Antidepressant-induced switch of beta 1-adrenoceptor trafficking as a mechanism for drug action.

Reduction in surface beta(1)-adrenoceptor (beta1AR) density is thought to play a critical role in mediating the therapeutic long term effects of antidepressants. Since antidepressants are neither agonists nor antagonists for G protein-coupled receptors, receptor density must be regulated through processes independent of direct receptor activation. Endocytosis and recycling of the beta1AR fused to green fluorescent protein at its carboxyl-terminus (beta1AR-GFP) were analyzed by confocal fluorescence microscopy of live cells and complementary ligand binding studies. In stably transfected C6 glioblastoma cells, beta1AR-GFP displayed identical ligand-binding isotherms and adenylyl cyclase activation as native beta1AR. Upon exposure to isoproterenol, a fraction of beta1AR-GFP (10-15%) internalized rapidly and colocalized with endocytosed transferrin receptors in an early endosomal compartment in the perinuclear region. Chronic treatment with the tricyclic antidepressant desipramine (DMI) did not affect internalization characteristics of beta1AR-GFP when challenged with isoproterenol. However, internalized receptors were not able to recycle back to the cell surface in DMI-treated cells, whereas recycling of transferrin receptors was not affected. Endocytosed receptors were absent from structures that stained with fluorescently labeled dextran, and inhibition of lysosomal protease activity did not restore receptor recycling, indicating that beta1AR-GFP did not immediately enter the lysosomal compartment. The data suggest a new mode of drug action causing a "switch" of receptor fate from a fast recycling pathway to a slowly exchanging perinuclear compartment. Antidepressant-induced reduction of receptor surface expression may thus be caused by modulation of receptor trafficking routes.

␤ 1 -adrenoceptor subtype (5,6). ␤AR down-regulation is accompanied by decreased receptor-stimulated cAMP formation (7). The two major effects at the molecular level become apparent in vivo after 10 -20 days of drug administration and coincide with the onset of clinical antidepressant response in humans. Therefore ␤AR down-regulation and diminished cAMP response to catecholamines may relate to the therapeutic action of antidepressants.
It has been proposed that the reduction in the number of functional ␤AR could be a regulatory response to the enhanced presence of norepinephrine in the synaptic cleft after acute inhibition of norepinephrine reuptake or of monoamine oxidase activity by antidepressants (1,8). Some clinically effective antidepressants, however, neither influence norepinephrine reuptake nor inhibit monoamine oxidase activity but still cause a ␤AR down-regulation. Furthermore, this model fails to explain the observed time lag between the rapid drug-induced increase in intrasynaptic neurotransmitter concentrations and the delayed receptor down-regulation.
Decreased ␤AR densities following antidepressant treatment can also be seen in cell culture systems lacking a presynaptic input, e.g. in cultured rat C6 glioblastoma cells (9,10). Thus, ␤AR down-regulation may directly result from postsynaptic actions of the antidepressants. Previous studies have shown that chronic treatment of cultured cells with the tricyclic antidepressant desipramine (DMI) induces phospholipidosis by inhibition of lysosomal phospholipid degradation, leading to changes in membranous and total phospholipid compositions (11). Altered membrane properties may directly influence receptor function or have an effect on vesicular membrane traffic and therefore on ␤AR endocytosis and recycling.
Desensitization of ␤AR after agonist exposure induces receptor endocytosis, protecting cells from overstimulation. It is mediated by uncoupling the activated ␤AR from the G S protein and internalization of receptor from the plasma membrane to early endosomes (12)(13)(14)(15). While ␤ 2 -adrenoceptor desensitization is well studied, much less is known about the mechanism of ␤ 1 -adrenoceptor desensitization (16,17).
Within seconds of agonist binding, ␤ 1 -and ␤ 2 -adrenoceptors are phosphorylated by G protein-coupled receptor kinases and cAMP-dependent protein kinases (18,19). This phosphorylation promotes the binding of the cytosolic protein ␤-arrestin, which inhibits the ability of the ␤AR to couple to G S (20). ␤-Arrestin then targets the phosphorylated receptor to clathrin-coated pits for endocytosis (21). The modes of internalization and recycling of ␤ 1 -adrenoceptors (␤1AR) are not known in detail. ␤ 2 -Adrenoceptors (␤2AR) are internalized via a dynamindependent mechanism through clathrin-coated pits similar to the transferrin receptor (22,23). There is a significant difference in the rate of internalization between the ␤ 1 -and ␤ 2adrenoceptor subtypes. ␤2AR are efficiently endocytosed following agonist stimulation, whereas ␤1AR undergo only slight internalization. The low affinity of the activated ␤1AR for ␤-arrestin may provide an explanation for the small extent of internalization (24). In early endosomes, ␤AR are dephosphorylated and subsequently recycle back via a perinuclear compartment to the plasma membrane in a fully resensitized state (25)(26)(27). Prolonged agonist treatment for several hours causes down-regulation of total receptor number (28). The currently accepted model for ␤AR down-regulation postulates that during chronic agonist exposure, endocytosed ␤AR do not recycle to the plasma membrane but are sorted to lysosomes, where they are degraded by proteases (29,30). However, Jockers et al. (31) have recently reported that ␤2AR down-regulation is fully maintained upon inhibition of receptor endocytosis and blockade of the lysosomal and ubiquitin proteasome pathway, suggesting that the primary inactivation step may occur at the plasma membrane. There is evidence in some cell lines that ␤AR mRNA levels are decreased during long term exposure to agonist, resulting in a reduced receptor synthesis (32,33). Whether antidepressant-induced ␤1AR down-regulation requires one of these molecular mechanisms is unknown.
The aim of the present study was to determine whether the reduction of ␤1AR cell surface density observed in cultured rat C6 glioblastoma cells following chronic DMI treatment is caused by drug-induced changes in receptor trafficking. Endocytosis and recycling of the ␤1AR fused to green fluorescent protein (␤1AR-GFP) were therefore compared in untreated and chronically DMI-treated C6 cells using a complementary approach of confocal fluorescence microscopy and ligand binding studies.

EXPERIMENTAL PROCEDURES
Materials-Materials for cell culture were supplied by Sigma and Invitrogen. Unless otherwise mentioned, all chemicals (analytical grade), were from Sigma or from Merck.
Plasmid Construction-Wild-type human ␤1AR DNA was retrieved from pSP65-␤1AR (a gift from Dr. Susanna Cotecchia, University of Lausanne) with EcoRI and ApaI and subcloned into pcDNA3 (Invitrogen) at the respective restriction sites.
Enhanced green fluorescent protein (EGFP; Clontech) was fused to the carboxyl terminus of the human ␤1AR. The 3Ј-end sequences of ␤1AR including the stop codon were removed with TfiI. To restore the 3Ј-end of the coding sequence and to create a suitable BamHI restriction site at the 3Ј-end for fusion to EGFP, the synthetic complementary oligonucleotides 5Ј-AATCCAAGGTGGATCTGCAG-3Ј and 5Ј-GATCCT-GCAGATCCACCTTGG-3Ј were used for the ligation to pEGFP-N1 (Clontech). The predicted nucleotide sequence of the fusion region was confirmed by dideoxy sequencing of the resultant construct, ␤1AR-GFP. The synthetic linker sequence between ␤1AR and EGFP encodes the amino acids ESKVDLQ, whereby the amino acids ESKV restore the original carboxyl terminus of the ␤1AR.
Cell Culture and Transfection-Rat C6 glioblastoma cells (European Collection of Animal Cell Cultures) were maintained in Eagle's high glucose minimum essential medium supplemented with 10% fetal bovine serum, 3.7 g/liter NaHCO 3 , 200 units/ml penicillin G, and 10 g/ml chlortetracycline at 37°C in a humidified atmosphere of 95% air and 5% CO 2 . Cells were passaged once a week and replated at a density of 2 ϫ 10 6 cells/10-cm dish.
For transfection, cells were grown to 80% confluence in 10-cm dishes. The ␤1AR-GFP plasmid (6 g) was linearized with AflIII and mixed with 20 l of LipofectAMINE reagent (Invitrogen) according to the manufacturer's instructions. Cells were exposed to the DNA-Lipo-fectAMINE mixture for 8 h. Two days after transfection cells were trypsinized, diluted, and seeded into 96-well plates at a density of 15,000 cells/well. Stably transfected cells were selected with Geneticin G418 (500 g/ml; Calbiochem) that was added to the culture medium 3 days after transfection. Wells with cells that were resistant to the antibiotic were identified 2 weeks after transfection. Cells were then reseeded into 96-well plates at limiting dilutions to establish clonal C6-␤1AR-GFP cell lines. Cell lines with highest expression levels of the receptor were identified by Western blot and fluorescence microscopy.
COS-1 cells (American Type Culture Collection) were transiently transfected with 5 g of either wild-type ␤1AR or ␤1AR-GFP DNA, essentially as described earlier (34). For the cAMP assay, COS-1 cells were seeded into 24-well dishes at a density of 15,000 cells/well and transiently transfected with 3 g of ␤1AR or ␤1AR-GFP DNA per 24 wells. Experiments were performed 48 h after transfection.
Radioligand Binding Assay-The number of cell surface ␤AR was determined in intact C6 and COS-1 cells using the hydrophilic ␤AR antagonist The number of total cellular ␤AR was determined using the membrane-permeant radioligand [ 3 H]dihydroalprenolol (Amersham Biosciences). Cells were harvested in Hanks' solution, pelleted, lysed in 5 mM Tris/HCl, 2 mM EDTA, 5 g/ml leupeptin, 0.5 mM phenylmethylsulfonyl fluoride, pH 7.4, and subjected to two freeze/thaw cycles. This cell lysate was diluted 1:12 with binding buffer (75 mM Tris/HCl, 12 mM The binding reaction was terminated by adding 1 ml of ice-cold Hanks' solution. 1 ml of the incubation mixture was immediately filtered through presoaked glass fiber filters (Whatman GF/C) at constant vacuum pressure. The filters were washed three times with 5 ml of 0.9% NaCl solution, dried, and placed in counter vials. Cells were then digested overnight using a tissue solubilizer. After neutralization and the addition of 5 ml of liquid scintillation mixture (ULTIMA GOLD; Packard Instruments), radioactivity was determined in a liquid scintillation counter. Protein concentrations were determined using the method of Lowry. Binding curves were fitted by nonlinear regression using GraphPad Prism (version 3.0).
cAMP Assay-Transiently transfected COS-1 cells were serumstarved for 2 h and then incubated at 37°C in 500 l of Hanks' solution containing 1 mM 1-methyl-3-isobutylxanthine and varying concentrations of isoproterenol (10 Ϫ11 to 10 Ϫ6 M) or no agonist (basal). The reaction was stopped after 10 min by exchanging the Hanks' solution for 500 l of ice-cold 0.1 N HCl. Following a freeze/thaw cycle to lyse the cells, the cytosolic extracts of each well were collected, lyophilized, resolved in 50 mM Tris-HCl, 4 mM EDTA, pH 7.5, and assayed for cAMP using the Amersham Biosciences cyclic AMP assay kit. The cAMP produced in response to agonist exposure was calculated as the cAMP accumulation in stimulated cells minus the cAMP level in unstimulated cells from the same 24-well dish. Protein concentrations were determined using the method of Lowry.
␤-Adrenoceptor Endocytosis and Recycling-COS-1 cells transiently expressing either wild-type ␤1AR or ␤1AR-GFP and C6-␤1AR-GFP cells were stimulated with 10 M isoproterenol at 37°C for the indicated times. Subsequently, the medium was removed, cells were chilled on ice, washed 10 times with ice-cold Hanks' solution followed by further incubation for 1 h at 37°C in fresh medium to allow receptor recycling. Cell surface ␤AR densities were determined in parallel cultures before agonist addition, after desensitization as well as at the end of the resensitization period as described above (radioligand binding assay).
To prevent receptor internalization and recycling in the intact cells during exposure to radioligand, the temperature was lowered to 13°C. Triplicates (200 l each) of the cell suspension were mixed with 50 l of radioligand solution and incubated for 3 h to ensure equilibrium binding. B max values were determined with plateau concentrations of the radioligand, which were previously determined in complete saturation binding experiments.
Total cellular binding sites (surface and internal) were measured in C6-␤1AR-GFP cells before agonist addition and after resensitization using [ 3 H]dihydroalprenolol as described above.
In some experiments, the protein synthesis inhibitor cycloheximide (200 M) was used. C6-␤1AR-GFP cells were preincubated with the inhibitor for 60 min before agonist stimulation. Cycloheximide was also present during all subsequent steps.
Desensitization Studies-C6-␤1AR-GFP cells at 90% confluence were incubated at 37°C for 30 min in Hanks' solution in the absence or presence of 10 M isoproterenol. Cells were then washed five times and maintained in agonist-free Hanks' solution containing 1 mM 1-methyl-3-isobutylxanthine for 10 min at 37°C prior to stimulation for the same period of time with 10 M isoproterenol. The cyclic AMP production in response to the 10-min challenge with isoproterenol was determined as described above (cAMP assay).
Effect of Chronic Treatment with Desipramine on ␤AR Trafficking-C6-␤1AR-GFP cells (7 ϫ 10 5 cells) were seeded into 10-cm dishes 8 days prior to the experiment. 48 h later, pharmacological treatment was started. The cells were exposed to 10 M DMI (Novartis) for 6 days or left untreated. The culture medium was changed 2, 4, and 5 days after the beginning of the treatment.
Confocal Laser-scanning Microscopy-Cells were observed using a laser-scanning confocal microscope (Zeiss Axiovert 100M; LSM 510) equipped with a Zeiss Plan-Apo 63 ϫ 1.4 NA oil immersion objective). GFP was excited at 488 nm (argon laser), and fluorescence emission was detected using a 505-530-nm band pass filter. Rhodamine-conjugated probes (transferrin, dextran) were excited at 543 nm (neon/helium laser). Red fluorescence was detected with a long pass (560 nm) filter. The pinholes were set to 1.1 airy units.
C6-␤1AR-GFP cells were grown on glass coverslips that were placed in 12-well dishes. Cells were seeded at a density of 4 ϫ 10 4 cells/well. 48 h later, medium was removed, and coverslips were rinsed with Hanks' solution and mounted on a thermostated imaging chamber. To study receptor endocytosis, the cells were stimulated with 10 M isoproterenol at 37°C in Hanks' solution. Receptor recycling was then observed at 37°C in fresh Hanks' solution after agonist washout.
To study the effects of DMI on receptor trafficking, C6-␤1AR-GFP cells were seeded at a density of 1 ϫ 10 4 cells/well. 24 h later, antidepressant treatment was started. To label lysosomal compartments, cells were incubated for the final 24 h of DMI treatment with rhodaminedextran (1 mg/ml; Molecular Probes, Inc., Eugene, OR). Excess lysosomal marker was then removed, and the cells were incubated in fresh medium in the absence or presence of isoproterenol (10 M) for 2 h at 37°C before imaging. For the transferrin experiments, cells were incubated with 10 M isoproterenol for 2 h at 37°C. Rhodamine-labeled transferrin (200 g/ml; Molecular Probes) was added for the last 60 min of the incubation period. Before imaging, the cells were thoroughly washed with Hanks' solution. To study ␤AR recycling after agonist removal, cells were maintained in fresh medium, returned to a 37°C incubator for 60 min, and then viewed under the microscope.
For three-dimensional modeling of fluorescence distribution in single live cells, stacks of confocal images were recorded before and 30 min after stimulation with 10 M isoproterenol. Sampling distances were ϳ100 nm in the lateral and Յ150 nm in the axial direction to meet the Nyquist criterion.
Three-dimensional Image Restoration and Image Quantification-Stacks of confocal fluorescence images were subjected to image restoration using the iterative maximum likelihood estimate algorithm of the Huygens System 2 (Scientific Volume Imaging, Hilversum, The Netherlands) and a theoretical point spread function. Subsequently, three-dimensional models of fluorescent objects were generated with the Imaris 3.2 Surpass and Measurement Pro software modules (Bitplane AG, Zü rich, Switzerland) capable of separating objects based on surfaces of equal intensities (isosurfaces). Appropriate threshold values for the generation of isosurfaces were applied to separate intracellular objects and objects representing fluorescence at the cell surface. Quantification of the voxel densities enclosed in these objects then provided a measure for the fluorescence of intracellular and cell surface-associated objects.

Expression of ␤1AR-GFP Fusion Proteins in Cultured Cells-
The ␤1AR was fused to the amino terminus of green fluorescent protein to provide a tool that would allow us to follow ␤1AR trafficking by confocal fluorescence microscopy in live cells during agonist-induced internalization and the subsequent resensitization phase.
First, plasmids encoding the fusion construct (␤1AR-GFP) or the wild-type ␤1AR were transiently transfected into COS-1 cells, and the expressed proteins were analyzed by immunoblotting (Fig. 1). An antibody to the ␤1AR detected two main bands with molecular masses of 50 and 70 kDa in cells expressing wild-type ␤1AR (Fig. 1A, lane 2) but not in untransfected cells (Fig. 1A, lane 1). ␤1AR are glycoproteins that contain one N-glycosylation site in the receptor N terminus (35). It is likely that these two bands detected by the antibody represented nonglycosylated (50 kDa) and glycosylated (70-kDa) ␤1AR. As expected from fusion to the 28-kDa GFP protein, the recombinant receptors showed a lower mobility than wild-type ␤1AR (Fig. 1A, lane 3). The 80-kDa band represented the unglycosylated fusion protein with a predicted molecular mass of 79 kDa. Similar to the wild-type receptor, the band with the lower mobility most likely represented a glycosylated form of the GFP-tagged ␤1AR, since the intensity of the 100-kDa band was strongly reduced by treatment of the cells with the glycosylation inhibitor tunicamycin, whereas the intensity of the 80-kDa band was augmented (not shown). The bands with a molecular weight of about 20 kDa (lane 2) and 52 kDa (lanes 3 and 5) seem to represent a carboxyl-terminal receptor fragment, since the difference in mobility shift of the two bands corresponds to the molecular mass of GFP. An anti-GFP antibody detected the receptor-GFP fusion proteins exclusively (Fig. 1A, lane 5). Expression of the GFP-tagged ␤1AR was also observed in membrane fractions of rat C6 glioblastoma cells stably transfected with ␤1AR-GFP DNA (Fig. 1A, lane 6).
Functional Analysis of the ␤1AR-GFP Fusion Protein-A number of experiments were performed to demonstrate that the ␤1AR-GFP fusion protein maintained the biochemical and signal transducing properties of the native receptor.
[ 3 H]CGP-12177 binding of the ␤1AR was not affected by attaching GFP to its COOH-terminal tail. Binding isotherms were nearly identical for wild-type and GFP-tagged receptors, as shown in Fig. 1B, with equilibrium dissociation constants of 5.3 Ϯ 1.3 nM for ␤1AR and 3.6 Ϯ 1.0 nM for ␤1AR-GFP.
The ability of wild-type and GFP-tagged ␤1AR to couple to G S proteins and to activate adenylyl cyclase upon the addition of agonist was also examined. Transfected cells were stimulated with varying concentrations of isoproterenol to initiate second messenger formation. Isoproterenol-stimulated cAMP responses were comparable in cells expressing wild-type ␤1AR or ␤1AR-GFP at similar levels (Fig. 1C). Thus, both forms of the receptor activated adenylyl cyclase with similar efficacy. Cells transfected with empty vector alone also exhibited increased cAMP levels when stimulated with high agonist concentrations. COS-1 cells appear to express sufficient endogenous ␤AR to efficiently activate adenylyl cyclase at high agonist concentrations.
The ability of the fusion construct to internalize upon agonist binding was assessed in transiently transfected COS-1 cells. Incubation with isoproterenol for 30 min led to a decrease in surface receptor density of 9.9 Ϯ 1.9%, as determined by radioligand binding (Fig. 1D). Under identical conditions, transiently expressed wild-type receptors internalized to a similar extent (13.5 Ϯ 1.7%). Agonist was then removed to study receptor recycling. The return of the ␤1AR-GFP to the cell surface was quantitatively similar to that of the wild-type ␤1AR. 60 min after agonist washout a complete recovery of surface receptor numbers was detected in ␤1ARand in ␤1AR-GFPexpressing cells.
Receptor trafficking was further investigated in C6-␤1AR-GFP cells using laser-scanning confocal microscopy ( Fig. 2A). In the absence of isoproterenol, GFP-tagged ␤1ARs were mainly localized in the plasma membrane. No detectable receptor endocytosis was observed. Challenging the cells with isoproterenol caused a profound change in receptor distribution. Receptors were internalized and appeared as green fluo-rescent speckles in the cytosol. The first intracellular receptors could be observed as early as 3 min after agonist addition. Receptor accumulation in intracellular structures became progressively more pronounced during 30 min of agonist stimulation. The density of endocytosed receptors sharply decreased following agonist washout and a 30-min incubation of cells in fresh medium at 37°C (Fig. 2B), suggesting that receptors relocalized to the plasma membrane upon removal of agonist. Radioligand binding studies revealed that a 30-min exposure of C6-␤1AR-GFP cells to isoproterenol prompted a 21.9 Ϯ 2.6%  2 and 4) and GFP-tagged ␤1AR (lanes 3 and 5) were subjected to SDS-polyacrylamide gel electrophoresis and analyzed by immunoblotting using antibodies to ␤1AR or GFP. Similarly, membrane fractions of untransfected C6 glioblastoma cells (lane 7) or of stably transfected C6 cells expressing GFP-tagged ␤1AR (lane 6) were probed for GFP. B, COS-1 cells transiently expressing wild-type or GFP-tagged ␤1AR were exposed to increasing concentrations of the radioligand [ 3 H]CGP-12177 to compare antagonist binding. Results represent means Ϯ S.D. of three independent experiments. C, cAMP levels were measured in COS-1 cells transiently transfected with either wild-type ␤1AR or ␤1AR-GFP DNA or with empty vectors (pcDNA3 and pEGFP, respectively). 48 h later, cells were exposed to increasing concentrations of isoproterenol for 10 min, and cAMP formation was determined. The results of each experiment were normalized to the maximal cAMP accumulation measured in cells expressing wild-type ␤1AR. Expression levels of wild-type and GFP-tagged ␤1AR were comparable within the experiments. Data (means Ϯ S.D.) are from three experiments (except for pEGFP (two experiments)). D, receptor internalization and recycling were studied in COS-1 cells transiently expressing wild-type or GFP-tagged ␤1AR. The cells were incubated for 30 min at 37°C in the presence of 10 M isoproterenol, washed, and maintained in agonist-free medium for 60 min at 37°C. Cell surface ␤AR density was determined before the agonist addition (unstimulated), immediately after agonist exposure (agonist-stimulated), and at the end of the resensitization period (resensitized). Data were normalized to the B max value obtained in unstimulated cells. Results represent the mean Ϯ S.D. of three different experiments.

Alternative Receptor Trafficking by Antidepressants
reduction in cell surface ligand binding sites. Furthermore, rapid recovery of surface receptors was observed in these cells following removal of agonist from the culture medium, confirming the ability of internalized ␤1AR-GFP to recycle to the plasma membrane (Fig. 3A). In order to examine whether this reduction was indeed due to a concomitant increase in intracellular receptor, the amount of internal fluorescence was quantified in complementary experiments using three-dimensional models reconstructed from stacks of confocal images (Fig. 3B). Fluorescence in intracellular objects that could be clearly separated from the cell surface ranged from 6.9 to 12.5% (9.2 Ϯ 2.4%, mean Ϯ S.D., n ϭ 6) of total fluorescence in single, isoproterenol-treated cells, compared with 0.3 Ϯ 0.4% in control cells (Fig. 3C). These data confirm that internalization of receptor occurs to an extent that is consistent with the biochemically detected reduction of ligand binding sites at the cell surface.
We then studied whether the reduced surface receptor density after exposure of C6-␤1AR-GFP cells to 10 M isoproterenol had functional consequences on receptor signaling. The cells were incubated for 30 min at 37°C in the absence or presence of agonist, washed, and rechallenged with 10 M isoproterenol for 10 min. Receptor-mediated cAMP production in response to the second stimulus was then determined. Agonist pre-exposure led to a significant decline in the isoproterenol-stimulated cAMP response. The second messenger production in prestimulated cells was reduced by 31.8 Ϯ 5.9% compared with cAMP levels in cells not pre-exposed to agonist. Thus, the observed reduction in cell surface receptor density results in a more profound decrease of receptor-stimulated second messenger production.
Effects of Chronic Treatment with Desipramine on ␤-Adrenoceptor Trafficking-We next examined whether receptor endocytosis and recycling were changed in C6-␤1AR-GFP cells that were chronically treated with the tricyclic antidepressant DMI in comparison with untreated control cells using a radioligand binding assay (Fig. 4A). Isoproterenol stimulation promoted endocytosis of about 15% of the receptor sites in control and DMI-treated cells. Thus, ␤1AR endocytosis was not influenced by chronic DMI treatment. 1 h after removal of agonist from the culture medium, complete restoration of surface receptor numbers was observed in control cells. The recovery of surface receptor density resulted from the return of internalized ␤AR back to the plasma membrane and not from the biosynthesis of new receptor protein, since identical restoration occurred in cells incubated in the presence of the protein synthesis inhibitor cycloheximide (not shown). In contrast, in the plasma membrane of antidepressant-treated cells, ␤AR density was still reduced at the end of the resensitization period, indicating that under these conditions endocytosed ␤1AR-GFP failed to recycle efficiently. The recycling block in DMI-treated cells appeared to be irreversible, because even after a prolonged resensitization period (180 min of incubation at 37°C) cell surface receptor densities did not recover (not shown).
To identify the step where the recycling process was interrupted by chronic DMI treatment, ␤1AR trafficking was assessed by laser-scanning confocal microscopy. Isoproterenol exposure of untreated and DMI-treated C6-␤1AR-GFP cells caused a receptor redistribution from the cell surface into intracellular compartments (compare Fig. 4B, top and center  rows, respectively). The extent of ␤1AR-GFP endocytosis was similar in untreated and DMI-treated cells, and receptors appeared to accumulate in perinuclear compartments. However, substantial differences in ␤1AR-GFP localization were detected between untreated and DMI-treated cells after resensitization. Only a small number of internalized receptors was detected in control cells at the end of the resensitization period. By contrast, numerous intracellular structures containing internalized GFP-tagged ␤1AR were observed in antidepressanttreated cells (compare Fig. 4B, bottom row, right and left panels). These results suggest that in control cells internalized ␤1AR-GFP returned back to the plasma membrane upon removal of agonist, whereas in DMI-treated cells a large fraction of endocytosed receptors was retained in intracellular compartments. These findings are consistent with the data of the binding assay and indicate that chronic DMI treatment inhibits recycling of endocytosed receptors back to the plasma membrane.
One might assume that internalized receptors in DMItreated cells failed to recycle, because they were sorted to lysosomes for degradation. To test this possibility, we performed double fluorescence studies with rhodamine-dextran in combination with ␤1AR-GFP. Dextran is known to specifically accumulate in late endosomes and lysosomes (36). Chronically DMI-treated C6-␤1AR-GFP cells were preincubated with rhodamine-labeled dextran and subsequently stimulated with isoproterenol to induce receptor endocytosis. Agonist exposure did not result in detectable colocalization of ␤1AR-GFP with the lysosomal marker. Internalized ␤1AR-GFP appeared in intracellular structures that were completely devoid of dextran labeling (Fig. 5), suggesting that the receptors were retained in early endosomes. The cells were then further incubated at 37°C in agonist-free medium to allow receptor recycling. At the end of the resensitization period, still no ␤1AR-GFP was detected in lysosomes (not shown). To rule out the possibility that proteolytic receptor degradation prevented the detection of ␤1AR-GFP that might have been diverted to lysosomes, the FIG. 4. Chronic DMI treatment impairs recycling of GFPtagged ␤1AR. C6-␤1AR-GFP cells were grown for 6 days in the absence (control) or presence of 10 M DMI (DMI-treated). Then ␤AR endocytosis was induced by stimulating the cells with 10 M isoproterenol at 37°C for 2 h. After removal of the agonist, cells were incubated for 60 min in fresh medium at 37°C to allow for receptor recovery. A, internalization and recycling of ␤1AR-GFP as assessed by radioligand binding. Surface ␤AR densities were measured using [ 3 H]CGP-12177 before (unstimulated) and immediately after the 2-h exposure to isoproterenol (agonist-stimulated) as well as 1 h after removal of agonist (resensitized). For each experiment, the B max value of unstimulated cells was designated 100%, and all other data were normalized to this value. Results are presented as mean Ϯ S.D. from five independent experiments. *, p Ͻ 0.01 compared with the mean values of unstimulated cells (one-way analysis of variance). B, localization of intracellular ␤1AR-GFP was determined before agonist stimulation (top row), after agonist exposure (middle row), and at the end of the resensitization period (bottom row). Representative images of confocal midsections of cells are shown. Three separate experiments produced similar results. Internalized ␤1AR-GFP are indicated by arrows.
FIG. 5. Chronic DMI treatment does not result in lysosomal targeting of internalized ␤1AR-GFP. Chronically DMI-treated C6-␤1AR-GFP cells were incubated overnight with rhodamine-dextran to label lysosomes. Subsequently, cells were stimulated with 10 M isoproterenol for 2 h at 37°C. Agonist exposure did not result in a colocalization of endocytosed ␤1AR-GFP (green) with rhodamine-labeled dextran (red), since no overlapping structures were detected in the merged image.
same experiment was performed in the presence of the protease inhibitor leupeptin (100 M). Even under these conditions, no ␤1AR-GFP could be detected in lysosomes (not shown). Radioligand binding experiments with the membrane-permeant antagonist [ 3 H]dihydroalprenolol confirmed that in DMI-treated cells internalized ␤1AR were not subject to degradation, since no decrease in the number of total (surface and internal) ␤AR was observed after long term desensitization (not shown).
As an alternative explanation, we hypothesized that the ␤1AR-GFP of DMI-treated cells were delivered to a nonrecycling, nonlysosomal compartment. Therefore, we compared the localization of internalized ␤1AR-GFP with that of transferrin receptors. ␤AR and transferrin receptors are internalized via distinct primary endocytic vesicles to the same early endosomes (37). Transferrin receptors then return via a recycling compartment back to the plasma membrane (38,39). If, indeed, internalized ␤1AR-GFP were delivered to another compartment in DMI-treated cells, one would expect that in control cells internalized ␤1AR-GFP and transferrin receptors would colocalize in early endosomes, whereas in DMI-treated cells no colocalization would occur. Experimentally, the trafficking of the transferrin receptor may be followed by rhodamine-labeled transferrin, because transferrin remains associated with transferrin receptors during internalization and constitutively recycles with the receptors back to the cell surface (22,40). As illustrated in Fig. 6 (central panel), in chronically DMI-treated cells a significant portion of the internalized ␤1AR-GFP following stimulation with isoproterenol colocalized with endocytosed transferrin receptors (colocalization shown in yellow), confirming that ␤1AR-GFP were present in early endosomal compartments following a 2-h exposure to agonist. Intracellular distribution of ␤1AR-GFP following agonist stimulation was also studied in untreated cells (Fig. 6, left panel). The extent of colocalization of endocytosed ␤1AR-GFP and transferrin receptors in control cells was comparable with that in DMI-treated cells, suggesting that chronic DMI treatment did not affect the primary subcellular localization of internalized ␤1AR-GFP. DMI-treated C6-␤1AR-GFP cells labeled with rhodamine transferrin were then allowed to recover in fresh medium at 37°C for 1 h and were subsequently examined for receptor distribution. Only a small number of internalized transferrin receptors was detected at the end of this recovery period, whereas endocytosed ␤1AR-GFP remained intracellular (Fig.  6, right panel). Thus, in DMI-treated cells transferrin receptors were able to return to the plasma membrane but not ␤1AR-GFP. This observation suggests that ␤1AR-GFP are internalized to an early endosomal compartment like transferrin receptors from where recycling to the cell surface is possible in principle. However, it seems that DMI treatment causes posttranslational modifications of the ␤1AR itself or of some accessory proteins, leading to an impaired receptor recycling.

DISCUSSION
Tricyclic antidepressants are well known for their potential to modulate the density of functional neurotransmitter receptors such as ␤1AR and serotonin receptors in the brain (1-3, 41, 42) as well as in cultured cells (4,9). The mechanisms for this reduction in receptor numbers are not completely understood. Importantly, the onset of down-regulation and clinical effectiveness requires 10 -20 days of antidepressant treatment (43). We have previously shown that in cultured C6 cells chronic exposure to DMI results in a progressive reduction in ␤1AR surface receptor density within similar periods of time (10). C6 cells thus provide a valid model system to examine potential mechanisms that might underlie the clinical and experimental long term observations.
A chimeric protein of GFP fused to the carboxyl terminus of the ␤1AR was used as a tool to visualize the processes of receptor internalization and recycling in live cells and to combine these observations with the biochemical assessment of ␤1AR on the cell surface and the total receptor numbers in whole cells. The feasibility of such an approach depends largely on the compatibility of the fusion partner at the carboxylterminus with the functional integrity of the receptor. Several studies with carboxyl-terminal fusions of G protein-coupled receptors (GPCR), such as the cholecystokinin A receptor (44), the thyrotropin-releasing hormone receptor (45), the A1 adenosine receptor (46), and the ␤2AR (31,47), demonstrated that ligand binding affinities, G protein coupling, and downstream effector activation are not significantly affected. Similarly, fusing GFP to the carboxyl terminus of ␤1and ␤2-adrenoceptors only marginally affects ligand binding (48). It may result in quantitative but not in qualitative changes in receptor internalization and recycling properties. However, recent studies have shown that the carboxyl terminus of the ␤2AR interacts with a number of proteins that seem to play an important role in determining the fate of endocytosed receptors (49,50). Elimination or alteration of a single amino acid at the extreme carboxyl terminus of the ␤2AR has been reported to disrupt such interaction and to influence receptor sorting (49 -52). Whether the addition of GFP to the intact COOH-terminal tail of GPCR also prevents these protein-protein interactions has not been explored. In this study, ligand binding, activation of adenylyl cyclase, and receptor internalization as well as recycling properties were similar for wild-type and for the chimeric FIG. 6. Transferrin receptor recycling in DMI-treated C6 cells. Untreated (control) and chronically DMI-treated C6-␤1AR-GFP cells were desensitized with 10 M isoproterenol for 2 h at 37°C. During the second hour of the stimulation period, cells were loaded with rhodamine-labeled transferrin. DMI-treated cells were then washed and incubated for an additional 1 h in fresh medium at 37°C to allow for resensitization. Localization of ␤1AR-GFP (green) and transferrin receptors (red) was examined by laser-scanning confocal microscopy after desensitization and at the end of the resensitization period. receptors. Thus, there is no evidence that the addition of GFP affects receptor function or trafficking. However, our results cannot exclude the possibility that fusion of GFP to the carboxyl terminus of the receptor might modulate other aspects of ␤1AR regulation.
Studying agonist induced trafficking of ␤1AR in C6-␤1AR-GFP cells, significant differences were found between cells that were chronically exposed to DMI and untreated control cells. Whereas chronic exposure to DMI had no effect on agonistinduced receptor internalization or on the localization of internalized receptors, the fate of the internalized receptors seemed to be quite different under these conditions. In antidepressanttreated cells, internalized receptors after agonist stimulation remained trapped intracellularly, whereas in control cells receptors recycled to the cell surface within the first hour after agonist removal. It appeared that receptors were internalized normally but were then redirected to a slowly recycling or even nonrecycling compartment. Chronic application of tricyclic antidepressants may thus modulate the fate of the internalized receptors. The long lasting intracellular receptor accumulation would be consistent with trafficking along a "long cycle." Such a trafficking route had been postulated for the V2 vasopressin receptor (53).
Contrary to most other GPCR, stimulation of the ␤1AR results only in partial receptor internalization, possibly as a consequence of a weak association with ␤-arrestins (24,54,55). In C6-␤1AR-GFP cells, a small but reproducible, agonist-induced reduction of the plasma membrane-associated receptors of ϳ15% was observed in ligand binding experiments (Fig. 4A). These properties of the model system provided the basis for studying cellular receptor trafficking after antidepressant treatment. Accumulation of internalized receptors could be readily observed by confocal fluorescence microscopy and was quantified in three-dimensional models of single cells that were calculated from restored image stacks. The lower estimate for internalized receptors obtained from the model data could be due to the difficulty of separating internalized from plasma membrane-associated material in cases where internalized material was located in close proximity to the cell surface or in budding vesicles. This led to the unavoidable exclusion of some material from the estimate of internalized receptor. However, we cannot exclude the possibility that a subpopulation of stimulated receptor is unavailable for ligand binding after agonist washout but may still be present at the cell surface. The reduction of receptors in the plasma membrane could not be visualized, but it could be derived from ligand binding assays, which demonstrated the initial decrease and the subsequent reappearance of cell surface receptors after a resensitization period. The combination of the two approaches in this study proved to provide a powerful method to study limited receptor trafficking and revealed a defect in receptor recycling to the cell surface in DMI-treated cells.
Our desensitization studies showed that exposure of C6-␤1AR-GFP cells to isoproterenol led to a marked decline in receptor-dependent cAMP production, although ligand binding to ␤AR at the cell surface was reduced by only 15%. A large portion of lost binding sites can be attributed to internalization of ␤1AR-GFP as our estimates from the three-dimensional models show. Therefore, reduced signaling in desensitized cells may be explained at least in part by receptor internalization. Chronic treatment of cells with DMI leads to a comparable reduction in surface ␤AR density. Hence, our results suggest that the inability of internalized ␤1AR-GFP in DMI-treated cells to recycle back to the plasma membrane is likely to cause a substantial decrease in receptor signaling.
Interestingly, chronic exposure of ␤1AR-GFP cells to DMI in the absence of agonist stimulation also led to reduced ␤AR numbers in a majority of cases. In contrast to cells that were challenged with agonist, intracellular receptor accumulation could not be found under these basal conditions. In agonisttreated cells, no indication for proteolytic degradation was obtained within a resensitization period of up to 3 h. This does not exclude the possibility that receptors could still be subject to lysosomal degradation, albeit using a slow cycling pathway. As shown for mannose 6-phosphate receptors by Gonzalez-Noriega et al. (56), definitive fate decisions may be made as late as 3-4 h after internalization, because receptors could still be salvaged within this time period, whereas intervention at later time points could not prevent degradation in the lysosomal compartment anymore. We previously reported that chronic DMI exposure results in alterations of the cellular and membranous phospholipid patterns (11) and promotes lysosomal phospholipid accumulation (57). The modification of the physicochemical membrane characteristics does not interfere with receptor binding and internalization. However, it may be the reason for an inefficient membrane recycling and thus for incomplete receptor recycling. Slowing down receptor recycling may lead to a new steady state distribution in the absence of any changes in the rate of de novo synthesis of receptors. Since the recycling and its modification only relate to a fraction of the surface receptors, one would predict that such a mechanism would take considerable time for a new steady state of total receptor distribution to be reached. This and the prolonged time required for changes in the phospholipid patterns might also explain the observed lag in receptor down-regulation and possibly in the onset of the clinical effectiveness in the treatment of depressive states. Future experiments will address the question whether DMI-induced changes in phospholipid composition could provide a molecular basis for the altered receptor recycling.
The antidepressant-induced switch redirecting a GPCR from a fast recycling pathway to a slowly recycling or degradative pathway presents a new mode for drug action. It appears to engage a cellular machinery that may divergently sort receptors of different families after their endocytosis (58). As the underlying molecular mechanisms, two major principles may be envisioned. Antidepressant-induced changes in the phosphorylation pattern of specific carboxyl-terminal regions of the receptor may induce strong retention of endocytosed receptors in intracellular compartments. Such a phosphorylationdependent mechanism is exemplified by the vasopressin 2 receptor (59). Alternatively, association of receptors with cellular targets may be regulated by drug-induced changes of cellular components other than the receptor. In such a model, the receptor would not be the direct target of the chronic drug treatment. Rather, the cellular handling of the internalized receptor would differ in accordance with the drug treatment. In that respect, it should be noted that a substantial collection of receptor-associated proteins, including ␤-arrestins, scaffold proteins, and protein kinases, co-migrate with the receptor during internalization and recycling (60,61). These components represent attractive candidates for the antidepressantinduced switch of receptor fate. The present data would be consistent with either of the two potential mechanisms. The identification of the molecular correlate(s) responsible for the divergence from the normal recycling pathway is likely to link well known signaling components to the action of antidepressants.
In summary, our results demonstrate that chronic exposure to the tricyclic antidepressant DMI impairs ␤1AR recycling in rat C6 glioblastoma cells. Agonist-stimulated ␤1AR are internalized normally but are then redirected to a slowly recycling or even nonrecycling compartment. Antidepressant-induced reduction of ␤1AR density may thus be caused by a modulation of receptor trafficking routes.