The Conserved Glutamate (Glu136) in Transmembrane Domain 2 of the Serotonin Transporter Is Required for the Conformational Switch in the Transport Cycle*

The alternate access model provides the theoretical framework for understanding how transporters translocate hydrophilic substrates across the lipid bilayer. The model postulates at least two conformations of a transporter, an outward and an inward facing conformation, which seal the translocation pathway to the interior and exterior of the cell, respectively. It is not clear how the conformational switch is triggered in neurotransmitter/sodium symporters, but Na+ is likely to play an essential role. Here, we focused on Glu136 of the serotonin transporter (SERT); this residue is conserved in transmembrane domain 2 of neurotransmitter/sodium symporters and related proteins. Three substitutions were introduced, resulting in SERT-E136D, SERT-E136Q, and SERT-E136A, which were all correctly inserted into the plasma membrane. SERT-E136Q and SERT-E136A failed to support substrate influx into cells, whereas SERT-E136D did so at a reduced rate. Binding experiments with the inhibitor 2β-[3H]carbomethoxy-3β-(4-iodophenyl)tropane (β-[3H]CIT) supported the conjecture that the mutant transporters preferentially adopted the inward facing conformation: β-[3H]CIT interacted with SERT in a manner consistent with binding to the outward facing state. Accordingly, the Na+-induced acceleration of β-[3H]CIT association was most pronounced in wild-type SERT, followed by SERT-E136D > SERT-E136Q > SERT-E136A. Similarly, SERT-E136Q supported substrate efflux in a manner indistinguishable from wild-type SERT, whereas SERT-E136A was inactive. Thus, in the absence of Glu136, the conformational equilibrium of SERT is shifted progressively (SERT-E136D > SERT-E136Q > SERT-E136A) to the inward facing conformation.

porters for serotonin (SERT), dopamine (DAT), and norepinephrine (NET), belong to the NSS family, which also comprises, among others, the transporters for glycine, ␥-aminobutyric acid, and choline (4). Most, if not all, members of this family exist as constitutive oligomers in the plasma membrane (5); there is substantial experimental evidence to support the hypothesis that oligomerization is required for export of the transporters from the endoplasmic reticulum (6 -12). It has therefore been of interest to understand how oligomers are assembled; several candidate contact sites have been proposed (5,13).
GABA transporter-1 (GAT1) (6) and DAT (7) have a leucine heptad repeat in transmembrane domain (TM) 2 that is apparently involved in stabilizing the oligomeric interface. However, despite a conserved overall topology and extensive homology, TM2 of SERT differs substantially from those of GAT1 and DAT. Specifically, SERT does not contain a canonical leucine heptad repeat; the second leucine in GAT1 is Ala 125 in SERT. In addition, if TM2 is assumed to adopt an ␣-helical conformation, there is a glycine residue (Gly 128 ) in SERT one helical turn below Ala 125 . Together, these two residues create a large cavity that precludes any leucine heptad repeat-stabilized oligomeric interaction. Accordingly, the oligomeric structure of SERT is stabilized by an interface of SERT comprising TM11 and TM12 (12). Thus, in this work, we surmised that, by contrast with GAT1 (6,11), DAT (7), and NET (14), mutations in TM2 of SERT would be tolerated without impairing the expression of the protein on the cell surface. This allowed us examine the role of the highly conserved glutamate residue (Glu 136 ) in TM2 of SERT without the confounding effect on oligomeric assembly and endoplasmic reticulum export. Our experiments show that, in SERT, Glu 136 plays a pivotal role in regulating the transport cycle because it is required to afford the outward facing conformation. Substitutions with closely related amino acids sufficed to impair the translocation process, and the resulting mutant transporters accumulated in a state that was consistent with their being in the inward facing conformation.
Epifluorescence Microscopy-The subcellular localization of SERT molecules tagged with YFP was examined by fluorescence microscopy (11). In brief, 0.3 ϫ 10 6 HEK293 cells were seeded onto a 15-mm glass coverslip coated with poly-D-lysine. The next day, cells were transfected with a plasmid encoding either YFP-tagged wild-type SERT or appropriate mutants thereof; after 24 h, YFP-tagged proteins were visualized in living cells using a Zeiss Axiovert M200 epifluorescence microscope.

H]Serotonin Uptake and Release Experiments-
The procedures for uptake and superfusion experiments with radioactively labeled serotonin have been described previously (6). Uptake of serotonin was determined as follows. Cells (5 ϫ 10 5 ) stably expressing SERT were incubated for 1 min at 22°C in Krebs-HEPES buffer (120 mM NaCl, 3 mM KCl, 2 mM CaCl 2 , 2 mM MgSO 4 , 20 mM dextrose (final pH 7.35), and 10 mM HEPES (pH adjusted to 7.4 with NaOH)) with 2 Ci of [ 3 H]serotonin in the absence (total uptake) and presence (nonspecific uptake) of 1 M paroxetine. The specific activity was adjusted with unlabeled serotonin to vary between 15 cpm/fmol and 50 cpm/pmol. After 1 min, the medium was aspirated, and the cells were rapidly rinsed with 1 ml of ice-cold phosphate-buffered saline. Uptake experiments with radiolabeled [ 3 H]serotonin were done in parallel in triplicate under identical conditions (cells seeded in parallel and same incubation time and washing procedure).
The procedure for measuring release has been described in detail (16). In brief, cells were grown overnight on poly-D-lysine-coated round glass coverslips and preincubated in the presence of 0.4 M [ 3 H]serotonin for 60 min at 37°C in a final volume of 0.1 ml of buffer. Coverslips were then transferred to superfusion chambers (0.2 ml) and superfused with Krebs-HEPES buffer at 25°C and a perfusion rate of 0.7 ml/min.
After a washout period of 45 min to establish stable efflux of radioactivity, the experiment was started with collection of 2-min fractions. The buffers were switched to sodium-free buffer after 6 min (NaCl replaced with choline HCl) as indicated in the figure legends. The pH of the sodium-free buffer was adjusted with Tris base. After 18 min, p-chloroamphetamine (PCA; 10 M, dissolved in H 2 O) was added to each buffer. At the end of the experiment, cells were lysed in 1% SDS, and the amount of tritium remaining in the cells was determined. The release of 3 H is expressed as fractional rate, i.e. the radioactivity released during a fraction is expressed as a percentage of the total radioactivity present in the cells at the beginning of that fraction. All comparisons of wild-type and mutant SERTs were done in parallel.
Cell-surface expression was assayed by staining the cell surface with the fluorescent dye trypan blue as described by Oksche and co-workers (17). Cyan fluorescent protein (CFP)-tagged constructs were detected with a band pass filter (475-525 nm) using the 458-nm laser line at 30 -45% input power. The plasma membrane was visualized after addition of 20 l of trypan blue (0.05% in phosphate-buffered saline) at an excitation wavelength of 543 nm with a long pass filter (585 nm). CFP and trypan blue images were captured sequentially, and the overlay images were produced with Zeiss imaging software.
Fluorescence recovery after photobleaching was also recorded on a Zeiss LSM 510 confocal laser scanning microscope. Circular regions of interest were specified for bleaching (bleach regions). Before bleaching, a whole frame scan was obtained; the area of interest was then bleached (70 iterations) at maximum laser power (at 558 nm). Subsequently, pictures were captured (40 scans). The recorded fluorescence intensities were digitized and averaged over the bleach region using Image J equipped with plugins from the Wright Cell Imaging Facility (available at www.uhnresearch.ca/facilities/wcif/index.htm) after correction for nonspecific bleaching of the remaining cell area and background fluorescence. The corrected values obtained were plotted and subjected to nonlinear curve fitting to an equation describing the monoexponential rise to a maximum: y ϭ bottom ϩ (top Ϫ bottom)⅐(1 Ϫ e Ϫkx ), where "bottom" refers to the relative fluorescence intensity immediately after bleaching (before recovery) and "top" refers to the value reached at equilibrium (after recovery). The half-time for fluorescence recovery in the bleach region was calculated from the estimate of k. ␤-[ 3 H]CIT Binding Assays-Equilibrium binding of ␤-[ 3 H]CIT was determined as described (12) in a reaction volume of 0.1 or 0.2 ml with minor modifications. For association kinetics measurements, the membrane preparations (10 g/assay) were incubated with the radioligand (2-10 nM ␤-[ 3 H]CIT) in 20 mM Tris-HCl and 1 mM MgCl 2 (pH 7.4) with NaCl, KCl, and choline HCl at the concentrations indicated in the figure legends. Nonspecific binding was defined in parallel incubations done in the presence of 1 M paroxetine; at 10 nM radioligand and 120 mM NaCl, it was Ͻ10% of total equilibrium binding regardless of whether wildtype or mutant SERT was investigated. The incubation time was varied between 2 and 120 min, and the reaction was terminated by rapid filtration over glass-fiber filters. The radioactivity trapped on the filter was quantified by liquid scintillation counting. The on-rate (k app ) was determined by fitting the data by nonlinear regression to an equation describing a single exponential association. To assess the kinetics of ␤-[ 3 H]CIT dissociation, association of the radioligand (2-10 nM) was allowed to proceed for 1 h. Thereafter, the reaction mixture was diluted 100-fold in 20 mM Tris-HCl and 1 mM MgCl 2 (pH 7.4) with NaCl, KCl, and choline HCl at the concentrations indicated in the figure legends, and aliquots of the diluted mixture were filtered at successive time points. The k off values were determined by fitting the data to an equation describing monoexponential decay. All data are reported as means Ϯ S.E. Statistical analysis was performed using ANOVA followed by Scheffe's post hoc test (where appropriate, as indicated below or in the figure legends).

Cell-surface Expression of the Mutant Serotonin Transporters-Glu 136 in
SERT is invariant in all Na ϩ -dependent neurotransmitter symporters and is also found in many distantly related transporters (including archaeal and bacterial transporters) (18). In GAT1, mutation of this residue results in loss of oligomer formation; hence, the mutant transporters are retained in the endoplasmic reticulum. TM2 of SERT differs substantially from that of GAT1. To explore the role of Glu 136 in SERT, we replaced it with a residue of equal charge, an isosteric residue, and a residue devoid of any hydrogen bond-donating side chain, resulting in mutants SERT-E136D, SERT-E136Q, and SERT-E136A, respectively. Upon expression in HEK293 cells, the three YFP-tagged mutant transporters were visualized at the cell surface by fluorescence microscopy (Fig. 1A). In fact, cells expressing mutant SERT were indistinguishable from those expressing wildtype SERT. Specific uptake of [ 3 H]serotonin was detectable in cells expressing SERT-E136D (Fig. 1B, OE), but not in those expressing SERT-E136A and SERT-E136Q (Fig. 1B, and ࡗ, respectively).
Epifluorescence microscopy is well suited to visualize the distribution of fluorescently tagged proteins. However, it is not possible to rule out that the SERT mutants were trapped in a submembranous compartment and that defective membrane insertion accounted for the reduction or loss of substrate influx. We addressed this issue by two approaches: (i) confocal microscopy using trypan blue (17) and (ii) fluorescence recovery after photobleaching. The membrane-impermeable dye trypan blue binds to the outer leaflet of the cell membrane and emits a strong fluorescence at Ͼ585 nm (17). The differences in the fluorescence spectrum between trypan blue and CFP are large. Thus, trypan blue is ideally suited to visualize the cell surface of cells that have been transfected to express surface molecules tagged with CFP. In confocal images, we observed extensive co-localization of the fluorescence emitted by trypan blue (Fig. 2, middle panels) with that of wild-type and mutant SERTs at the cell surface (upper panels); this is readily evident from the overlays shown in Fig. 2

(lower panels).
When compared with transporters confined to vesicles (submembranous), transporters residing in the plasma membrane are predicted to have extensive lateral mobility. We tested this by bleaching the CFP moiety attached to SERT (Fig. 3A), SERT-E136A (Fig. 3B), SERT-E136D (Fig. 3C), and SERT-E136Q (Fig. 3D) at the cell surface. As can be seen from the summary of half-lives (Fig. 3E), the kinetics of fluorescence recovery were virtually identical for all transporters. We ruled out that fluorescence recovery was due to photo recovery by bleaching transporters (wild-type and mutant) trapped in intracellular vesicles. Fluorescence emission from intracellular transporters did not recover to any appreciable extent over the time course of the experiments (Fig. 3F). Incidentally, these control experiments also showed that the transporters seen at the cell surface must be in the cell membrane rather than in submembranous vesicles because transporters in vesicles could not support fluorescence recovery within the observation period. There is the caveat that the transporters that initially resided in vesicles may have been reinserted into the membrane and that their fluorescence had recovered but was no longer visible in a vesicular form but was present as surface fluorescence. Although we cannot formally rule out this possibility, it appears unlikely: control experiments were also done with transporters with mutated C termini. These were trapped within the cell and never reached the plasma membrane (10). Fluorescence emitted by these transporters also failed to recover within the observation period.

Binding of ␤-[ 3 H]CIT as a Conformational
Probe-Given that wildtype and mutant SERTs co-localized with trypan blue and had comparable lateral mobility, we surmised that all transporters were present at the cell surface. Hence, we hypothesized that SERT-E136A, SERT-E136Q, and possibly SERT-E136D were trapped in a conformation that was not conducive to transport of substrate. The substrate translocation process is described by the alternate access model, which assumes a minimum of two distinct conformations of the transporter, i.e. an outward and an inward facing conformation. Inhibitors bind to the outward facing conformation, and their binding is contingent on the presence of Na ϩ (19,20). Using membrane preparations from cells expressing wildtype SERT, we therefore tested whether binding of the radioligand inhibitor ␤-[ 3 H]CIT is useful as a probe for the conformational cycle of the transporter. The apparent association rate of ␤-[ 3 H]CIT binding was too slow to be measured reliably in the total absence of Na ϩ (ionic strength being preserved by choline chloride) (see Fig. 6B). In the presence of a low concentration of Na ϩ (5 mM, i.e. within the range of the [Na ϩ ] i ), binding of ␤-[ 3 H]CIT was slow (k app ϭ 0.08 Ϯ 0.01 min Ϫ1 , n ϭ 3) (Fig. 4A, OE). When the concentration of Na ϩ was raised to 120 mM (i.e. within the extracellular range), the apparent association rate was dramatically accelerated, and equilibrium was reached within 5 min (k app ϭ 0.36 Ϯ 0.06 min Ϫ1 , means Ϯ S.E., n ϭ 3) (Fig. 4A, f). Binding was also accelerated by K ϩ (k app ϭ 0.17 Ϯ 0.01 min Ϫ1 , n ϭ 3) (Fig. 4A, ), but the observed k app was substantially lower than that achieved in the presence of equimolar Na ϩ , although the concentrations of Na ϩ and K ϩ were lower than and exceeded normal intracellular [Na ϩ ] i and [K ϩ ] i , respectively. Thus, both ions promoted the accumulation of the conformation that supported binding of the radioligand inhibitor, but Na ϩ was substantially more effective.
Na ϩ and K ϩ also differed in their effects on the dissociation kinetics of ␤-[ 3 H]CIT binding. If ␤-[ 3 H]CIT was allowed to bind to wild-type SERT in the presence of 120 mM Na ϩ and if the reaction mixture was subsequently diluted 100-fold in sodium-free (choline-containing) solution, the complex dissociated with a half-life in the range of ϳ6 -10 min (Fig. 4B, F; and Table 1). If the reaction mixture was diluted in a solution containing Na ϩ , the lifetime of the complex was prolonged by a factor of ϳ2.5 (Fig. 4B, f; and Table 1). In contrast, dilution in K ϩ -containing solution did not affect the dissociation rate of ␤-[ 3 H]CIT to an appreciable extent (Table 1;    It is evident that Na ϩ had a more pronounced effect on the association rate compared with the dissociation rate of ␤-[ 3 H]CIT binding. Sodium and the inhibitor may bind in a sequential order: in this model, Na ϩ binds independently of the inhibitor and induces or stabilizes a conformation that is prone to interact with ␤-[ 3 H]CIT. Alternatively, the radioligand and Na ϩ may bind concomitantly and stabilize their binding to SERT in a reciprocal manner. The latter model predicts that the concentration-response curve of sodium depends on the radioligand concentration. This was not the case; as shown in Fig. 4C, the half-maximum concentration of Na ϩ required to promote binding of ␤-[ 3 H]CIT was comparable at a low concentration (ϽK D ) and at a quasi-saturating concentration of the radioligand. Thus, Na ϩ binding was independent of inhibitor binding, whereas inhibitor binding was dependent on Na ϩ binding. This is consistent with the interpretation that sequential binding is the favored reaction: Na ϩ binding occurs first and promotes the conformational switch that allows for binding of

Binding of ␤-[ 3 H]CIT to SERT Mutated at Position 136-When
CIT binding was assessed under conditions that afforded equilibrium binding in wild-type SERT, it was evident that only the SERT-E136D mutant bound the radioligand with an affinity comparable with that of wild-type SERT (K D ϭ 1.36 Ϯ 0.31 and 3.24 Ϯ 1.05 nM for wild-type SERT and SERT-E136D, respectively; means Ϯ S.E., n ϭ 3) (Fig. 5A, f and , respectively). In contrast, binding to SERT-E136Q (Fig. 5A, ࡗ) and SERT-E136A (Fig. 5A, OE) was of low affinity; hence, a reliable calculation of the dissociation constant was not possible. The loss of affinity was not confined to the binding of the inhibitor ␤-[ 3 H]CIT; the natural substrate serotonin also bound with an affinity that was reduced by Ͼ1 and 2 orders of magnitude to SERT-E136Q and SERT-E136A, respectively, compared with wild-type SERT and SERT-E136D (Fig. 5B). Because K D values could not be reliably calculated for SERT-E136Q and SERT-E136A, IC 50 values (summarized in the legend to Fig. 5) were not converted to K i values. However, it is obvious that the concentration of ␤-[ 3 H]CIT employed in the competition assay was far FIGURE 3. Fluorescence recovery after photobleaching of wild-type and mutant SERTs. HEK293 cells transiently expressing wild-type (wt) SERT (A) and the indicated mutants (C-D) were seeded onto coverslips, and fluorescence recovery after photobleaching was recorded using a confocal scanning microscope as described under "Materials and Methods." Left panels, recorded fluorescence intensity plotted against time; right panels, representative fluorescence images at different time points (in seconds), where the zero time point denotes images taken immediately before bleaching. White arrows point to the bleached regions. rel., relative. The mean half-times of fluorescence recovery are summarized in E. All data are given as means Ϯ S.E. and represent 6 -12 measurements from at least two experimental days (i.e. independent transfections). There is no significant (n.s.) difference between groups according to one-way analysis of variance followed by Scheffe's post hoc test (level of significance, p Ͻ 0.05). F shows representative prebleaching images (zero time point) and lack of fluorescence recovery after bleaching of wild-type SERT (left panels) and SERT-E136Q (right panels) trapped in an intracellular compartment; the observation period was for up to 60 s (60-s time point).
below its K D for SERT-E136Q and SERT-E136A; thus, the difference in IC 50 reflects a reduced affinity of these mutant transporters for their substrate.
Similarly, although wild-type SERT and the three mutants differed markedly in the extent to which ␤-[ 3 H]CIT binding was promoted by Na ϩ , the effect of Na ϩ on dissociation of ␤-[ 3 H]CIT binding was comparable ( Fig. 7 and Table 1). Regardless of whether wild-type SERT (Fig.  7A), SERT-E136A (Fig. 7B), SERT-E136D (Fig. 7C), or SERT-E136Q (Fig. 7D) was investigated, the lifetime of the ␤-[ 3 H]CIT⅐transporter complex was about twice as long if the reaction was diluted in solution containing 120 mM NaCl rather than choline chloride. This ratio (k off,Na ϩ/k off,choline ϭ 2-3) was comparable between wild-type and mutant SERTs (Table 1; see also Fig. 4B).
Similarly, if dissociation of prebound ␤-[ 3 H]CIT was initiated by diluting the reaction mixture in medium containing 120 mM K ϩ , the off-rate was comparable with that observed after dilution in medium containing choline for wild-type SERT (Fig. 8A), SERT-E136D (Fig. 8C), and SERT-E136Q (Fig. 8D). In contrast, K ϩ caused a modest, albeit statistically significant, decrease in the dissociation rate constant for ␤-[ 3 H]CIT binding to SERT-E136A ( Fig. 8B and Table 1). At present, it is not clear why the E136A mutation renders the transporter susceptible to this effect of K ϩ . Regardless of the mechanistic basis for this phenomenon, the available data are consistent with a model in which Glu 136 is required for the Na ϩ -driven conformational transition to/stabilization of the outward facing conformation of the transporter.
Outward Transport of Substrate by the Mutant Transporters-Neurotransmitter transporters of the NSS family are capable of bidirectional transport, i.e. they can support efflux of substrate if the driving force is inverted by reversal of the Na ϩ gradient and provided that the substrate

TABLE 1 Dissociation rate constant (k off ) for ␤ -͓ 3 H͔CIT binding to wild-type SERT and SERT mutated at Glu 136
The reaction was carried out as described in the legends to Figs. 2B, 5, and 6. Data are means Ϯ S.E. from three to five independent experiments performed in duplicate.   concentration within the cell is high enough. Furthermore, in monoamine transporters (i.e. SERT, NET, and DAT), transport reversal can be induced by amphetamine derivatives; these compounds promote substrate efflux, and this accounts for their biological effects. Mutations at Glu 136 allow for the accumulation of the resulting SERT mutants at the cell surface but reduce (SERT-E136D) or abolish (SERT-E136Q and SERT-E136A) inward transport by the mutant proteins. Our model assumes that the mutant transporters fail to support efficient inward transport because, at steady state, they accumulate predominantly in an inward facing conformation. This model predicts that at least some of these SERT mutants should be capable of binding and translocating substrate if the substrate is present on the cytoplasmic side. We exploited the fact that cells can also be passively loaded with [ 3 H]serotonin; after prolonged incubations, substantial amounts are trapped within the cells even in the absence of a functional transporter (16,22). It is noteworthy, however, that the slope of basal release from nontransfected cells was comparable in magnitude to that of basal release from SERT mutant-expressing cells (and this also holds true for basal release expressed in absolute terms (22)). The cells expressing wild-type SERT or SERT-E136Q were superfused with buffer until a stable base line was established. Release of preloaded [ 3 H]serotonin was triggered either by addition of PCA (Fig. 9A) or by reversal of the Na ϩ gradient (Fig. 9B), which abrogates the driving force for inward transport and favors extrusion of the substrate. As shown in Fig. 9A, PCA-promoted efflux was of comparable magnitude in cells expressing wild-type SERT (Fig. 9A, Ⅺ) and SERT-E136Q (छ). Similarly, if external Na ϩ was replaced with choline to reverse the driving force for substrate translocation, increased levels of [ 3 H]serotonin were recovered in the superfusate of cells expressing wild-type SERT (Fig. 9B, Ⅺ) and SERT-E136Q (छ). As expected, omission of Na ϩ was less effective than PCA in promoting [ 3 H]serotonin efflux (Fig. 9, cf. A and B). More important, we confirmed that the action of PCA in the absence of external Na ϩ was specific because the compound failed to release [ 3 H]serotonin in the sole presence of choline (Fig. 9B). We focused on a comparison of wildtype SERT and SERT-E136Q because SERT-E136Q did not show any detectable uptake of substrate (cf. Fig. 1B). However, similar results were also obtained with SERT-E136D; in contrast, we failed to observe [ 3 H]serotonin efflux upon stimulation with PCA or withdrawal of Na ϩ in cells expressing SERT-E136A (data not shown).

DISCUSSION
The alternate access model was formulated some 40 years ago to describe how transporters translocate a hydrophilic substrate through the membrane (23). One of the central tenets of the model is that the transporter undergoes a conformational cycle through at least two states: the outward facing conformation, which is open to the extracellular milieu and sealed off on the cytoplasmic side, and the inward facing conformation, which has the opposite orientation. In addition, there may be a transition state, in which the substrate permeation pathway is blocked on both sides, resulting in a locked state. Structures that are consistent with an inward facing conformation (24,25) and with the locked state (26) were recently visualized for bacterial transporters. It is not clear how the conformational switch is brought about. Here, we have identified Glu 136 as a residue that, in SERT, participates in the switch between outward and inward facing conformations. We conclude that removal of Glu 136 shifts the conformational equilibrium in favor of the inward facing state of SERT. This conclusion is based on the following observations. (i) The mutant transporters SERT-E136D, SERT-E136Q, and SERT-E136A were all visualized at the cell surface, and there was no indication of appreciable intracellular retention. Thus, the structural change was subtle enough to allow the mutant transporters to pass the stringent quality control in the endoplasmic reticulum. Despite their presence at the cell surface, the mutant transporters poorly supported substrate influx. It was particularly striking that conservative substitutions (aspartate for glutamate and glutamine for glutamate) sufficed to greatly reduce (SERT-E136D) or to abolish (SERT-E136Q) detectable inward transport. (ii) Despite the pronounced decrease in inward transport, SERT-E136Q and SERT-E136D readily supported outward transport when challenged with an amphetamine derivative or with a reversal of the Na ϩ gradient. Substrate efflux through these mutants was comparable in magnitude to that mediated by wild-type SERT. (iii) The mutants were less susceptible to the Na ϩpromoted conformational change, which resulted in accelerated association of the inhibitor ␤-[ 3 H]CIT. (iv) NSSs can, in principle, operate bidirectionally, but substrate influx is the favored mode; accordingly, in SERT, the K m for inward transport is substantially lower than that for outward transport (16). Thus, under basal conditions (i.e. in the absence of substrate and in the presence of high extracellular Na ϩ and high intracellular K ϩ concentrations), SERT is predicted to exist predominantly in the outward facing conformation. Accordingly, in binding experiments, the substrate serotonin competed with ␤-[ 3 H]CIT binding to wild-type SERT with an affinity reasonably similar to its K m for inward transport. In contrast, the affinity of serotonin for the SERT mutants was markedly reduced for those mutants that failed to support substrate influx (SERT-E136Q and SERT-E136A).
It is worth noting that we failed to detect outward transport of [ 3 H]serotonin with SERT-E136A, whereas this was detectable with SERT-E136Q. This seeming discrepancy can be resolved by taking into consideration that the transporter not only needs to be in the inward facing conformation to support outward transport, it must also be capable of adopting the outward facing conformation and of undergoing the transport cycle. On the basis of our binding data, we consider SERT-E136A to be the mutant least capable of undergoing the transport cycle; and thus, it does not support release to an extent that is detectable by superfusion.
Our interpretation relies in part on the assumption that high affinity binding of the inhibitor ␤-[ 3 H]CIT can be used as a conformational probe. This assumption is justified because inhibitor binding to SERT is strongly affected by Na ϩ : consistent with previous reports (27)(28)(29), we found that, in the absence of Na ϩ , binding of ␤-[ 3 H]CIT was very low and that addition of Na ϩ accelerated the association rate of ␤-[ 3 H]CIT binding. However, there was no reciprocal modulation of Na ϩ affinity by inhibitor binding. This indicates that Na ϩ drives SERT into a conformation that allows for ␤-[ 3 H]CIT binding. Thus, conceptually, inhibitor binding to SERT can be envisaged to be analogous to the binding of an open channel blocker to a channel: conditions that allow for channel opening increase the association rate of the channel blocker. Accordingly, conditions that promote the outward facing conformation enhance binding of ␤-[ 3 H]CIT to SERT. This interpretation is supported by analogous experiments in a series of mutants of DAT that were trapped in the inward facing conformation unless the switch to the outward facing state was triggered by the addition of Zn 2ϩ (30,31). It has been repeatedly pointed out that the alternate access model fails to account for the electrophysiological properties of NSS in general and for SERT specifically; instead, a model was proposed in which SERT was considered as a channel rather than as a transporter (32,33). Of note, as pointed out by Accardi and Miller (34), "the gap between transporters and ion channels is an exceedingly fine line"; this notion is exemplified by the orthologous SERT from Drosophila melanogaster (35), which fulfills the criteria of a bona fide ion channel (36,37). Our experiments were obviously not designed to address this controversy. However, it is worth pointing out that our interpretation remains valid in a channel-line model, provided that opening of the outer gate of SERT is equivalent to its adopting the outward facing conformation.
The mutants differed substantially in the extent to which Na ϩ accelerated the association rate of ␤-[ 3 H]CIT binding. This can be rationalized by a model in which the Glu 136 mutants of SERT accumulate predominantly in the inward facing conformation and are thus unavailable to Na ϩ -promoted binding of ␤-[ 3 H]CIT. Alternatively, Glu 136 may serve as a cation-binding site by providing a counterion within the permeation pathway (21). This explanation appears unlikely for several reasons. (i) Neither glutamine nor alanine can serve as a counterion; thus, they are predicted to be similar. However, the two mutants differed substantially in their rates of Na ϩ -promoted ␤-[ 3 H]CIT binding. (ii) Similarly, both aspartate and glutamate provide a negative charge; thus, SERT-E136D is expected to be similar to wild-type SERT. This was not the case. (iii) Na ϩ not only accelerates the association rate of inhibitor binding (28) but also slows the dissociation rate. This effect of Na ϩ was maintained in all mutants at a similar magnitude regardless of the nature of the substitution. (iv) Finally, we observed that K ϩ also accelerated the association rate of ␤-[ 3 H]CIT binding, albeit to a substantially lower extent compared with Na ϩ . By contrast with Na ϩ , K ϩ did not discriminate between wild-type and mutant SERTs. In SERT, intracel- lular K ϩ was originally proposed to be transported outwards during the transport cycle (38) and to thereby facilitate the return to the outward facing conformation (39). Alternatively, in the model that considers SERT as a channel, K ϩ is thought to bind to an internal site to stabilize the open state (32).
Previous studies reported on mutational surveys that also included replacement of the conserved glutamate residue in TM2 of SERT and other monoamine transporters. In DAT, the homologous E117Q mutation results in a protein that is retained within the cell and that binds radioligand poorly (40,41). In contrast, the analogous mutations in NET (NET-E113Q, NET-E113D, and NET-E113A) are expressed to some extent at the cell surface, albeit with substantially higher proportions of intracellularly retained protein compared with the SERT mutants (14). Contrary to SERT, NET-E113D does not support uptake of substrate, whereas influx of substrate mediated by SERT-E113Q approaches 50% of that supported by wild-type NET (14). In SERT, the mutation E136C results in a transporter that binds ␤-CIT poorly and supports only low levels of substrate uptake (42). Additional information is not available because the mutant was not further investigated. Nevertheless, based on the available data, SERT-E136C is phenotypically similar to our mutants. Because even aspartate fails to substitute for glutamate at position 136, we therefore conclude that no amino acid other than glutamate is compatible with full transporter function.
The recently determined structure of a bacterial NSS homolog (LeuT Aa ) (26) strongly supports our interpretation. LeuT Aa was crystallized together with the substrates (leucine and one chloride and two sodium ions). The conformation of LeuT Aa is not conducive to substrate translocation (i.e. it corresponds to a locked state). The substrate leucine is tightly bound and hindered from movement by the binding site-forming residues in TM1, TM3, TM6, and TM8. Notably, leucine associates with TM1 and TM6 at the sites of ␣-helix unwinding, forming multiple dipolar contacts with the peptide backbone and one of the co-substrate Na ϩ ions. Although the sequence similarity between mammalian NSS members and the bacterial leucine transporter is low, the overall architecture of the different members of this family may be well preserved. The sequences specifying domains in the immediate vicinity of Glu 136 of SERT are reasonably comparable with those in LeuT Aa . Accordingly, we inspected the environment of Glu 62 in LeuT Aa (which corresponds to Glu 136 in SERT), and this is shown in Fig. 10. It is composed of side chains from TM2, TM6, and TM10. Glu 62 (SERT Glu 136 ) points with its carboxylate to the middle of the protein, toward the substrate-binding site (highlighted by the presence of the substrate leucine in Fig. 10). Of note, only the middle part of TM6 serves as a barrier between Glu 62 (SERT Glu 136 ) and the putative substrate-binding cleft. Glu 62 is hydrogen-bonded to the amide of Gly 258 (SERT Gly 340 ) in TM6. This residue is part of a short amino acid stretch in which ␣-helix 6 unwinds (in both the crystal structure and our homology model) to take part in direct interactions with the substrate. In addition, Glu 62 is in close contact with the glutamate (Glu 419 ; SERT Glu 508 ) in TM10, one of the few TM10 residues conserved in both SERT and LeuT Aa . Moreover, the crystal structure captured two water molecules in immediate proximity to both Glu 62 and Glu 419 in LeuT Aa , homologous to Glu 136 and Glu 508 in SERT, respectively. One of these water molecules is shared by the carboxylate of Glu 62 and the main chain carbonyl of Ser 256 (a glycine in SERT), the latter being part of the substrate-binding site. Both water molecules are in close vicinity of Gly 260 ; replacing the homologous residue in SERT with alanine resulted in an inactive transporter. 4 Glu 62 itself may be viewed as a key component of the extensive TM10 -TM2-2H 2 O-TM6 hydrogen bond network, crucial for correct substrate binding and translocation (41). Intuitively, Glu 62 would ideally link the distal conformational rearrangements (e.g.  Transiently transfected HEK293 cells were grown on glass coverslips and preloaded with tritiated serotonin for 1 h. Thereafter, they were transferred to small superfusion chambers and superfused (for details, see "Materials and Methods"). The experiment was started after a washout period at t ϭ 0 with the collection of superfusate in 2-min fractions (A). In experiments employing zero sodium conditions, choline was used as a replacement, and the buffers were switched after 6 min as indicated (B). PCA was added to all buffers after 18 min. The collected radioactivity was counted and is expressed as a fractional efflux rate, i.e. as a percentage per 2 min of the cellular [ 3 H]serotonin content at that very time point. WT, wild-type.  (26), a close-up view is shown of the pertinent region (generated with DeepView/Swiss-PdbViewer Version 3.7, available at www.expasy.org/spdbv/), with selected residues projecting from the helical backbone. Red, oxygens; blue, nitrogens; yellow spheres, water; green dashed lines, putative hydrogen bonds. The indicated residues in LeuT Aa correspond to the following residues in SERT: LeuT Aa Glu 62 to SERT Glu 136 , LeuT Aa Glu 419 to SERT Glu 508 , and LeuT Aa Gly 258 to SERT Gly 340 . The helices are colored in gray, with the important ones highlighted as TM2, TM6, TM10. The substrate-binding site of LeuT Aa contains leucine (labeled LEU in green) in close vicinity to Glu 62 (Glu 136 in SERT); the purple spheres adjacent to the substrate represent two co-crystallized co-substrate sodium ions. movements of TM10 -intracellular loop 5-TM11) with the transporter reorientations, openings or closings of the substrate-or inhibitor-binding pocket. Being tightly connected to TM6 main chain atoms, Glu 62 in LeuT Aa or Glu 136 in SERT would ensure a concerted movement of TM6 during conformational alterations associated with substrate translocation in or out of the cell.
Thus, provided the environment of Glu 136 in SERT does indeed closely resemble that of Glu 62 in LeuT Aa , the carboxyl moiety of Glu 136 may be engaged in close interactions with at least four partners, viz. the two water molecules, Glu 508 , and Gly 340 . This puts a number of constraints on the geometry of the residue occupying position 136. Hence, it is conceivable that any substitution at this site would tend to impair the transport process. And indeed, the most detrimental substitution (E136A) renders the transporter nonfunctional. The conformational rearrangements conveyed by Glu 136 are triggered by Na ϩ binding. Without direct contact between Glu 136 and Na ϩ ions, the only obvious link between them is the unwound TM6 segment. Thus, in this model, Na ϩ -promoted inhibitor binding depends on whether the TM6 region is stabilized by Glu 136 . The unwound region of helix 6 extends to Gly 344 in SERT, which is likely to provide a hinge region. If this residue was mutated to alanine (SERT-G344A), Na ϩ -promoted binding of ␤-[ 3 H]CIT to SERT was abolished. 4 Thus, this model allows us to rationalize the experimental findings in SERT and the related transporter NET. Nevertheless, despite the conserved nature of the glutamate residue, there must be substantial differences in the structural rearrangements that accompany the translocation process even in closely related transporters that share overlapping specificities for substrates and inhibitors. This is exemplified by the observation that NET-E113Q rather than NET-E113D affords inward transport (14), whereas SERT-E136E but not SERT-E136Q allows for substrate influx. Similarly, replacing a segment of the extracellular loop of SERT with the corresponding segment of NET suffices to trap the resulting SERT/NET chimera in the outward facing conformation (43).