Ligand Effects on Cross-linking Support a Conformational Mechanism for Serotonin Transport*

Serotonin transporter (SERT) is responsible for the re-uptake of 5-hydroxytryptamine (5-HT) from the synaptic cleft after release from serotonergic neurons. We show here that cysteine residues at positions in transmembranes 1 and 3 of SERT, like the corresponding positions in the γ-aminobutyric acid transporter, can be cross-linked using copper(II)(1,10-phenanthroline)3. The presence of a cross-link was detected by a novel methionine mutagenesis strategy. A change in mobility for an N-terminal cyanogen bromide fragment accompanied disulfide cross-linking of the two cysteine residues. Cross-linking also inhibited transport, and this process was blocked by cocaine, which is expected to stabilize SERT in conformations where the two positions are separated, but cocaine did not decrease accessibility of either of the two cysteines to modification by 2-aminoethyl methanethiosulfonate. Cysteine was required at both positions on the same molecule for efficient cross-linking, indicating that the reaction was intramolecular.

The neurotransmitter:sodium symporter family (also known as SLC6) contains transporters for many neurotransmitters including serotonin (5-HT), 2 ␥-aminobutyric acid, dopamine, norepinephrine, and glycine (1). It also contains hundreds of bacterial and archaeal proteins including the amino acid transporters TnaT, Tyt1, and LeuT (2)(3)(4). Several crystal structures for LeuT in complex with amino acids and inhibitors have been published (3,(5)(6)(7). Crystal structures of transporters from several other families that were previously thought to be unrelated to the neurotransmitter:sodium symporter family were observed to adopt a very similar structure to that of LeuT (8 -11).
In the LeuT structures, especially in complex with the competitive inhibitor tryptophan (7), a permeation pathway is visible leading from the extracellular medium toward the site of amino acid and Na ϩ binding. However, none of the LeuT structures indicates the location of the permeation pathway leading from the binding site to the cytoplasm.
Studies employing cysteine scanning mutagenesis of serotonin transporter (SERT) identified a potential permeation pathway leading from the binding site to the cytoplasm, composed of TMs 1, 5, 6, and 8 (12)(13). This pathway is apparently accessible in a cytoplasm-facing conformation of SERT. Accessibility of reactive cysteine residues in this cytoplasmic pathway was modulated by 5-HT and SERT inhibitors (12,14). Cocaine, a competitive inhibitor of SERT, norepinephrine, and dopamine transporters, decreased accessibility of the cytoplasmic pathway residues (12-13) but increased reactivity of some positions in the extracellular pathway (15)(16). 5-HT and the noncompetitive inhibitor ibogaine increased accessibility in the cytoplasmic pathway and decreased it in the extracellular pathway (12)(13)(14).
We generated a model of LeuT in a cytoplasm-facing conformation that was consistent with accessibility information from the cysteine scanning studies (13). LeuT could interconvert between conformations represented by the crystal structures and the cytoplasm-facing model by tilting or rocking of a 4-helix bundle composed of TMs 1, 2, 6, and 7 within a scaffold formed by the rest of the protein. We proposed that this "rocking bundle" mechanism could account for alternating access in the neurotransmitter: sodium symporter transporter family (13).
These accessibility changes are consistent with cocaine stabilizing conformations of SERT similar to those of the LeuT crystal structures (with the bundle tilted so as to open the extracellular pathway and close the cytoplasmic pathway). A model was recently proposed for cocaine binding to dopamine transporter in a conformation similar to that of the occluded LeuT crystal structure (17). Ibogaine is expected to stabilize a conformation more like the cytoplasm-facing model of LeuT (with the bundle tilted so as to close the extracellular pathway and open the cytoplasmic one).
Information about proximity between residues within a polypeptide can be obtained by site-directed cross-linking experiments (18 -21). In the homologous ␥-aminobutyric acid transporter GAT-1, an interaction between TM1 and TM3 was proposed based on paired cysteine mutagenesis and crosslinking (22). In a mutant containing cysteines at positions 68 and 143, transport was sensitive to inhibition by Cd 2ϩ and copper(II)(1,10-phenanthroline) 3 (CuPh 3 ), which can chelate or cross-link, respectively, two cysteine thiol groups. Cross-linking by CuPh 3 was not shown chemically but only by inhibition of transport activity (22).
Detection of cross-linking events in membrane proteins has been challenging. In other systems, where disulfide cross-links were formed between polypeptide chains, detection by a change in mobility on SDS-PAGE was sufficient (21). Alterna-tively, protease-sensitive sites have been inserted into proteins to allow cleavage into peptides that would separate on SDS-PAGE unless cross-linked (23). However, to be useful, these sites must be inserted in locations where they are accessible to the specific protease and also do not interfere with protein function, requirements that may be difficult to satisfy. In this study, we employed a novel strategy of methionine mutagenesis to generate specific cleavage sites for cyanogen bromide.
According to our analysis of LeuT (13), cysteine crosslinking between Cys-68 and Cys-143 in GAT-1 represented disulfide formation between the 4-helix bundle and the scaffold comprising the remainder of the protein. We reasoned that the distance between these positions might change with the bundle orientation, and might be affected by binding of specific ligands that affect transporter conformation. Accordingly, we tested the corresponding positions in SERT, where cocaine and ibogaine are known to have different effects on protein conformation.
Mutagenesis-All rSERT mutants were generated by using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). After introduction of the desired mutations into pBluescript II SK(Ϫ) containing rSERT, the mutated region was excised by double digestion with suitable restriction endonucleases and subcloned individually back into the original pBluescript-rSERT through the same restriction sites. All mutants contained the C109A mutation and a c-myc epitope tag at the N terminus of the protein (residue numbers are for un-tagged rSERT). The C109A mutation was used because this single cysteine is the only residue that reacted with MTS (methanethiosulfonate) reagents applied to the cell exterior (25). X3M denotes M1V/M86V/M124V/L320M/C109A, with the first three methionines in the rSERT sequence changed to valine to remove all sites for CNBr cleavage in the N-terminal half of the protein except for Met-135 (see Fig. 1). This background facilitated analysis of CNBr cleavage products. W103C, I179C, and X3M2C designate X3M/W103C, X3M/I179C, and X3M/W103C/I179C, respectively. Leu-320 was replaced with a methionine to introduce a CNBr cleavage site in the third extracellular loop (between TM5 and TM6). The transport activity and sensitivity of these mutants to inhibitors is shown in Table  1. All mutations were screened by restriction mapping and confirmed by DNA sequencing.
Cells plated in 96-well culture plates (ϳ25,000 cells per well) were infected with recombinant VTF7-3 virus and transfected with plasmid bearing rSERT C109A or other mutant cDNA under control of the T7 promoter (24). Transfected cells were incubated for ϳ24 h at 37°C with 5% CO 2 before assaying transport. Transport Inhibition by Copper(II)(1,10-Phenanthroline) 3 and Effect of Inhibitors-To measure inhibition by CuPh 3 , HeLa cells transfected with the indicated constructs were washed once with 20°C PBSCM and then incubated with CuPh 3 for the indicated time. Then the cells were washed three times with PBSCM at 20°C before measuring transport. To measure the effect of inhibitors on CuPh 3 modification, cells were preincubated with the indicated concentration of cocaine or ibogaine for 10 min before incubating with CuPh 3 for the indicated time. Cells were then washed three times with PBSCM containing the same inhibitor concentration to remove CuPh 3 and then washed three times with PBSCM alone to remove the inhibitor before measuring transport. The CuPh 3 stock solution was freshly prepared for each experiment by Preparation of Lyophilized Protein Sample-HeLa cells expressing rSERT mutants were treated, where indicated, with CuPh 3 as described for transport inhibition experiments. Cells were washed once with ice-cold PBSCM, and then scraped into an appropriate volume of cell collecting buffer (20 mM Tris, 5 mM EDTA, pH 7.3) containing protease inhibitor mixture (Sigma) and transferred to a 1.5-ml tube. Cells were lysed by three cycles of freezing at Ϫ80°C and thawing at 20°C and then passed through a 26-gauge needle 10 times. Cell debris was removed by precipitation at 500 ϫ g for 10 min at 4°C and the membranes in the supernatant suspension were retained for analysis. Protein concentration was determined with the Micro BCA protein assay (Pierce). Protein samples containing ϳ50 g were placed in 1.5-ml tubes, evaporated to dryness in a Speed-Vac evaporator, and stored at Ϫ20°C until use.

5-HT Transport
Chemical Cleavage of rSERT by CNBr-For the cleavage reaction, 15 l of water was added to each lyophilized rSERT protein sample followed by 85 l of a saturated solution of CNBr in 88% formic acid (26). The sample was agitated in a Vortex mixer until the protein was completely solubilized in the CNBr solution. Samples were incubated 4 h in the dark at 20°C. Addition of 500 l of benzene to the sample tube allowed formic acid to evaporate as an azeotrope as the sample dried in a SpeedVac evaporator. For more complete removal of formic acid, an additional 500 l of benzene was added and the sample dried again. Cleavage products were dissolved in 50 l of Laemmli sample buffer (27) with or without dithiothreitol (DTT), and the samples were stored at Ϫ20°C until use.
Data Analysis-Nonlinear regression fits of experimental and calculated data were performed with Origin (OriginLab, Northampton, MA), which uses the Marquardt-Levenberg nonlinear least squares curve-fitting algorithm. The statistical analyses are from triplicate experiments. Data with error bars represent the mean Ϯ S.E. for triplicate measurements. Statistical analysis was performed using Student's paired t tests. Rate constants (k) for 2-aminoethylmethane thiosulfonate (MTSEA) modification were calculated from the concentration of MTSEA (IC 50 ) that led to half-maximal inactivation in 10 min (t1 ⁄ 2 ) using the equation: k ϭ ln(2)/(IC 50 ϫ t1 ⁄ 2 ).

Effect of CuPh 3 on SERT Mutants-In previous work with
GAT-1, Zomot et al. (22)showed that replacing Trp-68 in TM1 and Ile-143 in TM3 with cysteine caused the transporter to become sensitive to inactivation by CuPh 3 , which they interpreted as evidence for cross-linking the two cysteine sulfhydryl groups. We reasoned that if the structures of SERT and GAT-1 were sufficiently similar, introduction of cysteines at the corresponding positions of SERT would create a similar sensitivity to CuPh 3 . Accordingly, we mutated Trp-103 in TM1 and Ile-179 in TM3 in rSERT, to cysteine. Mutants with each of these mutations separately and with both together were created in the X3M/C109A background (see "Experimental Procedures" for details) to make single or double cysteine replacement mutants. Table 1 shows that these mutants all transported 5-HT at ϳ70% the rate of SERT C109A (which itself was similar to wild-type SERT (25)). As shown in Fig. 2, CuPh 3 strongly inhibited the 5-HT influx activity of HeLa cells transfected with the double cysteine replacement mutant X3M2C (filled circles). This inhibition was half-maximal at 0.24 Ϯ 0.06 M and saturated at ϳ70% inhibition at the highest CuPh 3 concentrations tested (5-10 M). For rSERT X3M (triangles) or C109A (on which it is based, filled squares), or I179C (containing one of the two cysteines, open circles), CuPh 3 had no effect. CuPh 3 also weakly inhibited W103C (the mutant with the other inserted cysteine, open squares). The IC 50 for CuPh 3 inhibition of W103C was 2.9 Ϯ 0.01 M, over 10-fold higher than that of X3M2C. The activity was decreased 40% by incubation with 10 M CuPh 3.

TABLE 1
Relative uptake of 5-HT and IC 50 of pseudo-wild type (C109A) and mutants for cocaine, ibogaine, and CuPh 3 5-HT influx was measured as described under "Experimental Procedures" in the absence of inhibitors and after incubation with a range of concentrations of each agent for 10 min at 20°C. Half-maximal inhibition was determined from fitting the resulting rate versus concentration curves. The rate for C109A was 0.53 Ϯ 0.1 pmol min Ϫ1 mg Ϫ1 .

Mutants
Relative These data are similar to the corresponding results with GAT-1, indicating a similar structural relationship between the two positions in both transporters. The weak sensitivity of W103C also corresponds to the sensitivity of GAT-1 W68C to CuPh 3 . CNBr Cleavage of rSERT Mutants-It has been assumed that inhibition by CuPh 3 represents oxidation of the two cysteine residues to a cystine disulfide (18). To verify that this was indeed the case, we took advantage of the X3M background to measure disulfide linking of TM1 with other parts of SERT using CNBr cleavage. In X3M, and mutants generated in that background, the first methionine after the N-terminal c-myc tag is Met-135, near the cytoplasmic end of TM2, and the next endogenous methionine is in TM7 (see "Experimental Procedures" for mutant descriptions). CNBr is expected to generate a 14.9-kDa fragment containing the c-myc tag by cleaving these mutants at Met-135 (Fig. 1). Fig. 3 shows that a myc-positive fragment of approximately this mass was generated from CNBr cleavage of X3M. The pattern was not significantly affected by treatment with CuPh 3 or DTT. CNBr treatment of I179C and W103C, expressed together (Fig. 3) or separately (supplemental Fig.  S1), also yielded a single N-terminal product at 15-20 kDa that was not affected by CuPh 3 or DTT. A faint band at ϳ30 kDa was occasionally observed with W103C, especially if the cells expressing this mutant were treated with CuPh 3 before digestion (supplemental Fig. S1). DTT added after CNBr cleavage eliminated this band, suggesting that it resulted from disulfide formation (supplemental Fig. S1). In cells expressing the X3M2C mutant (with cysteines at both 103 and 179), a prominent band at ϳ40 kDa was found in CNBr digests. Treatment with CuPh 3 did not increase the intensity of this band, but it was absent in DTT-treated samples, indicating that it resulted from disulfide formation. If the N-terminal fragment was bonded through a 103-179 disulfide to the 185-amino acid peptide expected from cleavage at Met-135 and Met-320 (see Fig. 1), we would expect a product of 36.1 kDa (see supplemental Table  S1). The absence of this 40-kDa band when the two single cysteine replacement mutants W103C and I179C were co-expressed, indicates that the disulfide bonded peptide did not result from an intermolecular cross-linking event.
Conformation Dependence of Cross-linking Reaction-Previous studies suggest that SERT and related transporters exist in conformations that expose the binding site either to the extracellular or the cytoplasmic side of the membrane (12-15). The inhibitors cocaine and ibogaine apparently stabilize extracellular-and cytoplasm-facing SERT conformations, respectively, as demonstrated by their ability to increase or decrease accessibility of positions in the extracellular or cytoplasmic permeation pathways (12)(13)(14)16). In our proposed mechanism for transport (13), the distance between TMs 1 and 3 should increase and decrease during this process. Therefore, we examined the effect of cocaine and ibogaine on cross-linking between the cysteine residues inserted at positions 103 and 179. Cocaine, by stabilizing SERT in a conformation similar to that of the LeuT crystal structure, was expected to hold the two cysteine residues apart, thereby decreasing cross-linking. Ibogine was expected to stabilize SERT conformations similar to a cytoplasm-facing conformation, and therefore was not expected to inhibit crosslinking, possibly even facilitating the process.    4 shows that CuPh 3 inactivation of X3M2C was dramatically inhibited by the presence of cocaine, and much less so by ibogaine. At 1 M, CuPh 3 inactivated 60% of transport activity within 5 min (Fig. 4A). The presence of 100 M ibogaine did not noticeably change the time course of inactivation, but decreased the extent to less than 40%. Cocaine (7 M) completely blocked inactivation by this concentration of CuPh 3 . In a 2-min incubation, the extent of inactivation by CuPh 3 increased up to 5 M CuPh 3 (Fig. 4B). Again, inactivation was completely blocked by 5 M cocaine and to a much lesser extent by 100 M ibogaine, which decreased the extent of inactivation by roughly half. The concentrations of cocaine and ibogaine used in this experiment were equivalent to 5 times IC 50 in terms of their inhibitory potency for each mutant (see Table 1). In contrast to the effect of these agents on X3M2C, CuPh 3 did not inactivate I179C (see also Fig. 2) and there was no additional effect of either ibogaine (Fig. 5A) or cocaine (Fig. 5B). Although W103C was mildly sensitive to CuPh 3 (Fig. 2), neither ibogaine (Fig. 5C) nor cocaine (Fig. 5D) influenced this sensitivity.
Several additional inhibitors and substrates were tested for their ability to modify the reactivity of I179C, W103C, and X3M2C. As shown in Fig. 6, I179C was not sensitive to inactivation by CuPh 3 in the presence or absence of any of the tested ligands, including 5-HT, MDMA, imipramine, citalopram, fluoxetine, cocaine, or ibogaine. W103C was moderately sensitive to CuPh 3 , but this sensitivity did not change in the presence of any of the added ligands. X3M2C was even more sensitive to CuPh 3 , as shown also in Fig. 2. Although the transport substrates 5-HT and MDMA had no significant effect on this inactivation, imipramine, citalopram, and fluoxetine, like cocaine, all partially protected this mutant from inactivation. Of the inhibitors, only ibogaine did not protect against CuPh 3 .
Transporter ligands might influence cross-linking by altering the conformation of SERT or by sterically blocking one or both of the cross-linked residues. To assess the extent to which cocaine and ibogaine decrease access to Trp-103 or Ile-179, we measured inactivation by the cysteine reagent MTSEA. As previously described (16), W103C was accessible and sensitive to inactivation by MTSEA. This inactivation was unchanged in the presence of either cocaine or ibogaine (supplemental Fig.   S2). MTSEA also inactivated I179C (Fig. 7A), as previously shown for MTSET (30). Although ibogaine had no significant effect on this inactivation, cocaine significantly potentiated the effect of MTSEA, increasing the rate constant for inactivation by 3-fold (Fig. 7B). Thus, rather than blocking access to Cys-179, cocaine actually increases the reactivity of this residue.

DISCUSSION
Protein conformational changes are thought to underlie the alternate exposure of binding sites to opposite sides of the membrane during transport (31,32). We recently proposed a conformational mechanism for alternating access in the neurotransmitter transporter family (13). This mechanism proposed that tilting or rocking of a 4-helix bundle composed of TM helices 1, 2, 6, and 7 within a scaffold formed by the rest of the protein accomplished these accessibility changes. In the work described here, we provide additional experimental evidence in favor of that mechanism.
In the original LeuT crystal structure, proposed as a model for neurotransmitter transporter structure (3), the ␣-carbons of Leu-29 in TM1 and Ile-111 in TM3 are situated 10.6 Å from each other (Fig. 8) and ϳ10 Å closer to the extracellular medium than the bound leucine molecule. In this structure, the bound leucine at the substrate site was occluded from both sides of the membrane. A more recent LeuT structure with tryptophan bound was presented as a possible model for the conformation of the transporter with the substrate site exposed to the extracellular medium (7). In this "open-out" structure, the C-␣ positions of the same two residues were 13.6 Å apart (Fig. 8). In the ␥-aminobutyric acid transporter GAT-1, Kanner and co-workers (22) found that when both of the corresponding residues were replaced with cysteine, transport activity became sensitive to the sulfhydryl oxidizing agent CuPh 3 , and suggested that this inhibition resulted from disulfide formation between the two transmembrane helices. The C-␣ positions in the extended conformation of cystine are only 7.3 Å apart, significantly closer than the corresponding residues in either LeuT structure. Thus, CuPh 3 inactivation of GAT-1 either occurred by another mechanism, or a conformational change was required to bring the two cysteine residues together.
In the work presented here, we show that cysteine residues at the two corresponding positions in SERT led to chemical cross-linking between TM1 and TM3 (Fig. 3). This observation was facilitated by judicious mutation of methionine residues in SERT to create a background in which a single methionine in TM2 provided a CNBr cleavage site between the two reactive cysteine residues. We propose this method as a new application of CNBr cleavage to identify crosslinking reactions in proteins, especially in cases where the cross-linking event does not lead to a measurable change in activity. The presence of cysteine at positions 103 and 179 was required for efficient CuPh 3 inactivation (Figs. 2 and 4) and cross-linking (Fig. 3). However, the effect of CuPh 3 on crosslinking was not visible in Fig. 3, and the extent of cross-linking was not measurably sensitive to ligand binding, although inactivation was dependent on CuPh 3 (Figs. 2 and 4) and blocked by inhibitors (Figs. 4 and 7). The inability to detect a CuPh 3 effect by CNBr cleavage may reflect the high extent of spontaneous cross-linking. Proteins destined for surface expression are subject to spontaneous disulfide oxidation in the endoplasmic reticulum (33). We surmise that only a minor fraction of SERT expressed in these cells catalyzed 5-HT influx, and was sensitive to oxidation by CuPh 3 , whereas the majority of cell surface SERT was cross-linked during biosynthesis.
From our results with CNBr cleavage, we conclude that, at least in this system, the technique is useful to identify a pair of cysteine residues that are close enough to form a disulfide. However, we found that for determining the time course of cross-linking or measuring the effect of ligands on the process, it was more useful to follow changes in activity that accompanied the reaction.
The W103C mutant was weakly inactivated by CuPh 3 (Figs. 2 and 5) and showed some evidence of cross-linking as indicated by the occasional formation of a minor DTT-sensitive band (supplemental Fig. S1). The corresponding mutant in GAT-1 also showed some sensitivity to CuPh 3 (22). It is not clear what this cross-linking and CuPh 3 sensitivity represent, although it is likely to be a different process from that observed with X3M2C. The mobility of the occasionally observed W103C cross-linked product was higher than with X3M2C, and cocaine, which blocked CuPh 3 inactivation of X3M2C (Fig. 4), had no effect on W103C (Fig. 5). Whatever process was responsible for the reactivity of W103C, it is likely to represent a common feature of SERT and GAT-1. It is not clear why only W103C and not I179C was sensitive to inactivation and cross-linking by CuPh 3 . However, as part of the mobile 4-helix bundle, position 103  (13). TM1 (light) is shown in stick representation and TM3 (dark) as a schematic helix. Distances between ␣-carbon atoms of Leu-29 and Ile-111 are shown along with, for comparison, the corresponding distance in the extended form of cystine, which is the product of disulfide formation between two cysteine molecules.
might have more access to other cysteine residues in SERT or adjacent proteins.
In our proposed mechanism for alternating access in this family, the outer half of the bundle, including TM1, tilts toward the scaffold, including TM3, as the extracellular permeation pathway closes (13). At the same time, the cytoplasmic half of the bundle tilts away from the scaffold, exposing the cytoplasmic permeation pathway. In our model, the distance between the ␣-carbons of Leu-29 and Ile-111 in the cytoplasm-facing conformation is only 5.5 Å, well within the distance required to form a disulfide between cysteines at the corresponding positions in GAT-1 and SERT (Fig. 8). Disulfide formation, therefore, would be facilitated by the conformational change that we proposed to interconvert these transporters from their openout to cytoplasm-facing conformations. Moreover, covalent linkage of the corresponding positions in TMs 1 and 3 of GAT-1 and SERT would be expected to block transport by preventing the conformational change leading to the occluded or open-out conformations.
To provide further support for this interpretation, we utilized cocaine and ibogaine, two inhibitors that apparently stabilize different conformations of SERT. Using the reactivity of cysteine residues placed in the cytoplasmic permeation pathway, we previously showed that cocaine closed that pathway and ibogaine opened it (12,14). At the same time, cocaine increased exposure, and ibogaine decreased exposure, of cysteine residues placed in the extracellular permeation pathway, consistent with a concerted conformational change resulting from tilting of the 4-helix bundle (14 -15, 34). Fig. 4 shows that cocaine blocked the ability of CuPh 3 to inactivate the SERT double cysteine mutant X3M2C, as expected if cocaine binding restricted SERT to conformations similar to the occluded or open-out structures of LeuT (Fig. 8). The effect of cocaine was not simply due to steric occlusion of the cysteine residues at positions 103 and 179, because the single cysteine replacement mutants were either equally (W103C, supplemental Fig. S2) or more sensitive (I179C, Fig. 7) to inactivation by MTSEA in the presence of cocaine. The increased accessibility of I179C is likely due to cocaine stabilizing the extracellular-facing conformation of SERT. We previously showed that cocaine also increased accessibility of the corresponding residue in norepinephrine transporter (15).
Other inhibitors, imipramine, fluoxetine, and citalopram (all antidepressant drugs that inhibit SERT by binding from the cell exterior), also protected SERT X3M2C from inactivation by CuPh 3 (Fig. 6). In contrast, the transported substrates 5-HT and MDMA had little effect on inactivation. Ibogaine was the only inhibitor that did not significantly protect against CuPh 3 .
We interpret these results in the context of a conformational change associated with transport that brings positions 103 and 179 into close proximity. In the absence of ligands, the conformational flexibility of the transporter is high enough that the rate-limiting factor in cross-link formation is the rate of oxidation, itself dependent on CuPh 3 concentration (Figs. 2 and 4B), rather than the proportion of SERT in a cross-linkable conformation. The presence of substrates (5-HT or MDMA) initiates catalytic cycling of the transporter, but does not inhibit crosslinking because intermediate conformations during transport also favor cross-linking. However, inhibitors such as cocaine and antidepressants bind to and stabilize SERT conformations in which positions 103 and 179 are separated enough not to cross-link (Figs. 4 and 6) and are accessible to MTSEA (Figs. 7 and supplemental S2).
Ibogaine, despite being a SERT inhibitor, did not inhibit CuPh 3 -dependent inactivation to the extent observed with other inhibitors. Compared with cocaine, ibogaine was less than 10% as effective at decreasing the sensitivity to CuPh 3 (Fig.  4B) although both inhibitors were present at concentrations 5-fold higher than their IC 50 . Ibogaine apparently stabilizes a conformation of SERT different from that favored by other inhibitors. Thus, we conclude that when ibogaine is bound, SERT is in a conformation closer to that of the cytoplasm-facing intermediate, in which positions 103 and 179 are in close proximity. The observation that ibogaine slightly inhibited the rate of CuPh 3 -mediated inactivation relative to no addition suggests that the two cysteine residues at 103 and 179 may not be as close in the ibogaine-bound state as in the cytoplasm-facing conformation, or that the orientation of their sulfhydryl groups are constrained by ibogaine in a way that impedes their oxidation to cystine. Our model of the cytoplasm-facing conformation of LeuT (13) is not sufficiently accurate to predict the orientation of these residues in SERT, and we also do not know how closely the cytoplasm-facing conformation of SERT resembles the conformation with ibogaine bound.
The results presented here are consistent with a mechanism in which the extracellular half of the 4-helix bundle containing TM1 moves toward TM3 in the scaffold when the transporter is in the cytoplasm-facing conformation. Cocaine binding is proposed to block this movement and in our current results also prevented inactivation due to cross-linking. We anticipate that future structural and biochemical studies will provide further support for the rocking bundle mechanism.