The Metabotropic Glutamate Receptor mGluR5 Is Endocytosed by a Clathrin-independent Pathway*

Metabotropic glutamate receptors 5 (mGluR5) are members of the growing group C G protein-coupled receptor family. Widely expressed in mammalian brain, they are involved in modulation of the glutamate transmission. By means of transfection of mGluR5 receptors in COS-7 cells and primary hippocampal neurons in culture followed by immunocytochemistry and quantitative image analysis and by a biochemical assay, we have studied the internalization of mGluR5 splice variants. mGluR5a and -5b were endocytosed in COS-7 cells as well as in axons and dendrites of cultured neurons. Endocytosis occurred even in the absence of receptor activity, because receptors mutated in the glutamate binding site were still internalized as well as receptors in which endogenous activity had been inhibited by an inverse agonist. We have measured a constitutive rate of endocytosis of 11.7%/min for mGluR5a. We report for the first time the endocytosis pathway of mGluR5. Internalization of mGluR5 is not mediated by clathrin-coated pits. Indeed, inhibition of this pathway by Eps15 dominant negative mutants did not disturb their endocytosis. However, the large GTPase dynamin 2 is implicated in the endocytosis of mGluR5 in COS-7. mGluR5 is the first shown member of the group C G-protein coupled receptor family internalized by a nonconventional pathway.

Metabotropic glutamate receptors 5 (mGluR5) are members of the growing group C G protein-coupled receptor family. Widely expressed in mammalian brain, they are involved in modulation of the glutamate transmission. By means of transfection of mGluR5 receptors in COS-7 cells and primary hippocampal neurons in culture followed by immunocytochemistry and quantitative image analysis and by a biochemical assay, we have studied the internalization of mGluR5 splice variants. mGluR5a and -5b were endocytosed in COS-7 cells as well as in axons and dendrites of cultured neurons. Endocytosis occurred even in the absence of receptor activity, because receptors mutated in the glutamate binding site were still internalized as well as receptors in which endogenous activity had been inhibited by an inverse agonist. We have measured a constitutive rate of endocytosis of 11.7%/min for mGluR5a. We report for the first time the endocytosis pathway of mGluR5. Internalization of mGluR5 is not mediated by clathrin-coated pits. Indeed, inhibition of this pathway by Eps15 dominant negative mutants did not disturb their endocytosis. However, the large GTPase dynamin 2 is implicated in the endocytosis of mGluR5 in COS-7. mGluR5 is the first shown member of the group C G-protein coupled receptor family internalized by a nonconventional pathway.
Glutamate is the major excitatory neurotransmitter in the mammalian central nervous system. It exerts its effects by binding to ionotropic and metabotropic receptors. Metabotropic glutamate receptors (mGluRs) 1 belong to the seven-transmembrane domain G protein-coupled receptor (GPCR) family. Together with calcium-sensing receptors, ␥-aminobutyric acid receptor B (GABA B ) and pheromone, olfactory and taste receptors, mGluRs form the third family of GPCRs (group C) (for reviews, see Refs. 1 and 2). They have little homology with other GPCRs from groups A and B and are characterized by a large N-terminal extracellular domain that binds glutamate. Eight genes code for mGluRs that generate more than 15 proteins by alternative splicing. On the basis of their sequence similarity, pharmacology, and signal transduction mechanisms, mGluRs have been classified into three groups. Group 1 consists of mGluR1 and mGluR5 that are preferentially positively coupled to phospholipase C via G q/11 proteins. mGluRs of group 2 and 3 are negatively coupled to adenylate cyclase. Group 1 mGluRs are expressed in neuronal and glial cells within the brain, and receptor activation has been implicated in several forms of synaptic plasticity.
Currently, very little is known about the targeting and trafficking mechanisms of mGluRs in cells. During the last few years, attention has focused mainly on endocytosis of a subtype of ionotropic glutamate receptors: the ␣-amino-3-hydroxy-5methyl-4-isoazolepropionic acid (AMPA) receptors. Internalization of AMPA receptors in response to various stimuli is an important process for the rapid attenuation of neurotransmission during synaptic plasticity (3). By different approaches, it has been shown that AMPA receptors are endocytosed by a clathrin-coated pit-mediated process (4 -6).
We are interested in understanding the endocytic properties of mGluR5. Alternative splicing of the mGluR5 gene generates two splice variants: mGluR5a and mGluR5b, which differ only by a 32-amino acid insertion in the intracellular C tail domain of mGluR5b. There is no functional difference between them except that mGluR5b seems to be the predominant form expressed in brain of adult animals (7). Both splice variants as well as mGluR1a interact with the family of Homer proteins (2). The first identified member of this family, Homer1a, is a product of an immediate early gene expression, which is induced during intense neuronal activities. All Homers can bind to mGluR5 and mGluR1a by their amino terminus and also to other proteins of the postsynaptic density.
Although most GPCRs undergo endocytosis, the conditions and mechanisms of this process can vary from receptor to receptor. Many of them are endocytosed via the clathrin-coated pit pathway but some are not (8 -10). Some have an agonistinduced endocytosis, and some are continuously endocytosed even in the absence of stimulation. These different behaviors should serve distinct physiological functions.
By using transfection of epitope-tagged mGluR5 receptors in COS-7 cells and hippocampal cultured neurons we found that mGluR5a and -5b are endocytosed constitutively by a clathrinindependent pathway.

Constructs
The epitope-tagged rat mGluR5a and mGluR5b expression plasmids were constructed using, as a template, the pRKG5a and pRKG5b plasmids that encode mGluR5a and mGluR5b (12). An MluI restriction site was inserted just after the signal peptide for the N-terminal epitopetagged receptor (between the Ser 22 and Ser 23 codons of mGluR5a) using a PCR overlap extension method. Sense and antisense oligonucleotides coding the c-Myc epitope (EQKLISEEDL) or the HA epitope (YPYDVP-DYA) with MluI cohesive ends were synthesized and used to introduce the c-Myc or the HA into the N-terminal MluI site. Using this method, we generated the plasmids pRKG5aNmyc (mGluR5-myc in Ref. 13), pRKG5bNmyc, pRKG5aNHA, and pRKG5bNHA encoding mGluR-5aNmyc, mGluR5bNmyc, mGluR5aNHA, and mGluR5bNHA, respectively. By determining inositol phosphate accumulation in transfected cells as described previously (14), we verified that the epitope tagging changed neither the G protein coupling nor the pharmacological profile of the receptor (Table I). Plasmids encoding mGluR5aNHA mutated in the glutamate binding site were obtained by using the QuikChange strategy. Plasmids encoding the GFP-tagged dynamins, Dyn1-GFP, Dyn1K44A-GFP, Dyn2-GFP, and Dyn2K44A-GFP, were kindly provided by M. McNiven (15). To determine the role played by Eps15, we used two cDNAs kindly provided by A. Benmerah and A. Dautry-Varsat, one control GFP-DIII⌬2 (16) and one dominant negative mutant GFP-E⌬95/295 (17).
Rat Hippocampal Neurons-Hippocampal cells were prepared from 18-day-old rat embryos according to the method of Goslin and Banker (18). Briefly, the hippocampi were dissociated by trypsin and mechanical treatment, plated on poly-L-lysine-coated glass coverslips at a density of 9500 cells/cm 2 , and maintained in a serum-free medium (N2 medium) suspended above a glia feeder layer. Adding 5 mM cytosine arabinoside prevented proliferation of nonneuronal cells. All experiments were performed in cells kept 8 -12 days in culture. Neurons were transfected with designated plasmids using Effectene (Qiagen) according to the manufacturer's indications except that the cells were left only 1 h in contact with Effectene and DNA.

Endocytosis Assay and Immunocytochemistry
To follow endocytosis of tagged receptors, antibodies anti-Myc (5 g/ml) or anti-HA (1:100), diluted in conditioned culture medium, were incubated on living cells for 30 min at 37°C. When specified, Rh-Tf was added at a concentration of 300 nM. Cells were either directly fixed by incubation for 20 min at room temperature in 4% paraformaldehyde, 4% sucrose in PBS, or remaining surface antibodies were stripped away with an acid wash (4-min incubation at 4°C with Dulbecco's modified Eagle's medium, pH 2) before fixation.
After fixation, cells were incubated for 10 min at room temperature with 50 mM NH 4 Cl in PBS in order to quench aldehyde groups and then permeabilized in PBS, 0.2% bovine serum albumin, 0.05% saponin (permeabilizing buffer) for 40 min at 37°C. If necessary, the cells were incubated for 45 min at room temperature with anti-MAP2 (1:2000) in permeabilizing buffer. After three washes in permeabilizing buffer, cells were incubated for 45 min at room temperature with fluorochromecoupled secondary antibodies (Cy3 anti-mouse or anti-rabbit (1:500), fluorescein isothiocyanate anti-rabbit (1:100)), washed several times with permeabilizing buffer then in PBS, and finally mounted in Dabco/ Moviol as described by Hémar et al. (19).
When the remaining surface anti-Myc (1:100) was visualized after having been incubated on living cells for 30 min at 37°C, cells were incubated for 1 h at 4°C with Cy3 anti-mouse secondary antibodies and then fixed and mounted in Dabco/Moviol.
Immunofluorescence was visualized with a Zeiss Axiophot 2 microscope. Image acquisitions were done with a Quantix digital cooled CCD camera (Photometrix) driven by IP-labs (Scanalitics) software. Images were mounted with Adobe Photoshop.

Image Analysis
The detection and counting of endocytic vesicles was performed automatically by a computer program that uses an undecimated wavelet decomposition of the image. The program recognizes the vesicles by correlating and thresholding wavelet coefficients at different scales of analysis (20). Region masks defining the area occupied by the cell body and processes are determined by an automatic thresholding algorithm when processing fluorescently labeled neurons or by interactive drawing when analyzing COS-7 cells. Vesicles inside each cell are validated as the result of a logical AND operator between the vesicle image and the region masks. The results are presented as the mean Ϯ S.E. of the number of vesicles per cell surface unit. Statistical significance was assessed by using a Mann-Whitney test.

Biotinylation Assay of Receptor Endocytosis
The day before the experiment, COS-7 cells transfected with expression vector encoding untagged mGluR5a were pooled and plated in six-well plates in order to avoid variability in transfection efficiency between wells. Cells were washed twice in ice-cold PBS (pH 8) and incubated with 1 ml of sulfo-NHS-SS-biotin (Pierce) diluted in PBS (pH 8) for 30 min at 4°C. After two washes in PBS, two wells were kept at 4°C (t 0 ) in ice-cold PBS, and the others were incubated different times at 37°C in prewarmed PBS. Cells were then washed twice in PBS. For each incubation time, one well was kept at 4°C while the other was incubated three times for 15 min with 1 ml of cleavage solution (50 mM 2-mercaptoethanesulfonic acid, 100 mM NaCl, 2.5 mM CaCl 2 , and 50 mM Tris-HCl (pH 8.7 at 4°C)) and washed twice in ice-cold PBS. All of the wells were lysed by scraping in 1 ml of lysis buffer (25 mM Hepes, 150 mM NaCl, 1% Triton X-100, and "Complete" protease inhibitor mixture (Roche Molecular Biosciences)). After 20 min on ice, each lysate was spun at 8000 ϫ g for 10 min. 20 l of each clarified lysate was kept for an estimation of the total quantity of mGluR5 at each point. 850 l of each clarified lysate were incubated overnight at 4°C with 50 l of streptavidin-agarose beads. The beads were washed three times at 4°C in lysis buffer and once in 50 mM Tris-HCl (pH 7.4). The beads were extracted with SDS-sample buffer containing 1 M dithiothreitol. The lysates and the extracted beads were analyzed by Western blotting using anti-mGluR5 antibody (0.8 g/l). The ECL kit (Amersham Biosciences) was used for chemiluminescent detection. Quantitative analysis of Western blots was performed using Metamorph software (Universal Imaging). We measured the amount of mGluR5 at the surface at t ϭ 0 (i.e. biotinylated receptors at t ϭ 0), the amount of surface and intracellular receptors (i.e. total biotinylated receptors) at each time of incubation at 37°C, and the intracellular receptors (biotinylated receptors resistant to cleavage). We measured for each point the total amount of mGluR5 in order to calibrate these data. The percentage of internalized mGluR5 at time t was then calculated with the equation, where surface(t 0 ) Ϫ total(t) represents the degraded receptors or endocytosed receptors which underwent biotin reduction inside the cell.
The endocytosis rate was determined by fitting data to a single exponential function of the form y ϭ A(1 Ϫ e (t/) ), where A is constant, using Kaleidagraph. The endocytosis rate is 1/.

mGluR5a and -b Are Endocytosed in Transfected COS-7 Cells and in Rat Hippocampal
Neurons-To investigate internalization of mGluR5 receptors, recombinant receptors were transiently expressed in heterologous COS-7 cells or in hippocampal neurons. In transfected COS-7 cells, the receptors are detected at the cell surface and also in high proportion (90%) in intracellular compartments (putatively of the biosynthetic and endocytic pathways). 2 Since there is no available antibody against the extracellular part of mGluR5, we used mGluR5a and mGluR5b constructs containing the c-Myc or HA epitope tags at the extracellular N terminus, allowing specific labeling of surface-expressed receptors on living cells with antitag antibodies. Endocytosis immunoassay was performed as follows: 1) transfected cells were incubated for 30 min at 37°C with anti-tag antibodies; 2) remaining surface antibodies were or not stripped away with an acid wash; 3) finally, cells were fixed, permeabilized, and incubated with fluorescent secondary antibodies. Both splice variants, mGluR5aNmyc and mGlu-R5bNmyc, were endocytosed as shown by the presence of an acid wash-resistant punctiform labeling (Fig. 1, E and F). The labeled compartments are endosomal structures, because they contain antibodies that have been internalized from culture medium. The antibody was internalized by receptor-mediated endocytosis, since no intracellular labeling was observed in nontransfected cells on the same coverslips (Fig. 1, compare A with C and B with D). The comparison of the images obtained with or without acid wash shows that there are still receptors at the cell surface (Fig. 1, compare E with A and F with B). The remaining surface labeling appears very different from that of the endocytic compartments. Remaining surface antibodies were revealed, after the incubation at 37°C, under nonpermeabilizing conditions (i.e. at 4°C without fixation and permeabilization (Fig. 1, G and H)). The labeling appears as smaller puncta compared with the intracellular labeling (Fig. 1, compare G with E and H with F). This shows that the binding of the antibody has not induced a severe down-regulation of the receptor at the surface. In any case, the cross-linking of the receptors by the bivalent anti-Myc antibody does not seem to be necessary for their internalization, since we observed the same labeling when anti-Myc Fab fragments were used. 3 In order to determine whether the mGluR5 receptors are also endocytosed in neurons, we performed the same experiments in primary cultures of rat embryonic hippocampal dissociated neurons. In transfected neurons, mGluR5a and -b are both expressed and endocytosed in dendrites and axons (Fig. 2). Indeed, after incubation at 37°C, anti-Myc labeling appears as large fluorescent acid wash-resistant puncta in cell body and neurites (Fig. 2, B and C, and Fig. 2, E and F, with acid wash, compared with Fig. 2, A and D, without acid wash). The internalized receptors (in red) are found in dendrites labeled with the dendritic specific marker MAP2 (in green; Fig. 2, B and E) and in axons that do not contain MAP2 (Fig. 2, C and F).
mGluR5a Is Constitutively Endocytosed-In the experimental conditions we used, mGluR5a or mGluR5b receptors are endocytosed without the addition of glutamate. This suggested that the receptors could be endocytosed constitutively, which means in the absence of ligand binding and/or activity.
Although the first observations have been done in the absence of added agonist, one cannot rule out that such an agonist is produced by the cells and would induce the internalization of the receptor. To test this hypothesis, we studied the endocytosis of mutated receptors that have a reduced affinity for glutamate. We took advantage of the similarity of the glutamate binding site of another group 1 metabotropic glutamate receptor, mGluR1, to design the mutants. The binding site of mGluR1 has been crystallized, and the amino acids interacting with glutamate have been identified (21). We mutated into alanine the tyrosine at position 222 and the aspartate at position 304 of mGluR5a (mGluR5aNHA-Y222A and mGluR5aNHA-D304A, respectively). The EC 50 for glutamate in activating the Y222A and D304A mutated receptors is 30and Ͼ300-fold higher than that measured with the wild-type receptor, respectively (Table I). Both mutated receptors transfected in COS-7 cells were endocytosed (Fig. 3, A and B). We have evaluated the endocytosis level by image analysis. We quantified the number of acid wash-resistant fluorescent spots (endosomes) per cell surface unit (Fig. 3C). The numbers obtained for both mutants are not significantly different from that of the wild type (wild type: 21.4 Ϯ 1.2 spots/100 m 2 , 2.1 ϫ 10 4 m 2 explored in 47 cells analyzed; mGluR5aNHAY222A, 21.2 Ϯ 2.9 spots/100 m 2 , 6.2 ϫ 10 3 m 2 explored in 11 cells analyzed, p ϭ 0.8; mGluR5aNHAD304A, 20.7 Ϯ 3.3 spots/100 m 2 , 5.3 ϫ 10 3 m 2 explored in 10 cells analyzed, p ϭ 0.7). We conclude that the binding of a ligand is not necessary for the endocytosis of mGluR5a.
It has been shown that mGluR5 receptors display an endogenous constitutive activity that can be inhibited by the inverse 2 A. Hémar, unpublished data. 3 L. Fourgeaud and A. Hémar, unpublished data. agonist MPEP (22,23). MPEP also inhibits the activity stimulated by the agonist binding. In order to look for endocytosis of mGluR5aNmyc in the absence of activity, we treated cells with 10 M MPEP for 30 min before and during the anti-Myc incubation in our endocytosis assay. Fig. 3 shows that in COS-7 cells (Fig. 3D) and in neurons (Fig. 3E) We conclude that neither agonist binding nor intrinsic activity of the receptor is necessary for endocytosis of mGluR5a.
In order to evaluate the amount of receptors that undergo constitutive endocytosis we performed biotinylation assays of endocytosis on COS-7 cells transfected with untagged mGluR5a encoding vector. After biotinylation at 4°C with sulfo-NHS-SS-biotin, the cells were incubated at 37°C in PBS for different times. After purification on streptavidinagarose beads and Western blot revealed with anti-mGluR5 antibodies, we measured the total (without cleavage of surface biotin) and intracellular (after cleavage) amounts of biotinylated receptors. Fig. 4A shows the corresponding mGluR5 labeling on one Western blot. The amount of intracellular receptors increases with time while the total amount decreases. Since the experiment was done in PBS, this shows that the receptor is indeed constitutively endocytosed. We quantified the density of the bands and calculated the percentage of internalized receptors (Fig. 4B, see "Experimental Procedures" for details). The time course of endocytosis can be well described by a single exponential and a time constant () of 8.5 min. The rate of endocytosis is thus 11.7%/min. This indicates that the constitutive endocytosis of mGluR5 is not due to bulk flow endocytosis of the membrane, since this value is significantly higher than 1%/min (24).
mGluR5a Is Endocytosed by a Clathrin-independent Pathway-Several internalization pathways can coexist in a cell. The most extensively characterized mechanism of endocytosis is clathrin-mediated, but it has been recently evidenced that bona fide receptors can enter the cell by other pathways (reviewed in Refs. 9 and 10). In order to discriminate between these pathways, the best tools now available are dominant negative mutants of proteins implicated in clathrin-mediated endocytosis like Eps15 and the dynamins.
Overexpression of an EH-deleted dominant negative mutant of Eps15 (E⌬95/295) that still contains the binding site for the adaptor AP2 inhibits clathrin-coated pit assembly and blocks clathrin-dependent endocytosis (16,17). We used this mutant to see if it could inhibit endocytosis of mGluR5a in COS-7 cells and in hippocampal neurons. We cotransfected GFP-tagged E⌬95/295 and mGluR5aNHA and looked for the endocytosis of anti-HA antibodies. In order to control the efficiency of the Eps15 mutant, Rh-Tf endocytosis was monitored in the same cells. Whereas Rh-Tf endocytosis is abolished (Fig. 5G), Fig. 5, E and F (for COS-7 cells) and Fig. 6, C and D (for neurons) show that endocytosis of the receptor (Figs. 5F and 6D) is still observed in cells coexpressing GFP-E⌬95/295 ( Fig. 5E and 6C). Quantifications show that the endocytosis in the presence of dominant negative mutant of Eps15, GFP-E⌬95/295, is the same as in cells transfected with a control Eps15 construct, GFP-DIII⌬2, lacking the AP2 binding site (Fig. 5, A and B, for COS-7; Fig. 6, A and B, for neurons) (25), we conclude that mGluR5a is endocytosed through a nonclathrinmediated pathway. Nevertheless, after endocytosis, the receptor is found at least partly in the same endosomes as transferrin endocytosed by clathrin-coated pits (Fig. 5D). This has been observed also for the interleukin-2 receptor and more recently for the muscarinic acetylcholine receptor M2 (19,26,27). To confirm that the mGluR5a is not in clathrin-coated pits at the cell surface, we performed a double labeling of the receptor at 4°C and of ␣-adaptin after permeabilization (Fig. 7). ␣-Adaptin is a protein that is part of the adaptor complex AP2, which is associated with clathrin-coated pit at the plasma membrane (28). As expected, very little, if any, colocalization is observed between the mGluR5aNmyc cell surface clusters and ␣-adaptin. In accordance, no colocalization was observed with clathrin using anti-clathrin heavy chain antibodies (data not shown).
Dynamins 1 and 2 are large GTPases that have been implicated in clathrin and nonclathrin-mediated endocytosis (9, 10, 25, 29 -31). Dynamin 1 is a presynaptic protein that is used for the recycling of synaptic vesicles, whereas dynamin 2 is ubiquitously expressed (15). To dissect more precisely the machinery used by mGluR5 to be internalized, COS-7 cells or cultured hippocampal neurons have been co-transfected with mGluR5aNHA and dynamin 1 or dynamin 2 wild type or dominant negative mutants coupled to GFP.
In hippocampal neurons, although dynamin mutants seem to be well expressed and localized near the membrane, mGluR5a is still endocytosed with dynamin 1 or 2 dominant negative mutant ( Fig. 9; Dyn1-

DISCUSSION
Our results show by immunofluorescence and image quantification that the group 1 metabotropic glutamate receptor mGluR5a is endocytosed in COS-7 cells (Fig. 1) and in axons and dendrites of hippocampal neurons in culture (Fig. 2). mGluR5a is endocytosed constitutively, because mutated receptors that cannot bind glutamate are internalized (Fig. 3, A-C) and because the receptor is still internalized in the presence of the inverse agonist MPEP that inhibits the intrinsic and stimulated activity of the receptor (Fig. 3, D-F). We found that the endocytic rate is 11.7%/min as measured by a biochemical assay (Fig. 4). Using dominant negative mutants of proteins involved in clathrin-coated pit-mediated endocytosis (Eps15 and dynamins), we show that mGluR5a is endocytosed by a clathrin-independent pathway (Figs. 5, 6, 8, and 9).
Compartmentalization of Transfected Epitope-tagged mGluR5 Receptors-Previous studies on the in vivo localization of mGluR5 reported a mainly dendritic but also axonal presence of mGluR5 in hippocampus (33,34). The immunolocalizations were done with antibodies that could not discriminate between the two splice variants mGluR5a and -b that differ only by 32 amino acids inserted in the intracellular part of the mGluR5b (12, 35). One hypothesis is that one splice variant could be presynaptic and the other postsynaptic. For the first time, we report that both epitope-tagged mGluR5a and mGluR5b are expressed and endocytosed in axons and dendrites of transfected hippocampal neurons in culture. The nonpolarized expression of the transfected mGluR5 receptors in hippocampal neurons in culture contrasts with the polarized dendritic localization in hippocampal neurons of transfected mGluR1a, which shares 60% homology with mGluR5 (data not shown) (36,37). Hence, our results disfavor the hypothesis that mGluR5 splice variants have different polarized localizations.
Ango et al. (38) have shown that, transfected in cultured cerebellar granule cells, mGluR5a is expressed only in cell bodies. Overexpression of the immediate early gene product Homer1a leads to the nonpolarized expression of the receptor in dendrites and axons as we observed in hippocampal neurons. Overexpression of the long forms of Homer1 (Homer1b and -c) leads to the polarized dendritic expression of mGluR5a. In contrast, they found that, transfected in striatal neurons in culture that express endogenously Homer1b proteins, mGluR5a is expressed in neurites. In accordance with these FIG. 4. Constitutive endocytosis rate of mGluR5 in COS-7 cells. COS-7 cells transfected with pRKG5a were biotinylated with cleavable biotin at 4°C. The endocytosis was then performed at 37°C in PBS for the indicated times. At each time, a sample was cleaved in order to evaluate the intracellular level of biotinylated mGluR5, and a sample was not, in order to evaluate the total level of biotinylated receptors. Extracts of each sample were purified on streptavidin-agarose beads, and biotinylated mGluR5 was revealed by Western blot. mGluR5 labeling of one experiment is presented in A. After quantification, we calculated the amount of internalized receptors as described under "Experimental Procedures." A fit of the data with a single exponential is presented in B (correlation coefficient r ϭ 0.94). Two independent experiments were pooled.
FIG. 5. mGluR5aNHA is endocytosed in COS-7 cells in which clathrin-coated pit internalization is inhibited. COS-7 cells were transfected with pRKG5aNHA together with an expression vector encoding a mutated Eps15 protein, GFP-E⌬95/295 (E-H), or a control Eps15 construct, GFP-DIII⌬2 (A-D). After incubation for 30 min at 37°C with anti-HA antibody and Rh-Tf, remaining surface antibodies were acid-stripped, and the intracellular antibodies were detected by Cy5-coupled secondary antibodies. A and E, images of GFP labeling; B and F, images of Cy5 (mGluR5aNHA) labeling; C and G, images of rhodamine (transferrin) labeling; D and H, merged images of B and C and of F and G, respectively, where the mGluR5aNHA labeling is represented in red and Rh-Tf labeling is in green. The yellow that appears corresponds to the colocalized labelings. Scale bar, 10 m. data, we have observed by reverse transcription-PCR and Western blot that hippocampal neurons in culture express Homer proteins (13) (data not shown).
Endocytosis of mGluR5 has been reported in cell bodies of cultured enteric neurons (39). Here we report that both splice variants are endocytosed in cultured neurons from the central nervous system, in dendrites, and in axons. Axonal endocytosis of receptors is not well documented; to our knowledge, only one report shows the axonal endocytosis of a GPCR (dopaminergic receptor D1) (40). Recently, Garrido et al. (41) showed that the C-terminal tail of the voltage-gated sodium channel (Na v 1.2) contains a sequence able to target a reporter protein to axons  and dendrites of transfected hippocampal neurons, but it drives endocytosis of this chimera only in the somatodendritic domain. This dissymmetry leads to a polarized axonal expression. This exemplifies that axonal endocytosis should have different molecular requirements than that in dendrites.
Constitutive Endocytosis-By two different approaches, we show that mGluR5aNHA is endocytosed constitutively (i.e. in absence of ligand binding and phospholipase C activation). We obtained the same results for endocytosis of mGluR5b in the presence of MPEP. 3 By a biochemical approach, we showed that the wild type receptor (untagged) is also constitutively endocytosed at a rate around 12%/min. In agreement with our data, Hubert et al. have observed in the substancia nigra that 80% of the labeling of mGluR5 is intracellular, although they have not identified the type of intracellular compartments containing mGluR5, biosynthetic or endocytic (42). Although we show a constitutive endocytosis, it does not mean that endocytosis flux is not regulated by ligand or receptor activity. A regulation of mGluR5 endocytosis by ligand has been described in enteric neurons (39). It has been reported by several groups that the very close cousin of mGluR5, mGluR1a, is endocytosed in heterologous cells. Both a constitutive and a stimulated endocytosis have been observed (43)(44)(45)(46)(47). We show here that, surprisingly, the number of intracellular spots is increased after MPEP treatment. This could come from an accelerated endocytosis, a modification of the shape of endosomes, and/or changes in the intracellular pathway followed by mGluR5. We pursue the biochemical experiments in order to measure potential variations of endocytosis parameters (endocytosis rate, proportion of recycling/degradation) that we cannot obtain by image analysis. In the quantification, we count the number of spots/cell surface but not the content and size of each spot.
The constitutive endocytosis of mGluR5 creates an intracellular pool that could permit the cell to rapidly modulate the number of cell surface receptors.
Internalization Pathway-Using the dominant negative mutant of Eps15, a protein involved in clathrin-coated pit-mediated endocytosis, we show that mGluR5a is endocytosed by a nonclathrin-coated pit pathway. The same results have been obtained for mGluR5b (data not shown). Our results demonstrate for the first time that a member of GPCR group C is constitutively endocytosed by a nonconventional pathway.
If agonist-stimulated mGluR1a endocytosis in HEK293 cells seems to occur trough clathrin-coated pits (45,47), the constitutive endocytosis pathway is less well established (46). Authors have concluded the use of clathrin-mediated pathway, because the endocytosis was inhibited by hypertonic sucrose treatment and because the receptor is found in the same endosomes as transferrin. Hypertonic sucrose does not inhibit only clathrin-mediated endocytosis. For instance, it inhibits M2 muscarinic acetylcholine and GABA A receptors endocytosis that are not mediated by clathrin-coated pits (48,49). mGluR5 endocytosis is also inhibited by hypertonic sucrose in COS-7 cells (data not shown), and endocytosed receptors colocalized partly with endocytosed transferrin (Fig. 5). Furthermore, Dale et al. (46) showed that dynamine 1 dominant negative mutant does not inhibit mGluR1a endocytosis as we found for mGluR5. Thus, further experiments using Eps15 mutants on mGluR1a constitutive internalization will be more conclusive for the understanding of the constitutive internalization pathway followed by mGluR1a. mGluR5a (as well as mGluR5b) (data not shown) endocytosis is insensitive to dynamin 1 mutant in COS-7 cells and in neurons. However, in COS-7 cells, the nonconventional pathway followed by mGluR5 uses dynamin 2, the endogenous isoform in these cells. Surprisingly, dynamin 1 dominant negative mutant is not able to compete with dynamin 2 in this pathway, although it competes with dynamin 2 implicated in clathrin-coated pit endocytosis. Indeed, transferrin endocytosis is inhibited by both dynamin 1 and dynamin 2 dominant negative mutants in COS-7 cells (data not shown) (15). This points out that dynamins may have distinct functions in clathrin and nonclathrin endocytosis. This can be put together with the results of Altschuller et al. (50), showing a different function of dynamins in apical versus basolateral endocytosis in the polarized epithelial cell line Madin-Darby canine kidney cells. Furthermore, in a collaborative work with the laboratory of Mark McNiven (Mayo Clinic, Rochester, MN), we have shown that the dynamin 3aaa isoform (31) can inhibit mGluR5 endocytosis in COS-7 cells (data not shown).
In neurons, we obtained a different result in that dynamin 2 dominant negative mutants does not inhibit mGluR5 endocytosis, although transfected Dynamin 2 mutants are targeted to the membranes close to the site of mGluR5 endocytosis (Fig. 6). Perhaps the experimental paradigm is not adapted to neurons. This hypothesis cannot be ruled out, because there is only one publication using these tools in neurons. The authors show that FIG. 9. Effect of dynamin dominant negative mutants in hippocampal neurons. Neurons were transfected with pRKG5aNHA together with expression vector encoding mutated dynamin 1 coupled to GFP (A and B; Dyn1K44A in E) or mutated dynamin 2 coupled to GFP (C and D; Dyn2K44A in E). After incubation for 30 min at 37°C with anti-HA antibody, the anti-HA antibodies were detected by Cy3-coupled secondary antibodies. A and C, images of the GFP-coupled mutant (Dyn1K44A-GFP and Dyn2K44A-GFP, respectively); B and D, labeling of mGluR5aNHA in the same cell. In E is the histogram of the quantified results of the mutants compared with the control neurons transfected with the wild type dynamins. Insets, the entire cell from which the enlarged region has been taken. Scale bar, 10 m.
virally overexpressed dynamin 2 mutants inhibit the ionotropic glutamate receptor AMPA GluR2 subunit internalization (4). Another explanation for such an observation in hippocampal neurons is that there could exist an endocytosis pathway used by mGluR5 that does not depend on dynamins 1 and 2. In general, the role of dynamins in the different neuronal compartment where they are expressed is not clear except for dynamin 1 in synaptic vesicle recycling (51,52).
The involvement of dynamins in nonclathrin-coated pit-mediated endocytosis has already been described in nonneuronal cells (29,30,53,54). Endocytosis of some GPCRs is insensitive to dynamin 1 dominant negative mutants in heterologous cells (e.g. the muscarinic acetylcholine receptor M2 (55), dopaminergic receptor D2 (56), and angiotensin receptor AT 1A (57)). This suggests that they are not endocytosed by a clathrincoated pit-mediated endocytosis. It would be interesting to know if the pathway used by these receptors needs dynamin 2 as we show for mGluR5. An inhibitory effect of dynamin 1 dominant negative mutants has been observed in some nonclathrin-coated pit endocytosis: in the case of interleukin-2 receptor endocytosis using the regular K44A mutant (53) and in the case of muscarinic receptor M2 using dynamin 1 that lacks the complete GTP binding domain or the phosphatidylinositol 4,5-bisphosphate-stimulated GTPase activity (54). It is interesting to note that in that last case the dynamin 1 K44A mutant was not able to inhibit endocytosis. This suggests that perhaps dynamins do not interact with the same partners in clathrin-dependent and -independent endocytosis.
What could be this clathrin-independent pathway? The main studied nonclathrin pathway is caveolae-mediated endocytosis, but few transmembrane receptors have yet been found in such caveolin-coated structures. Moreover, real caveolae have not been observed in lymphocytes and in neurons (58 -60). A noncoated, raft-mediated endocytosis could be involved. Rafts are membrane microdomains enriched in sphingolipids and cholesterol in which there is preferential localization of glycosylphosphatidylinositol-anchored proteins and some signaling molecules (59). Interleukin-2 receptors have been localized in detergent-resistant membrane and observed in noncoated areas of the membrane. A first step to unravel the internalization pathway of mGluR5 will be to know if mGluR5 is present in rafts. There are two interesting observations: the G q , the G protein coupled to mGluR5, is present in rafts (61,62), and mGluR1a, endocytosed through clathrin-coated pits, is not found in rafts (63). Elucidation of this nonclathrin endocytosis is of great interest for understanding the peculiar role for endocytosis of receptors.