Glycosyl Modification Facilitates Homo- and Hetero-oligomerization of the Serotonin Transporter

The serotonin transporter (SERT) is an oligomeric glycoprotein with two sialic acid residues on each of two complex oligosaccharide molecules. In this study, we investigated the contribution of N-glycosyl modification to the structure and function of SERT in two model systems: wild-type SERT expressed in sialic acid-defective Lec Chinese hamster ovary (CHO) cells and a mutant form (after site-directed mutagenesis of Asn-208 and Asn-217 to Gln) of SERT, QQ, expressed in parental CHO cells. In both systems, SERT monomers required modification with both complex oligosaccharide residues to associate with each other and to function in homo-oligomeric forms. However, defects in sialylated N-glycans did not alter surface expression of the SERT protein. Furthermore, in heterologous (CHO and Lec) and endogenous (placental choriocarcinoma JAR cells) expression systems, we tested whether glycosyl modification also manipulates the hetero-oligomeric interactions of SERT, specifically with myosin IIA. SERT is phosphorylated by cGMP-dependent protein kinase G through interactions with anchoring proteins, and myosin is a protein kinase G-anchoring protein. A physical interaction between myosin and SERT was apparent; however, defects in sialylated N-glycans impaired association of SERT with myosin as well as the stimulation of the serotonin uptake function in the cGMP-dependent pathway. We propose that sialylated N-glycans provide a favorable conformation to SERT that allows the transporter to function most efficiently via its protein-protein interactions.

SERT, NET, and DAT proteins share extensive sequence homology (5,9,11,16,17) with several common structural characteristics, including homo-oligomeric properties (18 -25, 61) and multiple sites for N-linked glycosylation within a large extracellular loop between transmembrane domains 3 and 4 (9). N-Glycosylation of SERT (27)(28)(29)(30)(31)(32)(33), NET (34 -36), and DAT (24 -26) is important for their neurotransmitter uptake functions (9); however, neither the mechanism by which N-glycosyl groups contribute to the serotonin uptake function nor the degree of glycosyl modification required to maintain the efficient uptake function of SERT is known. Initially, Gielen and Viefhöfer (27), Dette and Wesemann (28), and Szabados et al. (29) reported that, at nerve ending membranes, sialic acidcontaining structures involved in the serotonin uptake process and sialic acid moieties contribute to the transport function of SERT. Later, during the purification of SERT from human platelets, Launay et al. (30) demonstrated that SERT has two sialic acid residues on each complex N-linked oligosaccharide molecule.
In general, N-glycosylation has an important role in the quality control pathway that ensures correct folding and processing of membrane proteins (37)(38)(39)(40)(41). Defects in the glycosylation pathway retain the glycoproteins in the endoplasmic reticulum (42,43). However, removal of glycosylation sites neither keeps SERT in the endoplasmic reticulum nor impairs its expression on the plasma membrane, but reduces its uptake function (31)(32)(33). In this study, using recombinant SERT overexpressed in Lec4 cells (44 -47) and N-glycosylation site mutants of SERT, QQ (31), overexpressed in Chinese hamster ovary (CHO) cells, we investigated the roles of sialylated Nglycans in SERT function. Lec4 cells are N-acetylglucosamine transferase V-defective, and Lec2 cells are nucleotide cytidine monophosphosialic acid (CMP-sialic acid) transferase-defective CHO cells. Glycoproteins and glycolipids synthesized in Lec2 and -4 mutants have defects in sialic acid content (44 -47). Our overall findings indicate that (i) functional oligomerization requires sialylated N-glycans; (ii) sialylated glycans are not required for normal cell-surface expression; and (iii) sialylated glycans are also required for normal cytoskeletal associations of SERT with myosin IIA. Based on these findings, we propose a structural role for sialylated N-glycans in conferring an optimum conformation to SERT that facilitates homo-oligomerization upon biosynthesis and, in turn, exposes the domain(s) for myosin cytoskeletal associations. To explore if the impaired oligomeric properties of SERT were due to the lack of sialylation or branching, we also used Lec2 cells and found the same result with those using Lec4 cells. Utilizing the Lec mutant cells in this study was a novel approach to elucidate the significance of sialic acid residues in SERT function.

EXPERIMENTAL PROCEDURES
Materials-The mutant form of rat SERT (rSERT; Asn 208 and Asn 217 of the N-linked glycosylation consensus sequences replaced with glutamine residues by site-directed mutagenesis), the QQ construct in plasmid pCGT148 (31), was a generous gift from Dr. C. G. Tate  Plasmids, Constructs, and Cell Line Expression Systems-QQ and rSERT were subcloned into the EcoRV/XbaI and KpnI/NotI sites, respectively, of the pcDNA3 expression vector, placing rSERT expression under the dual control of the cytomegalovirus and T7 RNA polymerase promoters, suitable for expression in CHO cells (52). A Myc or FLAG epitope was tagged to the N and C termini of rSERT and QQ, respectively.
Lec and CHO cells were maintained in ␣-minimal essential medium, and JAR cells in RPMI 1640 medium. All media were supplemented with 10% fetal bovine serum, 2 mM L-glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin. Neither CHO nor Lec cells endogenously express rSERT proteins. To measure whole cell and surface expression and protein-protein interactions, these cells were transiently transfected with rSERT constructs using a 1:2 ratio of LipofectAMINE 2000 reagent in Opti-MEM. Cells were used in biotinylation, immunoblotting, and immunoprecipitation (IP) assays 24 h post-transfection.
Anti-SERT Antibody-Blakely and co-workers (53) successfully raised an antibody (CT2) against the C-terminal 34 amino acids of SERT. Using this approach as a guide, Proteintech Group, Inc. (Chicago, IL) synthesized a peptide corresponding to the last 26 amino acids of the SERT C terminus (positions 586 -630) and generated a polyclonal anti-body against this synthetic peptide. This is a portion of the protein not affected by glycosylation mutation and thus should recognize wild-type and QQ transporters equally well. The synthetic peptide sequence (positions 586 -630) matches the C terminus of rSERT, a highly conserved region across different SERT species, but divergent from other gene family members (5).
We first purified this antibody by standard affinity purification via peptide-Sepharose procedures (54) and then biotinylated the purified antibody following the manufacturer's instruction (Amersham Biosciences). Briefly, the synthetic peptide corresponding to the last 26 amino acids of the C terminus of SERT was dissolved in coupling buffer (0.1 M NaHCO 3 and 0.5 M NaCl (pH 8.0)) and bound to CNBr-activated Sepharose 4B beads (Amersham Biosciences) that were previously washed with 1 mM HCl by overnight incubation at 4°C. First, the column was packed, and crude antibody was run through the column. After washing the beads with 10 column volumes of phosphate-buffered saline, the antibody was eluted sequentially with 0.1 M glycine (pH 3.0) to tubes containing 50 ml of 1 M Tris (pH 8.9). The protein concentrations of each fraction were obtained using the BCA protein assay kit, and protein-containing fractions were stored at Ϫ80°C.
Immunoblot Analysis-Following transfection, cells were solubilized in phosphate-buffered saline containing 0.44% SDS, 1 mM phenylmethylsulfonyl fluoride, and protease inhibitor mixture. The protease inhibitor mixture contained 5 mg/ml pepstatin and 5 mg/ml leupeptin, and 5 mg/ml aprotinin was included in each lysis buffer (18), which also contained the alkylating agent N-ethylmaleimide (NEM) at a final concentration of 5 mM to prevent oxidation and formation of nonspecific disulfide bonds during lysis and to retain the native monomeric structures in the gel (21). Samples were analyzed by 10% SDS-PAGE (55) and transferred to nitrocellulose membrane (56). Immunoblot analysis was performed first with anti-Myc (diluted 1:2500), biotinylated anti-FLAG (diluted 1:1000), or biotinylated anti-SERT (diluted 1:1000) antibody and then with horseradish peroxidase-conjugated streptavidin or horseradish peroxidase-conjugated anti-mouse or anti-rabbit secondary antibody (diluted 1:10,000), respectively. The signals were visualized using the ECL Western blotting detection system. Blots were visualized using a VersaDoc 1000 gel visualization and analysis system (Bio-Rad).
Co-immunoprecipitation-Protein-protein transporter interaction was demonstrated by coexpressing two different protein forms in CHO or Lec cells at a 1:1 ratio. Following transfection, CHO, Lec, or JAR cells were treated with 10 mM NEM for 30 min. Cells were lysed in IP buffer (55 mM triethylamine (pH 7.5), 111 mM NaCl, 2.2 mM EDTA, and 0.44% SDS ϩ 1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, and protease inhibitor mixture) (18) containing 5 mM NEM. Cell lysates were first precleared by incubation with nonimmune rabbit serum and protein A for 1 h and then centrifuged. The precleared lysate was combined with an equal volume of a 1:1 slurry of rabbit anti-mouse protein A-Sepharose beads as described previously for primary IP with the antibody (21) and then mixed overnight at 4°C. The lysate immune complexes were recovered by brief centrifugation in a bench-top microcentrifuge (Beckman), washed several times with high-and low-salt IP buffers, and eluted in Laemmli sample buffer (50 mM Tris-HCl (pH 6.8), 2% SDS, 0.1% bromphenol blue, 10% glycerol, and 1% ␤-mercaptoethanol). For the nonreducing condition, 5 mM NEM was included in Laemmli sample buffer containing no ␤-mercaptoethanol. Samples were separated on a 10% SDS-polyacrylamide gel (55). After electrophoresis, gels were analyzed by immunoblotting with either biotinylated monoclonal anti-FLAG antibody (diluted 1:1500) or polyclonal anti-myosin IIA antibody (diluted 1:1000) to demonstrate self-association or myosin IIA association, respectively. The signals were developed with the ECL detection system.
Cell-surface Biotinylation-Following transfection, cells were biotinylated with the membrane-impermeant biotinylating reagent NHS-SSbiotin as described previously (18,21). This reagent selectively modifies only the external lysine residues on membrane proteins. All SERT proteins used in this study, with or without epitope tagging, contained the same number of external lysine residues and were expected to react equally with NHS-SS-biotin.
Following biotinylation, cells were treated with 100 mM glycine to complete quenching of the unreacted NHS-SS-biotin and lysed in Trisbuffered saline containing 1% SDS, 1% Triton X-100, and protease inhibitor mixture/phenylmethylsulfonyl fluoride. The biotinylated proteins were recovered with streptavidin-agarose beads during overnight incubation. Biotinylated proteins were eluted and separated on SDSpolyacrylamide gel. Immunoblot analysis was performed using anti-SERT, anti-Myc, or anti-FLAG antibody. The primary antibody was detected using horseradish peroxidase-conjugated secondary antibodies and the ECL detection system.

PNGase F Treatment-CHO and Lec cells transiently transfected
with transporters were first lysed in IP buffer containing NEM at a final concentration of 5 mM. The cell lysate was then treated with rabbit anti-mouse protein A-Sepharose beads coated with biotinylated anti-SERT antibody. The lysate immune complexes were recovered, washed, and then eluted in sample buffer supplemented with 0.4 units/ml PNGase F. After a 3-h incubation at 37°C (34,35), the reaction mixture was separated by SDS-PAGE, and immunoblot analysis was performed using horseradish peroxidase-conjugated streptavidin and the ECL blotting system. Transport Assay-CHO and Lec cells in 24-well culture plates were infected with recombinant vTF-7 vaccinia virus, which works with T7 RNA polymerase (52), and transfected with plasmids bearing mutant SERT cDNA under the control of the T7 promoter as described previously (18). Transfected cells were incubated for 16 -20 h at 37°C and then used to assay serotonin uptake by incubation with 20.5 nM [1,2-3 H]serotonin (3400 cpm/pmol) in phosphate-buffered saline containing 0.1 mM CaCl 2 and 1 mM MgCl 2 for 10 min. The intact cells were washed quickly with ice-cold phosphate-buffered saline, lysed in SDS, transferred to scintillation vials, and counted (18). Cells with no plasmid DNA (mock-transfected) were negative controls.
Data Analysis-Nonlinear regression fits of experimental and calculated data were performed with Origin (MicroCal Software, Northampton, MA), which uses the Marquardt-Levenberg nonlinear least-squares curve-fitting algorithm. Each figure shows a representative experiment that was performed at least twice. The statistical analysis given under "Results" is from multiple experiments. Data with error bars represent means Ϯ S.D. of triplicate samples.

RESULTS
Glycosylation Pattern of SERT-We first evaluated expression of SERT proteins and their delivery to the cell surface in both model systems. After transfection, expression of wild-type SERT in sialic acid-defective Lec4 cells and the QQ mutant form in CHO cells was demonstrated by immunoblot analysis of cell-surface proteins labeled with extracellular NHS-SS-biotin and total cell extracts. Analysis of membrane proteins with anti-SERT antibody produced three major bands at 90, 65, and 60 kDa in the SERT construct-transfected CHO cells (Fig. 1, lane 1) and Lec4 cells (lane 5). Analysis of total cell extracts with biotinylated anti-SERT antibody also gave rise to the same resolution pattern for SERT in CHO cells (lane 2). The major observable difference between Lec4 and CHO cells is that the 60-kDa species was more abundant in Lec4 cells than in CHO cells, whereas the reverse was true for the 65-kDa species. Under our experimental conditions, sialic acid moieties do not contribute to the mobility of SERT upon SDS-PAGE. Control experiments in which either no biotinylation reagent was added or mock-transfected cells were used did not reveal the bands visible on these immunoblots (data not shown).
Removal of potential glycosylation sites during migration of transporter proteins was observed in two different approaches: (i) surface expression of QQ in CHO cells (Fig. 1, lane 4) and Lec4 cells (lane 7) and (ii) PNGase F treatment of SERT. PNGase F removed N-linked glycosylation sites from expressed proteins regardless of oligosaccharide residues.
PNGase treatment eliminated the bands of the 90-, 65-, and 60-kDa species of SERT and revealed a faster migrating single band at ϳ50 kDa (Fig. 1, lane 3), similar to the QQ transporter (lane 4). Migration of transporters from PNGase F-treated CHO-SERT (lane 3) and CHO-QQ (lane 4) confirmed that the 50-kDa species represents the non-glycosylated form of the SERT polypeptide. However, the 50-kDa species did not appear in the blotting of CHO-SERT cells. Consistent with previous studies (31)(32)(33), a defect in the glycosylation processing of SERT did not impair its membrane trafficking or assembly.
Comparison by Western blot analysis of the protein pool labeled with external NHS-SS-biotin (Fig. 1, lane 1) and total cell extracts (lane 2) demonstrated that, on the plasma membrane, the majority of wild-type transporters appeared at the 90-kDa band and that, in whole cell extracts, 65-and 60-kDa species were more abundant. However, all three of these SERT species were reduced to 50 kDa in the presence of PNGase F; therefore, we believe that SERT proteins are either differentially or partially glycosylated during their biosynthesis process in CHO cells.
Glycosylation Pattern and Alteration in Uptake Function of SERT-To examine the nature of the defect in the glycosylation mutant, serotonin uptake efficiencies of transporters were characterized in two model systems (QQ in CHO cells and wild-type SERT in Lec4 cells) and compared with the uptake efficiency of wild-type SERT expressed in CHO cells (Fig. 2). Overall, our results show that the wild-type transporter was 56% less active in Lec4 cells than in its parental cell line (CHO cells) and that SERT was 70% more active than QQ in CHO cells. However, in Lec4 cells, this difference was only 37%. The difference in uptake efficiency of SERT and QQ in CHO cells was much higher than in Lec4 cells. There was a gradual decrease in SERT uptake function between the model systems.
Neither the mutations at both glycosylation consensus sequences nor the defect in the sialic acid residues totally eliminated transport function.
To explore whether the location or number of sialylated N-glycans affects the serotonin uptake ability of SERT, mutant transporters with a single glycosylation site (Q1 (N208Q) or Q2 (N217Q)) were prepared by site-directed mutagenesis and characterized in CHO cells for their serotonin uptake capacities. Interestingly, as the data in Fig. 2 show, a single deletion of either asparagine did not change the transport function.
Next, we attempted to determine whether the lower activities of transporters are due to lower protein expression levels. Surface and whole cell expression of mutant transporters (QQ, Q1, and Q2 in CHO cells and wild-type SERT in Lec4 cells) was measured and compared with that of wild-type SERT in CHO cells using the standard curve (Fig. 3A). Standard curves were generated with Myc-SERT-FLAG (wild-type SERT tagged with Myc and FLAG epitopes at the N and C termini, respectively). Different amounts (in a broad range of 1-25 g of protein) of detergent-soluble total cell extracts from CHO or Lec4 cells expressing Myc-SERT-FLAG were separated by SDS-PAGE, and immunoblot analysis was performed with biotinylated anti-SERT antibody (Fig. 3A, inset). The integrated density  4 and 7) transporters were expressed in either CHO or Lec4 cells using the cytomegalovirus transfection system and labeled with NHS-SS-biotin. They were fractionated by adsorption to streptavidin-agarose beads, and the biotinylated fraction was resolved by 10% SDS-PAGE and immunoblotted with affinity-purified anti-SERT antibody as described under "Experimental Procedures." To compare whole cell and plasma membrane expression differences, soluble cell lysate proteins from SERTexpressing CHO cells (lane 2) were resolved and immunoblotted with affinity-purified biotinylated anti-SERT antibody. N-Glycosyl groups were stripped from the SERT proteins in CHO cells (lane 3) and Lec4 cells (lane 6) with PNGase F treatment. Immunoblot analyses were done with horseradish peroxidase-conjugated streptavidin as described under "Experimental Procedures." All lanes contain protein recovered from the same number of cells equivalent to 30% of one well from a confluent 24-well dish. Three wells for each condition were pooled, and an aliquot of this mixture was run on the gel. The positions of molecular mass standards run on the same gel are shown in kilodaltons.
value for each band was converted to an equivalent amount of Myc-SERT-FLAG for anti-SERT antibody using the VersaDoc 1000 analysis system. Using this standard curve, the whole cell and plasma membrane expression of the transporters was then determined by quantitative Western blotting as the relative amounts of Myc-SERT-FLAG proteins (Table I).
In other cell contexts (9,31,33) and for other transporters such as human NET (34 -36) and DAT (24,25), it has been demonstrated that more than one glycosylated form can be expressed at the surface and that the non-glycosylated form will increase at the surface if the glycosylation sites are mutated. Because of that, the density of each band was calculated as the combination of the 90-, 60-, and 65-kDa bands for SERT in CHO and Lec4 cells.
Glycosylation and Homo-oligomeric Property of SERT-SERT is an oligomeric protein (18 -21), and as suggested previously, this is the most efficient form for serotonin uptake function (20). Glycosyl modification has been suggested to be an important factor for oligomerization of transmembrane proteins (48,49). Based on this fact, we tested whether glycosylation is a prerequisite modification for SERT monomers to associate physically and functionally in an oligomeric form.
To determine the impact of sialylated N-glycans on the physical association of SERT monomers, different epitope tags (Myc-SERT, SERT-FLAG, Myc-QQ, and QQ-FLAG) were prepared and characterized for their serotonin uptake functions and whole cell and surface expression in CHO and Lec4 cells to explore whether epitope tagging alters expression or membrane assembly of transporters. Standard curves were generated with Myc-SERT-FLAG-transfected CHO and Lec4 cells. Western blot analysis was performed with anti-Myc antibody (Fig. 3B, inset) or anti-FLAG antibody (Fig. 3C, inset). The integrated density value for each band was converted to an equivalent amount of Myc-SERT-FLAG for anti-Myc and anti-FLAG antibodies using the VersaDoc 1000 analysis system. Table I shows that neither the epitope tag nor mutation changed the expression efficiency or membrane trafficking of transporter proteins and that their expression level was similar to that observed for unmodified wild-type SERT.
Next, the self-association property of Myc-SERT with SERT-FLAG was analyzed in co-IP assays in CHO and Lec4 cells by cotransfecting the cells with a 1:1 ratio of Myc-SERT and SERT-FLAG constructs. After transfection, cells were treated with 10 mM NEM for 30 min (21). A prominent FLAG band was seen upon IP of CHO cells coexpressing Myc-SERT/SERT-FLAG (Fig. 4, lane 2), but not of Lec4 cells (lane 5). These results indicate that the two forms of SERT associate in CHO cells, but lose this ability in Lec4 cells in the absence of sialic acids. The associations of SERT proteins were detected as two bands at ϳ65 and 60 kDa.
We also attempted to characterize if self-association occurs between the forms of SERT proteins with molecular masses of 60 and 65 kDa. Thus, co-IP assays were done under nonreducing conditions created by excluding ␤-mercaptoethanol from and including 5 mM NEM in the lysis buffer. Upon resolving the proteins associated with Myc-SERT under nonreducing conditions, one major FLAG band appeared at ϳ90 kDa (Fig. 4, lane  3). These results suggest that the 90-kDa species is the one that associates to form homo-oligomeric transporters (lane 2). Under similar nonreducing conditions, resolution of the SERT protein still gave rise to three species (data not shown), but upon co-IP, appeared only as 90-kDa bands.
To determine the effect of sialylated N-glycans (i.e. whether the individual monomers must be modified to associate in oligomeric form) in the physical association of SERT proteins, CHO cells were cotransfected (1:1 ratio) with one wild-type and one mutant form of the transporter (Myc-SERT/QQ-FLAG) (Fig. 4, lane 1

) or with both forms as mutant constructs (Myc-QQ/QQ-FLAG) (lane 4). After IP with anti-Myc antibody and resolution of the Myc-SERT-bound proteins under nonreducing conditions, a FLAG band with a very low intensity was detected in Myc-SERT/QQ-FLAG-transfected CHO cells (lane 1).
Our results imply that, in the pool of expressed QQ and SERT transporters, only wild-type SERT proteins associate. However, as they represent half of the total amount of expressed transporters, the intensity of the FLAG band detected by Western blotting was much weaker (lane 1) than that of both forms of glycosylated wild-type SERT protein interactions (lane 3). As a control for nonspecific precipitation of SERT-FLAG, we performed the same analysis using cells expressing only SERT-FLAG. A FLAG band was not detected on the blot (lane 6), indicating that SERT-FLAG did not bind nonspecifically to the beads or to anti-Myc antibody.
Glycosylation and Functional Association between SERT Proteins-Subsequently, the functional association between two SERT monomers was tested in CHO and Lec4 cells using our previous techniques (18,21), that were originally used by MacKinnon (57) with potassium channel. Sensitive (wild-type) and toxin-insensitive mutant forms were mixed, and the toxin sensitivity of the mixture was analyzed to determine the subunit stoichiometry. In a similar fashion, the functional consequences of homo-and hetero-oligomerization between SERT

FIG. 3. Standard curves for quantitation of SERT mutant expression. The indicated amounts of detergent-solubilized cell lysate from
Myc-SERT-FLAG-expressing CHO and Lec4 cells were separated by SDS-PAGE and visualized by Western blot analysis. The integrated density value for each band was converted to an equivalent amount of Myc-SERT-FLAG for antibodies (Ab) against peptide-purified and biotinylated SERT (A), Myc (B), and FLAG (C). In this way, the relative amount of SERT constructs was determined in both total cell lysate and the cell-surface pool isolated using streptavidin-agarose (Table I), and the expression levels for SERT forms could be compared even though they were detected with different antibodies. Immunoblots were quantitated using a the VersaDoc 1000 system.

TABLE I Transport function and expression characteristics of SERT and mutants in CHO and lec4 cells
SERT and the mutants were expressed in CHO and Lec4 cells under the control of the cytomegalovirus and T7 promoters. Preparation of mutants, transport measurement, and quantitation of expressed proteins were performed as described under "Experimental Procedures." Rate of uptake is expressed as means Ϯ S.D. of triplicate determinations from three independent experiments. They were obtained by nonlinear regression analysis using the Origin plotting program (MicroCal Software). CHO (18) and NET (21) proteins were demonstrated.
To study the effect of sialylated N-glycans on the functional association of SERT proteins, we used the previously reported SERT monomers: Res-FLAG (resistant to MTSEA inactivation and with a FLAG epitope tag at the C terminus) and Sens-Myc (sensitive to MTSEA inactivation and with a Myc epitope tag at the N terminus). MTSEA inactivates the serotonin uptake function of Sens-Myc, but not of Res-FLAG.
If the physical association between these two forms has no effect on transport function, the amount of inactivation in a mixture should be equal to the amount of activity contributed by Sens-Myc. If the activity of the two forms is completely independent, as expected for a monomeric transporter, cells expressing intermediate mixtures should retain intermediate levels of transport after MTSEA treatment, proportional to the amount of Sens-Myc expressed.
In Fig. 5, three possible results are plotted according to predictions of the amount of activity remaining after MTSEA treatment. If no functional association occurs between Res-FLAG and Sens-Myc, the amount of inactivation should be equal to the amount of activity contributed by Sens-Myc, i.e. 50% inactivation by MTSEA should be seen in a 1:1 mixture of Sens-Myc and Res-FLAG plasmids, and the resulting activities should fall on the straight dotted lines in Fig. 5. However, if SERT proteins functionally associate in a dimeric form, then a 75% inhibition would be expected from the same 1:1 mixture. The dashed lines assume random association of Res-FLAG and Sens-Myc into dimers and that all activity of all dimers containing Sens-Myc is sensitive to MTSEA. The solid lines assume random dimeric formation and that only dimers containing two Sens-Myc subunits would be sensitive to MTSEA.
To investigate the role of glycosylation in the functional interaction of SERT monomers, Res-FLAG and Sens-Myc were coexpressed first in CHO (Fig. 5A) and then in Lec4 cells (Fig.  5B) by mixing them in different ratios. The activity profile of the mixed transporters was then characterized in the presence of MTSEA. In Fig. 5A, the squares represent the amount of serotonin transport measured in the absence of MTSEA reagent. After treatment with MTSEA, less of the total activity (circles) was inactivated than expected based on the content of Sens-Myc, which means that the experimental points deviated noticeably from the dotted and dashed lines, an indication that the two forms of SERT functionally interact in CHO cells (Fig.   5A). However, in Lec4 cells, the experimental points fell on the dotted line, showing a linear relationship between these two forms of SERT and that they function independently (Fig. 5B). Our data from three independent experiments fit best with the solid line in CHO cells, but with the dotted line in Lec4 cells. Therefore, in CHO cells, the experimental points coincide with the prediction for a two-hit inactivation process in which both subunits need to be modified for inactivation to occur; but in Lec4 cells, SERT monomers function independently.
In Lec4 cells, in addition to SERT, also other sialic acidmodified glycoproteins are defective. This fact limits our interpretation of the direct effect of sialic acid residues on the functional association of SERT monomers. To address this issue, we prepared mutant forms of the transporter (Res-FLAG-QQ and Sens-Myc-QQ) and monitored them in CHO cells. In this way, we could compare the effect of glycosyl modification in a system where it is known that the only difference is the glycosylation of SERT. Res-FLAG-QQ and Sens-Myc-QQ were mixed in different ratios in CHO cells, and the serotonin uptake by the mixed transporters was then measured in the presence of 0.25 mM MTSEA (Fig. 5C). In the absence of sialylated N-glycans, the SERT proteins did not functionally interact. The functional association of SERT proteins was also tested in CHO cells by coexpressing Res-FLAG-QQ and Sens-Myc or Res-FLAG and Sens-Myc-QQ and by measuring their serotonin transport functions in the presence of 0.25 mM MTSEA. Under both conditions, SERT proteins showed no functional association (data not shown).
Glycosylation and Hetero-oligomerization of SERT-Our data support the notion that modification with sialylated Nglycans is a requirement for SERT proteins to associate with each other and to function as homo-oligomeric forms. As both potential glycosylation sites are located at the extracellular side of the plasma membrane, we sought a cytosolic partner for SERT. Myosin is reported to be one of the protein kinase G-anchoring proteins (50), and SERT is phosphorylated by cGMP-dependent protein kinase G (15,51,58), which interacts with protein kinase G-anchoring proteins. Based on this information, myosin IIA-SERT association was tested in both model systems using co-IP assays.
First, CHO cells were transiently cotransfected with SERT and myosin IIA (59). To prevent the possible formation of nonspecific disulfide bonds, the cells and extracts were treated with 10 mM NEM beginning 30 min prior to IP. Co-IP assay was performed in the presence of 10 mM MgATP, which dissociates actin cytoskeletal protein from myosin, as described previously by Wei and Adelstein (59). Myc-SERT and associated proteins were eluted from Myc-coated protein A beads in sample buffer with 1% ␤-mercaptoethanol (21), and immunoblot analysis was done with anti-myosin IIA antibody. A 200-kDa band appeared on the blot, showing the ability of myosin IIA and SERT to directly associate (Fig. 6A, lane 1).
To understand the involvement of glycosylation in myosin IIA-SERT association, co-IP assays were performed with Lec4 cells expressing SERT (Fig. 6A, lane 3)  However, studying protein-protein interactions in heterologous expression systems suffers from drawbacks such as possible overexpression problems. To further explore the possible obstacles of a heterologous expression system, we tested myosin IIA-SERT interaction in placental choriocarcinoma JAR cells, which constitutively express both proteins (Fig. 6B, lanes  1 and 2). Immunoblot analysis of JAR cell lysate with biotin-  1-4 and 6) and Lec4 cells (lane 5) were cotransfected with SERT-FLAG and Myc-SERT constructs at a 1:1 ratio where indicated, solubilized, and treated with mouse anti-Myc antibody (Ab)-coated rabbit anti-mouse protein A-Sepharose beads. The immunoprecipitates were separated and blotted with biotinylated anti-FLAG antibody as described under "Experimental Procedures." Control experiments in which only SERT-FLAGtransfected CHO cells were tested did not reveal a visible band (lane 6) on the immunoblot. All lanes contain protein recovered from the same number of cells equivalent to 30% of one well from a confluent 6-well dish. The positions of molecular mass standards run on the same gel are shown in kilodaltons. MSH, ␤-mercaptoethanol. ylated anti-SERT antibody produced three major bands at 90, 65, and 60 kDa, similar to those observed for the transiently transfected SERT-expressing CHO or Lec4 cells (Fig. 1, lanes 1  and 5). After confirming the endogenous expression of myosin IIA and SERT in JAR cells, a co-IP assay with anti-myosin IIA antibody was performed to test their association in JAR cells in the presence of 10 mM MgATP. Probing the myosin IIA-bound proteins with biotinylated anti-SERT antibody resulted in a SERT protein stain at ϳ65 kDa, indicating an endogenous association between myosin and SERT proteins (Fig. 6B, lane  4). Similarly, under reduced electrophoretic conditions, the 90and 60-kDa species appeared to be associated with myosin IIA (lane 7). We found that even though the 90-kDa species was predominantly stained by biotinylated anti-SERT antibody, the 60-kDa species also associated with myosin IIA.
Although transiently transfected Myc-QQ protein was easily detected in JAR cells using anti-Myc antibody (Fig. 6B, lane 3), it was not detected in the co-immunoprecipitate with myosin IIA (lane 5), an observation that strengthened our confidence in the results of glycosyl modification in myosin IIA-SERT coprecipitation. Our data strongly indicate glycosylation-dependent association of myosin IIA and SERT. To control for nonspecific precipitation of myosin IIA, the same analysis was done without anti-myosin IIA antibody (lane 6), and no protein band was labeled, demonstrating that myosin IIA did not bind nonspecifically to the protein A beads.
Glycosylation and Regulation of rSERT by cGMP-The functional consequence of the involvement of N-glycosyl modification on myosin IIA-SERT interaction was tested by measuring the serotonin uptake function of transporters upon stimulation with a cGMP donor, 8-bromo-cGMP, in both our model systems. CHO and Lec4 cells were transiently transfected with SERT or QQ constructs. The following day, they were preincubated in serum-free medium containing 1 mM 8-bromo-cGMP for 1 h at 37°C (60) and then assayed for serotonin uptake function in the dark in the presence of 8-bromo-cGMP. Fig. 7 shows that SERT expressed in CHO cells was also activated in the cGMP pathway. Addition of 8-bromo-cGMP increased the activity of heterologously expressed SERT by 66% of the control value in the absence of the cGMP donor. In contrast, neither the QQ mutant in CHO cells nor SERT in Lec4 cells was significantly affected by 8-bromo-cGMP. DISCUSSION The oligosaccharide residues of glycoproteins in general are believed to be important for a variety of functions, including correct protein folding, oligomerization, membrane trafficking, and targeting (37)(38)(39)(40)(41)(42). SERT is a homo-oligomeric glycoprotein (18 -21) with two sialic acid residues on each complex oligosaccharide residues. In two model systems, we demonstrated that defects in sialylated N-glycans do not alter whole cell or plasma membrane expression of SERT proteins, but do decrease serotonin uptake function. These findings suggest that glycosyl modification may contribute to the correct folding and oligomeric properties of SERT proteins, as reported for epidermal growth factors and insulin receptors (48,49). Several studies using physical measurements such as gel filtration, cross-linking, and fluorescence resonance energy transfer techniques also reported, as we did (18), that SERT is a homo-oligomeric protein with functional interactions between subunits that occur in intracellular compartments (18 -21, 61). We wondered if modifications with sialylated N-glycans directly participate in the serotonin uptake function or confer an optimum conformation to SERT proteins that facilitates their protein-protein interactions. Our data demonstrate that defects in glycosyl modification, specifically in sialic acid residues, impair homooligomerization of SERT proteins. The lower uptake rates of QQ in CHO cells and of SERT in Lec4 cells must be due to the transport capacity of the active SERT monomers.
To determine the impact of glycosyl modification with sialylated N-glycans, we studied the hetero-oligomeric interactions of SERT, specifically with myosin, under defective glycosylation conditions. Our rationale for choosing myosin is based on three related reasons. 1) Both glycosylation sites are located at the extracellular side of the plasma membrane; therefore, we sought a cytosolic partner for SERT. 2) Myosin is an abundant cytoskeletal protein and a strong candidate partner for SERT, as it is a protein kinase G-anchoring protein (50), and the uptake function of SERT is regulated in the NO-dependent pathway via cGMP-dependent protein kinase G (15) through interactions with anchoring proteins (50).
3) The impact of glycosylation could be observed in the functional hetero-oligomerization of SERT. By measuring the serotonin uptake function in the presence of the cGMP donor 8-bromo-cGMP, the effect of glycosylation on the myosin-SERT interaction could be monitored.
The overall findings from both model systems suggest that modification with sialylated N-glycans is required for myosin-SERT hetero-oligomerization. Glycosyl modification of SERT should occur earlier than the myosin-SERT association. If glycosylation directly contributes to SERT function rather than having an impact on its conformation, then the cGMP donor should still be able to stimulate the uptake function of SERT under the defective glycosylation conditions, and the myosin-SERT association should not be impaired. If glycosylation has an impact on the conformation (positioning) of SERT proteins, then myosin-SERT interactions and regulation of SERT function in the NO-dependent pathway should be impaired under defective glycosylation conditions.
On the basis of our present data and previous reports about the sialic acid residues of SERT (27)(28)(29)(30), we believe that modification with the sialylated N-glycans confers on SERT proteins an optimum position that exposes the homo-oligomeric domains on SERT for self-association. Defects in sialylated N-glycan modification impair SERT-SERT association, and as a result, the myosin-SERT binding domain(s) are not accessible for hetero-oligomerization or regulation in the NO/cGMP-dependent pathway. However, one other way to explain our results is that other protein(s) that regulate SERT function in the NO/cGMP-sensitive pathway may also be involved in myosin-SERT association, but could not function under defective glycosylation conditions. In the future, identification of these proteins will be invaluable in understanding the impact of N-glycans on SERT proteins.
The peptide (Sepharose)-purified biotinylated anti-peptide SERT antibodies reacted with three SERT species with molecular masses of 90, 65, and 60 kDa in heterologous and endogenous expression systems. This pattern was consistent with previous observations in HeLa cells (33) and Sf9 cells (31) transiently transfected with SERT. Once the N-glycosyl groups were removed by PNGase F treatment, there was a single 50-kDa core protein, similar to that of the QQ mutants. These results indicate that SERT proteins are differentially or partially glycosylated during biosynthesis. The 50-kDa non-glycosylated monomeric SERT species were not observed in wildtype SERT-expressing CHO, Lec4, or JAR cells.
Co-IP experiments under nonreducing conditions provided further information regarding the size of the SERT species that has a role in protein-protein interactions. The results indicate that a denaturing aggregation event was not responsible for the association of coexpressed SERT forms (Fig. 4, lanes 2 and  3; and Fig. 6B, lanes 3 and 6). Additionally, the migration difference between the QQ mutant and wild-type SERT demonstrated that glycosylated SERT species with molecular masses of 90, 65, and 60 kDa are involved in homo-and heterooligomeric formation. None of the co-IP products migrated at ϳ50 kDa, as observed for the QQ mutant. It is possible that the lower uptake functions of either wild-type SERT in Lec4 cells or QQ in CHO cells are due to the glycosylation-impaired SERT monomers. Based on the present findings and the reports on DAT (24,25), we believe that the involvement of N-linked glycosylation in the oligomeric expression of these biogenic amine transporters is a common phenomenon, as are their homo-oligomeric features.
The use of the Lec4 mutant cells in this study was a novel approach to elucidate the role of N-linked glycosylation in the structure and function of SERT and to demonstrate the significance of the sialic acid residues. Future studies using wildtype SERT in other Lec mutant cells should further advance our understanding of glycosyl modification of monoamine transporter structure, function, and regulation. Additionally, the application of fluorescence resonance energy transfer techniques in living cells will allow for more direct observation of the role of sialylated N-glycans in oligomeric states of SERT proteins. Identification and characterization of the factors associated with the regulation of SERT uptake function will provide vital information that might lead to new treatments for patients with major neuropsychiatric disorders.