Partitioning of the serotonin transporter into lipid microdomains modulates transport of serotonin.

The serotonin transporter (SERT) is an integral membrane protein responsible for the clearance of serotonin from the synaptic cleft following the release of the neurotransmitter. SERT plays a prominent role in the regulation of serotoninergic neurotransmission and is a molecular target for multiple antidepressants as well as substances of abuse. Here we show that SERT associates with lipid rafts in both heterologous expression systems and rat brain and that the inclusion of the transporter into lipid microdomains is critical for serotonin uptake activity. SERT is present in a subpopulation of lipid rafts, which is soluble in Triton X-100 but insoluble in other non-ionic detergents such as Brij 58. Disaggregation of lipid rafts upon depletion of cellular cholesterol results in a decrease of serotonin transport capacity (V(max)), due to the reduction of turnover number of serotonin transport. Our data suggest that the association of SERT with lipid rafts may represent a mechanism for regulating the transporter activity and, consequently, serotoninergic signaling in the central nervous system, through the modulation of the cholesterol content in the cell membrane. Furthermore, SERT-containing rafts are detected in both intracellular and cell surface fractions, suggesting that raft association may be important for trafficking and targeting of SERT.

The serotonin transporter (SERT) 1 is a member of a Na ϩ / Cl Ϫ -dependent family of integral membrane proteins that includes carriers for neurotransmitters, osmolytes, and nutrients (1)(2)(3). SERT, the norepinephrine transporter, and the dopamine transporter are most closely related, defining a subgroup characterized by a high degree of similarity both in sequence and pharmacological properties (4). SERT is responsible for the clearance of serotonin (5-hydroxytryptamine, 5HT) from the synaptic cleft following the release of the neurotransmitter (2).
5HT is believed to accumulate inside the cell through cotransport with Na ϩ and Cl Ϫ and countertransport with K ϩ (4). SERT is an important pharmacological target for substances of abuse, such as cocaine, 3,4-methylenedioxymethamphetamine (Ecstasy), and p-chloramphetamine, and a variety of therapeutic antidepressants (1,5,6).
Acute changes in SERT endogenous activity are likely to originate from local variations in 5HT concentration, cell surface distribution of SERT, interaction with regulatory proteins, or reversible post-translational modifications. To date SERT expression on the plasma membrane has been shown to be down-regulated by protein kinase C activation, following elevation of intracellular levels of Ca 2ϩ (7,8), or by phorbol 12myristate 13-acetate treatment (9 -12). The SNARE protein syntaxin 1A (Syn 1A) is one of the few proteins known to modulate SERT function via protein-protein interaction by regulating the number of SERT molecules on the plasma membrane (13,14).
To carry out clearance of extracellular 5HT, SERT must be inserted into the plasma membrane in an active form. This may require targeting the transporter to specific plasma membrane regions, but not others. The study of how asymmetry is generated and maintained in polarized cells has changed old concepts of cell membranes. According to the Singer-Nicholson fluid mosaic model the lipid bilayer is a neutral two-dimensional solvent for membrane proteins (15). However, in model lipid bilayers, lipids can exist in different phases: as gel, liquidordered (l o ), and liquid-disordered states (as in the Singer-Nicholson model). In the l o state, saturated hydrocarbonic chains of phospholipids are tightly packed with cholesterol, thus conferring a limited horizontal mobility in the membrane. This model has been confirmed at least partially in cell membranes: the exoplasmic leaflet contains lipid microdomains, which are enriched in sphingolipids and cholesterol and are segregated from glycerolipids. Such lipid microdomains, also known as "rafts," resist solubilization at 4°C in non-ionic detergents, such as Triton X-100 (16,17). The segregation of proteins into lipid microdomains is lost after depletion of cholesterol or sphingolipids. Lipid rafts appear to be involved in several cellular functions, from membrane trafficking to cell signaling (18).
Membrane cholesterol was shown to be required for the stabilization of the structure of SERT. It has been observed that cholesterol depletion reversibly reduces ligand binding to SERT expressed in Sf9 insect cells (19) and in HEK cells (20). In both cases the function could be partially rescued by replacement of cholesterol, but not by other sterols, such as pregnenolone, 5-cholestene, and ergosterol.
In the present study we show that SERT associates with lipid rafts and that, following the disaggregation of these lipid microdomains by cholesterol depletion, SERT, although retained on the cell surface, is no longer able to efficiently transport 5HT. We propose the existence of a novel mechanism of SERT regulation, based on lipid-protein interaction within plasma membrane lipid rafts.

EXPERIMENTAL PROCEDURES
Plasmids-Rat SERT cDNA obtained from Patrick Schloss (21) was used as a template to generate a PCR fragment containing the complete coding sequence but lacking the stop codon. The sequences of the primers used were 5Ј-CCGCTCGAGCAGGATGGAGACCACACC-3Ј and 5Ј-TTGGTACCACAGCATTCATGCGGAT-3Ј. The PCR product was digested with XhoI and KpnI and cloned the SalI-KpnI-digested vector pFLAG-CMV-5b (Sigma) resulting in plasmid pSERT-FLAG. The complete insert of the plasmid was checked by DNA sequencing. The pTLR4-FLAG plasmid was obtained from Marta Muzio (DRO-Oncology, Pharmacology Department, Pharmacia Corp., Nerviano, Italy).
Reagents-An antiserum against the N-terminal domain of SERT was raised using the core protein of the hepatitis B virus as an epitope carrier (22). A carrier vector was created from the hepatitis B core protein clone ⌬CW (23), by introducing a BamHI site into the DNA at codon 79 (in the immunodominant loop region of the protein), and an HpaI site ten codons later. The DNA coding sequence for an N-terminal rat SERT peptides corresponding to amino acid residues 5-24 was inserted between these sites. The recombinant protein was expressed in Escherichia coli strain BL21(DE3). Cells were lysed in 50 mM Tris-HCl, pH 8.0, 5 mM EDTA, 1 mM dithiothreitol, 0.1 mg/ml phenylmethylsulfonyl fluoride, 2 mg/ml lysozyme. Following lysis, 2 ml of 0.2 mg/ml DNase in 0.1 M MgCl 2 was added, and the extract centrifuged at 15,000 rpm for 30 min. (NH 4 ) 2 SO 4 was added to the supernatant to a final concentration of 50%, and the precipitated protein was pelleted (15,000 rpm, 10 min). The protein pellet was resuspended in running buffer containing 50 mM Tris-HCl, pH 7.5, 0.5 M NaCl, 1 mM dithiothreitol, and passed down a CL4B gel filtration column. Fractions containing protein were reprecipitated with 50% (NH 4 ) 2 SO 4 and resuspended in running buffer overnight. The protein was then layered onto 20 -40% sucrose gradients, with 70% sucrose cushions, and centrifuged in a SW28-Ti rotor at 28,000 rpm for 16 h. Fractions from the leading peak were pooled, dialyzed against running buffer, and concentrated to ϳ15 mg/ ml. The antiserum against the purified protein was raised in rabbit by Murex Biotech (Dartford, UK).
The monoclonal anti-transferrin receptor antibody was obtained from Zymed Laboratories; the monoclonal anti-flotillin 1 antibody and the anti-Munc 18-1 antibody were obtained from BD Transduction Laboratories; the anti-syntaxin 1A antibody was obtained from Santa Cruz Biotechnology. Unless otherwise indicated, chemicals and reagents were obtained from Sigma.
Cholesterol Depletion-Cells were washed twice with PBS and then incubated with the appropriate amount of methyl-␤-cyclodextrin (M␤C, Sigma) for 30 min at 37°C. The cell monolayer was washed three times with buffer and immediately processed for the indicated assay as described. An MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) assay was carried out using the procedure described by Carmichael et al. (25).
Lipid Analysis-Cholesterol content of cells was measured enzymatically using the Infinity Cholesterol Reagent (Sigma), according to the manufacturer's protocol. For phospholipids analysis, lipids were extracted using the Folch method (26), the lower lipid-containing phase was separated from the upper phase and used to analyze the total phosphorus using the Bartlett assay (27). Protein concentration was determined using the Markwell assay (28).
[ 3 H]5HT Uptake Assay-T-REx-SERT and HEK cells were plated on poly-L-lysine-coated (0.1 mg/ml) 24-well plates 3 days before experiments. If transient transfection was required, cells were transfected using GeneJuice transfection reagent according to the manufacturer's protocol (Novagen) 1 day after plating and grown for an additional 48 h. 5HT uptake assays were carried out as described previously (24). RN46A-B14 cells were seeded onto uncoated 12-well plates 4 days before experiments, and the assay was carried out as above. For the ion replacement experiments, Na ϩ ions were replaced with choline, and Cl Ϫ ions were replaced with acetate.
Immunofluorescence-Patching experiments were carried out as described by Harder and Simons (1999). Briefly, cells were grown on sterile glass coverslips (treated with 0.1 mg/ml poly-L-lysine) in 6-well plates. Cells in DMEM medium were incubated with 5 g/ml CTxB at 4°C for 10 min. Cells were washed twice with DMEM and incubated with anti-CTxB (1:500) for 10 min at 37°C. The cells were then placed on ice, washed with ice-cold PBS, and fixed with 4% (w/v) paraformaldehyde in PBS, pH 7.5, for 10 min at room temperature. The fixative was quenched by incubating cells with 50 mM NH 4 Cl for 10 min. Fixed cells were incubated with blocking solution (0.2% (v/v) Triton X-100, 5% (v/v) donkey serum in PBS) for 1 h at room temperature, followed by incubation with anti-FLAG M2 antibody (mouse, 1:1000, Sigma) diluted in blocking solution overnight at 4°C. Cells were washed with PBS and incubated with the anti-mouse fluorescein isothiocyanate-and antirabbit Texas Red-conjugated antibodies (donkey, Jackson ImmunoResearch), diluted in blocking solution for 1 h at room temperature. After washing, the cells were mounted in 2 g/l p-phenylenediamine in 1:1 (v/v) glycerol:PBS. Samples were imaged using a Zeiss LSM510 laser scanning confocal microscope.
Sucrose Flotation Gradients-At confluence, cells were washed twice with ice-cold PBS. If required, the cells were incubated with 3 g/ml CTxB for 10 min at room temperature. The cell monolayer was washed three times with PBS, and cells were lysed with 1 ml of TNE buffer (50 mM Tris HCl, pH 7.4, 150 mM NaCl, 5 mM EDTA, 1ϫ Complete TM Protease Inhibitor mixture) containing the indicated concentration of non-ionic detergent, for 30 min at 4°C. The lysates were homogenized with 15 strokes of a Dounce homogenizer (Thomas Scientific), transferred into a minifuge tube, and centrifuged at 4°C for 10 min at 830 ϫ g in an Eppendorf centrifuge (5417R). Supernatants were transferred into a fresh minifuge tube. The pellets (P 1 ) were resuspended in 1 ml of solubilization buffer (50 mM Tris, pH 8.8, 5 mM EDTA, 1% SDS) and stored at Ϫ20°C. 500 l of cell supernatant was mixed with an equal volume of 80% (w/v) sucrose in TNE buffer, transferred into a ultracentrifuge tube, then carefully overlaid with 1 ml of 30% (w/v) sucrose in TNE buffer and finally with 0.5 ml of 5% (w/v) sucrose in TNE buffer. The tube was placed in a cooled AH650 swinging-bucket rotor, and centrifugation was carried out in a Discovery100-Sorvall ultracentrifuge at 4°C, for 17 h at 134,400 ϫ g. Eight fractions were collected from the top to the bottom of the gradient. The remaining pellet (P 2 ) was resuspended following centrifugation in 500 l of solubilization buffer and stored at Ϫ20°C.
Cell Surface Biotinylation of Subcellular Fractions-Cholesterol was depleted from cells as described. Control cells and cells treated with M␤C were washed three times with ice-cold PBS ϩϩ (PBS containing 0.1 mM CaCl 2 and 1 mM MgCl 2 ) and incubated in ice-cold 1.5 mg/ml sulfo-NHS-biotin/PBS ϩϩ (Pierce) at 4°C for 30 min with gentle agitation. The cells were washed three times with ice-cold quench buffer (100 mM glycine in PBS ϩϩ ) and incubated in quench buffer for 30 min at 4°C with gentle shaking. The cell monolayer was rinsed three times with ice-cold PBS ϩϩ and then lysed in 1% Brij 58 as described above. Aliquots of lysates containing equal amounts of protein were centrifuged for 1 h at 101,300 ϫ g in a Discovery100-Sorvall ultracentrifuge at 4°C. The pellet ("rafts") was resuspended in the same lysis buffer used to lyse the cells, and the supernatant was placed in a fresh tube ("soluble" material). An aliquot of total, soluble, and rafts samples was incubated overnight at 4°C with equilibrated NeutrAvidin beads (60 g of protein/50 l of beads; Pierce), with end-over-end mixing. The supernatant (unbound fraction in, corresponding to cytoplasmic proteins) was saved in a fresh tube. The beads were washed three times in lysis buffer. Bound proteins (out, corresponding to cell surface proteins) were eluted by boiling the beads in SDS-sample buffer. Blots were acquired and analyzed using UVP BioImaging Systems and LabWorks software.

Cholesterol-dependent Association of SERT with Lipid Rafts
Is Important for 5HT Uptake Activity-We used the cyclic oligosaccharide methyl-␤-cyclodextrin (M␤C) to analyze the effect of cholesterol depletion on SERT function. T-REx-SERT cells were treated with various amounts of M␤C and incubated for 30 min at 37°C. Cell viability was not significantly affected by M␤C in the 1.25-20 mg/ml range (Fig. 1A), and 50% or more of cholesterol was sequestered by 10 -20 mg/ml M␤C (Fig. 1B). In T-REx-SERT cells [ 3 H]5HT uptake was reduced to less than 50% upon incubation with 10 -20 mg/ml M␤C (Fig. 1C). Saturation kinetic analysis of [ 3 H]5HT transport in these cells to obtain V max and K m values proved difficult, due to the high level of SERT expression in these cells, which results in a marked increase of K m values (24). Thus, we performed these experiments in HEK cells, which were transiently transfected with SERT. In accordance with previous results obtained under similar conditions (20), cholesterol depletion by M␤C in these cells resulted in a decrease of V max values in a concentrationdependent manner ( Fig. 1D: control, V max 8.4 Ϯ 0.2 pmol/min/ 10 6 cells; 10 mg/ml M␤C, V max 4.0 Ϯ 0.6 pmol/min/10 6 cells; 20 mg/ml M␤C, V max 3.0 Ϯ 0.6 pmol/min/10 6 cells). At the same time the affinity of [ 3 H]5HT uptake decreased slightly (control, K m 446 Ϯ 73 nM; 10 mg/ml M␤C, K m 819 Ϯ 452 nM; 20 mg/ml M␤C, K m 938 Ϯ 717 nM). 5HT uptake through rat SERT depends on the presence of Na ϩ and Cl Ϫ (29). To determine whether cholesterol depletion affected serotonin transport even at low sodium concentration, 5HT uptake was measured in the presence of either 150 or 10 mM Na ϩ . At 150 mM sodium treatment with M␤C reduced the transport of serotonin, whereas at 10 mM sodium 5HT uptake was not significantly affected by the reduction of cholesterol (Fig. 1E). As shown in Fig. 1F, 5HT transport rate was reduced in cholesterol-de-pleted cells both at high (150 mM) and low Cl Ϫ (10 mM) compared with the controls.
Cholesterol is a major component of lipid rafts. A feature of these microdomains is their resistance to extraction by nonionic detergents at 4°C, such as Triton X-100 (16,18). This property has been widely used to isolate rafts: due to the high lipid content, detergent-resistant complexes manifest high buoyancy upon centrifugation in a density gradient, and thus float in the upper layers of the density gradient. To examine whether the apparent importance of cholesterol for 5HT uptake activity reflects the association of SERT with lipid microdomains, we examined the solubilization profile of the transporter in non-ionic detergents. Cells were lysed at 4°C in the appropriate detergent, homogenized, and subjected to centrifugation at 830 ϫ g for 10 min to clear the lysates from nuclei and unbroken cells. Cell lysates were placed at the bottom of a discontinuous sucrose gradient (5% to 30% to 40%) and centrifuged at 134,400 ϫ g for 17 h at 4°C. Fully solubilized proteins remained in the bottom layers, while insoluble proteins associated with lipid microdomains floated in the upper layers at the 5-30% interface. SERT was fully solubilized by 1% Triton X-100, whereas about 40% of SERT was insoluble in 1% Brij 58 and therefore recovered in the upper layers of the sucrose density gradient ( Fig. 2A, top panel). The distribution of SERT along the sucrose gradient did not depend on the amount of transporter expressed by T-REx-SERT. Similar flotation pro- files where obtained when SERT expression was induced with either 10 ng/ml tetracycline (enough to achieve half-maximal 5HT uptake (24)) or with 1 g/ml tetracycline/5 mM sodium butyrate, which enhances SERT expression levels ϳ5-fold (data not shown). A similar flotation profile for the transporter was obtained when pSERT-FLAG was transiently transfected into HEK cells (data not shown). In contrast, another membrane protein, the Toll-like receptor (TLR)-4 transiently expressed in HEK cells, was fully solubilized by both detergents (Fig. 2A, bottom panel), indicating that SERT insolubility in Brij 58 is neither due to an incomplete solubilization of bulk membrane nor due to overexpression of a membrane protein per se. Similar flotation profiles for the serotonin transporter were obtained when Triton X-100 and Brij 58 were used in a range 0.5-3% w/v (data not shown), suggesting that variation of the detergent/protein mass ratio does not affect SERT solubility. The lipid microdomains marker flotillin-1 (30) was recovered in the upper fractions, whereas the transferrin receptor, TfR, a non-lipid rafts marker (31), was recovered at the bottom of the density gradient (Fig. 2B, middle and bottom panels).
If SERT insolubility in Brij 58 reflects its association with lipid rafts, the question arises whether cholesterol is needed to maintain the structure of these high buoyancy complexes. M␤C selectively depletes cholesterol from membranes, therefore disrupting lipid microdomains whose integrity depends on cholesterol. For this purpose, T-REx-SERT cells were incubated with 10 mg/ml M␤C for 30 min at 37°C and subsequently lysed in 1% w/v Brij 58 at 4°C. Flotation profiles for SERT on density gradient centrifugation revealed that cholesterol depletion by M␤C caused SERT to shift to the lower buoyant density frac-tions (Fig. 2B, top panel). Notably, flotillin-1 was still recovered in the upper fractions of the density gradient, suggesting that it may localize in a different subset of lipid microdomains, less sensitive to cholesterol levels. When cholesterol was sequestered from cells by a combination of 0.5% w/v Brij 58 and 0.2% w/v saponin, which is known to form insoluble complexes with cholesterol, SERT was effectively solubilized and no longer recovered in high buoyancy fractions of the sucrose gradient (Fig. 2C). Other tools to disaggregate lipid microdomains were investigated. However, similarly to what was previously reported by Schuck and coworkers (32), the metabolic inhibitors sphingomyelinase or fumonisin B 1 did not reduce the association of rafts markers with lipid microdomains in intact T-Rex-SERT cells (not shown).
Differential Solubilization of SERT in Various Detergents-The lipid and protein composition of lipid rafts has recently been shown to differ when different detergents are used (32,33). For this reason, we compared the solubilization of SERT in four non-ionic detergents having both different structure and hydrophilic-lipophilic balance (HLB) values (Table I): Triton X-100, Tween 80, Brij 58, and Tween 20. Critical micellar concentration (CMC) values should be taken into account when assessing the efficacy of different detergent in solubilizing a given protein. Thus, we compared the four detergents at concentrations corresponding to 10-, 100-, or 300-fold of their respective CMC. However, similar solubilization profiles were obtained for the three conditions. The subunit B of Cholera toxin binds to the ganglioside GM1 that localizes in lipid microdomains and is therefore widely used as a rafts marker (31). SERT was fully solubilized in Triton X-100, whereas GM1 was still recovered in the upper fractions. Similarly, SERT was fully solubilized in Nonidet P-40, a non-ionic detergent of similar structure to Triton X-100 (not shown). However, SERT was partially insoluble in Brij 58 (ϳ40%), in Tween 80, and in Tween 20 (Fig. 3A, left panel). The vast majority of actin was retrieved at the bottom of the gradient (Fig. 3A, right panel), supporting the conclusion that the insolubility of SERT is due to the association with low buoyancy lipid rafts and not with the cytoskeleton. The enrichment of GM1 in lipid microdomains varied as well as the distribution of SERT along the density gradient depending on the detergent used. Most (Ͼ50%) of CTxB was recovered in the upper fractions of the gradient when cells were lysed in Brij 58 (Fig. 3A, middle  panel). Notably, when cells were lysed in Tween 80 and Tween 20, a consistent amount of SERT and CTxB immunoreactivity was recovered in the pellet (P 1 ) resulting from partial cell lysis, even at detergent concentrations corresponding to 300-fold CMC (Fig. 3A).
To assess the condition at which SERT-lipid microdomains sediment, cells were lysed in 1% Brij 58, and cleared lysate was centrifuged at 15,000 ϫ g for 20 min or 101,300 ϫ g for 1 h. To monitor the sedimentation of lipid rafts, pellets (rafts) were resuspended in lysis buffer and supernatants were subjected to flotation experiment. Brij 58-resistant domains containing insoluble SERT and GM1 were only partially pelleted at 15,000 ϫ g for 20 min, whereas they sedimented after centrifugation at  101,300 ϫ g for 1 h (Fig. 2B). This indicates that SERT associates with Brij 58-resistant complexes of different sizes: a smaller type that sediments at 101,300 ϫ g (for 1 h) and a larger type that sediments at 15,000 ϫ g (for 20 min). Using this method, the majority of SERT was found in the raft fractions, whereas by density gradient centrifugation only ϳ40% SERT was found in lipid rafts fractions. Such a discrepancy between flotation density gradient compared with sedimentation has been previously noted for the NPC1 protein (34). It is possible that the association of SERT with lipid microdomains of different sizes led to an underestimation of the amount of transporter associated to lipid rafts, and longer centrifugations may be required for these rafts to reach the equilibrium in the sucrose density gradient.
It has been shown that detergents differ in their ability to enrich lipid-associated proteins in different cell lines, seemingly due to their ability to enrich different types of lipids (32). The phospholipids/cholesterol ratio of lipid microdomains isolated from T-Rex-SERT cells by Triton X-100 or Brij 58 was examined. Lipid raft membranes were prepared by pelletting lysates at 101,300 ϫ g for 1 h at 4°C. Fig. 3C shows that the lipid rafts membranes isolated with the two detergents contained cholesterol to a similar extent, whereas phospholipids levels were ϳ10-fold higher when Brij 58 was used compared with Triton X-100. Even though nothing can be said about the nature of these phospholipids (whether they are sphingomyelins and/or glycerophospholipids), they seem to be important for preserving the association of SERT to lipid rafts in fibroblasts.

Disruption of Lipid Rafts Results in a Decreased SERT Turnover Number and in the Redistribution of SERT along the Plasma Membrane-
The saturation analysis of 5HT transport reveals that cholesterol removal from cell membranes causes a decrease in the maximum transport velocity (V max ) and changes in affinity (K m ; Fig. 1D). The observed change of V max could be produced either by reducing the number of SERT on the plasma membrane or by decreasing the turnover number of transported 5HT. To discriminate between these two possibilities, we subjected cells to cholesterol depletion followed by cell surface biotinylation and lipid rafts isolation. T-REx-SERT cells were incubated with 10 mg/ml M␤C for 30 min at 37°C to deplete membranes of cholesterol. Cells were then placed on ice and incubated with 1.5 mg/ml impermeant sulfo-NHS-biotin for 30 min at 4°C to label proteins at the cell surface, followed by quenching of unbound biotin. Cells were subsequently lysed in ice-cold 1% Brij 58 as described under "Experimental Procedures" and lipid rafts were isolated by centrifugation at 101,300 ϫ g for 1 h.
Cholesterol depletion led to a partial solubilization of SERT, due to disaggregation of lipid microdomains (Fig. 4, A and D), as seen before. Raft-associated SERT was found in both the intracellular and the cell surface fractions. Cholesterol depletion did not affect the amount of SERT localized on the plasma membrane (Fig. 4, B and E). By lowering cellular cholesterol levels SERT was still present at the cell surface, but it was shifted into detergent-soluble membrane domains (Fig. 4, C  and F). Therefore, the V max reduction observed in [ 3 H]5HT transport following cholesterol depletion appears to be caused

FIG. 3. SERT is insoluble in various non-ionic detergents.
A, lysates were obtained using the indicated detergent at a concentration corresponding to 300-fold CMC (see Table I). Precleared lysates were then subjected to flotation experiments. The eight fractions, the pellets, and an aliquot of the precleared lysate (designated "tot") were analyzed by SDS-PAGE followed by immunoblotting for SERT (left panel), GM1 (anti-CTxB, 1:1,000, middle panel), and actin (anti-actin, 1:5,000, right panel). Fractions 1 and 2 (indicated in bold) contain detergent-insoluble complexes (rafts). B, T-REx-SERT cells were incubated with 3 g/ml CTxB and lysed in 1% Brij 58. Lysates were centrifuged at 15,000 ϫ g for 20 min or 101,300 ϫ g for 1 h. Resulting rafts pellets were re-solubilized in solubilization buffer and subjected to sucrose gradient centrifugation as described under "Experimental Procedures." Gradient fractions, pellet P, rafts, and an aliquot of precleared lysate (tot) were analyzed by immunoblotting for SERT and GM1. Western blots shown for A and B are representative of three independent experiments. C, phospholipid and cholesterol concentration in lipid rafts. T-Rex-SERT cells were lysed with Brij 58 or Triton X-100. Lipid rafts were pelleted by sedimentation and resuspended in solubilization buffer. Each sample was divided into two equal aliquots. The first aliquot was used to determine protein and cholesterol concentration. The second aliquot was used for lipid extraction, and total phosphorus was determined as described under "Experimental Procedures." The cholesterol and phospholipids content in each sample was normalized to the protein concentration of the sample. Plot shows the mean of two independent experiments Ϯ S.E. by a change in the turnover number and not by a decrease of transporter molecules at the cell surface.
The lipid raft hypothesis proposes that phase separation causes cholesterol, sphingolipids, and proteins to aggregate forming distinct microdomains within the plane of the membrane determining their insolubility in non-ionic detergents (35). Indeed the segregation of raft markers in distinct clusters on the plasma membrane is lost upon cholesterol depletion (31). Cross-linking with antibodies is used to stabilize lipid microdomains on the cell surface by clustering together raft components in larger patches of micron size, facilitating their microscopic detection (31,36). To test whether in intact cells SERT colocalizes with lipid rafts, we clustered GM1 with CTxB and anti-CTxB antibody, fixed the cells, and visualized the immunofluorescent staining for SERT and GM1 by confocal microscopy. We found that SERT labeling is localized preferentially to clusters in the areas of cross-linked GM1 patches (Fig. 5). In cells treated with 10 mg/ml M␤C, SERT and GM1 immunoreactivity seemed essentially diffusely distributed. When cholesterol was further reduced with 20 mg/ml M␤C, SERT and GM1 staining appeared uniformly localized along the plasma membrane. At this M␤C concentration cells show morphological changes, such as rounding-up and a loss of branches as well as a tendency of detaching more easily from the support (Fig. 3D,  third row). This is likely to be caused by loss of interaction between adhesion molecules associated with lipid microdomains and the extracellular matrix (37,38).
SERT Associates with Lipid Rafts in Brain-To investigate whether the serotonin transporter is associated with lipid microdomains in native tissues, we examined the solubility and flotation behavior of SERT in rat brain. Midbrain and forebrain regions, rich in serotonergic neurons, were homogenized and lysed with Triton X-100, Tween 80, Brij 58, or Tween 20 at concentrations corresponding to 300-fold CMC (see Table I).
After detergent-extraction, soluble and insoluble materials were separated by density gradient centrifugation. Samples were then analyzed by immunoblotting. SERT was identified as a single band of ϳ76 kDa and was only weakly associated with Triton X-100-resistant complexes (Fig. 6A). SERT was recovered in the raft fractions (1 and 2) upon lysis with Tween T-REx-SERT cells were incubated with the indicated amount of M␤C for 30 min at 37°C, washed, and processed for cell surface biotinylation as described under "Experimental Procedures," and rafts were isolated by differential centrifugation. A and D, soluble material (sol) and lipid raft fractions were analyzed by immunoblotting for SERT and actin. Actin was not detected in raft fractions, indicating that no soluble material was sedimented following differential centrifugation. B and E, an aliquot of each cleared lysate (tot) was then incubated with Neu-trAvidin beads to separate cell surface proteins (out) from intracellular proteins (in). C and F, similarly, soluble and raft fractions obtained by differential centrifugation were incubated with NeutrAvidin beads to monitor changes in SERT distribution upon cholesterol depletion. Actin was not detected in the cell surface (out) fractions indicating that only cell surface proteins were labeled by sulfo-NHS-biotin. Blots shown are representative of three independent experiments. The actin immunolabeling of the tot samples was used to normalize SERT signal, and the data from all three independent experiments were quantified, plotted, and are shown in D-F. 80, Tween 20, or Brij 58. Thus, SERT from rat brain tissue showed a similar pattern of association with lipid rafts compared with heterologously expressed SERT in cultured cells.
The SNARE protein syntaxin 1A has been shown to associate with lipid microdomains in PC12 cells, whereas its binding partner Munc 18-1 (nSec1) is not associated with rafts (39,40). Similarly, in rat brain Munc 18-1 was fully solubilized by each of the four detergents employed, whereas syntaxin 1A exhibited high buoyancy and was recovered in the upper fractions of the density gradient (Fig. 6A). The similar biochemical behavior of SERT and syntaxin 1A suggests that in brain they may localize to the same lipid environment, whereas Munc 18-1 appears to be spatially segregated in a different membrane environment. As seen in T-Rex-SERT cells, the raft-protein flotillin-1 was found in the upper fraction of the gradients, whereas the non-rafts protein TfR was recovered at the bottom (Fig. 6A). Functional consequences of cholesterol depletion on the activity of native SERT were investigated in the neuronal cell line RN46A-B14 (41). The 5HT transport rate was reduced by 40% in neurons treated with 10 mg/ml M␤C compared with the controls (Fig. 6B), similarly to results obtained in heterologous expression systems. DISCUSSION The lateral organization of membranes in rafts proposed by Simons and Ikonen (18) is of particular interest for membrane protein function, because these lipid microdomains have been shown to selectively incorporate or exclude proteins. Our data presented in this study provide biochemical and morphological evidence that SERT is associated with lipid microdomains and that this association is required for efficient 5HT transport activity.
By means of flotation experiments, we examined SERT resistance to solubilization in different non-ionic detergents. SERT is solubilized in Triton X-100, and it is partially insoluble in Brij 58, Tween 20, and Tween 80. From our data it seems that SERT detergent insolubility is associated with an HLB value of Ն15, similarly to the microvillar protein prominin (42). However, the structure of detergents may also affect SERT solubilization (i.e. presence of a phenol ring between the hydrophobic tail and the hydrophilic head of Triton X-100 and Nonidet P-40). The flotation profiles of both SERT and GM1 change markedly when different detergents are used. This may arise from a differential ability of detergents to insert into the tight acyl chain packing of lipids in rafts, thus resulting in the isolation of various subpopulations of microdomains, each characterized by a specific lipid composition (32). The lipid microdomains isolated using Brij 58 contain ϳ10-fold more phospholipids than the lipid microdomains obtained with Triton X-100. A more detailed analysis of the composition of these Brij 58 rafts is necessary to elucidate the importance of specific lipids. Interestingly, our data are consistent with previous findings of Talvenheimo and Rudnick (43). The authors tested the ability of several detergents in solubilizing the imipramine binding sites of SERT-expressing platelets. They found that solubilization in Triton X-100 inactivates SERT, whereas in other detergents, such as Brij 58, Tween 20, and Tween 80, SERT is not solubilized and imipramine-binding activity is recovered in particulate material that sediments at 226,000 ϫ g for 30 min, which is quite similar to the conditions that we found to be necessary to sediment lipid rafts (101,300 ϫ g for 1 h, Fig. 3B). The similarity between our data and the data of Talvenheimo and Rudnick (43) led us to hypothesize that the population of SERT molecules found to be associated with lipid microdomains were in an active form. This hypothesis has been corroborated by the finding that disaggregation of lipid rafts leads to the solubilization of SERT (Fig. 2, B and C) and to the concomitant impairing of 5HT transport (Fig. 1C).
Furthermore, we have shown here that disruption of rafts by cholesterol depletion affects the V max of 5HT transport (Fig.  1D). This could result either from a reduction of transporters located on the plasma membrane or from a decrease in the number of transported 5HT molecules per unit of time ("turnover number"). We show that, upon desegregation of lipid rafts, the relative amount of SERT located on the cell surface is unchanged. Thus, the V max reduction caused by cholesterol depletion appears to reflect a change in the turnover number of 5HT transport. Using a detergent-free method, we also present morphological evidence that in intact cells SERT is spatially segregated in lipid rafts and upon cholesterol depletion is redistributed evenly along the plasma membrane.
From our data we can conclude that SERT requires the association with lipid rafts to catalyze the transport of 5HT: the loss of interaction with lipid microdomains is detrimental to SERT function, possibly due to conformational changes that mainly affect the translocation of the substrate across the plasma membrane. The increase in K m for 5HT, although only slight, may provide an additional indication for a structural change of the transporter. The alteration of the physical properties of rafts (e.g. bilayer thickness and membrane curvature) may cause SERT to assume an inactive conformation, similarly to what was observed for some ion channels (44,45). Although 5HT can still bind with high affinity to SERT, transport is severely affected. Lipid raft association might therefore be important for conformational changes associated with the translocation process, involving steps subsequent to 5HT binding. Such changes could affect, for example, ion binding and could be caused either by a loss of direct interaction with cholesterol or by a loss of interaction between SERT and other raft proteins. Interestingly, the responsiveness to cholesterol depletion is lost at low extracellular Na ϩ concentration, whereas it is not affected by lowering the extracellular Cl Ϫ concentration. The lack of response to M␤C treatment at 10 mM Na ϩ suggests that at low extracellular Na ϩ concentration the 5HT transport may be independent of a conformational state stabilized or facilitated by lipid rafts association or by cholesterol itself.
In neurons, SERT functions in the reuptake of 5HT following its release at the presynaptic membrane. It has been shown that SERT is present all along the axolemma and in axon terminals but is present at very low levels at the plasma membrane of cell soma and dendrites (46,47). This suggests that SERT function is tightly linked to its position in specific regions of the plasma membrane. Thus, a regulated mechanism would be required to target SERT to the plasma membrane of neurons in a polarized manner. In the past few years a body of data has become available on the role of lipid rafts both in neuronal trafficking and in cell signaling (48 -50). We demonstrated that SERT associates with lipid rafts also in rat brain. Similarly to SERT expressed in T-REx-SERT cells, neuronal SERT is recovered in low buoyancy raft fractions when solubilization is carried out in Brij 58, Tween 20, and Tween 80 and to a lesser extent in Triton X-100 (Fig. 6A). Moreover, upon cholesterol depletion the velocity of 5HT transport is decreased in neurons (Fig. 6B). We hypothesize that the localization of SERT in neuronal lipid rafts physiologically regulates its function within the axolemma and at nerve terminals. Plasma membrane-bound SERT may be found in equilibrium between an active raft-associated state and an inactive state residing outside lipid rafts. The relative distribution between these two states might depend on the physiological situation; for example, the raft-associated transport-competent SERT conformation might be favored following neurotransmitter release.
The question is: what triggers the association with lipid microdomains? Cholesterol appears to regulate different steps of neurotransmitter release. In the central nervous system, cholesterol is released by glial cells through apoE-containing lipoproteins and is taken up by neurons, where it contributes to an increase in the number of synapses (51). Cholesterol also positively regulates the formation of synaptic vesicles (52) and enhances the assembly of SNARE proteins in defined synaptic vesicles-docking and fusion sites (40). Thus, by regulating the levels of cholesterol in neurons, SERT activity could be modulated at the synapse by a dynamic association with lipid rafts. Experiments are currently underway to investigate this hypothesis.
Our data also show that in cultured cells the population of SERT-containing rafts is not homogenous in terms of size. In fact, a fraction of lipid rafts is sedimented by centrifugation at 101,300 ϫ g for 1 h ("small" rafts) but not at 15,000 ϫ g for 20 min ("large" rafts). Similar raft subsets have been identified for the protein prominin and are involved in the transport of the protein from the trans-Golgi/trans-Golgi network to the apical plasma membrane (42). Our finding that SERT rafts are present in intracellular membranes (Fig. 3B) may suggest that lipid microdomains play a role in trafficking SERT to the plasma membrane and is consistent with a possible involvement of lipid rafts in the apical delivery of SERT to the axolemma and axon terminals. Further studies on the involvement of lipid microdomains in SERT trafficking and targeting are currently being carried out in our laboratory.
The modulation of transport activity, in particular the regulation of subcellular localization, has recently been studied intensively in several members of the neurotransmitter transporter family (for review see Refs. [53][54][55]. However, the molecular mechanisms of transporter regulation remain poorly understood. The modulation of SERT activity through lipid raft association reported here suggests a novel, previously unappreciated regulatory mechanism based on lipid-protein interaction. Another transporter, i.e. the ␥-aminobutyric acid transporter GAT1 was previously shown to require cholesterol for functional reconstitution (56). During the revision of the manuscript, Jayanthi and coworkers published evidences (57) regarding the association of the norepinephrine transporter with lipid rafts in trophoblasts. This interaction is shown to mediate the norepinephrine transporter internalization upon protein kinase C activation.
Thus, the functional regulation by lipid microdomains as shown here for SERT, may also prove relevant for related neurotransmitter transporters and might help to explain the underlying mechanism of previously observed regulatory processes, such as those involving syntaxin 1A. Together with other SNARE proteins, syntaxin 1A was shown to be associated with lipid rafts in PC12 cells (39,40). Our data suggest that SERT and syntaxin 1A may share the same lipid environment in rat brain. Syntaxin 1A has been shown to positively affect cell surface expression of SERT and to regulate two conducting states of the transporter in thalamocortical neurons via a direct interaction with the N terminus of SERT (13,58). Thus, syntaxin 1A might regulate SERT, and possibly related transporters, within lipid rafts.