Rapid agonist-induced desensitization and internalization of the A(2B) adenosine receptor is mediated by a serine residue close to the COOH terminus.

The G(s)-coupled rat A(2B) adenosine receptor (A(2B)-AR) was epitope-tagged at the NH(2) terminus with hemagglutinin (HA) and subjected to progressive deletions or point mutations of the COOH terminus in order to determine regions of the receptor that contribute to agonist-induced desensitization and internalization. When expressed stably in Chinese hamster ovary cells, a mutant receptor in which the final 2 amino acids were deleted, the Leu(330)-stop mutant, underwent rapid agonist-induced desensitization and internalization as did the wild type (WT) receptor. However, the Phe(328) and the Gln(325)-stop mutants were resistant to rapid agonist-induced desensitization and internalization. Co-expression of arrestin-2-green fluorescent protein (arrestin-2-GFP) with WT receptor or Leu(330)-stop mutant resulted in rapid translocation of arrestin-2-GFP from cytosol to membrane upon agonist addition. On the other hand, agonist activation of the Phe(328)-stop or Gln(325)-stop mutant did not result in translocation of arrestin-2-GFP from cytosol. A COOH terminus point mutant, S329G, was also unable to undergo rapid agonist-induced desensitization and internalization, indicating that Ser(329) is a critical residue for these processes. A further deletion mutant (Ser(326)-stop) unexpectedly underwent rapid agonist-induced desensitization and internalization. However, activation of this mutant did not promote translocation of arrestin-2-GFP from cytosol to membrane. In addition, whereas WT receptor internalization was markedly inhibited by co-expression of dominant negative mutants of arrestin-2 (arrestin-2-(319-418)), dynamin (dynamin K44A), or Eps-15 (EDelta95-295), Ser(326)-stop receptor internalization was only inhibited by dominant negative mutant dynamin. Taken together these results indicate that Ser(329), close to the COOH terminus of the rat A(2B)-AR, is critical for the rapid agonist-induced desensitization and internalization of the receptor. However, deletion of the COOH terminus also uncovers a motif that is able to redirect internalization of the receptor to an arrestin- and clathrin-independent pathway.

Prolonged or repeated exposure of G protein-coupled receptors (GPCRs) 1 to agonist usually results in a decrease in subsequent receptor responsiveness, a process termed desensitization (1). The mechanisms underlying rapid desensitization often involve GPCR phosphorylation by a family of G proteincoupled receptor kinases (GRKs; Ref. 2). Receptor phosphorylation by GRKs promotes binding of arrestins (3), which triggers desensitization by uncoupling the receptor from G-protein (4). Following desensitization, many GPCRs appear to internalize by an arrestin-dependent process via clathrin-coated pits (5). In most cases this leads to the eventual intracellular dephosphorylation of the receptor, and its reinsertion into the cell membrane in a resensitized state (6). However, in some cases GPCR internalization appears to contribute to desensitization (7). More prolonged agonist activation generally leads to the redirection of internalized receptor to a lysosomal compartment with subsequent down-regulation (8), although some GPCRs appear to be targetted for down-regulation after relatively short agonist treatment times (9). Thus although a general picture is emerging of how GPCR responsiveness is regulated in the presence of prolonged agonist treatment (1)(2)(3)(4), the molecular signals that determine the pathways of desensitization, internalization, resensitization, or down-regulation remain to be clarified.
The endogenous nucleoside adenosine is now known to regulate cellular function via activation of four adenosine GPCRs, the A 1 , A 2A , A 2B , and A 3 -ARs (10). Both the A 2A -and A 2Badenosine receptors are G s coupled and stimulate cAMP formation, and the A 2B -AR subtype is thought to regulate such diverse processes as vascular tone, neurosecretion, and mast cell activation (11). Agonist-induced desensitization of A 2B -ARs has been reported in a number of cell types (12)(13)(14)(15)(16) and tissues (17). Using a combination of molecular biological techniques we have shown that the mechanism of A 2B -AR desensitization likely involves GRK2 and non-visual arrestins. Thus, expression of a dominant negative mutant GRK2 construct in NG108-15 cells (12) or the antisense-induced reduction of nonvisual arrestin levels in HEK293 cells (15) reduced the rate of agonist-induced desensitization of endogenous A 2B -AR responsiveness. Very recently, we have shown that agonist activation of A 2B -ARs transiently expressed in HEK293 cells promotes arrestin-2-GFP translocation from cytosol to cell membrane (18). Furthermore, the A 2B -ARs subsequently undergo internalization to compartments which co-localize with the endosomal markers transferrin and rab-5 (18), as well as with arrestin-2-GFP itself. Another recent study (19) indicates that the A 2B -AR internalizes in an arrestin-and dynamin-sensitive fashion. Together, these results indicate that A 2B -AR undergoes desensitization and internalization in a GRK-and arrestindependent manner. For other GPCRs, the COOH terminus of the receptor is often implicated as a site for GRK-mediated phosphorylation and arrestin interaction (2). For example, mutation of a threonine residue in the COOH terminus of the canine A 2A -AR inhibited the agonist-induced phosphorylation and desensitization of this receptor (20). At present nothing is known of the molecular determinants of the desensitization and internalization of the A 2B -AR. In the present study we investigated the role of the COOH terminus of this receptor in these processes using a series of COOH terminus deletion and point mutants. The results indicate that a serine residue close to the end of the COOH terminus of A 2B -AR is critical for rapid agonist-induced desensitization and internalization of the receptor. Furthermore, we find that altering the COOH terminus can alter the arrestin dependence of agonist-induced receptor internalization.

EXPERIMENTAL PROCEDURES
Materials-The pcDNA3 and pEGFPN1 vectors were obtained from Invitrogen and CLONTECH, respectively. The construct EGFP-C2-Eps15 (E⌬95-295) was a kind gift from Dr. A. Benmerah. Fugene-6 and restriction enzymes were obtained from Roche Molecular Biochemicals, ligase from New England Biolabs, and DNA preparation kits from Qiagen. DMEM and DMEM/F-12 (50:50) were from Life Technologies, fetal calf serum from Harlan Sera Labs, and geneticin (G418 sulfate) from Calbiochem. The anti-HA monoclonal HA.11 antibody was from BabCO, the goat anti-mouse secondary antibody conjugated with alkaline phosphatase from Sigma, the rhodamine-conjugated mouse monoclonal antibody 12CA5 from Roche Molecular Biochemicals, and the goat anti-mouse rhodamine-conjugated secondary antibody from Molecular Probes, as was the Slow-Fade Mounting Medium. Colorimetric alkaline phosphatase substrate was from Bio-Rad. [ 3 H]cAMP was obtained from Amersham Pharmacia Biotech Int. All other biochemical reagents were obtained from Sigma.
Cloning, Hemagglutinin (HA) Epitope Tagging, and Site-directed Mutagenesis of the Rat A 2B -AR-The rat A 2B -AR was cloned from rat brain by reverse transcriptase-polymerase chain reaction, using the sense primer 5Ј-GACTCTAGAATGCAGCTAGAGACGCAGG-3Ј and the antisense primer 5Ј-GACTCTAGATCACAAGCTCAGACTGAAAGT-3Ј, which include a site for the restriction enzyme XbaI (underlined). The cDNA obtained was subcloned into the XbaI site of HA-pcDNA3. The expression vector HA-pcDNA3 was derived from pcDNA3 and pJ3H (21). Briefly, a HindIII-EcoRI DNA fragment from pJ3H containing the sequence for the HA epitope tag and part of the vector Multiple Cloning Site was subcloned into the unique XbaI site of pcDNA3. COOH-terminal truncated A 2B -AR cDNA clones were obtained by polymerase chain reaction amplification, using WT A 2B -AR cDNA as template, and sense and antisense primers containing a site for the restriction enzyme XbaI, for subsequent subcloning in HA-pcDNA3. Point mutants were obtained by polymerase chain reaction amplification as for deletion mutants, with an antisense primer containing the desired mutation and a site for the restriction enzyme EcoRI at the 5Ј end, to facilitate subsequent cloning in HA-pcDNA3. The sequence of all constructs was checked by automated DNA sequencing. The WT A 2B -AR was GFPtagged at the COOH terminus by subcloning the coding sequence of the rat A 2B -AR into the vector pEGFPN1.
Cell Culture and Transfections-CHO-K1 cells were cultured in DMEM/F-12 (50:50) medium, 10% fetal calf serum, 100 units ml Ϫ1 penicillin, and 100 g ml Ϫ1 streptomycin. Stably transfected CHO cells were cultured in the above medium supplemented with 600 g ml Ϫ1 geneticin. For stable transfections, 60 -80% confluent CHO cells in 2 ml of medium in a 60-mm culture dish were transfected with 1 g of DNA linearized with BglII, using 3 l of Fugene-6, according to the manufacturers instructions. After 24 h, cells were split (1:20) into medium containing 600 g ml Ϫ1 geneticin. Individual cell clones were allowed to develop for 1-2 weeks and transferred to 24-well dishes using a pipette tip. Clones functionally expressing receptor constructs were identified by measuring 10 M NECA-stimulated cAMP accumulation, as described below.
Whole Cell cAMP Accumulation-CHO cells were seeded into 24-well multitray culture plates and grown at 37°C under 6% CO 2 and humidified conditions. In pretreatment experiments, medium was replaced by 0.5 ml of prewarmed medium containing NECA (10 M) or water (control). At various time points after this, the medium was removed and wells washed 3 times with 1 ml of ice-cold PBS. Following this, 0.5 ml of prewarmed fresh medium (without fetal calf serum) containing the phosphodiesterase inhibitor 4-(3-butoxy-4-methoxybenzyl)-imidazolidin-2-one (Ro 201724, 250 M) was added to each well, followed immediately with either NECA (10 M) or water (control). This was incubated for 20 min at 37°C, and the reaction was terminated by the addition of 20 l of ice-cold trichloroacetic acid (100%). In non-pretreatment experiments, 0.5 ml of prewarmed fresh medium (without fetal calf serum) containing the phosphodiesterase inhibitor Ro 201724 (250 M) was added to each well, followed 15 min later with either NECA (10 M) or water (control). This was incubated for 20 min at 37°C, and the reaction was terminated by the addition of 20 l of ice-cold trichloroacetic acid (100%).
Fifty l of the supernatant was transferred to a fresh tube containing 50 l of NaOH (1 M) and 200 l of Tris (50 mM, pH 7.4) and EDTA (4 mM) (TE buffer). A further 100 l of this mixture was transferred to a fresh tube containing 50 l of TE buffer, 100 l of [ 3 H]cyclic AMP in TE buffer (about 20,000 counts/min), and 100 l of cyclic AMP-binding protein (to give a final concentration of ϳ750 g of protein/ml, prepared from bovine adrenal cortex). Tubes containing 50 l of cyclic AMP (0.125-40 pmol) were used to construct a standard curve. After 90 min incubation at 4°C, 200 l of TE buffer containing charcoal (Norit GSX: 50 mg/ml final concentration) and bovine serum albumin (2 mg ml Ϫ1 final concentration) was added to each tube. After 15 min bound and non-bound [ 3 H]cAMP were separated by centrifugation at 2,900 ϫ g for 15 min at 4°C. The resulting supernatant was transferred into vials for liquid scintillation spectroscopy. Standard curve data were fitted to a logistic expression (GraphPAD Prism) and the unknowns read off. Protein content of the cell monolayers was determined (22) and cAMP accumulation expressed as picomole of cAMP mg Ϫ1 protein, or expressed as a percent of the respective control NECA-stimulated cAMP accumulation.
Agonist-Induced Cell Surface Loss of A 2B -AR Constructs-A 2B -AR cell surface loss was assessed by ELISA as described previously (23). Briefly, cells plated at a density of 6 ϫ 10 5 cells per 60-mm dish were split after 24 h transfection if required (see below) into 24-well tissue culture dishes coated with 0.1 mg ml Ϫ1 poly-L-lysine. 24 h later, cells were incubated with DMEM containing the A 2B -AR agonist NECA (0.1 M to 1 mM) for 0 -60 min at 37°C. Reactions were stopped by removing the media and fixing the cells with 3.7% formaldehyde in TBS (20 mM Tris, pH 7.5, 150 mM NaCl, 20 mM CaCl 2 ) for 5 min at room temperature. Cells were washed three times with TBS, incubated for 45 min with TBS containing 1% BSA (TBS/BSA), then incubated with a primary antibody (anti-HA monoclonal HA.11 antibody, 1:1000 dilution in TBS/BSA) for 1 h at room temperature. Cells were washed three times with TBS, reblocked with TBS/BSA for 15 min at room temperature, and incubated with secondary antibody (goat anti-mouse secondary antibody conjugated with alkaline phosphatase; 1:1000 dilution in TBS/ BSA) for 1 h at room temperature. Cells were washed three times with TBS and a colorimetric alkaline phosphatase substrate added. When adequate color change was achieved, 100 l of sample were added to 100 l of 0.4 M NaOH to terminate the reaction, and the samples were read at 405 nm using a microplate reader.
In order to estimate relative cell surface expression of the different constructs, color development was carried out as above followed by protein assay of cells in each well and results expressed as arbitrary absorbance units mg Ϫ1 protein, with the value for empty pcDNA3-transfected cells (background reading) being subtracted from each of the values obtained for construct-transfected cells.
Immunofluorescence Microscopy and Single Cell Imaging-To assess the cellular distribution of WT and mutant A 2B -AR, CHO cells stably transfected with these constructs were grown in 6-well plates on coverslips. Cells were then incubated with primary antibody (anti-HA monoclonal, 1:1000 dilution) for 1 h at 4°C in DMEM supplemented with 1% BSA. Cells were washed twice with PBS and then incubated at 37°C Ϯ NECA (10 M; 30 min) in DMEM with 0.5% BSA. The cells were then fixed with 3.7% formaldehyde/PBS for 15 min at room temperature, washed with PBS, and permeabilized with 0.05% Triton X-100/ PBS for 10 min at room temperature. Nonspecific binding was blocked with blotto (0.05% Triton X-100/PBS containing 5% nonfat dry milk) for 30 min at 37°C. Goat anti-mouse rhodamine-conjugated secondary antibody was then added at a dilution of 1:150 in blotto for 1 h at 37°C. The cells were then washed six times with 0.05% Triton X-100/PBS and the last wash left for 37°C for 30 min. Finally the cells were fixed again with 3.7% formaldehyde as described. Coverslips were mounted using Slow-Fade mounting medium and examined by microscopy on an upright Leica TCS-NT confocal laser scanning microscope attached to a Leica DM IRBE epifluorescence microscope with phase-contrast and a Plan-Apo 40 ϫ 1.40 NA oil immersion objective. In experiments assessing WT, Gln 325 -stop, and A 2B -AR-GFP receptor redistribution, CHO cells stably expressing WT or Gln 325 -stop receptor were transfected as described above with 5 g of EGFPN1-A 2B -AR. Cells were then split onto coverslips in 6-well plates and 24 h later receptor redistribution assessed as described above.
In experiments assessing arrestin-2-GFP redistribution, CHO cells stably expressing WT or mutant A 2B -AR were transfected as described above with 0.5 g of arrestin-2-GFP and grown on poly-L-lysine-coated coverslips. To assess WT or mutant A 2B -AR distribution, cells were incubated for 30 min at 4°C with rhodamine-conjugated mouse monoclonal anti-HA antibody 12CA5. Cells were then washed 3 times with PBS prior to imaging and coverslips mounted on a heated imaging chamber through which media and drugs could be added. Cells were examined by microscopy on an inverted Leica TCS-NT confocal laser scanning microscope attached to a Leica DM IRBE epifluorescence microscope with phase-contrast and a Plan-Apo 40 ϫ 1.40 NA oil immersion objective. All images were collected on Leica TCS-NT software for two-and three-dimensional image analysis and processed on Adobe Photoshop 5.5.
Data Analysis-Log concentration-effect curves, time-dependent desensitization, and standard curves for cAMP and protein assays were analyzed by the iterative fitting program GraphPAD Prism (GraphPAD Software). Log concentration-effect curves, and standard curve data for cAMP and protein assays were fitted to a logistic expression for singlesite analysis. Values of t 0.5 for agonist-induced desensitization were obtained by fitting data to a single exponential curve.

RESULTS
The wild type (WT) rat A 2B -AR and various COOH terminus deletion mutants (Fig. 1) were constructed by reverse tran-scriptase-polymerase chain reaction and epitope tagged at the NH 2 terminus with a HA sequence, as described under "Experimental Procedures." These constructs were then stably transfected into CHO cells, however, we were unable to obtain CHO cells stably expressing a functional Tyr 299 -stop receptor mutant. The ability of the adenosine receptor agonist NECA to stimulate cAMP formation in the cells was then determined ( Fig. 2A). In CHO cells stably expressing WT receptor as well as those expressing COOH terminus deletions, NECA stimulated cAMP formation with an EC 50 of between 0.25 and 0.64 M, although there was variation in the maximum response observed. CHO cells stably transfected with the pcDNA3 plasmid vector alone did not respond to NECA ( Fig. 2A). Initial experiments to label A 2B -AR with the A 2 -AR radioligand [ 3 H]ZM 241385 (24) failed to detect specific binding in membranes prepared from CHO cells stably transfected with WT A 2B -AR, under conditions where specific binding to membranes prepared from CHO cells stably transfected with WT A 2A -AR was readily observed (data not shown). Therefore, in the ab- sence of a suitable A 2B -AR radioligand, an estimation of the relative expression level of WT and deletion mutant A 2B -AR was obtained using an ELISA assay (23), taking advantage of the HA-epitope tag contained in the NH 2 terminus of the receptor constructs. This analysis indicated that in the absence of agonist stimulation there was little difference in cell surface expression of these constructs (arbitrary absorbance units mg Ϫ1 cell protein: WT 1.37 Ϯ 0.19; Leu 330 -stop 1.12 Ϯ 0.15; Phe 328 -stop 1.19 Ϯ 0.14; Ser 326 -stop 1.32 Ϯ 0.12; Gln 325 -stop 1.07 Ϯ 0.17; n ϭ three separate experiments in each case).
The ability of the A 2B -AR constructs to undergo agonistinduced desensitization was then investigated. Cells were pretreated with 10 M NECA for 1 h, washed, and subsequently incubated with 10 M NECA for 20 min, after which cAMP accumulation was determined. In CHO cells expressing WT receptor, NECA pretreatment for 1 h produced an ϳ50% desensitization in subsequent NECA responsiveness, as compared with cells not preincubated with NECA (Fig. 2B). Similarly, in cells expressing the Leu 330 -stop receptor mutant, there was marked desensitization of the NECA response. On the other hand, cells expressing the Gln 325 -stop or Phe 328 -stop receptor mutants did not display any desensitization subsequent to the 1-h NECA pretreatment. In order to confirm that the differences in desensitization were not due to variations in receptor expression levels, CHO cells were transiently transfected with WT or Gln 325 -stop A 2B -AR constructs. In this case, agonist-stimulated cAMP formation in non-pretreated cells in response to NECA was greater in WT than Gln 325 -stop cells, but there was still no desensitization in the latter (NECAstimulated cAMP formation in non-pretreated WT cells was 185 Ϯ 36 and in pretreated cells 62 Ϯ 10 pmol of cyclic AMP mg Ϫ1 protein; NECA-stimulated cAMP formation in non-pretreated Gln 325 -stop cells was 86 Ϯ 23 and in pretreated cells 85 Ϯ 29 pmol of cyclic AMP mg Ϫ1 protein). Taken together these results indicate that the Ser 329 -Leu 330 motif close to the COOH terminus of the rat A 2B -AR is critical for rapid agonistinduced desensitization. However, the Ser 326 -stop receptor mutant was also subjected to agonist-induced desensitization, which was unexpected since this mutant does not contain the Ser 329 -Leu 330 motif. We next compared the rates of agonistinduced desensitization for the WT, Leu 330 -stop and Ser 326stop receptors (Fig. 2C). This indicated that the WT receptor desensitizes with a t 0.5 of 12.0 min, with the Leu 330 -stop receptor desensitizing faster (t 0.5 of 7.5 min) and the Ser 326 -stop receptor slower (t 0.5 of 22.0 min).
The ability of the WT and deletion mutant A 2B -AR constructs to undergo agonist-induced internalization was next investigated by ELISA (23). In CHO cells expressing WT receptor, addition of 10 M NECA induced a time-dependent loss of cell surface receptor, reaching about 50% loss after 1 h (Fig. 3A). Similarly, in cells expressing the Leu 330 -stop receptor mutant, there was marked agonist-induced loss of cell surface receptor, which appeared to be somewhat more rapid than WT. However, in cells expressing the Gln 325 -stop or Phe 328 -stop receptor mutants there was very little agonist-induced loss of cell surface receptor over a 1-h period. On the other hand the Ser 326stop receptor mutant did show agonist-induced loss of cell surface receptor, with around 35% loss after 1 h of agonist treatment. We next determined whether the poor internalization of the Gln 325 -stop receptor mutant could be rescued by overexpression of non-visual arrestins (Fig. 3B). Transient overexpression of arrestin-2-GFP or arrestin-3-GFP in WT receptor cells produced a modest increase in agonist-induced receptor loss at 15 min, but did not affect the overall extent of receptor loss. However, overexpression of non-visual arrestins did not appear to affect the rate or extent of agonist-induced loss of cell surface Gln 325 -stop receptor mutant.
To confirm our findings in the ELISA assays, imaging of the receptors by confocal microscopy was undertaken using an anti-HA primary antibody and rhodamine-linked secondary antibody (Fig. 4). Under basal conditions, the WT and mutant receptors all localized to the plasma membrane of CHO cells (left panels). After 30 min treatment with 10 M NECA, the cellular distribution of WT receptor and Leu 330 -stop receptor mutant had clearly changed to a punctate intracellular location. The internalized WT and Leu 330 -stop constructs co-localized with transferrin (data not shown), consistent with internalization to an endosomal compartment. On the other hand, a 30-min NECA pretreatment of the Gln 325 -stop or Phe 328 -stop receptor mutants did not lead to any detectable internalization. Interestingly, the Ser 326 -stop receptor mutant, which displayed agonist-induced cell surface loss in the ELISA assay (Fig. 3A), did not undergo translocation to endosomal compartments deep within the cell as had been observed with the WT and Leu 330 -stop receptor mutant. Instead there appeared to be redistribution of the mutant receptor very close to the cell surface. To confirm the difference in internalization between the WT and Gln 325 -stop receptor mutant, either WT or Gln 325stop CHO cells were transiently transfected with an A 2B -AR-GFP construct which did not contain the NH 2 terminus HAepitope tag (in separate experiments the ability of NECA to stimulate cAMP accumulation in non-transfected CHO cells transiently transfected with either A 2B -AR-GFP or HA-tagged WT A 2B -AR construct was the same indicating that the epitope tags did not affect acute receptor coupling; EC 50 for NECA was 0.31 Ϯ and 0.23 Ϯ M, and maximal cAMP stimulation was 268 Ϯ 36 and 287 Ϯ 64 pmol mg Ϫ1 protein for A 2B -AR-GFP or HA-tagged WT A 2B -AR constructs, respectively; n ϭ four separate experiments). Under basal conditions WT, Gln 325 -stop and A 2B -AR-GFP all localized to the cell surface (not shown). In cells stably transfected with WT receptor and transiently transfected with A 2B -AR-GFP, both constructs internalized and co-localized to the same intracellular compartment upon agonist treatment (top panel of Fig. 5, co-localization shown by orange overlay color in the top right panel). On the other hand, in cells stably transfected with Gln 325 -stop receptor and transiently transfected with A 2B -AR-GFP, agonist treatment led to the internalization of A 2B -AR-GFP, but not Gln 325 -stop.
It is possible that the lack of internalization of the Gln 325stop and Phe 328 -stop receptor mutants reflects their inability to interact with arrestin proteins. To investigate this, we assessed the ability of WT and mutant receptors to promote translocation of arrestin-2-GFP from cytosol to cell membrane upon agonist activation. In WT and Leu 330 -stop receptor CHO cells transiently transfected with arrestin-2-GFP, agonist activation promoted the rapid translocation of arrestin-2-GFP from cell cytosol to plasma membrane (Fig. 6), clearly observable within 30 s of agonist addition. On the other hand, agonist activation of Gln 325 -stop, Phe 328 -stop, and Ser 326 -stop receptor mutants for up to 4 min did not lead to any observable arrestin-2-GFP translocation.
To determine the mechanism of agonist-induced internalization of WT and Ser 326 -stop receptors, CHO cells expressing these receptor constructs were transiently transfected with dominant negative mutant constructs of arrestin-2, dynamin, or Eps-15 (arrestin-2-(319 -418), dynamin K44A, and E⌬95-295, respectively), to determine the role of endogenous arrestins, dynamin or Eps15 proteins in agonist-induced internalization of these receptors, as assessed by the ELISA assay. Interestingly, whereas agonist-induced loss of cell surface WT receptor was blocked by all three DNM constructs, loss of cell surface Ser 326 -stop receptor was only blocked by dyn-DNM ( Fig. 7; note that coexpression of dyn-DNM did not affect basal cell surface expression of Ser 326 -stop; arbitrary absorbance units mg Ϫ1 cell protein: WT 1.84 Ϯ 0.31 and 1.64 Ϯ 0.25 and Ser 326 -stop 1.31 Ϯ 0.24 and 1.32 Ϯ 0.23 in the absence or presence of transient dyn-DNM expression, respectively, n ϭ three separate experiments). This indicates that WT and Ser 326 -stop receptors internalize by different mechanisms; WT receptor internalizes by an arrestin/dynamin/clathrin-dependent mechanism, whereas the Ser 326 -stop receptor mutant appears to internalize in a dynamin-dependent but arrestin-and clathrin-independent mechanism.
On the basis of the results obtained with the deletion mutants, we engineered and investigated the function of two COOH terminus point mutant receptor constructs, S329G and S326A (Fig. 1). In CHO cells stably expressing these point mutants, NECA stimulated cAMP formation with EC 50 values of 0.30 and 1.0 M for S329G and S326A, respectively (n ϭ 3 separate experiments), similar to the other constructs studied. Furthermore, relative cell surface receptor expression as assessed by ELISA was similar to WT A 2B -AR expression (arbitrary absorbance units mg Ϫ1 cell protein: WT 1.37 Ϯ 0.19; S329G 1.26 Ϯ 0.18; S326A 1.22 Ϯ 0.10; n ϭ three separate experiments for each). As compared with WT A 2B -AR, the S329G mutant was resistant to rapid agonist-induced desensitization (Fig. 8A) and internalization (Fig. 8, B and C). On the other hand, the S326A mutant underwent rapid agonist-induced desensitization and internalization (Fig. 8). Furthermore, the internalized S326A construct co-localized with transferrin (data not shown), indicating that the phenotype of this point mutant is the same as the WT receptor.

DISCUSSION
The agonist-induced rapid desensitization and internalization of a number of GPCRs, such as the ␤ 2 -adrenoreceptor (25), the AT 1A -angiotensin receptor (26), and the protease-activated receptor 1 (27) are dependent upon the integrity of the COOH terminus tail of the receptor. In many cases it is likely that GRK-mediated phosphorylation of serine or threonine residues in the receptor leads to non-visual arrestin binding with consequent uncoupling of the receptor from G protein and targeting of the receptor to clathrin-coated pits for internalization (1)(2)(3)(4)(5). Although the agonist-induced GRK2-and arrestin-dependent desensitization and internalization of the A 2B -AR has been reported previously (12,15,18), it is not known which regions of the receptor are responsible for triggering these processes. Since the intracellular COOH terminus tail plays a critical role in the agonist-induced regulation of a number of GPCRs, we chose in the present study to investigate the role of this region of the receptor in the rapid agonist-induced desensitization and internalization of the rat A 2B -AR.
The deletion mutants (see Fig. 1) were designed primarily to investigate the role of the terminal few amino acids in these processes, since the final 7 amino acids contain 4 potential phosphoacceptor sites (serine/threonine residues), and phos- phorylation appears to be critical for the desensitization and internalization of many GPCRs (2). In addition, we also examined the Tyr 299 -stop mutant where the final 33 residues of the COOH terminus had been deleted. However, we were unable to obtain stably transfected CHO cells functionally expressing the latter mutant. When transiently expressed in HEK293 cells, the Tyr 299 -stop mutant was visualized by confocal microscopy at the cell surface but did not internalize in response to agonist. 2 Other GPCRs with large COOH terminus deletions have also been reported to be non-functional or poorly expressed (28). Apart from the Tyr 299 -stop mutant, we were able to isolate CHO cell lines stably expressing the WT A 2B -AR, as well as the deletion mutants Leu 330 -stop, Phe 328 -stop, Ser 326 -stop, and Gln 325 -stop. In response to the adenosine receptor agonist NECA, these receptors all coupled to cAMP formation with an EC 50 of just under 1 M, close to the value of 1.4 M reported for NECA acting at WT human A 2B -AR stably expressed in CHO cells (29). However, the maximum response generated for each receptor construct was somewhat different, which may reflect differences in receptor expression between various cell clones. We attempted to assess this with a ligand binding assay using [ 3 H]ZM 241385, a ligand with high affinity for the A 2A -AR but which can also label A 2B -AR at higher concentrations (K d 33 nM (24)). However, in preliminary experiments on WT A 2B cells we found that very high levels of nonspecific binding made detection of A 2B -ARs by this method impossible. Instead, an ELISA assay was employed to assess the relative cell surface expres-

FIG. 5. Agonist-induced cellular redistribution of transiently transfected GFP-tagged WT A 2B -AR in CHO cells stably transfected with either WT or Gln 325 -stop A 2B -AR. CHO cells stably expressing WT or Gln 325 -stop
A 2B -AR were transiently transfected with 2 g of eGFP-tagged WT A 2B -AR. Two days later cells were preincubated with an anti-HA antibody at 4°C for 1 h. Subsequently cells were incubated at 37°C for 30 min in the presence of agonist (NECA; 10 M). Receptor localization was determined by immunofluorescence in fixed cells as described under "Experimental Procedures." In these figures HAtagged and GFP-tagged receptor are red and green, respectively. Marked co-localization (orange/yellow) of stably expressed WT but not Gln 325 -stop A 2B -AR with transiently expressed GFP-A 2B -AR is clearly visible in the overlay after agonist stimulation.

FIG. 6. Analysis of agonist-induced translocation of arrestin-2-GFP in CHO cells stably expressing WT or COOH terminus deletion mutant A 2B -AR.
CHO cells stably expressing WT or COOH terminus deleted mutant A 2B -AR were transiently transfected with 0.5 g of arrestin-2-GFP and then split onto poly-L-lysine-coated coverslips. A 2B -AR was visualized using a rhodamine-conjugated anti-HA antibody (all cells shown expressed receptor). Prior to stimulation and viewing, coverslips were mounted in a chamber at 37°C as described under "Experimental Procedures." The initial diffuse cytoplasmic distribution of arr-2-GFP is shown prior to agonist stimulation (0 s). NECA (10 M) was added and the redistribution of arrestins was monitored in real time. Images were obtained at the times shown after agonist addition. Agonist-induced translocation of arrestin-2-GFP was only observed in WT and Leu 330 -stop mutant receptors. sion of the various constructs. This indicated no significant variation in expression of any of the constructs studied, and thus any differences in the desensitization/internalization profile behavior of constructs in response to agonist cannot be due to differences in cell surface receptor expression levels, but must instead be a property of the expressed receptor protein.
In agonist-induced desensitization experiments, the WT receptor response was desensitized by over 50% after a 1-h pretreatment with agonist, consistent with previous studies (12)(13)(14). Likewise the Leu 330 -stop mutant was desensitized by over 50% after a 1-h of NECA pretreatment. This suggests that residues Ser 331 and the terminal Leu 332 are not necessary for rapid agonist-induced desensitization of the receptor. On the other hand, the Phe 328 -and Gln 325 -stop mutants were resistant to rapid agonist-induced desensitization, suggesting that the motif critical for agonist-induced desensitization of the A 2B -AR is Ser 329 -Leu 330 . Since the Phe 328 -and Gln 325 -stop mutants were both resistant to desensitization, but showed, respectively, the lowest and highest maximum responses to NECA (Fig. 2B), then it seems highly unlikely that variations in maximum receptor response could explain the differences in desensitization observed between the various constructs. This was confirmed in CHO cells transiently transfected with WT or Gln 325 -stop A 2B -AR constructs. In this case the NECA-stimulated cAMP in non-pretreated cells was greater for WT than Gln 325 -stop, but the latter remained refractory to desensitization. Unexpectedly, the Ser 326 -stop mutant, which also does not contain the Ser 329 -Leu 330 motif, was subject to rapid agonistinduced desensitization. When directly compared (Fig. 2C), the desensitization of the Ser 326 -stop mutant was slower and less extensive than the WT receptor. However, the WT receptor desensitized less rapidly but to the same extent as the Leu 330stop mutant, indicating that deletion of the COOH terminus of the receptor can determine not only whether or not rapid agonistinduced desensitization occurs, but also its rate and extent.
We next examined the ability of the A 2B -AR constructs to undergo agonist-induced internalization, initially employing an ELISA assay. The results obtained reflected the behavior of the constructs in the desensitization assays. Whereas the agonist-induced surface receptor loss of WT receptor and the Leu 330 -and Ser 326 -stop mutants was rapid and reached around 50% after 1 h of NECA treatment, the Phe 328 -and Gln 325 -stop mutants were resistant to internalization. Thus for the A 2B -AR the same region of the COOH terminus (Ser 329 -Leu 330 ) is critical for both rapid agonist-induced desensitization and internalization. Comparison of desensitization (Fig. 2C) and internalization (Fig. 3A) time courses for WT receptor, Leu 330 -stop mutant, and Ser 326 -stop mutant underlines this relationship since the Leu 330 -stop mutant desensitizes and internalizes somewhat more rapidly than the other two constructs, whereas the Ser 326 -stop mutant is the slowest. Since the internalization of the A 2B -AR is arrestin-dependent (15), then it is possible that the inability of the Phe 328 -and Gln 325 -stop mutants to undergo rapid agonist-induced desensitization and internalization reflects an inability to interact with arrestins. In previous studies with other GPCRs the ability of mutant receptors to undergo these processes can sometimes be rescued by overexpressing non-visual arrestins in the same cell (30). Consequently, to assess whether the internalization of the Gln 325stop receptor could be rescued by non-visual arrestins, WT or Gln 325 -stop cells were transiently transfected with GFP-tagged arrestin-2 or arrestin-3. This did not increase the extent of agonist-induced surface receptor loss of WT receptor, indicating that in CHO cells, arrestin levels are not rate-limiting for the maximum extent of internalization. Furthermore, coexpression of GFP-tagged arrestin-2 or arrestin-3 failed to enhance the internalization of the Gln 325 -stop mutant, suggesting that the distal COOH terminus of the A 2B -AR represents a primary site for arrestin interaction.
To further substantiate these results, we directly visualized the receptor constructs by confocal microscopy. Following treatment with agonist for 30 min, and in agreement with the ELISA assays, we found that the WT and Leu 330 -stop constructs internalized extensively, taking on a punctate distribution consistent with endosomal location, as we have previously reported for the WT receptor (18). On the other hand, the Phe 328 -and Gln 325 -stop mutants did not undergo any internalization over the time period examined. The difference between the WT and Gln 325 -stop mutant was underlined by transiently transfecting WT or Gln 325 -stop cells with an A 2B -AR-GFP construct. Whereas in the WT cells, both WT receptor and A 2B -AR-GFP underwent agonist-induced internalization to similar intracellular compartments, in the Gln 325 -stop cells the A 2B -AR-GFP construct underwent internalization whereas the Gln 325 -stop receptor remained at the cell surface. On the other hand, visualization of the Ser 326 -stop mutant indicated that this construct underwent limited internalization to a compartment close to the plasma membrane, distinct from that seen with the WT and Leu 330 -stop constructs.
The ability of GPCR activation to promote the translocation of arrestin-2-GFP from cytosol to cell membrane has been reported for many GPCRs (19) including the A 2B -AR (18). To further investigate the interaction of the A 2B -AR constructs with non-visual arrestins, we transiently transfected WT and mutant A 2B -AR cells with arrestin-2-GFP and visualized translocation of the latter following agonist activation of the receptor. Under basal conditions, arrestin-2-GFP was evenly distributed throughout the cytosol of WT and mutant receptor cells. However, upon agonist activation with 10 M NECA, arrestin-2-GFP rapidly translocated from cytosol to cell membrane in WT and Leu 330 -stop cells only. No redistribution of arrestin-2-GFP was observed upon agonist activation of the Phe 328 -, Ser 326 -, and Gln 325 -stop mutants, and thus these constructs appear unable to interact with non-visual arrestins. In order to further identify the region of the A 2B -AR COOH terminus which is critical for receptor regulation, we constructed the point mutant S329G. Since this construct was resistant to rapid agonist-induced desensitization/internalization, we can conclude that Ser 329 , four residues from the extreme COOH terminus of the A 2B -AR, is the critical residue for rapid agonistinduced regulation of this receptor. Interestingly, a single threonine residue in the COOH terminus of the canine A 2A -AR was found to be critical for rapid agonist-induced desensitization, although internalization was not examined in that study (20). Although the Ser 326 -stop receptor mutant undergoes rapid agonist-induced desensitization and internalization as with the WT receptor and Leu 330 -stop receptor mutant, it apparently does so by an arrestin-independent mechanism. This was confirmed by experiments in which WT and Ser 326 -stop receptor mutant cells were transiently transfected with dominant negative mutant forms of arrestin-2, dynamin, and Eps15. Arrestin-2-(319 -418) contains the clathrin-binding domain of arrestin-2, but not the putative receptor interaction site; it constitutively interacts with clathrin and blocks the internalization of a number of GPCRs including the ␤ 2 -AR (31). Dynamin K44A does not express the GTPase activity required in this protein to promote the formation of clathrin-coated vesicles (32), and inhibits the internalization of many GPCRs (19). The Eps15 protein binds to the clathrin adaptor protein AP-2 and is required for normal formation of clathrin-coated pits (33); the E⌬95-295 mutant blocks clathrin-coated pit formation and the internalization of transferrin receptors (34). When these constructs were transiently expressed in WT receptor cells, each markedly reduced the agonist-induced loss of cell surface receptor, confirming the arrestin/dynamin/clathrin dependence of this process, in line with previous findings by ourselves (18) and others (19). Interestingly, although the internalization of the Ser 326 -stop mutant was blocked by dynamin K44A, it was unaffected by arrestin-2-(319 -418) and E⌬95-295, indicating that the agonist-induced internalization of the Ser 326 -stop mutant is not mediated by arrestins or clathrin-coated pits. Other mechanisms of internalization that are clathrin-independent but dynamin-dependent could include caveolae-mediated internalization (35). To further determine the importance of Ser 326 in A 2B -AR regulation, we constructed a S326A point mutant. However, when stably expressed in CHO cells this mutant displayed the same desensitization/internalization phenotype as the WT and Leu 330 -stop receptor constructs, indicating that Ser 326 only directs the receptor into the non-arrestin/clathrin internalization pathway when the TFSLSL motif distal to Ser 326 is absent. Importantly, these results indicate that care must be taken when interpreting the results of internalization experiments with GPCR COOH terminus mutants; modifications of the COOH terminus can alter not only the rate and extent of internalization, but also the mechanism, as demonstrated here. In conclusion, we have shown that the rapid agonist-induced desensitization and internalization of the A 2B -AR is dependent upon a serine residue (Ser 329 ) close to the end of the receptor COOH terminus. It is likely that this residue represents an arrestin interaction site, and future studies will be directed toward characterizing the possible interaction of non-visual arrestins with this region of the A 2B -AR.