studies have suggested important residues that may contribute to ligand binding and monoamine transport. In addition, two significant observations that have contributed VMAT2 structural information are the discovery that a lysine in predicted TM 2 and an aspartate in predicted TM 11 interact functionally as an ion pair (20), and the identification in rat VMAT2 of peptide domains near or at the ketanserin and tetrabenazine binding sites, and an amino acid residue (Lys 20 ) near or at the ketanserin binding site, from photoaffinity labeling studies performed in our laboratory (7,21). Although the reasonable inferences from molecular biology and the biochemical structural information from photoaffinity labeling have provided significant and novel information, a more general approach that can be applied to obtain structural information from the entire molecule and to map ligand binding sites has been sought.
With a few notable exceptions, large membrane proteins like VMAT2 have represented an enormous challenge to structural biology, and in most cases still pose serious technical difficulties to analysis by crystallography or NMR spectroscopy. Recently, a creative combination of protein engineering and biochemical methods has been applied to large membrane proteins including receptors, transporters, and ion channels, to obtain information about membrane protein structure, function, and ligand binding sites. Many of these methods utilize native or engineered cysteines as sites of unique chemical reactivity within the membrane protein, and require either a molecule that is devoid of native cysteine residues or a molecule with a non-reactive background of remaining cysteines for the assays being applied (22)(23)(24)(25)(26)(27)(28). Cysteine-based assays, including sitedirected labeling with chemical probes (i.e. sulfhydryl-reactive methanethiosulfonate reagents) and intramolecular cysteine cross-linking to determine helix proximity and orientation, have proven useful and general approaches. A few of the most notable examples of transporters studied by these engineered cysteine-based methods include lactose permease (22) and other bacterial transporters (23,24), monoamine transporters (25)(26)(27), and P-glycoprotein (28).
Inspired by the example of a number of functional cysteineless membrane proteins and the flexibility and utility of cysteine-based methods, we sought to apply such methods to the study of human VMAT2. In this paper, we report a study of the role of the 10 native cysteines in human VMAT2 expression, ligand binding, and monoamine transport, and demonstrate by cysteine replacement, derivatization, and protection experiments that certain cysteines are in important positions for human VMAT2 [ 3  ; anti-HA antibody (Covance Research Products, Denver, PA); cysteamine, polyethyleneimine, ATP, 2-mercaptoethanol, and anti-Flag M2 antibody (Sigma); goat anti-mouse horseradish peroxidase antibody, electroporation cuvettes, precast electrophoresis ready gels, SDS sample buffer, and prestained molecular weight markers (Bio-Rad); Supersignal West pico chemiluminescent substrate and Superblock PBS blocking buffer (Pierce); Hyperfilm ECL (Amersham Pharmacia Biotech); GF/B filters (Whatman via Fisher; or Brandel, Gaithersburg, MD); 0.22-and 0.45-m filters (Millipore via Fisher); transfection-competent Escherichia coli (Stratagene; Novagen, Madison, WI); emulsifier Safe scintillation mixture (Packard, Meriden, CT). Other reagents (sucrose, HEPES, KCl, MgSO 4 , NaOH, KOH, Tris, DTT, additional restriction enzymes, etc.) were from reputable and readily available sources.
Construction of Human VMAT2 Mutants-A cDNA encoding human VMAT2 was the kind gift of Dr. Robert H. Edwards (University of California San Francisco School of Medicine, San Francisco, CA). 5Ј-Forward and 3Ј-reverse oligonucleotides with desired restriction sites (5Ј BglII, 3Ј HindIII) were used to amplify the cDNA by PCR, the restriction-cleaved PCR product was ligated into the plasmid pBlueBac 4.5, and the entire VMAT2 coding region was checked for mutations by DNA sequencing. The 3Ј reverse PCR primer incorporated a FLAG epitope, enterokinase cleavage site, and a 6-histidine epitope at the C terminus of the protein, adding the amino acids -Y-K-D-D-D-D-K-H-H-H-H-H-H. The correct FLAG-and 6-histidine-tagged human VMAT2 cDNA was then subcloned from pBlueBac 4.5 into the mammalian cell expression plasmid pCDNA 3.1-at the XhoI and HindIII restriction sites. An N-terminal FLAG-tagged construct was compared with the C-terminal tagged construct in a single high concentration [ 3 H]TBZOH binding experiment and found to have less overall ligand binding (data not shown). Therefore, all human VMAT2 constructs used in this study were derived from the initial C-terminal FLAG-and 6-histidine tagged construct, which hereafter will be referred to as "WT." Except for the C-terminal epitope tags, the sequence of this base construct is identical to the sequences for human VMAT2 deposited in GenBank under accession nos. L14269 and L23205.
Each of the 10 individual cysteine residues in human VMAT2 were replaced by one of several site-directed mutagenesis methods. In the incremental process of creating a cysteine-less human VMAT2, all of the cysteines were changed to serines. However in generating 10 separate single cysteine mutants, some cysteine replacements were to serine and some to alanine (as indicated in Fig. 1C). The methods employed in the replacement of cysteines included the GeneEditor (Promega) and QuickChange (Stratagene) methods, a PCR-based method, and a mutually priming oligo extension method. In every case, the resulting constructs were thoroughly sequenced to ensure the absence of any unwanted mutations in the human VMAT2 protein. To facilitate assessment of relative expression levels by Western blot, each of the four consensus glycosylation sites were mutated in the third position, from N-X-(S/T) to N-X-(V/A), and an HA epitope was added (Y-P-Y-D-V-P-D-Y-A) after arginine 94 in the large lumenal loop (see Fig. 2B). (The four mutations were Thr 58 3 Ala, Thr 86 3 Val, Thr 93 3 Val, and Ser 111 3 Ala.) Constructs which had this set of mutations and the HA insertion were designated "GϪHA," for "glycosylation-negative, HA epitope" Additional desired mutations introduced by one of the above methods included "silent" mutations (which did not change the encoded protein sequence) to introduce restriction sites. Among others, these included a particularly useful KpnI restriction site between predicted TM segments 2 and 3, and sites added with the replacement of many of the cysteine residues to facilitate screening by restriction digests.
Transient Expression and Harvesting from COS-7 Cells-The large quantities of DNA needed for transfection and assay of human VMAT2 constructs in mammalian cells were obtained by growth in LB medium of transformed E. coli followed by DNA purification on Qiagen DNA Prep columns. COS-7 cells were grown on 150-mm plates at 37°C in DMEM (Life Technologies, Inc. catalog no. 11965-084) with 9% cosmic calf serum (HyClone catalog no. SH30087.02), penicillin/streptomycin/ glutamine (10 ml/liter; Life Technologies catalog no. 10378-016), in a 5% CO 2 atmosphere. On some occasions the medium was also supplemented with less than 2 mg/liter anti-mycotic fungizone (amphotericin B), which at the concentration used did not seem to have any noticeable effect on the growth or appearance of the COS cells, or on expressed human VMAT2 function. For most experiments, three 150-mm plates of COS-7 cells were transfected for each VMAT construct being tested. As an internal control, WT (C-FLAG/His) human VMAT2 was transfected and assayed on each occasion of COS cell expression. Transfection was performed by electroporation of COS cells in PBS with 15-50 g of DNA (equal for all constructs in a given experiment) in 0.4-cm gap width cuvettes with the voltage of the Bio-Rad GenePulser/capacitance extender unit set at 0.226 mV and the capacitance set at 0.950 microfarads. Cells were harvested 1.5-4 days after transfection, the cell culture medium usually being changed once within that time period. Cells were detached with trypsin, washed with 300 mM sucrose, 10 mM Hepes (pH 7.2), protease inhibitor-containing buffer (generally 20 g/ml leupeptin, 100 M phenylmethylsulfonyl fluoride, 100 M benzamidine, 10 g/ml soybean trypsin inhibitor), and homogenized with a custom-built steel ball homogenizer, or "cell cracker" (29). After 30 -50 passages through the cell cracker, the homogenate was centrifuged for 5 min at 735 ϫ g, and the vesicle containing supernatant was collected for use in assays. Protein concentrations for COS vesicle preparations were assessed by the Bradford method using the Bio-Rad protein assay reagent, and were generally in the range of 2-5 mg/ml. Thus, assays that used 50 l of COS homogenate generally contained between 0.1 and 0.25 mg of total protein. Assays were performed on freshly prepared samples (not frozen) of COS cell homogenates, and results were normalized per milligram of COS homogenate protein.
Radioligand Synthesis-The radioligand used in binding assays was [ 3 H]dihydrotetrabenazine ([ 3 H]TBZOH). This was synthesized by reaction at room temperature of equal volumes (0.5-1 ml) of a 1.5 mg/ml solution of TBZ in 100% methanol with 5-25 mCi of [ 3 H]sodium borohydride (50 -75 Ci/mmol, in 0.1 M NaOH). Reduction occurs at the ketone, incorporating the 3 H label. After Ͼ3 h, the reaction mixture was concentrated under a stream of nitrogen, diluted with 100% methanol, reconcentrated a number of times to remove water, and then streaked onto a silica gel TLC plate and developed with a solvent mixture of 5 parts ethyl acetate to 1 part methanol. The major [ 3 H]TBZOH product, in a chemical amount visible upon illumination with a hand-held UV light, migrated with an approximate R F of 0.75, and was confirmed by comparison with the nonradioactive standard compound. The [ 3 H]TBZOH was extracted from the silica gel with five 1-ml aliquots of methanol and quantitated by scintillation counting. The resulting [ 3 H]TBZOH was chemically stable for use during at least 2 years when stored at Ϫ20°C.

H]TBZOH Binding and [ 3 H]Serotonin
Uptake Assays-Ligand binding curves were performed by incubation of 50 l of COS homogenate (generally 2-5 g/l protein by the Bradford protein assay) with 200 l of 300 mM sucrose, 10 mM Hepes (pH 7.2) buffer containing the [ 3 H]TBZOH. Curves were derived from binding measurements at four to seven different concentrations of [ 3 H]TBZOH, generally in the range of 1-50 nM. Specific binding was assessed by addition of excess nonradioactive TBZ (ϳ50 M). In addition to some nonspecific or non-TBZprotectable binding, it was found that non-transfected COS cells themselves have a low level of "specific," or TBZ-protectable, [ 3 H]TBZOH binding. This was controlled for by performing parallel analysis of non-transfected or mock-transfected COS cells. In instances where this was not done (some of the many binding curves summarized in Fig. 1C), a calculated adjustment was made by subtracting an average COS cell specific binding background value (derived from multiple experiments) at each ligand concentration. Ligand binding at 30°C reached a maximum level by ϳ10 min (data not shown), but was generally allowed to proceed for Ͼ20 min. Monoamine uptake experiments were performed by incubation at 30°C of 50 l of COS homogenate in the presence or absence of ϳ50 M TBZ (5 M proton ionophore FCCP inhibited uptake to the same extent, data not shown) with 200 l of buffer (sucrose/ HEPES/protease inhibitors/ATP and salts) for final concentrations of 50 nM [ 3 H]serotonin, 5 mM ATP, 5 mM MgSO 4 , 4 mM KCl. From time-course studies, 3 min was considered to be in the initial linear range, and steady state accumulation was reached by times shortly after 5 min. Human VMAT2-containing vesicles with bound [ 3 H]TBZOH or accumulated [ 3 H]serotonin were collected on filters (usually Whatman GF/B, occasionally Millipore nitrocellulose/cellulose acetate) by vacuum filtration on Millipore 12-well filtration manifolds or on a Brandel M-48 BIT cell harvester. Ligand binding and uptake assays, which were vacuumfiltered on the Millipore manifolds, were performed in glass test tubes, and the test tubes were generally rinsed with 2 ϫ 4 ml of buffer (which was also vacuum-filtered). Assays filtered using the Brandel cell harvester were performed in 1.2-ml well volume polypropylene 96-well plates, and the wells were rinsed with 5 ϫ 1 ml of buffer. Radioactivity was quantitated by scintillation counting in a Packard model 1600CA or 2000CA scintillation counter.
Assessment of Relative Protein Expression Levels by Western Blot-Bradford protein assays were used to quantitate protein concentrations to ensure equal loading of COS homogenate for each construct in Western blots. Protein samples were dissolved in SDS sample buffer at a protein concentration of 0.1-0.5 g/l, and electrophoresed on SDSpolyacrylamide gel electrophoresis minigels (12%). Proteins were transferred to a 0.2-m nitrocellulose membrane, and nonspecific antibody binding sites were blocked by incubating the blot 1 h or longer in Superblock PBS (Pierce) ϩ 0.05% Tween 20. Anti-HA antibody was diluted 1:850 -1:1000 in 15 ml of Superblock ϩ 0.05% Tween and allowed to bind for 1 h or longer, followed by at least six washes of 5 min or longer with Tris-buffered saline, 0.05% Tween 20. The secondary antibody conjugate was goat anti-mouse horseradish peroxidase (Bio-Rad), diluted 1:50,000 -1:65,000 in 15 ml of Superblock, 0.05% Tween, and was allowed to bind for at least 1 h, followed by a second set of six 5-min washes in Tris-buffered saline, 0.05% Tween. Pierce ECL reagents were used following the manufacturer's recommendations, and then the blots were imaged using Hyperfilm ECL (Amersham Pharmacia Biotech).

Assessment of the Effect of Cysteine Derivatization on [ 3 H]TBZOH
Binding-Sulfhydryl modifying MTS reagents or N-ethylmaleimide (NEM) (added from concentrated solution stocks) were allowed to react with transfected COS cell vesicles (generally 0.10 -0.25 mg in a 50-l volume) containing WT human VMAT2 and cysteine replacement constructs at sulfhydryl-reagent concentrations of 0.5-5 mM, as indicated in the figures. Parallel reactions were performed on mock-transfected COS cells and were subtracted as background. Reaction occurred for 10 min at 30°C, in a volume of 50 l, after which the reaction was diluted 10-or 20-fold (to a volume of 500 or 1000 l) by addition of [ 3 H]TBZOHcontaining sucrose/HEPES buffer to terminate (or greatly reduce the rate of) any further cysteine modification reaction. Supporting this role for the dilution step, treatment with low concentrations of MTSEA (0.2 mM) for 10 -20 min at 30°C has essentially no effect on [ 3 H]TBZOH binding by WT human VMAT2 (data not shown). In an experiment assessing the ability of NEM to block reaction at the MTSEA primary site, pretreatment was performed sequentially with the two cysteinemodifying reagents (i.e. first 1.5 mM NEM, then 3 mM MTSEA) for 10 min each prior to dilution. Dilution with [ 3 H]TBZOH-containing buffer initiated the ligand binding portion of the experiment, which was for Ͼ15 min at 30°C, and was followed by filtration and scintillation counting. In experiments assessing the ability of 100 mM DTT, ␤-mercaptoethanol, or cysteamine to reverse the effects of MTSEA derivatization, 100 mM reducing agent was added to the COS vesicles with the [ 3 H]TBZOH-containing dilution buffer (after 10-min pretreatment of the COS vesicles with 2 mM MTSEA) and was present throughout the ligand binding incubation period. For the experiment demonstrating the requirement for sulfhydryl reactivity for MTSEA efficacy, the concentrated stock solution of MTSEA (25 mM) was incubated for 10 min at 30°C with an 8-fold molar excess of ␤-mercaptoethanol (200 mM) prior to addition to the homogenized COS vesicles. During MTSEA reaction with human VMAT2-containing COS vesicles, the MTSEA concentration was 2 mM, and the ␤-mercaptoethanol concentration was 16  For all conditions, a control experiment was performed with mocktransfected COS cells and subtracted as background. As with the other assays, results were normalized for the amount of COS homogenate protein, which for this set of experiments varied slightly between constructs in the range of 0.11-0.16 mg/assay. For the "unprotected" condition, MTSEA was preincubated with COS cell homogenate in a volume of 50 l, and after 10 min a 20-fold dilution was made with sucrose/HEPES buffer containing [ 3 H]TBZOH (final concentration of [ 3 H]TBZOH ϳ50 nM). For the "protected" condition, [ 3 H]TBZOH was added 5 min prior to addition of MTSEA (the [ 3 H]TBZOH concentration at the time of MTSEA incubation was 500 nM), and after the 10-min MTSEA incubation, 20-fold dilution was made without adding any additional radioligand, for a final [ 3 H]TBZOH concentration of 25 nM. Incubation was performed for Ͼ20 min at 30°C, followed by filtration and scintillation counting. Specific binding was assessed by addition of excess non-radioactive TBZ to a final concentration of ϳ50 M, and specific binding was easily measurable with similar results whether the non-radioactive TBZ (50 M) was added to the "protected" condition before or after addition of the 500 nM [ 3 H]TBZOH.
Statistical Analysis-Graphing, non-linear curve fitting (K d and B max calculations), linear regressions for protein assays, and statistical analysis were performed using the computer program Prism (GraphPad, San Diego, CA). Statistical significance of observed effects was assessed based on scintillation counting data using this program by a two-tailed, paired t test and is indicated in the figures as follows: *, p Ͻ 0.1; **, p Ͻ 0.01; ***, p Յ 0.001.

Replacement of Individual Cysteines, or All Non-TM Cysteines, Has Little Effect on [ 3 H]TBZOH Binding-
The positions of the 10 native cysteines in human VMAT2 are indicated on a diagram of the proposed secondary structure (Fig. 1A). As shorthand throughout this paper, the 10 cysteines are distinguished with numbers and bold text (Cys 1 -Cys 10 ) based on the order in which they appear in the protein sequence from the N terminus to C terminus. Their actual positions in the human VMAT2 primary amino acid sequence are (indicated in parentheses): Cys 1 (residue 126), Cys 2 (176), Cys 3 (207), Cys 4 (311), Cys 5 (333), Cys 6 (369), Cys 7 (383), Cys 8 (439), Cys 9 (476), and Cys 10 (497). The effect of mutation of each individual cysteine on [ 3 H]TBZOH binding was assessed (Fig. 1C). To control for biological and other sources of variability between sets of binding curves, a complete binding curve for WT was measured on each and every occasion. Compilation of values from all WT curves indicates a K d of 12.5 Ϯ 2.3 nM (or Ϯ18.4%, 95% confidence interval). Comparison was made between K d values for cysteine replacement constructs and WT human VMAT2 within each set of experiments. As seen in Fig. 2B, replacement of any individual cysteine, or replacement of all non-TM cysteines together (Cys 1 , Cys 5 , Cys 9 , and Cys 10 to Ser (C1.5.9.10S)), had no more than a modest effect on binding affinity. Beyond this general and qualitative conclusion, it is interesting to note the cysteine that, when replaced with alanine, consistently resulted in the largest decrease in binding affinity (2.5-fold increase over WT K d ) was Cys 3 in putative TM 4.
A Cysteine-less Human VMAT2 Is Expressed but Not Functional-Although no individual cysteine was found critical for [ 3 H]TBZOH binding, a cysteine-less human VMAT2 (in which all the cysteines were replaced by serines) does not bind [ 3 H]TBZOH ( Fig. 2A) or transport [ 3 H]serotonin (data not shown) above the low level of mock-transfected COS-7 cells. Three possible explanations for this finding are: 1) that there is a lack of protein expression, 2) that there is a requirement for a subset of cysteines in combination, or 3) that serine is a poor replacement at one or more positions, and that a cysteine-less construct with different cysteine replacements might be func-tional. To address the question of whether total lack of ligand binding and monoamine uptake was a result of lack of protein expression, we prepared human VMAT2 constructs with epitope tags for Western blotting. The four consensus glycosylation sequences present in the large lumenal loop were mutated, and an HA epitope tag was inserted into human VMAT2 (described under "Experimental Procedures" and Fig. 2B) at a site closely corresponding to the position of a readily detectable HA tag that had been reported previously (30). This set of mutations we designated "GϪHA," for "glycosylation-negative, HA epitope." (For human VMAT2 Western blotting, we found the anti-HA antibody/lumenal HA epitope combination much superior to the M2 anti-FLAG antibody/C-terminal FLAG epitope combination.) Comparison of WT with WT GϪHA for A, human VMAT2 is predicted to have 12 membrane-spanning domains, with the N and C termini on the cytoplasmic face of the vesicle membrane. All human VMAT2 constructs used in this study had C-terminal FLAG and 6-histidine epitope tags. This common epitope-tagged background without any additional amino acid changes is referred to throughout this paper as WT, for wild type. B, structure of tetrabenazine. C, K d values were obtained from ligand binding curves (as described under "Experimental Procedures"), and are plotted as percentage of the WT human VMAT2 K d from a control binding curve in every set of experiments (Ͼ20). Error bars show the S.E. of values from multiple binding curves (indicated in parentheses). Double cysteine to serine replacements included C1.5S and C9.10S, and C1.5.9.10S was a quadruple cysteine to serine replacement mutant. The cysteine replacement that reproducibly showed the largest decrease in relative [ 3 H]TBZOH binding affinity (2.5-fold) was Cys 3 to alanine in predicted TM 4. H]serotonin transport (data not shown) above the level of mock-transfected COS-7 cells. B, to facilitate detection by Western blot, the four consensus glycosylation sequences were mutated and an HA epitope tag was added, as described under "Experimental Procedures." This set of mutations was designated GϪHA. C, Western blots of cell homogenates from COS cells transfected with the WT GϪHA and Cys-less GϪHA constructs showed equivalent levels of human VMAT2 GϪHA immunoreactivity. An equal amount of transfected COS homogenate (2 g, assessed by the Bradford protein assay) was loaded for each lane. These data demonstrated that the total lack of ligand binding or transport by Cys-less human VMAT2 was not due to significant differences in protein expression. Wild type COS cells showed no immunoreactivity at the position of GϪHA human VMAT2, and very little HA immunoreactivity overall (data not shown).

FIG. 2. A cysteine-less human
the WT GϪHA and Cys-less GϪHA constructs (Fig. 2C) show equivalent levels of immunoreactivity. These data demonstrate that the total lack of ligand binding or transport by Cys-less human VMAT2 is not due to significant differences in protein expression. (Wild type COS cells showed no immunoreactivity at the position of GϪHA human VMAT2 and very little HA immunoreactivity overall.) Therefore, having ruled out protein expression as responsible for lack of detectable function, we focused on identifying cysteine subsets that would restore the function that was lost in the cysteine-less VMAT2.
Subsets of Cysteines That Support Human VMAT2 [ 3 H]TBZOH Binding-As noted in Fig. 1C, replacement of all the non-TM cysteines with serines did not significantly affect K d for [ 3 H]TBZOH binding. This construct (designated C1.5.9.10S), also had essentially WT B max (Fig. 3B) and expression (Western blot, Fig. 3C). At this time, useful information regarding TM cysteines came from the sequence of a recently discovered Caenorhabditis elegans VMAT contain-ing only three cysteines, all in predicted TM segments (32). The number and position of the C. elegans VMAT cysteines suggested a cysteine combination that might restore function to the cysteine-less human VMAT2 background. As seen in Fig. 3A, amino acid replacements in C. elegans VMAT at the corresponding positions of cysteines in human VMAT2 include a threonine (Cys 4 ), a glycine (Cys 5 ), a methionine (Cys 9 ), and four serines (Cys 1 , Cys 6 , Cys 7 , and Cys 10 ). When the human VMAT2 construct (designated ϩ2.3.8, see Fig. 3A) which has only these three "C. elegans" cysteines was assayed for functional activity, significant [ Fig. 4 (A and B)). Like cysteine-less (GϪHA) and C1.5.9.10S (GϪHA) human VMAT2, ϩ2.3.8 (GϪHA) was expressed at a level equivalent to WT as assessed by Western blot (Fig. 3C), leading to the general conclusion that replacement of 7 or all 10 cysteines in VMAT2 had little effect on total expression levels. One possible explanation for how ϩ2.3.8 could have a WT K d , with a decrease in B max but wild type levels of Western blot immunoreactivity, is a change in the ratio of expression of native transporters versus non-native transporters (which do not bind [ 3 H]T-BZOH with high affinity).
To test whether the combination of Cys 2 , Cys 3 , and Cys 8 is necessary as well as sufficient for ligand binding and transport, we replaced this set of the three C. elegans cysteines in WT human VMAT2 and tested [ 3 H]TBZOH binding and [ 3 H]serotonin uptake. Consistent with our initial observation that no individual cysteine is critical for function, this combination of three cysteines together is also non-essential for function. This construct, designated ϩ1.4.5.6.7.9.10 (see Fig. 3A), was found to have a high affinity [ 3 H]TBZOH binding K d which was nearly identical to both WT and ϩ2.3.8 (not more than a modest 2-fold change, Fig. 3B). Although replacement of the non-TM cysteines does not affect the number of high affinity binding sites (B max ), replacement of either set of three TM cysteines (Cys 4 , Cys 6 , and Cys 7 , in ϩ2.3.8, or Cys 2 , Cys 3 , and Cys 8 in ϩ1.4.5.6.7.9.10) leads to a similar 75% reduction in B max (Fig. 3B). Presumably, an alteration in the ratio of native transporters versus non-native transporters (which do not bind [ 3 H]TBZOH with high affinity) similar to the situation for ϩ2.3.8 also occurs for the ϩ1.4.5.6.7.9.10 construct. This effect may suggest a small but additive role for individual cysteines in TM segments in preserving the ligand binding competence and structural integrity of the transporter.

Comparison of [ 3 H]Serotonin Uptake between WT and Cysteine Replacement
Mutants-Serotonin uptake time course experiments were performed comparing TBZ-protectable [ 3 H]serotonin-specific uptake for WT, ϩ2.3.8, and ϩ1.4.5.6.7.9.10 human VMAT2 constructs (Fig. 4A). When normalized for the number of [ 3 H]TBZOH binding sites (to control for the reduction in high-affinity ligand binding transporters), WT and ϩ2.3.8 accumulate serotonin at the same rate. However, ϩ1.4.5.6.7.9.10 accumulates serotonin at a much faster rate than WT. In terms of total [ 3 H]serotonin accumulated, WT and ϩ1.4.5.6.7.9.10 are nearly equal, although ϩ1.4.5.6.7.9.10 has only one-fourth the number of native transporters (measured as high affinity [ 3 H]TBZOH sites). Since specific uptake is defined as TBZ-protectable uptake, if there were to be any VMAT molecules that transport but that do not bind TBZ, they would not contribute to the measured values, but would be subtracted with the nonspecific background. The human VMAT2 construct ϩ1.4.5.6.7.9.10 therefore appears to transport [ 3 H]serotonin significantly more efficiently than WT. Uptake for ϩ2.3.8 and ϩ1.4.5.6.7.9.10 relative to WT was reproducible in multiple experiments (Fig. 4B). To investigate whether one of the three cysteine replacements (Cys 2 , Cys 3 , or Cys 8 ) was primarily responsible for enhancement of uptake by the ϩ1.4.5.6.7.9.10 construct, the levels of TBZ-protectable [ 3 H]serotonin uptake/[ 3 H]TBZOH binding sites for the single cysteine mutants C2S, C3A, and C8A were compared with WT, ϩ2.3.8, and ϩ1.4.5.6.7.9.10 (Fig. 4C). From this analysis, it appears that enhancement of uptake is not due primarily to replacement of any one of these three cysteines, but is more likely an effect of the combination of replacement of these three cysteines together. The molecular differences between WT and ϩ1.4.5.6.7.9.10 are only three atoms; replacement of two sulfur atoms (at Cys 2 and Cys 3 ) with oxygen atoms, and the removal of a third sulfur atom (at Cys 8 ). Although details regarding the mechanism of this enhancement of [ 3 H]serotonin transport remain to be elucidated, residues near these positions may interact functionally in the transport process and define a portion of the transport channel.
[ 3 H]TBZOH binding by WT is more sensitive (statistically significant) to MTSEA than any of the other constructs, and ϩ1.4.5.6.7.9.10 is much less sensitive than any of the other constructs. This suggested that MTSEA sensitivity has two unequal "components," or sets of cysteine targets. In ϩ1.4.5.6.7.9.10, the primary MTSEA target (responsible for ϳ80% of binding inhibition) appears absent but the secondary MTSEA target (responsible for ϳ20% of binding inhibition) is present. In C1.5.9.10S and ϩ2.3.8, the primary target is present but the secondary target is absent. In WT, both the primary and secondary cysteine targets are present, conferring maximal sensitivity. From the data in Fig. 5A, it appears that the primary MTSEA target (with respect to [ 3 H]TBZOH binding inhibition) is one or more of Cys 2 , Cys 3 , or Cys 8 , and that the secondary target is one or more of the non-transmembrane cysteines Cys 1 , Cys 5 , Cys 9 , and Cys 10 .
To probe the environment in which these target cysteines reside, the ability of additional cysteine-modifying reagents of differing charge and hydrophobicity (Fig. 5B) to inhibit [ 3 H]TBZOH binding by WT human VMAT2 was examined. Three methanethiosulfonate derivatives (MTSEA, MTSET, and MTSES) and NEM were used in this study. MTSEA can be either charged or uncharged, depending upon pH and environment, and is considered membrane-permeable. MTSET is bulkier and carries a permanent positive charge, whereas MTSES carries a permanent negative charge. Both MTSET and MT-SES are considered membrane-impermeable. NEM is uncharged but primarily reacts with cysteines in an aqueous (non-hydrophobic) environment. As seen for WT human VMAT2 in Fig. 5C (set 1), MTSEA reaction at both primary and secondary targets results in Ͼ90% inhibition of [ 3 H]TBZOH binding by 3 mM MTSEA. MTSET (5 mM) has a much smaller but statistically significant effect (Fig. 5C, sets 2 and 3), with inhibition at a level consistent with and suggestive of modification at only the MTSEA secondary target (one or more of Cys 1 , Cys 5 , Cys 9 , and Cys 10 ). MTSET carries a permanent positive charge and is not expected to cross vesicle membranes. If COS-expressed VMAT2 molecules are correctly oriented with respect to sealed vesicle membranes (which we believe to be the case from proteolysis experiments; data not shown), only two of these cysteines (Cys 9 and Cys 10 ) would be predicted to be accessible to MTSET, whereas the others would be protected inside the vesicle lumen. NEM (2 mM) has no effect on [ 3 H]T-BZOH binding (Fig. 5C, set 4), nor can pretreatment with 1.5 mM NEM block subsequent inhibition of ligand binding by MTSEA (Fig. 5C, set 5). MTSES (3 mM) also has no effect on [ 3 H]TBZOH binding (Fig. 5C, set 6). Taken together, these data support the conclusion that the primary MTSEA target is in a hydrophobic and perhaps sterically hindered environment (i.e. in a transmembrane segment, as has been predicted for Cys 2 , Cys 3 , and Cys 8 ), that the MTSEA secondary sites are in hydrophilic regions of the transporter, and that the MTSET target appears to be a secondary MTSEA target on the cytoplasmic face of the vesicle membrane (perhaps Cys 9 and Cys 10 ). Since inhibition was seen by MTSEA and MTSET but not by MTSES or NEM, modification with a positively charged reagent at a secondary (with respect to MTSEA inhibition) cysteine target may be important for the observed effect of modest [ 3 H]TBZOH binding inhibition.
To investigate the relative contributions of Cys 2 , Cys 3 , and Cys 8 as the primary MTSEA targets, MTSEA pretreatment/ [ 3 H]TBZOH binding experiments were performed on the single cysteine mutants (Fig. 5D). Replacement of Cys 2 with serine or Cys 3 with alanine did not remove either the primary or secondary MTSEA targets, resulting in the WT level of ϳ90% ligand binding inhibition. In contrast, replacement of Cys 8 with alanine (net removal of a single sulfur atom from the entire transporter) resulted in loss of MTSEA sensitivity and recovery of ϳ80% of maximal [ (Fig. 6A), demonstrating the requirement for sulfhydryl reactivity for the MTSEA inhibition of [ 3 H]TBZOH binding.
We considered the possibility that a conformational change in VMAT2 structure in the presence of a proton electrochemical gradient might change the MTSEA reactivity of transporter cysteines (Fig. 6B). COS cell homogenates prepared as de- scribed under "Experimental Procedures" did not contain the levels of ATP/Mg 2ϩ /Cl Ϫ necessary to generate a proton electrochemical gradient and transport serotonin (data not shown). Therefore, to assess whether binding of [ 3 H]TBZOH would be either more or less sensitive to MTSEA in the presence of a proton electrochemical gradient (indicative of a conformational change involving MTSEA target residues), sensitivity to MT-SEA inhibition was assessed in the presence and absence of 5 mM ATP, 5 mM Mg 2ϩ , 4 mM Cl Ϫ (conditions used for uptake experiments). MTSEA sensitivity was found to be independent of the presence of a proton gradient; therefore, no change in cysteine accessibility in the presence of a proton gradient is indicated from this experiment.
To further probe the environment of the MTSEA primary target cysteine and secondary target cysteine or cysteines, we performed disulfide reversal experiments with the reducing agents ␤-mercaptoethanol, dithiothreitol (DTT), and cysteamine (Fig. 6, C and D). At concentrations as high as 100 mM, these three reducing agents themselves had little or no effect on [ 3 H]TBZOH binding. However, none of these reagents at this concentration reversed the effect of the MTSEA reaction corresponding to the primary target site (Cys 8 ). ␤-Mercapto-ethanol did show a statistically significant effect of reversal of MTSEA inhibition at a level consistent with reversal of modification at the secondary MTSEA target cysteine or cysteines (i.e. Cys 9 and Cys 10 ). Although the observed lack of disulfide reversal at the primary MTSEA target is somewhat surprising, the identity of the primary MTSEA target as Cys 8 is firmly established, and also the requirement for the highly cysteinereactive methanethiosulfonate leaving group for MTSEA inhibition of [ 3 H]TBZOH binding. The most likely explanation for the lack of observed reversal is that Cys 8 may be in a very hydrophobic (and perhaps sterically hindered) environment. This is consistent with the prediction for a very hydrophobic tetrabenazine binding site from a correlation study of the apparent affinity and hydrophobicity (over several orders of magnitude) for a series of tetrabenazine derivatives (33). After treatment of COS vesicles containing WT human VMAT2 with 2 mM MTSEA (as described under "Experimental Procedures"), incubation with 100 mM DTT for Ͼ15 min at 30°C had no reversal effect, whereas 100 mM ␤-mercaptoethanol (under the same conditions) had a small but statistically significant effect. The small effect of ␤-mercaptoethanol is consistent with sulfhydryl reversal at the secondary, but not the primary (Cys 8 ), MTSEA target site. D, the effect of adding ␤-mercaptoethanol (separate experiment from panel C) or cysteamine on recovery of [ 3 H]TBZOH binding after inhibition by MTSEA. After treatment of COS vesicles containing WT human VMAT2 with 2 mM MTSEA (as described under "Experimental Procedures"), incubation with 100 mM cysteamine for Ͼ15 min at 30°C had no reversal effect, whereas 100 mM ␤-mercaptoethanol (under the same conditions) had a small but statistically significant effect. The small effect of ␤-mercaptoethanol is consistent with sulfhydryl reversal at the secondary, but not the primary (Cys 8 ), MTSEA target site.
ognizing the significant technical challenges of completely removing a hydrophobic, high affinity ligand from a cell homogenate, a novel protection assay was developed. This assay utilized [ 3 H]TBZOH to both protect the tetrabenazine binding site during the MTSEA reaction, and as radioligand during the ligand binding experiment (described under "Experimental Procedures"). Consistent with an estimate of the TBZ dissociation half-life of less than 1 min, based on the common on-rate value for a small molecule ligand binding to a protein of 10 6 M Ϫ1 s Ϫ1 , and a K d of 12.5 nM (34), the ligand exchange rate is fast enough that, during the 20-min ligand binding incubation period, excess non-radioactive TBZ exchanges with bound [ 3 H]T-BZOH to allow detection of specific [ 3 H]TBZOH binding. Human VMAT2 constructs tested included WT, ϩ2.3.8, ϩ1.4.5.6.7.9.10, and the single cysteine replacements Cys 2 3 Ser, Cys 3 3 Ala, and Cys 8 3 Ala. As seen in Fig. 7 (A, B, D,  and E), MTSEA inhibition of [ 3 H]TBZOH binding by the four constructs that possessed the primary target cysteine (Cys 8 ) was significantly and reproducibly protected against MTSEA inhibition by the presence of [ 3 H]TBZOH during the MTSEA reaction. The amount of binding activity protected at 5 mM MTSEA was between 65% and 80% of maximal (compared with 10% or less binding in the unprotected condition), consistent with protection of the primary MTSEA target (Cys 8 ), but not the secondary MTSEA target cysteine(s). This supports a model in which the primary MTSEA target mediating [ 3 H]T-BZOH binding inhibition, Cys 8 (Cys 439 ) in TM 11, is at the TBZ binding site. In the two constructs that lacked Cys 8 (ϩ1.4.5.6.7.9.10, and the single cysteine replacement Cys 8 Ͼ Ala, Fig. 7 (C and F)) [ 3 H]TBZOH binding was insensitive to the primary effect of MTSEA, but was inhibited up to 20% by reaction at the secondary MTSEA target. The 20% effect of MTSEA inhibition mediated by the secondary cysteine tar-get(s) was not protected by [ 3 H]TBZOH, consistent with conformational effects rather than direct steric blocking at these sites. Although it is difficult to absolutely rule out the possibility that protection at Cys 8 may be due to a conformational effect of [ 3 H]TBZOH binding, the observed effect of protection is specific to Cys 8 and not shared with the secondary MTSEA target cysteines. Further, TBZ-protectable Cys 8 is positioned centrally with regard to previously available data from photoaffinity labeling (particularly compelling data for a direct effect), mutagenesis, and chimera analysis (as reviewed below under "Discussion"), clearly defining a region involved in TBZ binding.

DISCUSSION
In instances where x-ray crystallography or NMR studies have not yet been feasible (especially the case for the majority of integral membrane proteins), creative approaches to the study of structure, function, and ligand binding sites have been applied that take advantage of the unique chemical reactivity of native or engineered cysteines. An important prerequisite for the application of these methods is either a functional cysteineless background, or a background of cysteines that does not interfere with the particular assay (i.e. derivatization, crosslinking, etc.) being applied. We sought to apply such methods to the study of human VMAT2, which has 10 naturally occurring cysteines (Fig. 1A). We first examined the effect of replacement of human VMAT2 cysteines on [ 3 H]TBZOH binding. None of these cysteines individually are critical for human VMAT2 expression or ligand binding (Fig. 1C). However, although a cysteine-less VMAT (all 10 cysteines replaced with serines) is expressed at levels comparable to wild type (Fig. 2C), it neither binds [ 3 H]TBZOH ( Fig. 2A) nor transports serotonin (data not shown). Based on this observation, we sought to identify sub- sets of native cysteines that would allow high affinity ligand binding and monoamine transport and, at the same time, provide the essential non-reactive cysteine background for cysteine-based biochemical structure and function studies. The number and position of the C. elegans VMAT cysteines suggested a cysteine combination that might restore function to Cys-less human VMAT2. Indeed, ϩ2.3.8 human VMAT2 with a subset of only 3 (out of 10) cysteines, at the corresponding positions of the three cysteines in C. elegans VMAT (Fig. 3A), was found to have wild-type binding affinity (K d , Fig. 3B) and serotonin transport (uptake normalized to the number of [ 3 H]TBZOH binding sites, Fig. 4 (A and B)), with a reduction (by ϳ75%) in high affinity ligand binding sites (B max , Fig. 3B), but not significantly in total expression (assessed by Western blot, Fig. 3C). This subset of cysteines (present in C. elegans VMAT, and in human, rat, mouse, and Bos taurus VMAT2) is sufficient but not necessary to restore VMAT2 function. Replacement in wild type human VMAT2 of this set of three cysteines still resulted in a similarly functional transporter (near wild type [ 3 H]TBZOH binding K d , decreased B max , Fig.  3B). Compared with wild type, this latter construct (lacking the three transmembrane cysteines present in C. elegans VMAT, see Fig. 3A) was found to have two very interesting properties, i.e. significant enhancement of the rate of tetrabenazine-protectable serotonin transport (normalized to the number of [ 3 H]TBZOH binding sites, Fig. 4) and pronounced insensitivity of [ 3 H]TBZOH binding to inhibition by the thiol-modifying reagent MTSEA (methanethiosulfonate ethylamine, Fig. 5A). Cysteine derivatization experiments on a series of human VMAT2 cysteine replacement mutants allowed identification of Cys 8 in TM 11 (Cys 439 ) as the primary target mediating ϳ80% of the MTSEA inhibition of [ 3 H]TBZOH binding. Permanently charged or more hydrophilic sulfhydryl-reagents did not react at Cys 8 (Figs. 5 (C and D) and 6 (C and D)). A secondary component of MTSEA inhibition (responsible for ϳ20% of inhibition) is due to reaction at one or more of the non-TM cysteines Cys 1 , Cys 5 , Cys 9 , and Cys 10 (Fig. 5A). No significant difference in MTSEA inhibition of [ 3 H]TBZOH binding was detected in the presence or absence of a proton gradient (Fig. 6B). The secondary target cysteine(s) appeared to be accessible for reaction with the permanently charged sulfhydryl-modifying reagent MTSET (Fig. 5C) and ␤-mercaptoethanol (Fig. 6, C and D), but were not protected from MTSEA reaction by bound [ 3 H]TBZOH (Fig. 7). However, preincubation with 500 nM [ 3 H]TBZOH blocked modification at Cys 8 (Cys 439 in TM 11) by a 10,000-fold molar excess of MTSEA. The most straightforward explanation of this result is that Cys 439 , the primary target mediating MTSEA inhibition of [ 3 H]TBZOH binding, is positioned at the tetrabenazine binding site.
Previously available evidence regarding the location of the TBZ binding site of VMAT2 has been obtained from four lines of experimentation: estimation of the hydrophobicity of the TBZ binding site (33), photoaffinity labeling (21), site-directed mutagenesis (15), and analysis of VMAT2/VMAT1 chimeras (16 -19). The earliest of these studies utilized VMAT present in bovine chromaffin granules, and a series of five transporter substrates and seven TBZ derivatives of differing physicochemical properties to examine the hydrophobicity of the TBZ binding site on VMAT2. The apparent partition coefficient between octanol and buffer, which correlated with measurements of membrane bilayer/water partitioning, was used as a measure of drug hydrophobicity. Over a range of 5 orders of magnitude, ability to displace [ 3 H]TBZOH binding by the TBZ derivatives or transport substrates correlated well with hydrophobicity as estimated from octanol/water partition coefficients. The authors suggested that this correlation should be interpreted as a demonstration that bound TBZ (or derivatives) is in equilibrium with TBZ concentrated in the lipid membrane, and that TBZ binds in a hydrophobic site, presumably within the TM segments (33). The photolabeling, mutagenesis, and chimera experiments utilized cloned VMAT2 and VMAT1 isoforms from rat. Photoaffinity labeling studies from our laboratory (21) using the TBZ photolabel [ 125 I]azidoiodophenylpropionyl tetrabenazine ([ 125 I]TBZ-AIPP) demonstrated that the majority of labeling (60%) occurred on a small peptide containing the cytoplasmic loop between TM10 and TM11, and most of TM11. Significantly, there was some [ 125 I]TBZ-AIPP photoinsertion (40%) into an N-terminal peptide that contained lysine 20, which is the predominant (Ͼ90%) site of photoinsertion by the ketanserin photolabel [ 125 I]7-azido-8-iodoketanserin at the predicted TM 1 membrane/cytoplasm interface. [ 125 I]TBZ-AIPP photolabeling of two peptides distant in primary sequence suggests that these portions of the transporter may be juxtaposed in the folded protein structure. This is also suggested by identification of a functional ion pair between charged residues in TM 2 and TM 11 (20). VMAT1 isoforms have a significantly lower affinity for TBZ than VMAT2 isoforms. Mutagenesis studies of aspartate residues in transmembrane segments of rat VMAT1 identified an aspartate in TM10 (also an aspartate in VMAT2 isoforms), which, when mutated to glutamate, resulted in gain of a measure of TBZ binding affinity (15). Analysis of transporter chimeras between rat VMAT1 and VMAT2 allowed identification first of domains that contributed to VMAT2 higher affinity TBZ binding (TM 5-8 and TM 9 -12; Ref. 16), and then of individual amino acids (17,18) in VMAT2, which, when replaced by the VMAT1 residue, led to an incremental, additive transition from the higher VMAT2 tetrabenazine affinity to the lower VMAT1 affinity. Similarly, analysis of human VMAT1/VMAT2 chimeras identified VMAT1 domains, which, when substituted into a VMAT2 background, led to a large reduction in TBZ affinity, manifest by a large increase in IC 50 for transport inhibition (19).
Cys 439 in TM 11, discovered through this work to be the primary, TBZ-protectable target mediating MTSEA inhibition of human VMAT2 [ 3 H]TBZOH binding, is positioned centrally with regard to almost all previously available information on the tetrabenazine binding site (Fig. 8). Cys 439 is positioned in a hydrophobic environment (as predicted for the tetrabenazine binding site; Ref. 33), in the center of four mutations previously found to affect tetrabenazine binding (15,17,18), on a peptide that is photolabeled by a tetrabenazine photolabel (21), and on a transporter domain involved in TBZ binding identified from rat (16) and human (19) VMAT1/VMAT2 chimera studies. The cysteine derivatization and protection experiments described in this report support these earlier findings and significantly strengthen them against argument of the inherent uncertainties in interpreting site-directed mutagenesis data, or the possibility of photolabel insertion at a site somewhat distant from pharmacophore binding. Multiple independent experimental methods all identify the same transporter domain, converging on a tetrabenazine binding site for VMAT2, which directly involves TM 10 -12, with important contributions from the N-terminal portion of the molecule (TM 1), and the lumenal loop between TM 7 and TM 8. Since human VMAT2 with a cysteine 439 to alanine replacement is both functional and insensitive to MTSEA, this now opens the possibility of future detailed mapping studies of this and other VMAT2 ligand binding sites by derivatization and protection experiments of "engineered" cysteines at different locations throughout the transporter which may be distant in primary sequence, but nearby in tertiary structure at the TBZ binding site.
Secondary MTSEA targets responsible for a small percentage of the ligand binding inhibition were discovered to be one or a combination of the four non-TM cysteines Cys 1 , Cys 5 , Cys 9 , and Cys 10 . Two of these, Cys 5 or Cys 9 , are positioned near mutations previously found to affect TBZ binding (Fig. 8). The position of Cys 10 on the C-terminal tail in the folded protein structure is unknown, but it may be near the N terminus of the protein, which is photolabeled by [ 125 I]TBZ-AIPP. For membrane-impermeable MTSET, only the cytoplasmic cysteines Cys 9 and Cys 10 seem likely targets responsible for the modest amount of [ 3 H]TBZOH binding inhibition seen after treatment with this reagent, suggesting that the MTSET target (and perhaps the secondary MTSEA target as well) may be one or more of the two cytoplasmic cysteines, Cys 9 and Cys 10 .
Although future studies will be required to elucidate details of the mechanism of enhancement of TBZ-protectable [ 3 H]serotonin uptake per [ 3 H]TBZOH binding sites by replacement of three cysteines (net removal/replacement of only three atoms) in the ϩ1.4.5.6.7.9.10 human VMAT2 transporter, the suggestion that residues near these positions may interact functionally or physically as components of a transport channel is supported by information from previous VMAT mutagenesis studies. Transport of monoamines presumably entails interaction of amine nitrogens (both charged, primary amines, and the aromatic ring nitrogens present in serotonin and histamine) with TM charged or aromatic amino acid residues, and hydroxyl group interactions with one or more serine residues. Candidate serine residues were identified by the observation that replacement of three serines near the membrane/lumen interface of TM 3 with alanines eliminates [ 3 H]serotonin transport by rat VMAT2 (14). Interestingly, Cys 2 and Cys 3 are located on TM 3 and TM 4, respectively, with this set of three serines positioned between them. Mutation of rat VMAT2 Tyr 434 in TM 11 (only 3 amino acids away from the primary MTSEA target Cys 8 ) and Asp 461 in TM 12 appear to affect recognition of aromatic ring nitrogens, since transport of both serotonin and histamine is inhibited, but not transport of dopamine (17). Taken together, these mutagenesis data suggest the near proximity of the TM 3/TM 4 loop with TM 11, and by inference, the near proximity of Cys 2 , Cys 3 , and Cys 8 and the TM segments on which they reside.
An interesting comparison can be drawn between VMAT and the vesicle acetylcholine transporter (VAChT), which share a degree of sequence and functional similarity. Many aspects of this similarity have recently been reviewed (35). Identification in human VMAT2 of two distinct sets of cysteine targets mediating inhibition of ligand binding by cysteine-modifying reagents demonstrates yet another similarity between VMAT and VAChT. Although the positions of the 8 cysteines in Torpedo californica VAChT are generally not the same as the 10 cysteines in human VMAT2, two classes of cysteine targets mediating inhibition of vesamicol binding, with differing sensitivities to different sulfhydryl reagents, were identified. Although the precise identities of the cysteines in these two classes were not reported, the authors presented a logical argument that VAChT Cys 376 at the membrane/lumen interface of TM 10 was the likely candidate for cysteine class-2, deep in the transporter (36). This would be positioned at essentially the same level, one TM segment over from human VMAT2 Cys 8 in TM 11. If the authors' reasoning is correct that VAChT Cys 376 is the reactive "class-2" cysteine (which supports the data in this paper), this may suggest a similar manner by which the high affinity inhibitors tetrabenazine and vesamicol bind to and inhibit their respective transporter targets.
In conclusion, we have demonstrated by replacement, derivatization, and protection experiments that certain cysteines are in important positions for human VMAT2 [ 3 H]TBZOH binding (Cys 8 , Cys 9 , and/or Cys 10 , possibly Cys 1 and/or Cys 5 ) and [ 3 H]serotonin transport (Cys 2 , Cys 3 , and Cys 8 together). The cysteine replacement constructs developed as part of this research will prove useful in future cysteine-based studies of the structure, function, and ligand binding sites of VMAT2. In particular, since human VMAT2 with a cysteine 439 to alanine replacement is both functional and insensitive to MTSEA, this now opens the possibility of future detailed mapping studies of this and other VMAT2 ligand binding sites by placement and chemical modification of "engineered" cysteines at different locations throughout the transporter.